Person Authentication Using Multiple Sensor Data Fusion

This paper proposes a real-time system for face authentication, obtained through fusion of Infra Red (IR) and visible images. In order to identify the unknown person authentication in highly secured areas, multiple algorithms are needed. The four well known algorithms for face recognition, Block Independent Component Analysis(BICA), Kalman Filtering(KF) method, Discrete Cosine Transform(DCT) and Orthogonal Locality Preserving Projections (OLPP) are used to extract the features. If the data base size is very large and the features are not distinct then ambiguity will exists in face recognition. Hence more than one sensor is needed for critical and/or highly secured areas. This paper deals with multiple fusion methodology using weighted average and Fuzzy Logic. The visible sensor output depends on the environmental condition namely lighting conditions, illumination etc., to overcome this problem use histogram technique to choose appropriate algorithm. DCT and Kalman filtering are holistic approaches, BICA follows feature based approach and OLPP preserves the Euclidean structure of face space. These recognizers are capable of considering the problem of dimensionality reduction by eliminating redundant features and reducing the feature space. The system can handle variations like illumination, pose, orientation, occlusion, etc. up to a significant level. The integrated system overcomes the drawbacks of individual recognizers. The proposed system is aimed at increasing the accuracy of the person authentication system and at the same time reducing the limitations of individual algorithms. It is tested on real time database and the results are found to be 96% accurate.