Face direction estimation and face recognition using multiple cameras for communication in a virtual environment

This paper proposes a hierarchical method of face direction estimation and face recognition based on the "four directional features" with linear discriminant analysis. We expand this method to an approach with multiple cameras. Experiments show that our proposed method can achieve a recognition rate 92.6% for the faces that took in the extent of about 300/spl deg/ and the co-operation of multiple cameras can recognize occluded face images.