Multimodal Biometric Identification using Feature Fusion

Abstract

The Biometric features are already proven to be robust for forensic and security purposes. The fusion of multiple biometric features in a systematic way looks more promising. This paper addresses combining multiple biometric modals using proposed fusion technique with focus on face and palm print features. Five databases are used for experimentation on face, Face94, Face95 and Face96, FERET and FRGC and two dataset used for the palm print, PolyU and IITD database. Transform based features used are extracted from these databases using Gabor transform, Radon transform, Ridgelet transform and Radon-Gabor transform,, FPLBP, TPLBP. Feature level fusion has been applied using algorithms FFVM, FFVW. As per our study accuracy for fusion using TPLBP is 100 %, for FFVW method. Thus, the above feature level fusion technique is recommended, based on better accuracy and robustness