Fundamentals of Biometrics for Personal Verification/Identification

1. Fundamentals of Biometrics for Personal Verification/Identification Chaur-Chin Chen Department of Computer Science Institute of Information Systems and Applications National Tsing Hua University E-mail: cchen@cs.nthu.edu.tw Tel/Fax: (03) 573-1078/ (03) 572-3694

2. Outline • What is Biometrics? • Motivation by Evidence Iris Image Pattern Analysis Handwriting/Handprinting Verification Personal Signature Verification Hand Geometry Verification Voice (Speech) Pattern Recognition Face Image Recognition Fingerprint Image Verification/Identification Palmprint, Ear shape, Gesture, … • Fingerprint Classification and Verification • Opportunities and Challenges

3. What and Why is Biometrics? • What is Biometrics? Biometrics is the science and technology of interactively measuring and statistically analyzing biological data, in particular, taken from live people. • Why Biometrics? (1) The banking industry reports that false acceptance rate (FAR) at ATMs are as high as 30%, which results in financial fraud of US$2.98 billion a year. (2) In U.S., nearly half of all escapees from prisons leave through the front door, posing as someone else. (3) Roughly 4000 immigration inspectors at US ports-of-entry intercepted and denied admission to almost 800,000 people. There is no estimate of those who may have gotton through illegally. (4) Personal verification/identification becomes a more serious job after the WTC attack on September 11, in the year 2001. • The evidence indicates that neither a PIN number nor a password is reliable.

4. Some Biometric Images

5. Iris Image Pattern Analysis • The iris is the portion of texture regions surrounding the pupil of an eyeball. • The iris image can be sensed by a CCD camera under a regular lighting environment. • An ancient French criminologist Berthillon did exploratory work linking iris pattern to prisoner identity. • In 1980’s, ophthamologists Leonard Flom and Aran Safar posited that no two irises were alike. • In 1994, Professor John Dougman develop algorithms using 2D Gabor filters according to Flom and Safar’s concept to extract iris features for the use in human authentication. • IrisCode, the feature vector of an iris, consisting of 512 bytes is recorded and stored in the database for future recognition/matching. It takes less than 2 seconds in a Pentium III machine to compute an IrisCode. • Potential applications for iris scanning biometrics are widespread and installations have been undertaken in the financial sectors for CityBank ATMs as well as in some international airport for passenger identification. • http://www.astrontech.pl/html/body_iridian_merged.html

6. Handwriting/Handprinting Verification Personal Signature Verification • Handwritings and Signaturesare behavioral biometrics rather than anatomical biometrics such as an iris pattern or a fingerprint. • People handwrite digits or their names in their own special manners. An ancient Chinese calligrapher Wang, Xizhi (AD 306~365) produced many beautiful writings such that his signature would be paid for in gold. • Based on the mechanics of how we write is something very personal and often quite distinctive, biometrics handwriting and/or signature seeks to analyze the dynamics inherent in writing the digits, characters, letters, words, and sentences. • The features include how a person presses on the writing surface, how long a person takes to sign his name, how a person struggles to maintain verticality, angularity in letter forms and along the baseline, plus narrow letters. • http://www.handwriting.org/main/hwamain.html • Biometrics is the science and technology of interactively measuring and statistically analyzing biological data, in particular, taken from live people

7. Hand Geometry Verification • Hand geometry systems work by taking a 3D view of the hand in order to determine the geometric shape and metrics around finger length, height, and/or other details. • A leading hand geometry device measures and computes around 90 parameters and stores in a record of 9 bytes, providing for flexibility and storage transmission. • http://cse.msu.edu/rgroups

