Feature Fusion of Palm and Face Based on Curvelet Transform

This paper presents feature level fusion approach using multi resolution Curvelet transform for face and palm biometrics.In this paper feature extraction has been done by taking the curvelet transform of bit quantized images.The curvelet coefficient thus obtained acts as feature set for classification.The five sets of coefficients from five different versions of images are used to train five SVMs.During testingthe results of SVMs of palm and face are fused in a single column feature vector to determine the final classification.The results of fusion are compared with that of unimodal biometric system of palm and face separately.Using a common feature extraction method for both unimodal and fusion helps in analyzing the efficiency of recognition.The experimental results show that the proposed scheme out performs the unimodal biometrics using curvelet transform.All the experiments are carried out on two well-known data bases AT&T for face and POLY U for palm print images.