FBI — Palm Print Recognition

Palm Print RecognitionIntroductionPalm print recognition inherently implements many of the samematching characteristics that have allowed fingerprint recognitionto be one of the most well-known and best publicized biometrics.Both palm and finger biometrics are represented by theinformation presented in a friction ridge impression. Thisinformation combines ridge flow, ridge characteristics, and ridgestructure of the raised portion of the epidermis. The datarepresented by these friction ridge impressions allows adetermination that corresponding areas of friction ridgeimpressions either originated from the same source or could nothave been made by the same source. Because fingerprints andpalms have both uniqueness and permanence, they have beenused for over a century as a trusted form of identification.However, palm recognition has been slower in becomingautomated due to some restraints in computing capabilities andlive-scan technologies. This paper provides a brief overview ofthe historical progress of and future implications for palm printbiometric recognition.HistoryIn many instances throughout history, examination of handprintswas the only method of distinguishing one illiterate person fromanother since they could not write their own names. Accordingly,the hand impressions of those who could not record a name butcould press an inked hand onto the back of a contract became anacceptable form of identification. In 1858, Sir William Herschel,working for the Civil Service of India, recorded a handprint on theback of a contract for each worker to distinguish employees fromothers who might claim to be employees when payday arrived.This was the first recorded systematic capture of hand and fingerimages that were uniformly taken for identification purposes.1The first known AFIS system built to support palm prints isbelieved to have been built by a Hungarian company. In late1994, latent experts from the United States benchmarked thepalm system and invited the Hungarian company to the 1995International Association for Identification (IAI) conference. Thepalm and fingerprint identification technology embedded in thepalm system was subsequently bought by a US company in 1997.Page 121 of 166

Palm Print RecognitionIn 2004, Connecticut, Rhode Island and California establishedstatewide palm print databases that allowed law enforcementagencies in each state to submit unidentified latent palm prints tobe searched against each other's database of known offenders.2,3Australia currently houses the largest repository of palm prints inthe world. The new Australian National Automated FingerprintIdentification System (NAFIS) includes 4.8 million palm prints.The new NAFIS complies with the ANSI/NIST international standardfor fingerprint data exchange, making it easy for Australian policeservices to provide fingerprint records to overseas police forcessuch as Interpol or the FBI, when necessary.4Over the past several years, most commercial companies thatprovide fingerprint capabilities have added the capability forstoring and searching palm print records. While several state andlocal agencies within the US have implemented palm systems, acentralized national palm system has yet to be developed.Currently, the Federal Bureau of Investigation (FBI) CriminalJustice Information Services (CJIS) Division houses the largestcollection of criminal history information in the world. Thisinformation primarily utilizes fingerprints as the biometricallowing identification services to federal, state, and local usersthrough the Integrated Automated Fingerprint IdentificationSystem (IAFIS). The Federal Government has allowed maturationtime for the standards relating to palm data and live-scan captureequipment prior to adding this capability to the current servicesoffered by the CJIS Division. The FBI Laboratory Division hasevaluated several different commercial palm AFIS systems to gaina better understanding of the capabilities of various vendors.Additionally, state and local law enforcement have deployedsystems to compare latent palm prints against their own palmprint databases. It is a goal to leverage those experiences andapply them towards the development of a National Palm PrintSearch System.In April 2002, a Staff Paper on palm print technology and IAFISpalm print capabilities was submitted to the IdentificationServices (IS) Subcommittee, CJIS Advisory Policy Board (APB). TheJoint Working Group then moved “for strong endorsement of theplanning, costing, and development of an integrated latent printcapability for palms at the CJIS Division of the FBI. This shouldproceed as an effort along the same parallel lines that IAFIS wasdeveloped and integrate this into the CJIS technicalcapabilities .”5Page 122 of 166

