Multimodal biometric verification and identification using face and hand

Biometric identifiers – physiological or behavioral characteristics of a human – are more trustworthy and more suitable than knowledge-based and token-based techniques in distinguishing between a confirmed person and an impostor [1]. In this study, face and hand geometry have been chosen as biometric identifiers that identify and verify users in unimodal and multimodal cases. Face recognition and verification modalities of this study are built upon 2D Gabor wavelets where Gabor kernels having eight different orientations and five different frequencies have been convolved with the images of the users [2]. Hand recognition and verification modalities use ICA-based features [3]. Segmentation, hand registration and finger registration are the significant phases of this study that pre-process the hand image before the extraction and recognition steps. The fusion performance of the system has been analyzed by using the methods of Borda Count, Fixed Arithmetic Combination and Confidence-aided Fusion.