hypothesis about fingerprints

Every print subscription comes with full digital access

Science News

How fingerprints form was a mystery — until now.

A theory proposed by mathematician Alan Turing in the 1950s helps explain the process

Three of the most common fingerprint shapes — arch, loop and whorl (traced in purple) — can be explained in part by a theory proposed by British mathematician Alan Turing.

J. Glover et al / Cell 2023

Share this:

By McKenzie Prillaman

February 9, 2023 at 11:00 am

Scientists have finally figured out how those arches, loops and whorls formed on your fingertips.

While in the womb, fingerprint-defining ridges expand outward in waves starting from three different points on each fingertip. The raised skin arises in a striped pattern thanks to interactions between three molecules that follow what’s known as a Turing pattern, researchers report February 9 in Cell . How those ridges spread from their starting sites — and merge — determines the overarching fingerprint shape.

Fingerprints are unique and last for a lifetime. They’ve been used to identify individuals since the 1800s. Several theories have been put forth to explain how fingerprints form, including spontaneous skin folding, molecular signaling and the idea that ridge pattern may follow blood vessel arrangements.

Scientists knew that the ridges that characterize fingerprints begin to form as downward growths into the skin, like trenches. Over the few weeks that follow, the quickly multiplying cells in the trenches start growing upward, resulting in thickened bands of skin.

Since budding fingerprint ridges and developing hair follicles have similar downward structures, researchers in the new study compared cells from the two locations. The team found that both sites share some types of signaling molecules — messengers that transfer information between cells — including three known as WNT, EDAR and BMP. Further experiments revealed that WNT tells cells to multiply, forming ridges in the skin, and to produce EDAR, which in turn further boosts WNT activity. BMP thwarts these actions.

To examine how these signaling molecules might interact to form patterns, the team adjusted the molecules’ levels in mice. Mice don’t have fingerprints, but their toes have striped ridges in the skin comparable to human prints. “We turn a dial — or molecule — up and down, and we see the way the pattern changes,” says developmental biologist Denis Headon of the University of Edinburgh.

Increasing EDAR resulted in thicker, more spaced-out ridges, while decreasing it led to spots rather than stripes. The opposite occurred with BMP, since it hinders EDAR production.

That switch between stripes and spots is a signature change seen in systems governed by Turing reaction-diffusion, Headon says. This mathematical theory, proposed in the 1950s by British mathematician Alan Turing, describes how chemicals interact and spread to create patterns seen in nature ( SN: 7/2/10 ). Though, when tested, it explains only some patterns ( SN: 1/21/14 ).

Mouse digits, however, are too tiny to give rise to the elaborate shapes seen in human fingerprints. So, the researchers used computer models to simulate a Turing pattern spreading from the three previously known ridge initiation sites on the fingertip: the center of the finger pad, under the nail and at the joint’s crease nearest the fingertip.

By altering the relative timing, location and angle of these starting points, the team could create each of the three most common fingerprint patterns — arches, loops and whorls — and even rarer ones. Arches, for instance, can form when finger pad ridges get a slow start, allowing ridges originating from the crease and under the nail to occupy more space.

“It’s a very well-done study,” says developmental and stem cell biologist Sarah Millar, director of the Black Family Stem Cell Institute at the Icahn School of Medicine at Mount Sinai in New York City.

Controlled competition between molecules also determines hair follicle distribution, says Millar, who was not involved in the work. The new study, she says, “shows that the formation of fingerprints follows along some basic themes that have already been worked out for other types of patterns that we see in the skin.”

Millar notes that people with gene mutations that affect WNT and EDAR have skin abnormalities. “The idea that those molecules might be involved in fingerprint formation was floating around,” she says.

Overall, Headon says, the team aims to aid formation of skin structures, like sweat glands, when they’re not developing properly in the womb, and maybe even after birth.

“What we want to do, in broader terms, is understand how the skin matures.”

More Stories from Science News on Health & Medicine

California droughts may help valley fever spread

The first face transplant to include an eye shows no rejection a year later

50 years ago, some of plastic’s toxic hazards were exposed

A new drug shows promise for hot flashes due to menopause

A next-gen pain drug shows promise, but chronic sufferers need more options

What is ‘Stage 0’ breast cancer and how is it treated?

People with food and other allergies have a new way to treat severe reactions

Extreme heat and rain are fueling rising cases of mosquito-borne diseases

Subscribers, enter your e-mail address for full access to the Science News archives and digital editions.

Not a subscriber? Become one now .

The Myth of Fingerprints

Police today increasingly embrace DNA tests as the ultimate crime-fighting tool. They once felt the same way about fingerprinting

Clive Thompson

At 9:00 a.m. last December 14, a man in Orange County, California, discovered he’d been robbed. Someone had swiped his Volkswagen Golf, his MacBook Air and some headphones. The police arrived and did something that is increasingly a part of everyday crime fighting: They swabbed the crime scene for DNA.

Normally, you might think of DNA as the province solely of high-profile crimes—like murder investigations, where a single hair or drop of blood cracks a devilish case. Nope: These days, even local cops are wielding it to solve ho-hum burglaries. The police sent the swabs to the county crime lab and ran them through a beige, photocopier-size “rapid DNA” machine, a relatively inexpensive piece of equipment affordable even by smaller police forces. Within minutes, it produced a match to a local man who’d been previously convicted of identity theft and burglary. They had their suspect.

DNA identification has gone mainstream—from the elite labs of “CSI” to your living room. When it first appeared over 30 years ago, it was an arcane technique. Now it’s woven into the fabric of everyday life: California sheriffs used it to identify the victims of their recent wildfires, and genetic testing firms offer to identify your roots if you mail them a sample.

Yet the DNA revolution has unsettling implications for privacy. After all, you can leave DNA on everything you touch—which means, sure, crimes can be more easily busted, but the government can also more easily track you. And while it’s fun to learn about your genealogy, your cheek samples can wind up in places you’d never imagine. FamilyTreeDNA, a personal genetic service, in January admitted it was sharing DNA data with federal investigators to help them solve crimes. Meanwhile consumer DNA testing firm 23andMe announced that it was now sharing samples sent to them with the pharmaceutical giant GlaxoSmithKline to make “novel treatments and cures.”

What happens to a society when there’s suddenly a new way to identify people—to track them as they move around the world? That’s a question that the denizens of the Victorian turn of the century pondered, as they learned of a new technology to hunt criminals: fingerprinting.

For centuries, scholars had remarked on the curious loops and “whorls” that decorated their fingertips. In 1788, the scientist J.C.A. Mayers declared that patterns seemed unique—that “the arrangement of skin ridges is never duplicated in two persons.”

It was an interesting observation, but one that lay dormant until 19th-century society began to grapple with an emerging problem: How do you prove people are who they say they are?

Carrying government-issued identification was not yet routine, as Colin Beavan, author of Fingerprints , writes. Cities like London were booming, becoming crammed full of strangers—and packed full of crime. The sheer sprawl of the population hindered the ability of police to do their work because unless they recognized criminals by sight, they had few reliable ways of verifying identities. A first-time offender would get a light punishment; a habitual criminal would get a much stiffer jail sentence. But how could the police verify whether a perpetrator they hauled in had ever been caught previously? When recidivists got apprehended, they’d just give out a fake name and claim it was their first crime.

“A lot of that is the function of the increasing anonymity of modern life,” notes Charles Rzepka, a Boston University professor who studies crime fiction. “There’s this problem of what Edgar Allan Poe called ‘The Man of the Crowd.’” It even allowed for devious cons. One man in Europe claimed to be “Roger Tichborne,” a long-lost heir to a family baronetcy, and police had no way to prove he was or wasn’t.

Faced with this problem, police tried various strategies for identification. Photographic mug shots helped, but they were painstakingly slow to search through. In the 1880s, a French police official named Alphonse Bertillon created a system for recording 11 body measurements of a suspect, but it was difficult to do so accurately.

The idea of fingerprints gradually dawned on several different thinkers. One was Henry Faulds, a Scottish physician who was working as a missionary in Japan in the 1870s. One day while sifting through shards of 2,000-year-old pottery, he noticed that the ridge patterns of the potter’s ancient fingerprints were still visible. He began inking prints of his colleagues at the hospital—and noticing they seemed unique. Faulds even used prints to solve a small crime. An employee was stealing alcohol from the hospital and drinking it in a beaker. Faulds located a print left on the glass, matched it to a print he’d taken from a colleague, and—presto—identified the culprit.

How reliable were prints, though? Could a person’s fingerprints change? To find out, Faulds and some students scraped off their fingertip ridges, and discovered they grew back in precisely the same pattern. When he examined children’s development over two years, Faulds found their prints stayed the same. By 1880 he was convinced, and wrote a letter to the journal Nature arguing that prints could be a way for police to deduce identity.

“When bloody finger-marks or impressions on clay, glass, etc., exist,” Faulds wrote, “they may lead to the scientific identification of criminals.”

Other thinkers were endorsing and exploring the idea—and began trying to create a way to categorize prints. Sure, fingerprints were great in theory, but they were truly useful only if you could quickly match them to a suspect.

The breakthrough in matching prints came from Bengal, India. Azizul Haque, the head of identification for the local police department, developed an elegant system that categorized prints into subgroups based on their pattern types such as loops and whorls. It worked so well that a police officer could find a match in only five minutes—much faster than the hour it would take to identify someone using the Bertillon body-measuring system. Soon, Haque and his superior Edward Henry were using prints to identify repeat criminals in Bengal “hand over fist,” as Beavan writes. When Henry demonstrated the system to the British government, officials were so impressed they made him assistant commissioner of Scotland Yard in 1901.

Fingerprinting was now a core tool in crime-busting. Mere months after Henry set up shop, London officers used it to fingerprint a man they’d arrested for pickpocketing. The suspect claimed it was his first offense. But when the police checked his prints, they discovered he was Benjamin Brown, a career criminal from Birmingham, who’d been convicted ten times and printed while in custody. When they confronted him with their analysis, he admitted his true identity. “Bless the finger-prints,” Brown said, as Beavan writes. “I knew they’d do me in!”

Within a few years, prints spread around the world. Fingerprinting promised to inject hard-nosed objectivity into the fuzzy world of policing. Prosecutors historically relied on witness testimony to place a criminal in a location. And testimony is subjective; the jury might not find the witness credible. But fingerprints were an inviolable, immutable truth, as prosecutors and professional “fingerprint examiners” began to proclaim.

“The fingerprint expert has only facts to consider; he reports simply what he finds. The lines of identification are either there or they are absent,” as one print examiner argued in 1919.

This sort of talk appealed to the spirit of the age—one where government authorities were keen to pitch themselves as rigorous and science-based.

“It’s this turn toward thinking that we have to collect detailed data from the natural world—that these tiniest details could be more telling than the big picture,” says Jennifer Mnookin, dean of the UCLA law school and an expert in evidence law. Early 20th-century authorities increasingly believed they could solve complex social problems with pure reason and precision. “It was tied in with these ideas of science and progressivism in government, and having archives and state systems of tracking people,” says Simon Cole, a professor of criminology, law, and society at the University of California, Irvine, and the author of Suspect Identities , a history of fingerprinting.

Prosecutors wrung high drama out of this curious new technique. When Thomas Jennings in 1910 was the first U.S. defendant to face a murder trial that relied on fingerprinted evidence, prosecutors handed out blown-up copies of the prints to the jury. In other trials, they would stage live courtroom demonstrations of print-lifting and print-matching. It was, in essence, the birth of the showily forensic policing that we now see so often on “CSI”-style TV shows: perps brought low by implacably scientific scrutiny. Indeed, criminals themselves were so intimidated by the prospect of being fingerprinted that, in 1907, a suspect arrested by Scotland Yard desperately tried to slice off his own prints while in the paddy wagon.

Yet it also became clear, over time, that fingerprinting wasn’t as rock solid as boosters would suggest. Police experts would often proclaim in court that “no two people have identical prints”—even though this had never been proven, or even carefully studied. (It’s still not proven.)

