Face Recognition Based On Support Vector Machines

Facial expression recognition becomes a hotspot in scientific research field in recent decades,which is to analyze and detect the special expression state from given expression images or video frames and then to ascertain the subject's specific inborn emotion, achieving smarter and more natural inter-action between human beings and computers. Facial expression recognition has potential application values in many fields, including psychics study, image understanding, synthetical face cartoon, video retrievaling, robot technology, virtual reality and the study and develop of new human-computer-interface environment based on facial expression.In this work, we firstly discuss the background and then analyze the main expression recognition algorithms presented, emphasizing on Principle Components Analysis(PCA), wavelet transformation, Optical Flow Models(OFM) and Hidden Karkov Models(HMM).For the Gabor wavelet transform can extract statistic expression images'distortion features effectively, and the Support Vector Machine has strong classification ability. Then we present the expression feature extraction algorithm based on Gabor wavelet transform and geometry. At last, classify the expression using Support Vector Machine(SVM). The methods are discussed as below:Firstly, after reading reference literatures on facial expression recognition carefully, we summarize and decide to study single static image's expression recognition. Then download the basic expression database——JAFFE.Secondly, we point out manually 34 fiducial points on each image in our expression database(together is 213),and record them.Thirdly, we construct Gabor wavelet bank in 3 scales, 6 orientations. Each image is convolved with the 18 Gabor functions. After Gabor filtering, the amplitude values at selected fiducial points on the face images are used as Gabor coefficients. That is,we use a row vector of 612(=34*3*6) dimensions as feature value which is used for next classification.Lastly, we use Support Vector Machine(SVM) as classifier. The 213 images in the expression database was divided randomly into ten roughly equal-sized parts, from which the data from nine parts were used for training the classifiers and the last part was used for testing. This procedure was repeated ten times.In this article, the feature extract algorithm of Gabor wavelet transform is combined with the classification algorithm of Support Vector Machine to distinguish seven kinds of basic expressions on static images. We designed the Gabor filter bank based on the filters used perviously for texture segment and image retrieval, which furthest shields illumination effect and the difference of personal feature. And applied the method of Support Vector Machine(SVM) to expression recognition. The above methods obtained a high recognition rate when we used the basic expression database—JAFFE to train and test. Our experiment results indicate that the method is a kind of effective expression recognition method.