A Model Study for Face and Iris Feature Fusion and Recognition
Jianhu Gao · Dianzi xuebao · 2007
Multimodal biometric fusion and identification technique can improve the performance of identification system.Combined with 2-Dimensional Fisher Linear Discriminant Analysis(2DFLD),a model for face and iris feature fusion and recognition is presented in this paper.Firstly,compression is done to face and iris image respectively,and the two corresponding original feature matrixes are obtained.Secondly,the two original feature matrixes from face and iris image are integrated into one matrix,and a combined feature matrix is formed.Then feature extraction to the combined feature matrix is done by 2DFLD,and a fused feature matrix is constructed.Finally,Nearest Neighbor Decision(NND) rule is used in recognition.Experimental results on ORL(Olivetti Research Laboratory) face database and CASIA(Chinese Academy of Sciences,Institute of Automation) iris database show not only small sample effects can be solved,but also a high correct recognition rate can be gained,and demonstrate that the valid model is supplied for multimodal biometric identification.