Although that idea was plausible, “people just asserted it,” Mnookin notes; they were eager to claim the infallibility of science. Yet quite apart from these scientific claims, police fingerprinting was also simply prone to error and sloppy work.

The real problem, Cole notes, is that fingerprinting experts have never agreed on “a way of measuring the rarity of an arrangement of friction ridge features in the human population.” How many points of similarity should two prints have before the expert analyst declares they’re the same? Eight? Ten? Twenty? Depending on what city you were tried in, the standards could vary dramatically. And to make matters more complex, when police lift prints from a crime scene, they are often incomplete and unclear, giving authorities scant material to make a match.

So even as fingerprints were viewed as unmistakable, plenty of people were mistakenly sent to jail. Simon Cole notes that at least 23 people in the United States have been wrongly connected to crime-scene prints.* In North Carolina in 1985, Bruce Basden was arrested for murder and spent 13 months in jail before the print analyst realized he’d made a blunder.

Nonetheless, the reliability of fingerprinting today is rarely questioned in modern courts. One exception was J. Spencer Letts, a federal judge in California who in 1991 became suspicious of fingerprint analysts who’d testified in a bank robbery trial. Letts was astounded to hear that the standard for declaring that two prints matched varied widely from county to county. Letts threw out the fingerprint evidence from that trial.

“I don’t think I’m ever going to use fingerprint testimony again,” he said in court, sounding astonished, as Cole writes. “I’ve had my faith shaken.” But for other judges, the faith still holds.

The world of DNA identification, in comparison, has received a slightly higher level of skepticism. When it was first discovered in 1984, it seemed like a blast of sci-fi precision. Alec Jeffreys, a researcher at the University of Leicester in England, had developed a way to analyze pieces of DNA and produce an image that, Jeffreys said, had a high likelihood of being unique. In a splashy demonstration of his concept, he found that the semen on two murder victims wasn’t from the suspect police had in custody.

DNA quickly gained a reputation for helping free the wrongly accused: Indeed, the nonprofit Innocence Project has used it to free over 360 prisoners by casting doubt on their convictions. By 2005, Science magazine said DNA analysis was the “gold standard” for forensic evidence.

Yet DNA identification, like fingerprinting, can be prone to error when used sloppily in the field. One problem, notes Erin Murphy, professor of criminal law at New York University School of Law, is “mixtures”: If police scoop up genetic material from a crime scene, they’re almost certain to collect not just the DNA of the offender, but stray bits from other people. Sorting relevant from random is a particular challenge for the simple DNA identification tools increasingly wielded by local police. The rapid-typing machines weren’t really designed to cope with the complexity of samples collected in the field, Murphy says—even though that’s precisely how some police are using them.

“There’s going to be one of these in every precinct and maybe in every squad car,” Murphy says, with concern. When investigating a crime scene, local police may not have the training to avoid contaminating their samples. Yet they’re also building up massive databases of local citizens: Some police forces now routinely request a DNA sample from everyone they stop, so they can rule them in or out of future crime investigations.

The courts have already recognized the dangers of badly managed DNA identification. In 1989—only five years after Jeffreys invented the technique—U.S. lawyers successfully contested DNA identification in court, arguing that the lab processing the evidence had irreparably contaminated it. Even the prosecution agreed it had been done poorly. Interestingly, as Mnookin notes, DNA evidence received pushback “much more quickly than fingerprints ever did.”

It even seems the public has grasped the dangers of its being abused and misused. Last November, a jury in Queens, New York, deadlocked in a murder trial—after several of them reportedly began to suspect the accused’s DNA had found its way onto the victim’s body through police contamination. “There is a sophistication now among a lot of jurors that we haven’t seen before,” Lauren-Brooke Eisen, a senior fellow at the Brennan Center for Justice, told the New York Times .

To keep DNA from being abused, we’ll have to behave like good detectives—asking the hard questions, and demanding evidence.

*Editor's Note, April 26, 2019: An earlier version of this story incorrectly noted that at least 23 people in the United States had been imprisoned after being wrongly connected to crime-scene prints. In fact, not all 23 were convicted or imprisoned. This story has been edited to correct that fact. Smithsonian regrets the error.

Body of Evidence

Now science can identify you by your ears, your walk and even your scent Research by Sonya Maynard

Get the latest Science stories in your inbox.

Clive Thompson | | READ MORE

Clive Thompson is author of Smarter Than You Think: How Technology is Changing Our Minds for the Better and Coders: The Making of a New Tribe and the Remaking of the World . He is a contributing writer to the New York Times Magazine and Wired . Photo: Tom Igoe.

It’s a wonderful world — and universe — out there.

Come explore with us!  

Science News Explores

Experiment: are fingerprint patterns inherited.

Investigate whether fingerprint patterns are random or influenced by genetics

Fingerprints differ from person to person. This science activity can help you determine whether the patterns are random or inherited.

MirageC/Getty Images

Share this:

• Google Classroom

By Science Buddies

April 17, 2023 at 6:30 am

Objective : Collect, categorize and compare the fingerprints of siblings versus unrelated pairs of individuals to determine if fingerprint patterns are inherited.

Areas of science : Genetics & Genomics

Difficulty : Hard intermediate

Time required : 2–5 days

Prerequisites :

• Basic understanding of genetic inheritance
• Consent forms must be signed for each person participating in this experiment. You should inform people that although fingerprints can be used as forms of identification, you will assign their fingerprints a code and not use their name so that the fingerprints remain anonymous. For children under the age of 18, parents must grant consent.

Material availability : Readily available

Cost : Very low (under $20)

Safety : No issues

Credits : Sandra Slutz, PhD, Science Buddies; edited by Sabine De Brabandere, PhD, Science Buddies

During weeks 10 through 24 of  gestation  (when a fetus is developing inside of its mother’s womb, also called  in utero ), ridges form on the  epidermis , which is the outermost layer of skin, on the fingertips of the fetus. The pattern that these ridges make is known as a fingerprint and looks like the drawing shown in Figure 1 below.

Fingerprints are static and do not change with age, so an individual will have the same fingerprint from infancy to adulthood. The pattern changes size, but not shape, as the person grows. (To get a better idea of how that works, you can model the change in size by inking your fingerprint onto a balloon and then blowing up the balloon.) Since each person has unique fingerprints that do not change over time, they can be used for identification. For example, police use fingerprints to determine whether a particular individual has been at a crime scene. Although the exact number, shape and spacing of the ridges changes from person to person, fingerprints can be sorted into three general categories based on their pattern type: loop, arch and whorl, as shown in Figure 2, below.

The  DNA  that a person  inherits  from their parents determines many personal characteristics and traits, like whether someone is right- or left-handed or the color of their eyes. In this science project, you will examine fingerprints from siblings versus pairs of unrelated individuals to figure out if general  fingerprint  patterns are  genetic  or random. Have you ever looked at two girls and said, “You must be sisters”? We can often tell that two people are siblings because they appear to have several similar physical traits. This is because children receive half their DNA from each parent. All  biological siblings  are the mixture of both parents’ DNA. This results in a greater degree of matching traits between siblings than between unrelated individuals. Therefore, if DNA determines fingerprint patterns, then siblings are more likely to share the same fingerprint category than two unrelated individuals are.

Terms and concepts

• Gestation 
• Fingerprint patterns
• Biological siblings
• Fingerprint formation
• Inheritance
• What does it mean to be biologically related?
• What are fingerprints and how are they formed?
• What procedures do officials, like the police, use to record fingerprints?
• What are the different types or classes of fingerprints?

Materials and equipment

• Paper towel
• Moist towelettes for cleaning hands
• White printer paper, tracing paper or parchment paper
• Clear tape 
• Scissors 
• White paper 
• Sibling pairs (at least 15)
• Unrelated pairs of people (at least 15)
• Optional: Magnifying glass
• Lab notebook

Experimental procedure

1. To start this science project, practice taking reliable, clear fingerprints. First try the technique on yourself, then ask a friend or family member to let you learn by using his or her fingerprints.

• To make an ink pad variation, rub a pencil on a piece of printer paper, parchment paper or tracing paper several times until an area of about 3 by 3 centimeters (1.2 by 1.2 inches) is completely grey, as shown in Figure 3 (the paper on the left).
• Use a moist towelette to clean the person’s right index finger.
• Thoroughly dry the finger with a paper towel.
• Press and slide each side of the right index fingertip one time over the pad. 
• Then roll the grey fingertip onto the sticky side of a piece of clear tape. The result will look like the tape in Figure 3.
• Use another towelette to clean the person’s grey finger.
• Cut off the piece of tape containing the fingerprint and stick it onto a piece of white paper, as shown in Figure 3. 
• Perfect your technique until the fingerprints come out clear each time.
• When your prints start to fade, rub your pencil a couple of times over your pad and try again.

2. Make up a consent form for your science project. Because fingerprints can be used to identify people, you will need their consent to take and use their fingerprints. The Science Buddies resource on  Projects Involving Human Subjects  will give you some additional information on getting consent.

3. Collect fingerprints of pairs of siblings  and  of pairs of unrelated people.

• Make sure they sign a consent form  before  you take the fingerprint.
• Use the cleaning and printing system you developed in step 1 to take one fingerprint of each person’s right index finger.
• Label each fingerprint with a unique code, which will tell you which pair the fingerprint belongs to and whether that is a sibling pair or an unrelated pair. An example of an appropriate code would be to assign each pair a number and each individual a letter. Siblings would be labeled as subjects A and B, while unrelated individuals would be labeled as subjects D and Z. Thus, fingerprints from a sibling pair might carry the codes 10A and 10B while fingerprints from an unrelated pair might be labeled 11D and 11Z.
• Collect fingerprints from at least 15 sibling pairs and 15 unrelated pairs. For unrelated pairs, you can actually reuse your sibling data by pairing them up differently. As an example, you could pair sibling 1A with sibling 2B since these individuals are not related to each other. The more pairs you look at in your science project, the stronger your conclusions will be! For a more in-depth look at how the number of participants affects the reliability of your conclusions, see the Science Buddies resource  Sample Size: How Many Survey Participants Do I Need?

4. Examine each fingerprint and characterize it as a whorl, arch or loop pattern. You can use a magnifying glass if you have one. In your lab notebook, make a data table like Table 1, creating a separate row for each person, and fill it out.

In your lab notebook, make a data table like this one and fill it out using the fingerprint pattern data you collected. Be sure to make a separate row for each person.

5. To analyze your data, calculate the percentage of related pairs whose fingerprint patterns match and the percentage of unrelated pairs whose fingerprint patterns match. Advanced students can calculate the margin of error. The Science Buddies resource Sample Size: How Many Survey Participants Do I Need? can help you with this.

6. Make a visual representation of your data. A pie chart or bar graph will work well for this data. Advanced students can indicate the margin of error on their graph. 

7. Compare the percentage of related pairs whose fingerprint patterns match to the percentage of unrelated pairs whose fingerprint patterns match. 

• Are they the same? Is the difference significant taking the margin of error into account? Which one is higher? 
• What does this tell you about whether fingerprint patterns are genetic?
• Identical twins share (nearly) 100 percent of their DNA. Does your data include any identical twins? Do they have the same fingerprint pattern?
• How do your results change if you compare all 10 fingers rather than just one? Do all 10 fingers from the same person have the same fingerprint? 
• Toes also have ridge patterns. Do “toe prints” follow the same rules as fingerprints? 
• Are some patterns more common than others?
• If you make more quantitative measurements of the fingerprint patterns, can they be used to predict sibling pairs? With what degree of accuracy?
• If fingerprints are unique, why do misidentifications occur in forensics? How easy or hard is it to match a fingerprint with an individual?
• Read about statistics and use a mathematical test (like Fisher’s exact test) to determine if your findings are statistically relevant. To do this, you will need to make sure you understand p values and you will need to think about whether your sample size is large enough. Online calculators, like the one from  GraphPad Software , are good resources for this analysis.

This activity is brought to you in partnership with  Science Buddies . Find  the original activity  on the Science Buddies website.

