Exploiting feature space using overlapping windows for improving biometric recognition

Abstract

Biometrics is a highly researched topic due to its importance in security, surveillance, and authentication systems. Granulation is the procedure of partitioning data into windows. Two novel feature extraction techniques using overlapped granules based on the texture features of palm-prints are exploited here. The first category of attribute involves overlapping features that increase the information content by increasing granulation, resulting in additional features. The second attribute is differential information feature (DIF) which involves the first derivative of intensity representing feature dynamics. To further improve the performance, score level fusion is applied. The improvement in the identification values of 2 to 4 percentage points in general and up to 6 percentage points in some cases is seen. In most cases, score level fusion has shown better performance than unimodal methods. The ROC curves showed the superiority of the proposed method over other existing methods.