A Novel Dual-Modal Biometric Recognition Method Based on Weighted Joint Sparse Representation Classifaction

Chunxin Fang, Hui Ma, Zedong Yang

Abstract: The dual-modal biometric recognition based on feature-level fusion is an important research direction in identity recognition. To improve the performance of identity recognition, we propose a novel dual-modal biometric recognition method based on weighted joint sparse representation classification (WJSRC). The method introduces joint sparse representation classification (JSRC) to fuse fingerprint and finger-vein features at first. Then, a penalty function is constructed between the test and training samples to optimize the sparse representation. Finally, the image quality scores of samples are utilized to construct a weight function to optimize the decision-making. The experimental results on two bimodal datasets demonstrate that the proposed method has significant improvement for the accuracy and reliability of identity recognition.