An Optimal Content Matcher for Accurate Human Recognition in Multimodal Biometric Systems

The performance of multimodal biometric systems usually advances with increased use of number of modalities. A multimodal biometric system is designed to integrate extensively used reliable physiological traits namely face, finger and palmprint at feature level. The proposed multimodal system accuracy is enhanced by incorporating novel and efficient optimal content matcher adhering to Monge property and North West Corner transportation method. This optimal matcher is highly significant in large scale biometric identification systems in terms of maximizing the performance, reducing the response time with lesser iterations and cost effective. The fusion at feature level is considered to consolidate the evidences of face, finger and palmprint in contrast to matching or decision levels because the feature vector consists of wealthier data on the biometric origins. The experiments are carried out with real time datasets of physiological traits acquired from engineering graduates. The results proved that optimal content matcher accomplished with feature level fusion achieves higher recognition accuracy when compared with various multimodal systems employing other fusion levels and matchers for identification.