A Study Of Face Recognition Methods Based On Wavelet And SVM

Biometrics is a kind of science and technology using individual physiological or behavioral characteristics to verify identity. It provides a highly reliable and robust approach to the identity recognition. Automated face detection and recognition is one of the most attentional branches of biometrics and it is also the one of the most active and challenging tasks for computer vision and pattern recognition. It is not only widely applied in a variety of personal identification systems such as national security, public security, justice, government, finance, business and security facilities, but also can be used in the fields of human-computer interface and visual communication.This dissertation mainly studies the approaches to frontal face detection and recognition in gray images. The main contributions are as follows:1. For the frontal upright faces in gray images, a knowledge-based face detection method is proposed. A knowledge base is established according to prior knowledge of faces. The integral image method is used to calculate sub-block's statistic values and the computation burden is reduced significantly.2. A method of face detection based on support vector machine (SVM) is put forward. The features are extracted by applying the discrete cosine transform (DCT) to the preprocessing image. The DCT coefficients are inputted to the SVM and the SVM are trained using the cropped face samples and the "bootstrapping non-face" samples. The simulation results are satisfactory.3. A method of face recognition based on wavelet decomposition and LDA is proposed. The low frequency sub-images are obtained by utilizing two-dimensional wavelet transform for several times. The features are extracted by applying LDA to the sub-images. The nearest-neighbor classifier is designed to recognize the face. The simulation results show that the new method has better performance than the traditional LDA, and the computation burden is reduced greatly.4. A method of face recognition based on wavelet and DCT is proposed. The low frequency sub-image is transformed by DCT, and only a small set of coefficients is retained as the feature. A Euclidean distance nearest-neighbor classifier is designedto recognize the face. This method has small computation burden and high recognition rate.5. A method of face classification and recognition based on DCT and SVM is proposed. The features are extracted by DCT and a small set of DCT coefficients are inputted to SVM. The experiments show that the performance is satisfactory.