More Stories from Science News Explores on Genetics

The last woolly mammoths offer new clues to why this species died out

Scientists Say: Telomere

Freeze-drying turned a woolly mammoth’s DNA into ‘chromoglass’

Gene editing may help rice better withstand climate change

A protein in sweat may protect people from Lyme disease

Scientists Say: Genetic Engineering

A new type of immune cell may cause lifelong allergies

Paw-print DNA lets scientists track out-of-sight polar bears

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• RESEARCH HIGHLIGHT
• 06 January 2022

The surprising genes behind a fingerprint’s unique swirls

The same genes that build an animal’s limbs also encode the intricate patterns in fingerprints. Credit: Douglas Sacha/Getty

The arches, loops and whorls that make each person’s fingerprints unique are created by some of the same genes that drive limb development 1 .

Access options

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

24,99 € / 30 days

cancel any time

Subscribe to this journal

Receive 51 print issues and online access

185,98 € per year

only 3,65 € per issue

Rent or buy this article

Prices vary by article type

Prices may be subject to local taxes which are calculated during checkout

Nature 601 , 168 (2022)

doi: https://doi.org/10.1038/d41586-021-03850-3

Li, J. et al. Cell https://doi.org/10.1016/j.cell.2021.12.008 (2022).

Article   Google Scholar  

Download references

• Developmental biology

Self-organized tissue mechanics underlie embryonic regulation

Article 11 SEP 24

Human embryo models are getting more realistic — raising ethical questions

News Feature 11 SEP 24

Menopause age shaped by genes that influence mutation risk

News & Views 11 SEP 24

High-level Talent Recruitment dedicated to teaching & research

College of Water Sciences, Beijing Normal University

Beijing, China

OSU Neurology Clayton C. Wagner Parkinson’s Disease Research Professorship

Columbus, Ohio

The Ohio State University (OSU)

Professor/Associate Professor/Assistant Professor/Senior Lecturer/Lecturer

The School of Science and Engineering (SSE) at The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen) sincerely invites applications for mul...

Shenzhen, China

The Chinese University of Hong Kong, Shenzhen (CUHK Shenzhen)

LEO Lecturer III in Pharmaceutical Sciences

The College of Pharmacy at the University of Michigan, Ann Arbor, invites applications for a full-time (100%) LEO Lecturer III position in the Depa...

Ann Arbor, Michigan

University of MI- College of Pharmacy

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

• svg]:fill-accent-900">

The genes behind your fingerprints just got weirder

By Lauren J. Young

Updated on Jan 10, 2022 2:35 PM EST

Mapped across our fingertips is a diverse, intricate landscape. The swirling maze of tiny ridges, peaks, and valleys are used to help us flip pages of a book, feel and distinguish textures, and unlock our phones . But it turns out that fingerprints are not only unique signatures of our identity; they are signatures of our early stages of development. 

A recent study published in the journal Cell reveals that genes involved in limb development are influential in creating the shape of our fingerprints. Previously, researchers thought the genes active in skin formation were the most obvious connection to fingerprint patterns. 

“Fingerprints are an interesting feature of human biology that have been used for a number of practical purposes like individual identification and, more in the past, diagnosis of conditions,” says Denis Headon, a coauthor on the study and developmental biologist at the University of Edinburgh, who investigates the genetics of structure variations on skin in different species, like hair, feathers, and scales. “I think what we’re getting here is an insight into how the variation in fingerprints is arising and how it’s got to do with limb development processes.”

[Related: CRISPR breaks ground as a one-shot treatment for a rare disease ]

Early skin of an embryo is quite simple, Headon explains. On different regions of the body, the embryonic skin will develop hair follicles, teeth, or in this case, fingerprint ridges on the soles of the feet and palms of the hands. In a large DNA analysis of more than 23,000 people across ethnic groups, Headon and a team of international researchers explored the genetic underpinnings of the unique variations in the trait. 

The uniqueness of a fingerprint can trace back to some very fine features. One example is the exact positions of the ridges that run across the finger pads, says Headon. “You can get a split in the ridge, where it bifurcates into two—or a ridge can just stop. Sometimes you get very short ridges, like little islands.” Although not studied in this paper, sweat glands that decorate the tops of ridges also influence their path, Headon adds. However, many of our fingerprints generally fall under three common patterns: whorled, arched, or looped. 

“You can have more complicated prints with double whorls, you can have complex loops, but they still fall into these categories, arch, loop and whorl. And that’s what we’ve studied here with genetics,” Headon explains.

The team assessed DNA from their volunteers to find the genetic basis of these patterns, looking for a specific balance of genes that shift the chances of having an arch versus a loop versus a whorl. What they found was that many of the genes that underlie the formation of the three patterns were not primarily linked to skin formation but rather limb and finger formation. To test out this finding, the researchers homed in on EVI1, a known limb-development gene in humans that influences the three middle digits of the hand. They decreased the expression of EVI1 in mice, which don’t have fingerprints but do have distinct ridges that run across parts of the digits, and saw that the individuals with the altered gene also had changes in the ridge patterns. “It’s subtle … but we can see a shift in the shape of those transverse ridges,” Headon says. “This supports the idea that EVI1 is the gene responsible for influencing fingerprint typing in humans, as well as these simple ridges that we see in mice.”

What this is telling us [in humans] is that the distinction between arch, loop, and whorl is coming about from how the limb is growing and being shaped, and particularly how the fingers are being shaped and growing, in utero prior to birth.”

An embryo’s limbs start growing around week five of development. As different boney elements and parts of the digits in the hand begin to take shape, the genes that form limbs and also determine fingerprint patterns are already in action much earlier than the print begins to appear, Headon explains. Around week 8 to 10 of development, the hands and feet develop volar pads—temporary thick skin on palms and soles that are particularly prominent while fingertips and toes develop. Most humans lose their volar pads before birth, but some animals retain them, like primates. The bulbous pad “looks almost like a frog’s hand,” says Headon. 

“It was suggested in the 20th century that the shape of these pads—the height of them, the specific shape of them, and their size—influence where what type of fingerprint forms,” Headon explains. The team’s analysis indicated that finger length is genetically linked to the fingerprint patterns. So, the researchers infer that finger length and volar pad’s size, position, and formation all contribute to the resulting fingerprint arrangement. “The genetic influence on arch, loop, and whorl distinction is coming in at that earlier stage of building that field where the fingerprint will form.”

[Related: Scientists genetically engineered prehistoric proteins to detect diseases ]

The genes steering fingerprint formation can be helpful in learning about disease and genetic conditions, says Headon. While molecular tools are better for diagnosis, exploring the genetics behind these patterns could be used to better understand some chromosomal disorders where fingerprints and creases on the hands are altered, such Edwards Syndrome and Down’s Syndrome. But Headon points out that there are limitations and overuse of fingerprints in certain medical assessments. “In psychology, people have published before that certain fingerprint types are associated with, for example, schizophrenia,” he says. “I’m very skeptical of that research.” Still, he adds that these recent findings on the genetic origins of our fingerprints can reveal a lot about healthy human development. 

“These genetic approaches in large populations can give insights into developmental processes taking place in humans that are really largely hidden from view,” he says. “I’d hope this type of study can be applied to other characteristics in humans and give other information about how we develop and how we vary from one another.”

Latest in Science

Neanderthals’ social isolation may have sped up their extinction neanderthals’ social isolation may have sped up their extinction.

By Laura Baisas

How does Polaris Dawn fit into the history of spacewalks? How does Polaris Dawn fit into the history of spacewalks?

By Briley Lewis

Advertisement

What determines fingerprints?

24 August 2022

The Image Bank RF/Getty Images

What determines fingerprints? Is it a random process or is it something inherited from our parents?

Mike Follows Sutton Coldfield, West Midlands, UK

Fingerprints begin to form in the 10th week of gestation. Some researchers think they are caused by tension between the outer layer of the skin (the epidermis) and the layer underneath it (the dermis), which results in tiny folds in the skin’s surface.

An alternative hypothesis is that their formation might be the result of something called a Turing pattern , driven by biochemistry. Alan Turing is better known for breaking the Nazi Enigma code, but, in…

Sign up to our weekly newsletter

Receive a weekly dose of discovery in your inbox! We'll also keep you up to date with New Scientist events and special offers.

To continue reading, subscribe today with our introductory offers

No commitment, cancel anytime*

Offer ends 10 September 2024.

*Cancel anytime within 14 days of payment to receive a refund on unserved issues.

Inclusive of applicable taxes (VAT)

Existing subscribers

More from New Scientist

Explore the latest news, articles and features

One dose of a smallpox vaccine still gives good protection for mpox

Mathematics

The mathematics behind pouring a glass of wine.

Subscriber-only

New Scientist recommends new superhero drama Supacell

Good news for those whose lipstick isn't as soft as they hoped

Popular articles.

Trending New Scientist articles

• COP Climate Change
• Coronavirus (COVID-19)
• Cancer Research
• Diseases & Conditions
• Mental Health
• Women’s Health
• Circular Economy
• Sustainable Development
• Agriculture
• Research & Innovation
• Digital Transformation
• Publications
• Academic Articles
• Health & Social Care
• Environment
• HR & Training
• Stakeholders
• Asia Analysis
• North America Analysis
• Our Audience
• Marketing Information Pack
• Prestige Contributors
• Testimonials

• North America

The function of fingerprints: How can we grip?

Professor Gun-Sik Park, in the Department of Physics and Astronomy at Seoul National University explores the function of fingerprints from a lens of understanding the mechanism of our human ability to grip

The regulation of moisture levels in fingerprints has been found to play a crucial role in optimizing grip and maximizing friction. Our study revealed that the function of fingerprints is to control the hydration of the fingertips, allowing for precision gripping.

To understand the hydrodynamics of the fingerprint, we used electromagnetic waves in various frequency ranges, including megahertz, terahertz, infrared, and visible light. Our results showed that the fingerprint structure acts as a capillary channel for evaporation, leading to a steady-state hydration condition in either dry or wet conditions.

This research sheds light on the little-understood mechanism of grip in humans, primates, and koalas. Our findings suggest that the fingerprints function as a capillary evaporating channel, contributing to precision grip.

The function of fingerprints

The function of fingerprints has been a topic of interest for some time, with the long-standing belief being that fingerprints help increase friction for better grip. However, recent research has challenged this notion and has instead shown that the ridges of fingerprints actually reduce friction by reducing the contact area. The only established function of fingerprints has been found to be their modulation of friction when rubbed against a finely abraded surface, which enhances tactile sensitivity.

This discovery has led to the development of various tactile stimulating devices, including artificial fingers, used as friction sensors. Further studies have aimed to uncover the relationship between fingerprints, tactile perception, and gripping. It has been suggested that fingerprints greatly change the contact area and friction coefficient as grip force increases, potentially aiding in precision manipulation and gripping tasks.

Despite the advances in our understanding of the relationship between fingerprints and grip, there has been no conclusive evidence that fingerprints assist in gripping. There is still a hypothesis about the role of fingerprints in draining moisture to aid in grip in humid environments. Excessive moisture on the skin can act as a lubricant, reducing friction and causing slipping, however, the hydration level of fully occluded fingerprint skin does not exceed a certain value while gripping. It has been observed that the moisture control of the fingertips tends towards a value that maximizes friction during gripping.

The fingerprint acts as a capillary tube

Several studies have been published that focus on the friction and lubrication of human skin, with results indicating that there is still an unknown mechanism for regulating moisture to optimize the gripping of ridged skin.

In this study, we discovered that the fingerprint acts as a capillary tube where moisture evaporates at the meniscus between the object being held and the fingerprint. This helps maintain a steady-state moisture level, maximizing friction and providing optimal grip. We observed the dynamics of moisture over occlusion time using electromagnetic waves in various regimes, including terahertz and infrared waves and visible light. Our results showed that the hydration level reaches a steady state after around 180 seconds, regardless of whether the fingerprint is “wet” or “dry.” The steady-state hydration levels, friction forces, and occlusion times are dependent on the individual fingerprint’s structures and properties, as well as environmental factors like humidity. Our study also confirmed the tendency for hydration to increase exponentially in “dry” fingerprints and decrease linearly in “wet” fingerprints, as seen through IR waves with optical coherence tomography (OCT).

