Efficient person identification by fusion of multiple palmprint representations

The automatic person identification is a significant component in any security biometric system because of the challenges and the significant number of the applications that require a high safety. A biometric system based solely on one template (representation) is often not able to meet such desired performance requirements. Identification based on multiple representations represents a promising tendency. In this context, we propose here a multi-representation biometric system for person recognition using palm images and by integrating two different representations of the palmprint. Two ensembles of matchers that use two different feature representation schemes of the images are considered. The two different feature extraction methods are the block based 2D Discrete Cosine Transform (2D-DCT) and the phase information in 2D Discrete Fourier Transform (2D-DFT) that are complementing each other in terms of identification accuracy. Finally the two ensembles are combined and the fusion is applied at the matching-score level. Using the PolyU palmprint database, The results showed the effectiveness of the proposed multi-representation biometric system in terms of the recognition rate.