Bimodal Multi-Feature Fusion Based on Quaternion Fisher Discriminant Analysis

Because of its higher reliability, wider applicability and stronger security, multimodal biometrics has become a polar research direction of biometric recognition and attracts more and more research groups focusing on this area. Along with other fusion level of multimodal biometrics, feature level can reduce the redundant information to avoid calculation consumption, and simultaneously acquire the discriminative information to improve the system performance. The traditional methods of fusion level can only fuse two single feature modalities or one modality with two kinds of feature. This paper imported quaternion concept and proposed quaternion Fisher discriminant analysis that can fuse two modalities with four different features. Face and palm are selected as the experimental object and extracted the linear feature and the non-linear feature by PCA and KPCA respectively. Experimental results show the proposed algorithm achieves much better performance than four single feature recognition algorithms.