Moment Feature Extraction And Support Vector Machine-based Face Detection Algorithm

With the development of modern technology and the progress of the society, using traditional ID card or password for identification appear some single. The information they must use are only a string of Numbers or letters. On one hand, they have lower safety. On the other hand, they have not enough effective information for identification. Then the researchers present biometrics recognition methods.Face recognition is the most natural, direct, and nonintrusive method among biometrics recognition methods. Face detection and recognition has been widely applied in image recognition tasks, such as identity authentication, electronic commerce, video surveillance, and human machine interaction, and it has become a challenging and hot research point in pattern recognition and artificial intelligence domain. The research of face detection and recognition has large theoretical and practical values. With the separation from face recognition, the face detection has been developed as an independent subject, and has played an significant role in many fields. For example new man-machine interface, content-based video retrieval and visual supervise and so on.This paper studies the face detection techniques and the support vector machine (SVM) technology and the feature extraction algorithm for face detection and identification. First, the author made a brief introduction to the situation of studies and analyzes the advantages and disadvantages of various methods. Then a new testing method of feature extraction based on moment was proposed. In this paper face detection was divided into classification of face or non-face and face positioning. The key point was the former. The main contents of this paper as follows:(1) The H u moment and the Tchebyshev moment were introduced in detail. Meanwhile the author analyzed their effectiveness of characterization for presenting face. The results of experiment show the effectiveness for describing face image.(2) In the design for classification, this paper mainly considers the support vector machine (SVM). Support vector machine (SVM) is based on statistical learning theory and it is a small sample-studying classification method. The double sample training and single sample training were considered in the paper. So-called double sample training was training classifier using face and non-face samples. The required parameters were obtained in a cross-training method. But single sample training were done only using face image, which dispense with the facial training sample selection. In the experiment, the results for presenting face of H u moment and Tchebyshev moment contrasted. The latter is better than the former, which shows the orthogonal moment has advantages to describe image features. At the same time, it shows that the classifying performance of support vector machine classifier was better than the classifying performance of linear classifier, Bayes classifier, K-neighbor classifier. And a good classification performance of single sample training classification was showed compared with double sample training.Finally, a summary and outlook was done. And the author analyzes the insufficiency in research and further research.