Abstract
Biometrics aims at reliable and robust identification of humans from their personal traits, mainly for security and authentication purposes, but also for identifying and tracking the users of smarter applications. Frequently considered modalities are fingerprint, face, iris, palmprint and voice, but there are many other possible biometrics, including gait, ear image, retina, DNA, and even behaviours. This chapter presents a survey of machine learning methods used for biometrics applications, and identifies relevant research issues. The author focuses on three areas of interest: offline methods for biometric template construction and recognition, information fusion methods for integrating multiple biometrics to obtain robust results, and methods for dealing with temporal information. By introducing exemplary and influential machine learning approaches in the context of specific biometrics applications, the author hopes to provide the reader with the means to create novel machine learning solutions to challenging biometrics problems.
TopA General Look At Biometric Systems
The application area of biometrics with the grandest scale is in border control, typically an airport scenario. Within the national identity context, it is possible to conceive the storing and managing of the biometric information for the entire population of a country. A smaller scale application is access control, for instance securing the entrance of a building (physical access control) or securing a digital system (logical access control). In both applications, we have a verification (or authentication) problem, where the user has an identity claim, and a sampled biometric is checked against a stored biometric for similarity. In a sense, this is a one-class pattern classification problem.
The second important problem involves identification, where there is no identity claim, and the sampled biometric is matched against many stored templates. Checking passengers against a list of criminals, forensic applications, identification of individuals at a distance, or providing access in consumer products (e.g. fingerprint scanning on a laptop) would be typical applications. Depending on the application requirements, the problem may be sufficiently constrained to apply a discriminative approach.
The most important biometric modalities are fingerprint, face, iris, signature, palm print and voice. The biometric traits differ in their usability, convenience, security, and complexity. For providing access to a high-security facility, security is of primary importance, whereas a household appliance that identifies users via biometrics would strive to have maximum user convenience. Similarly, privacy can be a major determinant in the deployment of a particular biometric application. For this reason, a host of possible biometrics are considered for different applications, including DNA, gait, ear images, and even behaviours.
Key Terms in this Chapter
Biometric System: A biometric system involves a set of sensors to record a biometric from the users of the system a database of stored biometric templates, and an authentication algorithm by which the recorded biometric is compared to the template.
Biometric: A biometric is a personal or a behavioural trait that can be used to identify a person.
Authentication: Authentication is the decision process where a biometric is sampled from a person with an identity claim and the sampled biometric is compared to a biometric template stored previously for this person to validate the identity claim.
Biometric Fusion: The use of multiple biometric samples in a biometrics system. These samples can be collected through different modalities resulting in multimodal fusion, or multiple samples from a single modality or even a single sensor can be employed for fusion.
Biometric Template: The biometric template of a person is a pre-recorded biometric sample stored in a database for later authentication.