Palm Print RecognitionAs a result of this endorsement and other changing business needsfor law enforcement, the FBI announced the Next GenerationIAFIS (NGI) initiative. A major component of the NGI initiative isthe development of the requirements for and deployment of anintegrated National Palm Print Service. Law enforcementagencies indicate that at least 30 percent of the prints lifted fromcrime scenes — from knife hilts, gun grips, steering wheels, andwindow panes — are of palms, not fingers.6 For this reason,capturing and scanning latent palm prints is becoming an area ofincreasing interest among the law enforcement community. TheNational Palm Print Service is being developed on the basis ofimproving law enforcement’s ability to exchange a more completeset of biometric information, making additional identifications,quickly aiding in solving crimes that formerly may have not beenpossible, and improving the overall accuracy of identificationthrough the IAFIS criminal history records.ApproachConceptPalm identification, just like fingerprint identification, is based onthe aggregate of information presented in a friction ridgeimpression. This information includes the flow of the frictionridges (Level 1 Detail), the presence or absence of features alongthe individual friction ridge paths and their sequences (Level 2Detail), and the intricate detail of a single ridge (Level 3 detail).To understand this recognition concept, one must first understandthe physiology of the ridges and valleys of a fingerprint or palm.When recorded, a fingerprint or palm print appears as a series ofdark lines and represents the high, peaking portion of the frictionridged skin while the valley between these ridges appears as awhite space and is the low, shallow portion of the friction ridgedskin. This is shown in Figure 1.Figure 1: Fingerprint Ridges (Dark Lines) vs. Fingerprint Valleys (White Lines).Page 123 of 166

Palm Print RecognitionPalm recognition technology exploits some of these palmfeatures. Friction ridges do not always flow continuouslythroughout a pattern and often result in specific characteristicssuch as ending ridges or dividing ridges and dots. A palmrecognition system is designed to interpret the flow of the overallridges to assign a classification and then extract the minutiaedetail — a subset of the total amount of information available, yetenough information to effectively search a large repository ofpalm prints. Minutiae are limited to the location, direction, andorientation of the ridge endings and bifurcations (splits) along aridge path. The images in Figure 2 present a pictorialrepresentation of the regions of the palm, two types of minutiae,and examples of other detailed characteristics used during theautomatic classification and minutiae extraction gure 2: Palm Print and Close-up Showing Two Typesof Minutiae and Other Characteristics.HardwareA variety of sensor types — capacitive, optical, ultrasound, andthermal — can be used for collecting the digital image of a palmsurface; however, traditional live-scan methodologies have beenslow to adapt to the larger capture areas required for digitizingpalm prints. Challenges for sensors attempting to attain highresolution palm images are still being dealt with today. One ofthe most common approaches, which employs the capacitivesensor, determines each pixel value based on the capacitancemeasured, made possible because an area of air (valley) hassignificantly less capacitance than an area of palm (ridge). Otherpalm sensors capture images by employing high frequencyultrasound or optical devices that use prisms to detect the changePage 124 of 166

Palm Print Recognitionin light reflectance related to the palm. Thermal scannersrequire a swipe of a palm across a surface to measure thedifference in temperature over time to create a digital image.Capacitive, optical, and ultrasound sensors require onlyplacement of a palm.SoftwareSome palm recognition systems scan the entire palm, while othersrequire the palms to be segmented into smaller areas to optimizeperformance. Maximizing reliability within either a fingerprint orpalm print system can be greatly improved by searching smallerdata sets. While fingerprint systems often partition repositoriesbased upon finger number or pattern classification, palm systemspartition their repositories based upon the location of a frictionridge area. Latent examiners are very skilled in recognizing theportion of the hand from which a piece of evidence or latent lifthas been acquired. Searching only this region of a palmrepository rather than the entire database maximizes thereliability of a latent palm search.Like fingerprints, the three main categories of palm matchingtechniques are minutiae-based matching, correlation-basedmatching, and ridge-based matching. Minutiae-based matching,the most widely used technique, relies on the minutiae pointsdescribed above, specifically the location, direction, andorientation of each point. Correlation-based matching involvessimply lining up the palm images and subtracting them todetermine if the ridges in the two palm images correspond. Ridgebased matching uses ridge pattern landmark features such assweat pores, spatial attributes, and geometric characteristics ofthe ridges, and/or local texture analysis, all of which arealternates to minutiae characteristic extraction. This method is afaster method of matching and overcomes some of the difficultiesassociated with extracting minutiae from poor quality images.The advantages and disadvantages of each approach vary basedon the algorithm used and the sensor implemented. Minutiaebased matching typically attains higher recognition accuracy,although it performs poorly with low quality images and does nottake advantage of textural or visual features of the palm.Processing using minutiae-based techniques may also be timeconsuming because of the time associated with minutiaeextraction. Correlation-based matching is often quicker to processbut is less tolerant to elastic, rotational, and translationalvariances and noise within the image. Some ridge-based matchingcharacteristics are unstable or require a high-resolution sensor toPage 125 of 166