A microfluidic system at our fingertips

We found that only ridged skin, such as the skin on a fingerprint, can attain a steady-state hydration in the fully occluded state, not flat skin like that on the chest, forearm, or thigh. By using a charged-coupled device (CCD) camera and OCT, we directly visualized the movement of moisture in individual fingerprints over time and observed that evaporation leads to the shedding of excess water through the fingerprint and accumulation of water in the fingerprint valley. This “wet” system of fingerprint-moisture-glass can be considered a microfluidic system, where water films can be trapped by capillarity along the two sharp corners of the interface between the fingerprint and glass.

In conclusion, the function of fingerprints as a capillary channel naturally regulates moisture to maintain optimal friction for gripping. This insight could be useful in the design of tactile sensing devices and optimized haptic interfaces, particularly in developing artificial fingers for human-like touching and gripping.

• PNAS|December 15, 2020|vol.117|no. 50|31665–31673

Seoung-Mok Yum, In-Keun Baek, Dongpyo Hong, Juhan Kim, Kyunghoon Jung, Seontae Kim,Kihoon Eom, Jeongmin Jang, Seonmyeong Kim, Matlabjon Sattorov, Min- Geol Lee, Sungwan Kim,Michael J. Adams, and Gun-Sik Park

Contributor Details

LEAVE A REPLY Cancel reply

Save my name, email, and website in this browser for the next time I comment.

This article first appeared in Open Access Government April 2023 .

To view the article click here .

License: CC BY 4.0

This work is licensed under Creative Commons Attribution 4.0 International .

What does this mean?

Share - Copy and redistribute the material in any medium or format for any purpose, even commercially.

Adapt - Remix, transform, and build upon the material for any purpose, even commercially.

The licensor cannot revoke these freedoms as long as you follow the license terms.

Latest Publication

Open Access Government July 2024

• Terms & Conditions
• Privacy Policy
• GDPR Privacy Policy
• Marketing Info Pack
• Fee Schedule

• Free Resources
• Project Search
• Featured Projects
• Member Benefits

1059 Main Avenue, Clifton, NJ 07011

The most valuable resources for teachers and students

(973) 777 - 3113

[email protected]

1059 Main Avenue

Clifton, NJ 07011

07:30 - 19:00

Monday to Friday

123 456 789

[email protected]

Goldsmith Hall

New York, NY 90210

• Why We’re Unique

Fingerprinting

Introduction: (initial observation).

Look at the palm of your hands and fingertips. you will see patterns of ridges that are called fingerprints. The natural body oils and sweat collects in these ridges of skin so that when you touch things, you leave behind an impression of your print. Fingerprint ridges are useful because they give your hands a better grip when you pick something up. But since fingerprints are unique, they have also become a common and foolproof method of identification. They can be used to help solve crimes by identifying criminals. The two basic ideas scientists believe about fingerprints are:

• Fingerprints never change. Small ridges form on a person’s hands and feet before they are born and do not change for as long as the person lives.
• No two fingerprints are alike. The ridges on the hands and feet of all persons have three characteristics (ridge endings, bifurcations and dots) which appear in combinations that are never repeated on the hands or feet of any two persons. A ridge ending is simply the end of a ridge. A bifurcation is a Y-shaped split of one ridge into two. A dot is a very short ridge that looks like a “dot”.

Problem Statement: (Sample)

After each crime, detectives need to find and collect any thing that may possibly lead them to the person or persons who took part in the action. Finger prints are among the most important evidence that can be used to relate a person to a crime. The problem is that finger prints are not visible and we need to use special methods to revel and collect them. This project is an attempt on revealing and identifying finger prints.

This project guide contains information that you need in order to start your project. If you have any questions or need more support about this project, click on the “ Ask Question ” button on the top of this page to send me a message.

If you are new in doing science project, click on “ How to Start ” in the main page. There you will find helpful links that describe different types of science projects, scientific method, variables, hypothesis, graph, abstract and all other general basics that you need to know.  

Project advisor

Information Gathering:

More information:

The basic fundamentals in the science of fingerprint identification are permanence and individuality.

Permanence : Fingerprint ridges are formed during the third to fourth month of fetal development. These ridges consist of individual characteristics called ridge endings, bifurcations, dots and many ridge shape variances. The unit relationship of individual characteristics does not naturally change throughout life… until decomposition after death. After formation, an infant’s growing fingerprint ridges are much like drawing a face on a balloon with a ball-point pen and then inflating it to see the same face expand uniformly in all directions. Unnatural changes to fingerprint ridges include deep cuts or injuries penetrating all layers of the epidermis and some diseases such as leprosy.

Permanent scars, disease damage, and temporary changes such as paper cuts appear as jagged edges and sometimes “puckered” ridge detail in opposition to smooth flowing natural formations. Warts can come and go, but generally push apart an area of friction ridges and can disappear completely when the wart is gone because they are not a part of the friction ridge structure. Look at a wart with a magnifying glass and you will notice that the friction ridges “surround” the wart. Senile atrophy of friction skin due to old age causes the ridges to often almost flatten, causing fingerprints with many creases (creases are also unique but not always permanent) and poorly defined ridges. Oddly, newborn infants also often have more creases than clearly defined ridge detail in their barefoot prints. The creases are unique, but change relatively rapidly and often disappear as the infant grows. The best chance of seeing friction skin ridges on newborn infant footprints is to look carefully with a magnifying glass on and near the big toe.

Individuality : In the over 140 years that fingerprints have been routinely compared world wide, no two areas of friction skin on any two persons (including identical twins) have been found to contain the same individual characteristics in the same unit relationship. This means that in general, any area of friction skin that you can cover with a dime (and often with just a pencil eraser) on your fingers, palms, or soles of your feet will contain sufficient individual characteristics in a unique unit relationship to enable positive identification to the absolute exclusion of any other person on earth. Recent studies comparing the fingerprints of cloned monkeys showed that they, just like identical twin humans, have completely different fingerprints. When doctors state that twins have the same fingerprints, they are referring to the class characteristics of the general ridge flow, called the fingerprint pattern. These loop, arch and whorl ridge flow patterns have nothing to do with the individual characteristics used to positively identify persons. Before modern computerized systems, fingerprint classification was essential to enable manual filing and retrieval of fingerprints in large repositories.

Today there are more than 250 million criminal and civil fingerprints on record in the U.S.A., with that number increasing by about 34000 each day! All of these fingerprints used to be kept on file cards. Now all of the fingerprint cards are in the computer system, which matches prints in fraction of time.

When DNA evolved as a science, the term “DNA fingerprinting” was adopted to lend credibility to that science’s newcomer status which is in its infancy compared with the empirical validation of fingerprint identification world wide. DNA analysis as commonly practiced in forensic science laboratories cannot differentiate between identical twins, but fingerprints have always been able to differentiate identical twins.

Making fingerprints visible:

Fingerprints that are left behind on objects are called latent fingerprints. The word latent means “hidden,” and these fingerprints are often nearly invisible.

You can make these fingerprints visible by sprinkling them with one of the fingerprint powders. Commonly used fingerprint powders are talcum and graphite. Talcum is white and graphite is black. Any other fine powder may have similar results. Here is how to do it:

1. Make a good latent print with one of your own fingers.

2. Shake out a very small amount of fingerprint powder next to the fingerprint. You need only a very small amount. Remember, one of the most common mistakes is to use too much powder.

Brush softly, brush lightly, and keep brushing until every detail of the print is clear. The two commonest mistakes are brushing too hard and not brushing enough.

4. When you have finished brushing, blow lightly across the print to remove the excess powder.

Lifting the print:

You can pick up the print you have just finished brushing and preserve it on a piece of paper or on a card. Here’s how to do it.

1. From the sheets of tape squares, peel off one square (or cut a square from any clear adhesive tape). Touch only a tiny corner of the square when you handle it. You don’t want to get a fingerprints on it when you are taking it off the backing paper.

2. Position the tape over the fingerprint. Then lower it until it rests on top of the print. You may smooth it down gently, but don’t rub hard, as this will cause it to smudge.

3. Pick up the tape again by one corner. It will lift up the fingerprint with it.

4. Place the tape on a piece of paper. If the print was made with graphite, place the tape on a white background. If it was made with talcum powder, use a black background.

Classifying fingerprints:

All the fingerprints in the world consist of three basic patterns: Arch, loop and whorl. Within these three basic groups they are further divided into the following subcategories:

ARCHES (about 5% of all fingerprints):

LOOPS  (about 65% of all prints):

WHORLS (about 30% of all prints):

As a warm-up experiment try to classify your own fingerprints. Use a magnifying lens to examine them carefully. What do you see? Classify your fingerprints and record the result in a table like this:

Question/ Purpose:

What do you want to find out? Write a statement that describes what you want to do. Use your observations and questions to write the statement.

The purpose of this project is understanding and experimenting the science of fingerprinting. You may also test different powders to see which one can more clearly reveal fingerprints.

After you learn more about finger prints, you may also come up with additional questions such as:

• What is the most common thumb print? (Main Question )
• How similar are the finger prints of children to the finger prints of their parents?
• Do identical twins have identical finger prints?
• Do cats have finger prints?
• Do dogs have finger prints?
• Any of the above questions can be the subject of a different science project.

Need a graph?

Fingerprinting can help us to classify or identify people based on their fingerprints.

Science Project is a good opportunity for a student to learn about collecting data and making graphs. Following is a more specific question related to fingerprinting that requires collecting data and making a graph.

Which of the three basic fingerprint patterns (Arch, loop and whorl) are the most common for the thumb?

Identify Variables:

When you think you know what variables may be involved, think about ways to change one at a time. If you change more than one at a time, you will not know what variable is causing your observation. Sometimes variables are linked and work together to cause something. At first, try to choose variables that you think act independently of each other.

This is how you may define variables for the question “What is the most common thumb print?”.

Independent variable is the pattern of thumb prints.

Dependent variable is the number of people that have each specific thumb print.

Hypothesis:

Based on your gathered information, make an educated guess about what types of things affect the system you are working with. Identifying variables is necessary before you can make a hypothesis. Also make sure to view the information about defining a hypothesis in the “How to Start” section.

Following is a sample hypothesis for the question “What is the most common thumb print?”. This hypothesis is being tested in experiment number 3.

My hypothesis is that Arch is the most popular thumb print. My hypothesis is based on my observation of thumb prints in my family members.

If you are selecting any other question, you must come up with your own hypothesis.

Experiment 1:

Collect some fingerprints from classmates or family members. Then you find an object such as a glass cup and try to find out who has touched that glass cup. Use a card similar to this to collect and keep the prints.

Print and use this card with the ink pad to record fingerprints of friends and family. You will use these cards to identify fingerprints you discover, as well as for solving mysteries in the future. Begin by making your own fingerprints.

In order to make a clear fingerprint, wash and dry your hands before you start. The above card has space for two kind of impressions.

1. The Rolled Impression: Roll the top part of each finger on the ink-pad and then roll it on the card, in the appropriate space for that finger.

2. Plain Impression: These are used to verify the correct order of the rolled prints. You should take the four fingers and press them down on the ink-pad and then press on the card at the same time (large white space), then make a thumb print next to them.

After each fingerprinting, clean your hand with rubbing alcohol and paper towel.

Rubbing alcohol is flammable. Adult supervision is required.

Coding and Classification: (Additional information for advanced level students only)

After you collect fingerprints on cards, you may want to convert them to codes to create a computer file of your prints. National Crime Information Center has a coding system that you might use.

FINGERPRINT CLASSIFICATION (NCIC) CODE TABLE

If you have ever noticed on a criminal history return a field named FPC/ and followed by a 20 digit number then you have seen an NCIC classification before. Most all criminal histories and some NCIC wanted hits will have this information. Because of the probability of the wanted hit having an NCIC classification it may be an asset to understand just exactly what it means.

