Multiple Facial Features Representation for Real-Time Face Recognition

The combination of two face feature extraction methods for face recognition is proposed. The proposed approach treats the face recognition problem as a one-dimensional (1-D) problem rather than two-dimensional (2-D) geometry. The horizontal projection and the statistical distribution of facial gray image are adopted respectively as 1-D energy signal representation for each face image. To reduce the dimension of signal and improve the performance, the wavelet transform is proposed. Finally, the probabilistic neural network is used to recognize each individual. The performances of the proposed method are evaluated and compared with other proposed methods on ORL database and IIS database. The experiment results show that the performance of the proposed method is much better than the other methods. Besides, we developed a computer system that can capture face image in a complex background and recognize the person by comparing characteristics of the face to those of known individuals. The proposed algorithm is also evaluated on a real environment database and the results are encouraging. Experimental results show that the proposed method possesses excellent performance as well as low memory requirement.