8. Voice (Speech) Pattern Recognition • The basis for voice or speech technology was pioneered by Texas Instruments in the 1960’s. • The current voice recognition uses a standard microphone to record an individual’s voice and identity its unique characteristics. It attempts to analyze the physiological characteristics that produce speech, and not the sound or pronunciation. • A voice identification system requires that a “voice reference template” be constructed so that it can be compared against subsequent voice identification. Voice identification systems incorporate several variables or parameters in the recognition of one’s voice/speech pattern including pitch, dynamics, and waveforms. • It is estimated that the revenues from voice/speech identification systems and telephony equipments and services sold in America will increase from US$356 million in 1997 to US$22.6 billion in 2003. • Hidden Markov Model and Autoregressive Model • Fast Fourier Transform and Wavelet Analysis • http://www.buytel.com

9. Outline For Image Processing • A Digital Image Processing System • Image Representation and Formats 1. Sensing, Sampling, Quantization 2. Gray level and Color Images 3. Raw, RGB, Tiff, BMP, JPG, GIF, (JP2) • Image Transform and Filtering • Histogram, Enhancement and Restoration • Segmentation, Edge Detection, Thinning • Image Data Compression R.C. Gonzalez and R.E. Woods, Digital Image Processing, Prentice-Hall, 2002

10. Digital Image Analysis System • A 2D image is nothing but a mapping from a region to a matrix • A Digital Image Processing System consists of 1. Acquisition – scanners, digital camera, ultrasound, X-ray, MRI, PMT 2. Storage – HD (40GB+), CD (700MB), DVD (4.7GB), HD-DVD (20GB), Flash memory (256 MB +) 3. Processing Unit – PC, Workstation, PC-cluster 4. Communication – telephone, cable, wireless 5. Display – LCD monitor, laser printer, laser-jet printer

11. Image Processing System

12. Gray Level and Color Images

13. Pixels in a Gray Level Image

14. Gray and Color Image Data • 0, 64, 144, 196, 225, 169, 100, 36 (R, G, B) for a color pixel Red – (255, 0, 0) Green – ( 0, 255, 0) Blue – ( 0, 0, 255) Cyan – ( 0,255, 255) Magenta – (255, 0, 255) Yellow – (255, 255, 0) Gray – (128, 128, 128)

15. Image Representation (Gray/Color) • A gray level image is usually represented by an M by N matrix whose elements are all integers in {0,1, …, 255} corresponding to brightness scales • A color image is usually represented by 3 M x N matrices whose elements are all integers in {0,1, …, 255} corresponding to 3 primary primitives of colors such as Red, Green, Blue

16. Sensing, Sampling, Quantization • A 2D digital image is formed by a sensor which maps a region to a matrix • Digitization of the spatial coordinates (x,y) in an image function f(x,y) is called Sampling • Digitization of the amplitude of an image function f(x,y) is called Quantization

17. Gray Level and Color Images

18. Some Image File Formats • Raw – Raw image format uses a 8-bit unsigned character to store a pixel value of 0~255 for a Raster-scanned gray image without compression. An R by C raw image occupies R*C bytes or 8RC bits of storage space • TIFF – Tagged Image File Format from Aldus and Microsoft was designed for importing image into desktop publishing programs and quickly became accepted by a variety of software developers as a standard. Its built-in flexibility is both a blessing and a curse, because it can be customized in a variety of ways to fit a programmer’s needs. However, the flexibility of the format resulted in many versions of TIFF, some of which are so different that they are incompatible with each other • JPEG – Joint Photographic Experts Group format is the most popular lossy method of compression, and the current standard whose file name ends with “.jpg” which allows Raster-based 8-bit grayscale or 24-bit color images with the compression ratio more than 16:1 and preserves the fidelity of the reconstructed image • EPS – Encapsulated PostScript language format from Adulus Systems uses Metafile of 1~24-bit colors with compression • JPEG 2000

19. Image and Its Histogram

20. Edge Detection -1 -2 -1 0 0 0  X 1 2 1 -1 0 1 -2 0 2  Y -1 0 1 Large (|X|+|Y|)  Edge

21. Thinning and Contour Tracing • Thinning is to find the skeleton of an image which was commonly used for Optical Character Recognition (OCR) and Fingerprint matching • Contour tracing is usually used to locate the boundaries of an image which can be used in feature extraction for shape discrimination