A typical NCIC classification might look like this:

FPC/ 19PIPOPM1716PO18PICI

Obviously since most of us have 10 fingers and this is a 20 character string we can pretty safely assume that each two characters represent one finger on the person. The numbering always begins with the number 1 position which happens to be the right thumb and then traveling across the right hand to the right little finger which is #5. The thumb of the left hand is #6 and across to the left little finger which is #10.

Assuming you now understand the scheme of this character string lets move on to what each type of two character string represents. Anytime you see two numbers together, such as 02, 09, 11, 15 etc then you know that the pattern type is a loop pattern. In the case of the 1st two digits of the code above (which is the right thumb) you should know that the thumb is a loop pattern with a ridge count of 19. You can find more information about ridge counts on the loop definition page.

The second finger shown above is the right index finger, which is finger #2, and the code above is PI . This code represents two things, one is the pattern type denoted by the P which means plain whorl and the second character is the tracing of the pattern. In this case it is an inside tracing.

NCIC Codes for the fingers are as follows:

For more details visit: http://www.brazoria-county.com/sheriff/id/fingerprints/index.htm

Now find some latent fingerprints, make them visible and lift them using a clear tape. When you collect and lift fingerprint specimens, place them on a card like the following picture to keep track of possible suspects. If you have used a white powder, place it on the black square and if you have used black powder, place it on white square. Look at the print and determine which type of print they are and mark the appropriate box on the left. Also on the card write the location that the print was found.

Try to identify the latent fingerprints that you have found.

Experiment 2: Touch a smooth surface such as a glass surface with both hands. Then try to reveal the prints with different powders and compare the results. Some of the powders that you may test are: Talcum, graphite, cocoa, Titan (Titanium Oxide) or any other fine powder paint.

Experiment 3: Which of the three basic fingerprint patterns (Arch, loop and whorl) are the most common for the thumb?

Procedure :

• Prepare 20 small cards, one inkpad and some alcohol pads.
• Determine 20 different individuals that you can take their thumb prints.
• Have each person to push his/her thumb on the inkpad and then do the same on one blank card. Give the person an alcohol pad to remove excess ink from his/ her hand.
• After you collect 20 thumb prints on 20 cards, compare the type of print with samples provided above and determine if it is Arch, Loop or Whorl.
• Record your results in a table like this:

Make a Bar Graph:

A bar graph that you make for your results will have three bars. Bars will be named Arch, Loop and Whorl.

The height of each bar is the number of people in each type. For example if 7 people have arch thumb prints, then the arch bar will be 7 inches high.

Materials and Equipment:

Material that you need for fingerprinting are as follows:

• Magnifier , to help you see the details of a fingerprint
• Feather , Soft fluffy feather used as a brush
• Plastic tube , (straw) used as a handle for the brush
• Talcum powder (white)
• Graphite (black)
• Ink pad , for making fingerprint for your file
• Mirror , to help you see fingerprints on your toes and feet
• Rubbing alcohol, to clean your fingers after print
• Paper towel, to clean your fingers after print
• Sheets of clear tape , similar to sheets of labels, but clear
• White cards or papers

Material for fingerprinting can be purchased as a kit from MiniScience.com or individually from pharmacies and office suppliers.

Results of Experiment (Observation):

Experiments are often done in series. A series of experiments can be done by changing one variable a different amount each time. A series of experiments is made up of separate experimental “runs.” During each run you make a measurement of how much the variable affected the system under study. For each run, a different amount of change in the variable is used. This produces a different amount of response in the system. You measure this response, or record data, in a table for this purpose. This is considered “raw data” since it has not been processed or interpreted yet. When raw data gets processed mathematically, for example, it becomes results.

Summery of Results:

Summarize what happened. This can be in the form of a table of processed numerical data, or graphs. It could also be a written statement of what occurred during experiments.

It is from calculations using recorded data that tables and graphs are made. Studying tables and graphs, we can see trends that tell us how different variables cause our observations. Based on these trends, we can draw conclusions about the system under study. These conclusions help us confirm or deny our original hypothesis. Often, mathematical equations can be made from graphs. These equations allow us to predict how a change will affect the system without the need to do additional experiments. Advanced levels of experimental science rely heavily on graphical and mathematical analysis of data. At this level, science becomes even more interesting and powerful.

Conclusion:

We found out that fingerprints are not Hereditary as we know we also know that no two fingerprints are alike, and not even identical twins have the same fingerprints. People may have the same type of fingerprints in a family. This does not mean that they are hereditary. We also know how to lift and examine fingerprints

Related Questions & Answers:

Use a flashlight to search for fingerprints in your house at night. Are the prints easier or harder to find?

Use a mirror to examine your toe, foot and lip prints _ Do they have the same kind of patterns as your fingerprints?

What you have learned may allow you to answer other questions. Many questions are related. Several new questions may have occurred to you while doing experiments. You may now be able to understand or verify things that you discovered when gathering information for the project. Questions lead to more questions, which lead to additional hypothesis that need to be tested.

Related Activities:

If you determine that experimental errors are influencing your results, carefully rethink the design of your experiments. Review each step of the procedure to find sources of potential errors. If possible, have a scientist review the procedure with you. Sometimes the designer of an experiment can miss the obvious.

References:

Visit your local library and find available books and magazine articles related to fingerprinting. You may such references for additional information that you may need. List your references in your bibliography. To see a short bibliography Click here .

You may find a long bibliography at http://www.scafo.org/Online_Information/bibliography.htm

Following are some additional references.

http://www.brazoria-county.com/sheriff/id/fingerprints/index.htm

Photo: Courtesy of Tiffany K Graves, (Cody’s mom)

Question : My question for my fingerprinting project is “what method works best to identify a person’s fingerprints?”. Can you help me determine what other methods there are besides dusting? Or is dusting the only way and you use different types of dust to figure out the best outcome?

Answer : Since you have access to the people whom you want to compare their fingerprints, you don’t have to use dusting method. Just buy an inkpad, some rubbing alcohol and some paper towels. Ask every person to push his/her finger on the inkpad and then push the same finger on the paper you have prepared. They can later use rubbing alcohol and paper towel to clean the ink from their fingers.

It is always important for students, parents and teachers to know a good source for science related equipment and supplies they need for their science activities. Please note that many online stores for science supplies are managed by MiniScience.

Testimonials

" I called School Time and my husband and son came with me for the tour. We felt the magic immediately."

- Robby Robinson

" My husband and son came with me for the tour. We felt the magic immediately."

- Zoe Ranson

Contact Info

Our address, working hours.

Week Days: 07:00-19:00

Saturday: 09:00-15:00

Sunday: Closed

Science Project

• Science Projects
• Project Guides
• STEM Activities
• Lesson Plans
• Video Lessons

• Unanswered topics
• Active topics
• Board index Active Forums (Make all new posts here) Grades 6-8 Grades 6-8: Life, Earth, and Social Sciences

fingerprint hypothesis?

Moderators: AmyCowen , kgudger , bfinio , MadelineB , Moderators

Post by cuppe5 » Mon Nov 28, 2016 12:08 pm

Re: fingerprint hypothesis?

Post by PharmaMan » Mon Nov 28, 2016 3:27 pm

Return to “Grades 6-8: Life, Earth, and Social Sciences”

• Announcements
• Active Forums (Make all new posts here)
• ↳   Grades K-5
• ↳   Grades K-5: Life, Earth, and Social Sciences
• ↳   Grades K-5: Physical Science
• ↳   Grades K-5: Math and Computer Science
• ↳   Grades 6-8
• ↳   Grades 6-8: Life, Earth, and Social Sciences
• ↳   Grades 6-8: Physical Science
• ↳   Grades 6-8: Math and Computer Science
• ↳   Grades 9-12
• ↳   Grades 9-12: Life, Earth, and Social Sciences
• ↳   Grades 9-12: Physical Science
• ↳   Grades 9-12: Math and Computer Science
• ↳   Advanced Science Competitions (Intel ISEF, Intel STS, Siemens Competition, JSHS, etc.)
• ↳   Getting Started
• ↳   Preparing for the Competition
• ↳   STEM Activities
• ↳   Interview Requests
• ↳   Careers in Science
• ↳   Raspberry Pi Projects Kit
• ↳   Junior Solar Sprint
• ↳   SimPandemic
• ↳   Using SimPandemic
• ↳   General Discussion
• ↳   Preparing for the Science Fair
• ↳   Intel ISEF Preparation
• ↳   Science Teachers: Fairs, Projects, and General Support
• ↳   Grades K-5: Getting Ready for the Science Fair
• ↳   Grades 6-8: Getting Ready for the Science Fair
• ↳   Grades 9-12: Getting Ready for the Science Fair
• ↳   Physical Science
• ↳   Life, Earth, and Social Sciences
• ↳   Math & Computer Science Sponsored by Hyperion Solutions Corp
• ↳   General Questions
• ↳   InnovatioNation Teachers
• ↳   Grades 6-8: Wind Tunnel
• ↳   Grades 9-12: Wind Tunnel
• ↳   A Magnetic Primer Designer
• ↳   Bridge Design
• ↳   Build a Gauss Rifle!
• ↳   Build a Simple Circuit
• ↳   Build Your Own Crystal Radio
• ↳   Chemistry of Ice-Cream Making
• ↳   Do Oranges Lose or Gain Vitamin C After Being Picked?
• ↳   Electrolyte Challenge: Orange Juice vs. Sports Drink
• ↳   Enzyme-Catalyzed Reactions
• ↳   Forensic Science: Building Your Own Tool for Identifying DNA
• ↳   From Dull to Dazzling: Using Pennies to Test How pH Affects Copper Corrosion
• ↳   How Do Roots Grow...
• ↳   How Does Soil Affect the pH of Water
• ↳   Investigating the 'Mpemba Effect': Can Hot Water Freeze Faster than Cold Water?
• ↳   Is this connected to that? Use a homemade electronic tester to find out if electricity can flow between two objects.
• ↳   How Are Antibodies Used for Blood Typing?
• ↳   It's All in the Wrist
• ↳   Measuring the Sugar Content on a Liquid with a Laser Pointer
• ↳   Potions and Lotions
• ↳   Shaking Up Some Energy
• ↳   Spin Right 'Round with this Simple Electric Motor
• ↳   Sweet as Sugar
• ↳   The Strength of an Electromagnet
• ↳   Veggie Power! Making Batteries from Fruits and Vegetables
• ↳   Which Orange Juice Has the Most Vitamin C?
• Board index
• All times are UTC-07:00

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Sensors (Basel)
• PMC10819264

Fingerprint Recognition in Forensic Scenarios

Nuno martins.

1 Portuguese Military Academy, 1169-203 Lisbon, Portugal; [email protected]

2 Instituto Superior Técnico, Universidade de Lisboa, 1049-001 Lisbon, Portugal; tp.aobsilu.ocincet.rsi@xela

José Silvestre Silva

3 Military Academy Research Center (CINAMIL), 1169-203 Lisbon, Portugal

4 Laboratory for Instrumentation, Biomedical Engineering and Radiation Physics, Universidade de Coimbra (LIBPhys-UC), 3000-370 Coimbra, Portugal

Alexandre Bernardino

5 Institute for Systems and Robotics (ISR), 1049-001 Lisbon, Portugal

Associated Data

Data were obtained from FVC2000 database [ 11 ] and are publicly available upon request.

