Face and Iris Wavelet Feature Fusion through Canonical Correlation Analysis for Person Identification

The paper presents, a new technique for person identification using feature level fusion of face and iris biometric features. The fusion technique employed in the proposed work uses wavelet features of face and iris modalities. Canonical Correlation Analysis (CCA) technique is used to fuse face and iris wavelet features. CCA is a technique for extracting linearly correlated face or iris features from set of features of face or iris images. The technique is able to extract and enhance the discriminative power for high dimensional feature space for person identification from multi-feature information of face and iris. The experimental results on both synthetic and genuine multimodal data sets of face and iris features using SVM classifier validate the effectiveness of the proposed method. The proposed face and iris multimodal biometric system has achieved 100% recognition accuracy on VISA Face + VISA Iris, ORL+ VISA Iris, LFW + VISA Iris, VISA Face + CASIA, ORL+ CASIA and LFW + CASIA face and iris datasets.