Ensemble of multiple Palmprint representation

In this work, a new approach for personal authentication using palm image is presented. We design three ensembles of matchers which employ different feature representation schemes of the images: Discrete Cosine Coefficients; Invariant Local Binary Patterns; Gabor Filters. Each ensemble is obtained by varying the features used to train their matchers. Experimental results confirm that the three methods give complementary information which has been exploited by fusion rules. Finally, we combine our Palm based method system with other biometric characteristics that can be extracted from the hand (middle finger, ring finger, hand geometry), obtaining a further improvement of the performance.