Fingerprints are unique patterns used as biometric keys because they allow an individual to be unambiguously identified, making their application in the forensic field a common practice. The design of a system that can match the details of different images is still an open problem, especially when applied to large databases or, to real-time applications in forensic scenarios using mobile devices. Fingerprints collected at a crime scene are often manually processed to find those that are relevant to solving the crime. This work proposes an efficient methodology that can be applied in real time to reduce the manual work in crime scene investigations that consumes time and human resources. The proposed methodology includes four steps: (i) image pre-processing using oriented Gabor filters; (ii) the extraction of minutiae using a variant of the Crossing Numbers method which include a novel ROI definition through convex hull and erosion followed by replacing two or more very close minutiae with an average minutiae; (iii) the creation of a model that represents each minutia through the characteristics of a set of polygons including neighboring minutiae; (iv) the individual search of a match for each minutia in different images using metrics on the absolute and relative errors. While in the literature most methodologies look to validate the entire fingerprint model, connecting the minutiae or using minutiae triplets, we validate each minutia individually using n-vertex polygons whose vertices are neighbor minutiae that surround the reference. Our method also reveals robustness against false minutiae since several polygons are used to represent the same minutia, there is a possibility that even if there are false minutia, the true polygon is present and identified; in addition, our method is immune to rotations and translations. The results show that the proposed methodology can be applied in real time in standard hardware implementation, with images of arbitrary orientations.

1. Introduction

Fingerprints accompany all human beings from birth. They are a biometric key composed of unique patterns found on the distal phalanges of the fingers, distinct for each individual, and can be used for various purposes. Due to their unequivocal and invariant properties, fingerprints have gained importance in the field of forensic analysis, becoming a relative alternative to other traditional authentication methods [ 1 , 2 , 3 ]. Currently, there is a growing number of applications using fingerprint recognition systems, such as for accessing mobile phones, monitoring employee presence in a company and, in forensic investigations, to achieve the unequivocal identification of an individual. Technological advances in fingerprint processing enable capture, storage, and comparison methods to be more financially accessible today, allowing a significant portion of the population to use these technologies [ 4 , 5 , 6 ].

Fingerprints can be used to determine if two images are of the same finger, thereby identifying the individual to whom it belongs. Identification of fingerprints collected in a crime scenario is typically done manually, consuming a lot of time, which may be relevant and indispensable to solve the crime. Considering that many fingerprints are collected, a lot of time is spent on comparisons that a computer system can perform in seconds, and it must also be considered that the images collected in the field are sometimes of poor quality [ 7 ].

The main problem for which a solution is sought corresponds to the identification of a fingerprint based on comparison with others, with the aim of identifying those that correspond and belong to the same individual. It is essential to find a method that can be used to make reliable comparisons, generate precise outcomes, and requiring minimal computing power. This way, fingerprint comparison can be performed in the field in real time, with the help of a mobile computing device.

In this paper, the main contributions are:

• A methodology is proposed that can accurately and efficiently to compare two fingerprints and classify them as belonging to the same person or different individuals. The proposed method can be used on portable devices during field work providing real-time screening of collected fingerprints. The proposed methodology also looks forward to validating each minutia individually while in the literature most methodologies look up to validate the entire fingerprint model.
• A new process to validate extracted minutiae is proposed using the convex hull of the set of minutiae to create a region of valid minutiae, and an individual minutia validation is proposed by employing n-side polygons instead of triangulations, which is the approach most common in the literature.

This paper is organized into five sections. The first provides a brief introduction to the work, describing the problem and motivation for carrying out the work. Section 2 details a comparison of the databases used in the reviewed articles and a study of the state-of-the-art on fingerprint comparison methods which encompasses the gaps found. Section 3 defines the proposed methodology to achieve the objectives of the work. Section 4 contains the results of applying the proposed methodology in different databases, initially tuning the algorithm in one database, and then validating it in different ones. Section 5 is where the conclusions of the work are drawn, namely about the methodology applied and the produced results, the achieved objectives, and the proposals for future work.

2. Background

This section covers concepts relevant to the work to be developed, namely about fingerprint features and minutiae extraction.

2.1. Fingerprint Features

The features extracted from the fingerprints are organized hierarchically into global and local [ 8 ]. Global features are singular points in the fingerprint, namely the core and deltas. The core represents the point of convergence of the pattern. Deltas are points where the ridges diverge, and a point is formed that resembles the delta symbol ( Figure 1 ).

Global features: core and delta.

The local features of a fingerprint (see Figure 2 ) are minutiae and refer to the points at which the ridges join or end (bifurcations and terminations, respectively) and are of high relevance as they are used by most fingerprint matching algorithms (sometimes associated with global features). In a fingerprint image, depending on the quality and size, typically 10 to 200 minutiae can be found, and a good quality image should allow the identification of at least 50 to 100 minutiae. Each minutia is associated with a position and orientation, and its distribution is not uniform.

Local features: termination and bifurcation.

2.2. Crossing Numbers Method

The Crossing Numbers method is widely used to extract minutiae from a fingerprint image [ 9 ]. The termination and bifurcation extraction is reached through the analysis of the neighborhood of each pixel in the skeletonized image of the fingerprint using a 3 × 3 window centered in the reference pixel p :

The crossing number for the pixel p is computed through the difference between adjacent pixel values:

Each pixel is then labeled accordingly with its CN value following Table 1 , and a first set of local feature points is found.

Crossing number and type of minutia.

3. Related Work

In this section, selected works relevant to the topic are reviewed. The section is divided into subsections, analyzing the databases most mentioned in the reviewed works, the evaluation metrics of the proposed algorithms and the proposed methods, organized according to their main characteristic (based on minutiae, deep learning, and image texture).

3.1. Databases

In this section, a systematic study of the databases used in the reviewed articles is carried out. An analysis is made to the frequency of each database usage, its images, and their public availability.

In Figure 3 , it is observed that the databases that are most commonly used are generally those made available in the Fingerprint Verification Competition (FVC) [ 10 ]. Some authors used private databases, built for the purpose of the study, which are not included in the presented analysis. In total, there are references to 20 databases, four of which have already been discontinued, and the rest are publicly accessible, except for the FVC2006 set, which is only available for academic purposes. The FVC2000 DB4, FVC2002 DB4, FVC2004 DB4 and FVC2006 DB4 databases are composed of fingerprint images created using the synthetic fingerprint generation software SFinGe (FVC2000—not mentioned, FVC2002—SFinGe v2.51, FVC2004—SFinGe v3.0, FVC2006—SFinGe v3.0). The databases are not recent, but in the case of fingerprints, this situation is acceptable since despite being older, the images maintain their usefulness. The FVC2000, FVC2002, FVC2004, and FVC2006 datasets are the most used, being composed of images acquired through different sensors, presenting different challenges to fingerprint recognition and matching systems.

Databases used in the reviewed articles.

3.1.1. FVC2000

Table 2 summarizes the FVC2000 [ 11 ] database composed of four datasets, each with eight images of 110 fingers up to a total of 880 images.

FVC2000 Database, adapted from [ 11 ].

Different sensors represent various challenges for the algorithms.

Figure 4 shows some examples of FVC2000 images, one for each dataset.

FVC2000: ( a ) DB1, ( b ) DB2, ( c ) DB3, ( d ) DB4.

3.1.2. FVC2002

Table 3 summarizes the FVC2002 [ 12 ] database: 4 datasets with 8 images of 110 fingers each, up to a total of 880 images.

FVC2002 Database, adapted from [ 12 ].

Figure 5 shows some examples of FVC2002 images, one for each dataset.

FVC2002: ( a ) DB1 ( b ) DB2 ( c ) DB3 ( d ) DB4.

3.1.3. FVC2004

Table 4 summarizes the FVC2004 [ 13 ] database: 4 datasets and 8 images of 110 fingers each, up to a total of 880 images. This database contains images to which purposeful difficulties have been added to make the competition more challenging, such as rotating the angle of the image, asking volunteers to press harder on the sensor (introducing distortions), and collecting images from moistened fingers.

FVC2004 Database, adapted from [ 13 ].

Figure 6 shows some examples of FVC2004 images, one for each dataset.

FVC2004: ( a ) DB1 ( b ) DB2 ( c ) DB3 ( d ) DB4.

3.1.4. FVC2006

Table 5 summarizes the FVC2006 [ 14 ] database which has four datasets, each composed of 12 images of 150 fingers, up to a total of 1800 images per dataset.

FVC2006 Database, adapted from [ 14 ].

The availability of the FVC2006 database is exclusively for academic purposes, and no response was received from the University of Bologna to provide access to the images.

3.2. Evaluation Metrics

In this work, the most common metrics are used to evaluate the performance of fingerprint matching systems [ 15 ]. First, it is necessary to define the concepts that allow computing of those metrics. Considering the matching of two images, we define:

• TP—True Positive: should match and match.
• FP—False Positive: should not match and match.
• TN—True Negative: should not match and do not match.
• FN—False Negative: should not match and match.

The following metrics are computed:

Recall: is the metric used to evaluate the model’s ability to correctly identify all relevant positive cases in a dataset, that is, it is the proportion of true positives (positives that are correctly identified) compared to the total positive cases existing in the dataset. This metric is useful in situations where it is crucial to identify all positive cases, even if it means incorrectly classifying some negative cases as positive.

Precision: evaluates the proportion of positive identifications that were obtained correctly. This metric is useful in situations where the cost of a false positive case is high.

F1: recall and precision have a trade-off relationship, that is, increasing one may decrease the other. The F1 metric is used to combine both into a single metric that balances the previous ones.

False Match Rate (FMR): evaluates the probability that the system will identify two fingerprints that do not match as matching.

False Non-Match Rate (FNMR): evaluates the probability that the system will identify two fingerprints that match as non-matching.

Equal Error Rate (EER): measures the balance between FMR and FNMR. Considering different decision thresholds (values at which the system defines the images as corresponding or not), the EER is defined as the point at which these metrics are equal, that is, when the system makes a wrong decision with the same frequency (false positive and false negative).

All the mentioned metrics vary in the range 0 ,   1 . System performance will be optimal when the recall, precision, and F1 metrics have a value of one, and FMR, FNMR, and EER have a value of zero. Typically, the values might also be represented as a percentage.

3.3. Minutiae Extraction and Matching

In this section, recent articles on minutiae extraction and matching are reviewed. A division is made into subsections that group the mentioned articles, considering methods based on minutiae, based on deep learning, and based on image texture.

3.3.1. Minutiae Based Methods

Different methods are proposed based on the minutiae, especially in their geometric arrangement and the creation of a geometric model that represents the fingerprint.

Trivedi et al. [ 16 ] propose the creation of a non-invertible model to represent fingerprints. Images are pre-processed through normalization, binarization, skeletonization, and the application of Gabor filters. Minutiae triangles are formed, and those that cover a small area are discarded. Each validated triangle is described through feature vectors formed by the internal angles and the type of minutia that compose the triangulation. In the matching phase, for each triangle in one image, the triangle in the other image whose descriptor achieves the smallest Euclidean distance is selected. If the distance is below a threshold, the match is validated. The FVC2000, FVC2004 and FVC2006 databases were used, achieving EER of 5.57%, 2.56% and 0.48%, respectively.

Mohamed-Abdul-Cader et al. [ 17 ] propose the use of Delaunay triangulations. Images are pre-processed through histogram equalization, binarization, Gabor filters, and skeletonization. Delaunay triangulation is used to form triangles and for each a descriptor is built with the internal angles, the type of minutiae, and the minutiae orientations. Absolute and relative distances are used to select the best triangle match, and a matching score is computed using the number of corresponding minutiae (unique minutiae in the set of validated triangles) over the total minutiae in the model. A set of 158 images was selected from the FVC2002 DB1 database, achieving an EER of 6.68%.

Ghaddab et al. [ 18 ] propose the use of an expanded Delaunay triangulation, considering all possible triangulations if each minutia was removed from the set, increasing the robustness of the algorithm against false minutiae and increasing the complexity and computational effort. Each triangle is then associated with a feature vector that includes the elongation of the Steiner ellipse, the cosine of the largest angle, the perimeter, and the angle of rotation required to superimpose one vertex on the others. Considering two triangles that could possibly match, a threshold is applied to the distance between each feature in the feature vectors to validate the match. The FVC2000, FVC2002, and FVC2004 databases were used achieving EER of 2.25%, 1.62%, and 3.99%, respectively.

Surajkanta and Pal [ 19 ] also use Delaunay triangulation. Images are pre-processed by applying normalization, binarization, Gabor filters, and skeletonization. The feature vectors are built using the internal angles of each triangle, the type of minutiae, and their orientation. However, in contrast to other authors, the orientation of the minutiae is computed in relation to the core point of the fingerprint. In the FVC2000 database, an average EER of 6.68% was achieved across all four datasets.