22. Image Edge, Skeleton, Contour

23. Image Data Compression • The purpose is to save storage space and to reduce the transmission time of information. Note that it requires 6 mega bits to store a 24-bit color image of size 512 by 512. It takes 6 seconds to download such an image via an ADSL (Asymmetric Digital Subscriber Line) with the rate 1 mega bits per second and more than 12 seconds to upload the same image • Note that 1 byte = 8 bits, 3 bytes = 24 bits

24. Lenna Image vs. Compressed Lenna

25. Face Image Recognition • Face recognition technology works well with most of the shelf PC cameras, generally requiring 320*240 resolution at 3~5 frames per second. • Facial recognition software products range in price from US$50 to over US$1000, making one of the cheaper biometric technologies. • Four primary methods used to identify or verify users by means of facial features, including eigenface, PCA, 2D-PCA, LDA, 2D-LDA, wavelet analysis, neural network, and ad hoc methods. • Singular Value Decomposition and Pattern Recognition. • Fast Fourier Transform and Wavelet Analysis • http://facial-scan.com/facial-scan_technology.htm • http://www-white.media.mit.edu/vismod/demos/facerec

26. A Face Recognition Flowchart

27. Face Database • YALE • P. N. Belhumer, J. Hespanha, and D. Kriegman. Eigenfaces vs. fisherfaces: Recognition using class specific linear projection. IEEE Transactions on Pattern Analysis and Machine Intelligence, Special Issue on Face Recognition, 17(7):711--720, 1997. • YALE B • Georghiades, A.S. and Belhumeur, P.N. and Kriegman, D.J. From Few to Many: Illumination Cone Models for Face Recognition under Variable Lighting and Pose. IEEE Trans. Pattern Anal. Mach. Intelligence 23(6):643-660 (2001). • ORL • Ferdinando Samaria, Andy Harter. Parameterisation of a Stochastic Model for Human Face Identification. Proceedings of 2nd IEEE Workshop on Applications of Computer Vision, Sarasota FL, December 1994 • AR • A.M. Martinez and R. Benavente. The AR Face Database. CVC Technical Report #24, June 1998

28. Faces From The Same Person

29. Cumulative Distributions of Same Faces

30. Faces from Different Persons

31. Cumulative Distributions of Different Faces

32. Fingerprint Image Verification/Identification • Each fingerprint is a map of ridges and valleys in the epidermis layer of the skin. • The ridge and valley structures from unique geometric patterns. • A minutiae pattern consisting of ridge endings and bifurcations is unique to each fingerprint. • Most of the contemporary automated fingerprint identification and verification systems (AFIS) are minutiae pattern matching systems. • A modern AFIS is composed of 5 primary modules: (1) Image Enhancement, (2) Image segmentation and Thinning, (3) Minutiae Points Extraction, (4) Core and Delta Localization, and (5) Point Pattern Matching. • A fingerprint forum provided 5 sets of small databases for researchers to evaluate their identification/verification software. • SecuGen EyeD and Veridicom are two leading companies selling both commercial fingerprint identification/verification systems and sensors with resolution 500dpi. Veridicom FPS110 fingerprint reader sensed a 300*300 fingerprint image in a 2cm by 2cm area. • http://www.networkusa.org/fingerprint.shtml • http://bias.csr.unibo.it/fvc2000 • http://bias.csr.unibo.it/fvc2004 • http://www.fpusa.com

33. FINGERPRINTS.DEMON.NL

34. FVC 2004

35. FVC 2004

36. A Paradigm for Fingerprint Matching

37. Fingerprints and Their Histograms

38. Fingerprint Image Processing

40. Thank You • KoalaAngel wishes you have a wonderful university life • I am from Brisbane, Australia and sleep 16 hours each day but you should not • April 12, 2007

41. Are They From the Same Person?

42. Are They the Same Person?