Hybrid Method of Multimodal Biometric Authentication Based on Canonical Correlation with Normalization Techniques

Abstract

Multiple biometric models can be used to implement a highly reliable biometric authentication system. A biometric system that uses hybrid methods to authenticate a person typically from a wide range of sources divided into discrete groups based on their features. Combining different modalities, each modality is viewed from its own perspective in this system, resulting in a new perspective of biometric authentication. In this paper, a principal components analysis (PCA) is used to reduce the dimensionality of the face and hand image features. A combination of face, hand, and social behavioral twitter information is used to identify individuals. Despite the first two modalities being synchronized but user twitter information’s are not synchronized with them: this prevents using the data fusion to create feature-level fusion. A hybrid method can be used to obtain an appropriate feature-fusion technique. A Canonical correlation analysis (CCA) is used for the analysis of feature fusion of the face and hand images. These modalities are proposed in the current study as the feature-level approach must be fully exploited, in order to maximize its benefits and combines with tweeter social behavioral (SB) features using normalization techniques to get single feature vector for a registered database for matching. In this study, the results show that feature-level fusion using hybrid method allows good authentication of the individuals. The overall performance of the system is better than the existing one’s with respect to true acceptance rate.