3.3.2. Deep Learning Based Methods

Liu et al. [ 1 ] propose the use of a Convolutional Neural Network (CNN) based on the VGG-16 network. The images are not pre-processed, but alignment is performed, and transformations are subsequently applied to generalize the model. The number of convolutional layers and the cost function were adjusted for optimal performance. The proposed CNN achieved a precision of 98,42% in the FVC2000 database, compared to 97.85% produced by the VGG-16 CNN.

Li [ 15 ] proposes the use of a CNN to extract features and match noisy fingerprints. The proposed method achieved FMR 1.54% and FNMR 1.46% in the NIST DB4 database and was compared with a traditional method (based on coordinates and orientation of minutiae) that produced 28.82% and 28.78%, respectively.

Gorgel and Eksi [ 3 ] suggest pre-processing by applying normalization, binarization, and the Gabor Wavelet Transform (GWT) and then using a CNN to classify the images. The FVC2006 images were used, some images were replicated, and small variations were added to generalize the CNN. The proposed method achieved a precision of 91.50% and, using the same model without using the GWT 86.27% were achieved.

Engelsma et al. [ 20 ] propose a network composed of three sub-networks for image alignment, texture-based feature extraction, and minutiae extraction, resulting in an image representation whose dimensionality is reduced through a fully connected layer. The Euclidean distance between descriptors of fixed size is computed and, subsequently, the cosine similarity is computed, applying a threshold value to the result to validate or reject the correspondence. In the FVC2004 DB1 database, rank-1 accuracy of 99.5% and rank-100 accuracy of 100% were achieved.

Tang et al. [ 21 ] propose the use of a deep convolutional network to enhance the and extract the orientation field and minutiae map from latent fingerprints. The traditional methods are transformed to convolutional kernels and integrated as a shallow network with fixed weights and then the representation ability of deep learning is used to create a minutiae map. The model was trained on 8000 pairs of matched rolled fingerprints and experiments were conducted in NIST SD27 and FVC2004 databases outperforming other compared methods.

Cao and Jain [ 22 ] propose the use of a CNN for image enhancement and for ridge flow estimation and minutiae descriptor extraction. Complementary templates (minutiae and texture) are considered to represent the latent fingerprints. Experimental results on the NIST SD27 database achieved rank-1 identification accuracy of 64.7% while 75.3% were achieved in WVU latent database.

He et al. [ 23 ] propose a partial fingerprint verification network based on spatial transformer network and the local self-attention mechanism. The model was trained end-to-end and learned multi-level fingerprint features automatically. The network considers an image pair, performs its alignment and the matching, using affine transformations and fusing features. On the AES3400 dataset, EER of 8.6% was achieved and 3.2% in FVC2006 database.

Cui et al. [ 24 ] propose an end-to-end network to directly output pixel-wise displacement field between two fingerprints, including a siamese network for feature embedding, and a following encoder-decoder network for regressing displacement field. The algorithm consists of two steps: minutiae-based coarse registration and CNN-based fine registration. Compared to other image correlation methods, this method achieved better results in terms of FNMR and FMR.

3.3.3. Texture Based Methods

Monika and Kumar [ 25 ] propose the use of Local Binary Pattern (LBP) features as validation of minutiae matching. The images were pre-processed by binarization and skeletonization and the minutiae were extracted and matched using other authors’ techniques. On matched minutiae, LBP features in the neighborhood are computed and these features are also matched to validate the minutiae match. A subset of 20 images from the FVC2002 database was used. In both cases, the FNMR was 0%, with an FMR of 25% without the use of LBP validation, and an FMR of 20% with the proposed methodology, which resulted in an improvement.

Bakheet et al. [ 26 ] suggest a method based on the fusion of Speed-Up Robust Features (SURF) and Harris key points. Initially, the images were pre-processed through histogram equalization, normalization, segmentation, application of Gabor filters, and binarization. Then, Harris key points were extracted and SURFT descriptors in their neighborhood were extracted. Image matching is performed using the Euclidean distance between the descriptor vectors, using the RANSAC algorithm to refine the results and remove false matches. The FVC2000 DB1 and FVC2002 DB1 databases were used, achieving an accuracy of 92.5% and 95%, respectively.

Oriented Fast Rotated Brief (ORB) features are computed 100 times faster than SIFT features and 10 times faster than SURF features. Li and Shi [ 27 ] propose a methodology that compares a part of the fingerprint with the whole. To do this, the first step is to define a region of interest through the variance of the gray intensity. The ORB descriptors are then extracted and the best match is achieved by brute force using the hamming distance. Finally, the validation of the descriptor matching is done through the analysis of the second closest descriptor, the first being validated if the second presents a much greater distance. The FVC2004 database was used, and changes were made to the images to separate parts of the fingerprints. The proposed method achieved EER 2.83% (in 11 s) and was compared to a similar SIFT-based method that achieved 9.29% (in 123 s).

3.3.4. Summary

Fingerprint recognition and comparison has been studied for more than 40 years, but the design of an accurate and interoperable system that requires little computational power is still considered an open problem.

Minutiae based methods are the most intuitive in terms of what is expected by human interpretation of fingerprint matching, and there are some considerations to be made: the models created must not be invertible for security reasons (not allow reconstruction of the fingerprint). The results achieved on the noisiest images (FVC2004) verify that all methods are very sensitive to the quality of the image since they are highly dependent on the minutiae extracted. The features extracted to describe the minutiae and build a model are of extreme importance. The proposed methodologies are based on the creation of a model that represents the fingerprint, with no methods proposed that analyse minutia by minutia, approaching the work that is performed manually in the forensic area.

The use of deep learning for fingerprint matching has some drawbacks. Firstly, there is the need to train a specific model for each situation, making the use of generic models unfeasible, which requires not only time, but also the acquisition of many images necessary for the training stage. Secondly, databases are often not balanced, with more comparisons of images to reject than images to match, so an imbalance in the data could lead to a biased model. Finally, there is the difficulty of generalizing the model to different sensors and image conditions, such as variations in brightness, for example, making practical applicability difficult.

The SIFT, SURF, ORB, and LBP features are techniques generally applied to image processing and object recognition. However, when the problem deals with fingerprint matching, there are some disadvantages such as the computing of the key points, which can be computationally intensive, introducing restrictions for real-time systems. These techniques extract features from the image, which may be advantageous when the minutiae are not well defined, but when compared to systems based on minutiae, these techniques present worse performance, being more often used as a complement to other algorithms.

4. Proposed Methodology

This section describes the methodology implemented to process and search for matches between fingerprint images. The methodology is, in a larger picture, divided into four blocks (see Figure 7 ) which follow the common process. The new proposed methods are inserted into the pipeline.

Proposed methodology.

The first block concerns image pre-processing in which the orientation and frequency of the ridges are estimated to apply oriented Gabor filters. Then, minutiae are extracted using the Crossing Numbers method and validated by defining a region of valid minutiae, removing nearby minutiae by type, and removing minutiae clouds. The third step concerns the creation of a set of polygons that represent each minutia, and finally, the last step is the minutiae matching and, subsequently, fingerprint match.

Considering the main goals of our study, which consist of proposing a methodology that can be used in real time, on a portable device, in forensic areas. We have to keep in mind that it is important to have a low computational effort such that any common hardware can run the algorithm and to provide a method that can be validated and understood by an operator. This said, we cannot use deep learning since we do not want to train a model for every scenario and there is a small amount of images. The texture based methods, in general, improve the results achieved by other methods but increase the computational effort and, therefore, we will not be using any for now. Each minutiae based method proposed in the literature uses the set of extracted (and sometimes filtered) minutiae to create a model for the fingerprint which is further compared to find a match. We will be focusing on validating each minutia individually which is the work that is carried out manually in real forensic scenarios.

4.1. Pre-Processing

Image pre-processing was implemented through the algorithm proposed by Raymond Thai [ 28 ] which includes the state of the art pre-processing techniques: segmentation, normalization, ridge orientation and frequency estimation, the application of oriented Gabor filters and skeletonization.

Figure 8 shows the effect of the applied pre-processing techniques in two images of different databases (FVC2000 DB1 and FVC2002 DB1).

Image pre-processing: ( a , b ) are Original, ( c , d ) are Pre-processed, ( e , f ) are Skeletonized.

4.2. Minutiae Extraction and Validation

Minutiae are extracted using the Crossing Numbers (CN) method in the skeletonized image. Due to noise in the image and damage in the fingerprints (ex: scars) there are some spurious minutiae that are captured by the CN method.

Three techniques are implemented to create a set of valid minutiae from the extracted ones. The first step is to set a minimum distance δ between minutiae of the same type. Figure 9 represents the effect on the δ parameter in the excluded minutiae, having bifurcations in blue color and terminations in red.

Minimum distance for minutiae by type: ( a ) δ = 5 ( b ) δ = 10 ( c ) δ = 15 .

The second step is a novel definition of a region of interest (for valid minutiae) which is defined by the use of the convex hull. The convex hull was implemented using the incremental algorithm, and it defines the smallest convex polygon that contains all minutiae points in the set. Once the polygon is defined, an erosion of γ pixels is obtained through its interior (see Figure 10 ).

ROI definition: ( a ) Convex hull, ( b ) Convex border limits, ( c ) Erosion of γ = 10 pixels (red represents a termination while blue represents a bifurcation minutia).

This procedure excludes the border minutiae, which are not reliable for matching procedures, as shown in Figure 11 .

Set of minutiae: ( a ) Original ( b ) After ROI definition (red represents a termination while blue represents a bifurcation minutia).

The last step in minutiae validation is the removal of minutiae clouds, that is, when the algorithm finds a termination close to a bifurcation due to noise but only one is a real minutia ( Figure 12 ). Minutiae clouds are replaced by the average point, creating a new averaged minutia.

Minutiae Cloud Removal: ( a ) Original ( b ) Replaced (red represents a termination minutia, blue represents a bifurcation minutia, pink represents the new averaged minutia, the arrow and black circle emphasize the zoom).

4.3. Feature Extraction

Each valid minutia is individually represented through a set of polygons that are built using the neighboring minutiae as vertices. Then, each polygon is stored using its features, namely the edge size and the angles formed by adjacent vertices and the reference minutia. This means that each minutia will be associated with a set of fixed-length vectors that describe its associated polygons.

The first step in building the model to represent each minutia is to define n as is the number of desired vertices for each polygon. Then, a circumference centered on the reference minutiae is considered and its radius is incremented until the maximum radius is reached or a set of valid neighbor minutiae is found. The goal is to find at least n + 2 minutiae and guarantee that their distribution matches at least one per quadrant. To increase the number of minutiae registered through polygons, sets of n points were also allowed but only if they were found when the maximum radius for the circumference is achieved.

Figure 13 illustrates the search for neighbor minutiae to form the set of polygons that represents a minutia. After these minutiae are found, polygons with desired n vertices are formed using the points combinations. A polygon is valid if:

• There are no overlapping edges.
• There are no holes within the boundaries of the polygon.
• The starting point and the ending point coincide (closed polygon).
• The polygon contains the reference minutia.

Search for neighbor minutiae: ( a ) Invalid; ( b ) Valid (blue represents a detected minutiae, purple circle represents the search radius for neighbor minutiae, red represented the reference minutia).

Figure 14 shows examples of both invalid and valid polygons that represent a minutia (reference is green and polygon vertices are black).

The green dot is the reference minutia; invalid polygons: ( a , b ); valid polygons: ( c , d ). Blue represents the termination minutia, red represents the bifurcation minutia, green dots represents the reference minutia, black represents the polygon vertices, green lines represents the polygon edges.

A fixed-size feature vector is then built to represent each polygon using the sizes of the edges and the angles formed by the adjacent vertices and the reference minutia (see Figure 15 ).

