查看文章

In this paper, a hybrid approach for fusion of iris and face is presented using multi-instance and multi-modal sources of biometric information. Fusion is performed at feature level as well as match score level. It provides security at two steps. At first step, multiple instances of iris (left iris and right iris) are combined at feature level to generate a single merged feature set of iris. Instead of storing actual features in database, this resultant feature set obtained after fusion of left iris and right iris is stored as a template in the database. Thus, iris template becomes more secure and rich in information. At second step, match scores of merged iris and face are combined at match score level to generate the final decision whether the person is accepted or rejected. The proposed fusion strategy is blend of two fusion approaches (multi-instance and multi-modal) making it capable to overcome the drawbacks of unimodal biometric systems. Experimental results show that hybrid multibiometric system outperforms unimodal biometric systems.