Multimodal biometric systems use multiple biometrics traits to increase the recognition rate. The fusion module plays a key role in multi-biometric system performance. This paper presents a novel multimodal rank reinforcement approach based on the prior resemblance probability distribution of each identity in the training data. The resemblance probability distribution is used before the fusion to reinforce the rank list of each biometric matcher.
In this paper, we developed a multimodal biometric system based on the frontal face, the profiles face, and the ear. The experimental results show the ability of the prior reinforcement in increasing the accuracy of unimodal biometrics systems as well as increasing therecognition rate of various rank level fusion approaches.