Angle used to build the feature vector (blue represents the termination minutia, red represents the bifurcation minutia, green dot represents the reference minutia, black represents the polygon vertices, green lines represents the polygon edges, A and B are names of the vertices).

Figure 16 is an example of all valid polygons that represent a valid reference minutia. Each polygon feature will be further considered for matching.

Set of valid polygons for a reference minutia (blue represents the termination minutia, red represents the bifurcation minutia, green dot represents the reference minutia, black represents the polygon vertices, green lines represents the polygon edges, purple circle represents the search radius).

In this step, each minutia that was previously validated will be associated with a certain number of polygons that depend on its neighborhood minutiae. Each polygon is represented using a fixed-size vector that is not dependent on translations and rotations since it is always sorted in the same way (by keeping the larger edge size as most-left element and then keep the relative element order, and sorting the angles accordingly).

4.4. Matching

The matching has two steps: the minutiae matching and the fingerprint matching.

Minutiae matching is done through the polygons. Considering two minutiae, the set of polygons that represent each one is compared to find the closest match using the feature vectors. The best match is defined as the pair of feature vectors that has a smaller Euclidean distance. The pair that has smaller distance is not guaranteed to match (in totally different minutiae, there will always be a pair with smaller distance although it is a high distance compared to real matching minutiae). Thus, the pair of best matched polygons must be validated. The validation is performed using both an absolute and relative error criteria that is applied element-wise to the feature vectors. The thresholds applied to the edge sizes and the angles are different.

Fingerprint matching is conducted by defining a threshold T M C (Total Minutiae Corresponding) which is the minimum number of matching minutiae to consider the fingerprints as match.

5. Results and Discussion

A random set of 125 images from FVC2000 DB1 database [ 11 ] was chosen which resulted in 125 genuine comparisons and 7154 imposter matchings. An average of 23 minutiae (represented by polygons) were registered.

The first experiments were performed to understand the importance of the values chosen for the relative and absolute error criteria that are used to validate the best match polygons. To start with some configuration, the number of vertices was set to n = 5 , the absolute error threshold for the edges size was set to T H L = 5 , and the absolute error threshold for the angles was set to T H A = 10 . The threshold for fingerprint match was set as T M C = 12 and the relative error threshold T H R E L was varied. The results achieved are in Table 6 .

Results as a function of the relative error criteria.

The smaller the relative error threshold, the more polygons are validated using the absolute error threshold. This results in a more conservative system, which produces a higher number of false negatives, thus increasing the FNMR.

Then, the relative error threshold was fixed to T H R E L = 11 , and the other parameters were kept while varying T H L . The following results were achieved ( Table 7 ):

Results in function of the absolute error criteria for the edge sizes.

A smaller value of T H L means that it is harder for two polygons to match, and the system is conservative. The same absolute distance between two edge sizes produces a higher relative error if the edges are smaller. Thus, the absolute error criteria for edge sizes will have more importance to filter small edges that fail the relative error criteria.

Setting T H L = 5 and varying the absolute error threshold for the angles, the following results were achieved.

The results shown in Table 8 are important to understand that most of the angle verifications are done using the relative error criteria.

Results as a function of the absolute error criteria for the angles.

Using the best configuration found ( T H R E L = 11 , T H L = 5 and T H A = 10 ) the number of polygon vertices was changed.

Analyzing Table 9 we can state that, on one hand, the increase in the number of polygon vertices guarantees polygons with more distinct geometries and, as such, when a minutia is validated, the degree of confidence is greater, resulting in a decrease in the FMR since minutiae that do not hardly correspond are not matched. On the other hand, when increasing the number of polygon vertices, the algorithm becomes more dependent on the number of minutiae extracted, being more exposed to the quality of the database.

Results as a function of n.

Varying the threshold TMC allows us to understand how good is the border between the number of minutiae that are correctly and wrongly matched, and compute the EER.

An EER of 0.03% was achieved at T M C = 12 . In general, the system behaves as expected for a fingerprint matching system, that is, a smaller threshold TMC allows more false images to match while a higher TMC reduces the false matches ( Figure 17 ).

FMR and FNMR accordingly with TMC (FVC2000 DB1).

Keeping the algorithm parameters, more experiments were carried out using a random sample of 125 images from the FVC2002 DB1 database which led to an average of 35 registered minutiae (an increase of 52%). Initially, FNMR of 0% and FMR of 17.6% were achieved. These results suggest that the matching fingerprints have all more than 12 matching minutiae but 17.6% of the non-matching fingerprints are also reaching more than the TMC limit, which is expected since the high increase in the number of registered minutiae leads to hundreds of more possible polygons. Therefore, the error criteria were straightened and changed to T H R E L = 10 , T H L = 4 and T H A = 10 . With the new configuration, an FNMR of 21.65% and an FMR of 3.36% were achieved, suggesting that the TMC threshold was not good for the images and, therefore, a swipe in the TMC values was done (see Figure 18 ).

FMR and FNMR accordingly with TMC (FVC2002 DB1).

An EER of 7.8% was achieved for a threshold T M C = 9.5 but since the number of corresponding minutiae must be integer, using TMC = 10 FNMR of 9.10% and FMR of 6.74% were obtained.

A last set of experiments were carried out to challenge the algorithm and understand if it can deal with transformations applied to the images (such as translation and rotation) since the goal was to propose an algorithm that can be used in real-time without depending on the image alignment. From 15 random images chosen from FVC2000 DB1 database, a rotation of 45°, another rotation of 90° and a diagonal translation of 10% were applied, creating a new database composed by 60 images.

With the previous configuration n = 5 ,   T H R E L = 11 , T H L = 5 ,   T H A = 10 , and T M C = 12 , FNMR of 46.7% and FMR of 0.1% ( Figure 19 ) were achieved meaning that false matchings are being well rejected but the number of matching minutiae is smaller even in matching images. Thus, those results suggest that the threshold TMC is too high.

FMR and FNMR accordingly with TMC (transformed database).

Through the adjustment of TMC to 6.4, an EER of 3.6% was achieved, leading to an FNMR of 4.4% and an FMR of 2.6% using the integer threshold of 7, showing that the algorithm can identify matching minutiae without a previous image alignment. The results achieved show that the algorithm depends on the extracted and validated minutiae both in quality (to have the same geometrical relationship between minutiae in different matching images) and quantity (to ensure a higher number of registered minutiae), leading to the need of dynamically adjusting its parameters to achieve good results.

Compared to the state-of-the-art traditional methods, we can compare with the work from Mohamed-Abdul-Cader et al. [ 17 ], since they also choose a sample of around 150 images from FVC2002 DB1 and obtained EER of 6.68%. We achieved in 7.8% in a sample of the same database but we did not optimize the algorithm’s parameters in an exhaustive way. We overfitted the parameters in FVC2000 DB1 where we achieved 0.06% error and then performed only one iteration of tunning the parameters before attempt in FVC2002 DB1. This means that our results still have space for large improvement. Also, our results are in line with the state-of-the-art but our technique is a novelty since we are proposing a method that uses neighbor minutiae to validate the match of each individual minutia in opposition to what other authors do, using the set of minutiae as a whole.

Several authors have applied methods based on Deep Learning, such as He et al. [ 23 ] having achieved EER values between 3% and 6% (FVC2006 dataset) and between 7% and 9% in the AES3400 dataset. In the proposed method we achieved values of 7.8%, which are slightly higher than the values presented by He [ 23 ], justified by the fact that our initial objective was the development of a methodology able of running in an electronic device with low computational performance (example: cheap tablet) and capable of analyzing fingerprints in crime scenarios without internet access (that is, without access to the cloud).

6. Conclusions and Future Work

The design of a fingerprint matching system capable of acting on a large database is still an open problem and due to the unique characteristics of this biometric key, research continues to be carried out to propose more methods capable of handling fingerprints. This work aims at proposing a method that can act on a local database of fingerprints collected at a crime scene to carry out a selection of them and find those that are relevant to solving the crime in real time and without requiring a computational effort such that it can only be carried out on machines with high processing power and following the work that is done manually by forensic teams (this is, validating each minutia individually), so that the algorithm can be understood and validated by an operator.

In the state-of-the-art, methods were divided into three categories: methods based of deep learning, image texture, and minutiae. The first ones require high computational effort and the need to train models, which requires a high number of images and is not practical to be applied in real time. The second ones are based on image processing techniques that abstract themselves from the image and perform mathematical procedures and operation on the values of its pixels, being very sensitive to noise and presenting itself more as a validation complement of other methods. Finally, methods based on minutiae have the main disadvantage of depending on the reliability of minutiae extraction, but they are similar to the work carried out manually, allowing, for example, an operator to visualize and understand algorithm decisions and can be applied in real time without requiring a high computational effort (may be executed on portable devices).

The proposed methodology depends on the minutiae extracted from the image and aims to meet human reasoning, that is, to validate a minutia that is present in both images and check whether the number of validated minutiae reaches a level that allows us to guarantee that the fingerprints match. The algorithm is divided into four blocks: pre-processing, minutiae extraction and validation, feature extraction, and fingerprint matching. The first block was implemented with the application of the most mentioned state-of-the-art methods, resulting in images with substantial noise removal and clear distinction between ridges and valleys. In the second block, minutiae extraction was implemented using the Crossing Numbers method, which has proven results over the years, and minutiae validation was innovative, using the convex hull of the set of minutiae to define a region of valid minutiae. Feature extraction consists of the formation of polygons that represent a minutia, formed through neighboring minutiae. In the literature, most authors use minutiae triangulations to create models that represent the fingerprint, but with the aim of validating each minutia individually, the implement method associates each minutia with a set of unique polygons, with different shapes, depending on the arrangement, in space, of neighboring minutiae. Finally, fingerprint matching is intuitive; once a certain number of minutiae is associated in the two images, we consider the fingerprints to match.

The achieved results show that the implemented method is very dependent on the minutiae extracted, with different results and the need to make adaptations when changing databases. Image quality is a relevant factor, as it determines the details that are extracted. The concept of associating a set of polygons with a chosen number of vertices to each minutia proves to be challenging, since by increasing the number of polygon vertices we increase the certainty in the individual validation of each minutia, but more minutiae are required in the image and in the same relative position to ensure that there are enough of them to achieve an adequate number of minutia represented. Using dynamic parameters in the algorithm, namely the number of minutiae considered as a decision threshold to consider the fingerprints as match and the validation criteria of the polygon (maximum relative and absolute errors between the lengths of the edges and the angles formed between adjacent vertices and the reference minutia), FNMR 0% and FMR 0.06% were achieved in a random sample of 125 images from the FVC2000 DB1 database and FNMR 9.1% and FMR 6.7% in a random sample of 125 images from the FVC2002 DB1, results that are in line with the state of the art and suggest the possibility of transposing the method to real applications of the minutiae match. The application of the algorithm to images that underwent translation and rotation achieved FNMR 4.4% and FMR 2.6%, showing that the algorithm is capable of matching minutiae without requiring prior alignment of the images.

In future work, the dynamic adjustment of the polygon matching validation criteria would allow the automatic adaptation of the algorithm to different databases and the possibility of incorporating the use of minutiae that are valid but, because they are the extremes of the valid regions, they cannot be represented through polygons that surround them is proposed. Also, simple features of each polygon were chosen to allow the outlined objectives to be achieved, without increasing the computational effort but there is the possibility of exploring other features.

Funding Statement

This research was supported in part by the Military Academy Research Center (CINAMIL), and by UIDP/FIS/04559/2020 and UIDB/FIS/04559/2020 ( https://doi.org/10.54499/UIDB/04559/2020 accessed on 18 January 2024), funded by national funds through FCT/MCTES and co-financed by the European Regional Development Fund (ERDF) through the Portuguese Operational Program for Competitiveness and Internationalization, COMPETE 2020.

Author Contributions

J.S.S. and A.B. proposed the idea and concept; N.M. developed the software under the supervision of J.S.S. and A.B. All authors revised and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable (this study does not involve humans).

Data Availability Statement

Conflicts of interest.

The authors declare no conflicts of interest.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.