Biometric recognition using frequency domain methods

Verifying a user’s identity is critical in access control applications. Most current authentication systems are password based making them susceptible to problems such as forgetting the password and passwords being stolen. One way to overcome these problems is to employ biometric recognition which typically involves two stages; a training stage and a recognition stage. During the training, samples of a subject’s biometric signatures are used to create a biometric template. During recognition, the subject’s live biometric is matched against a database of stored templates to identify the subject. In a highly controlled setting where the training and recognition environments are very similar, biometric matching can be relatively easy leading to good recognition performance. However, the reality is that the biometrics from the same person may differ considerably between training and recognition. For example, a face image might appear very differently because the illumination levels between an indoor enrollment session and an outdoor verification session may be extremely different.