ITR: Fingerprint Feature Extraction and Matching

Michigan State University, East Lansing MI

Investigators

Abstract

Questions related to the identity of an individual such as "Is this the person who she claims to be?", "Has this applicant been here before", "Should this person be given access to our system?" are asked millions of time every day by organizations in financial services, health care, e-commerce, telecommunications, and government. In fact, identity fraud in welfare disbursement, credit card transactions, cellular phone calls, and ATM withdrawals totals several billion dollars each year. Recent concerns about homeland security have created additional interest in methods for personal identification. Personal identification is the process of associating a particular individual with an identity. Knowledge-based (e.g., password or PIN) and token-based (e.g., ID card or key) automatic personal identification approaches have been the two traditional techniques widely used. Since these approaches are not based on any inherent attributes of an individual, they suffer from the obvious disadvantages; tokens may be lost, stolen, or misplaced, and a PIN may be forgotten by a valid user or guessed by an impostor. As a result, these approaches are unable to differentiate between an authorized person and an impostor who fraudulently acquires the token or knowledge of the authorized person. Biometric identification refers to identifying an individual based on his distinguishing physiological and/or behavioral characteristics. Among the various such characteristics (e.g., face, voice, iris, fingerprints), fingerprints have been the most popular. Recent advances in fingerprint sensing technology allow these inexpensive sensors to be embedded in a variety of systems (e.g., laptops and PDAs). In spite of the deployment of large number of automatic fingerprint matching systems over the past 40 years, their performance is not perfect. The objectives of the proposed research are to (i) reliably extract fingerprint features (minutiae points), and (ii) develop robust fingerprint matching techniques based on these features even in the presence of noisy and distorted fingerprint images. We propose to use a number of statistical models both for feature extraction and matching. Proposed research will lead to more secure and robust fingerprint matching systems.