Application Research Of Facial Expression Recognition Based On Supported Vector Machine

As a new biometric identification technology, facial expression recognition gradually has become the focus of the field of pattern recognition and artificial intelligence at home and abroad in the recent years. Although the facial expression recognition technology develops very fast, the current expression feature extraction and expression feature recognition technology are still not perfect because of the special features of face as a flexible body and the diversity and complexity of expression change. The accuracy and robustness of traditional methods based on key point positioning is not enough. At the same time, the method based on Gabor wavelet,of which the classification accuracy is higher, but the characteristic vector dimension is very big and the computational efficiency is not high. At the same time, as the representative of traditional facial expression recognition theory, principal component analysis(PCA) cannot solve the small size and nonlinear problems. The expression feature extraction and feature recognition is the focus of the study,and the major works around these issues are as follows:① The paper choose 33 key points as facial expression feture points,and approve the precision of detection by using Face++ cloud service of key points detection.② Propose the expression feature extraction method based on the variation coefficient. Firstly, the statistics of the key point coordinate values of a large number of training samples must be done, and mean and coefficient of variation of all characteristics should be analyzed; furthermore, the characteristics of the higher "energy" is selected as an expression characteristics by sorting the active degree of the characteristics; lastly, the appropriate characteristics are selected as a standard to standardize the feature vector.③ Design expression recognition classifier test system through support vector machine(SVM) theory based on Matlab and Face++ platform. This system can recognize facial expression on static picture and video.And do the comparison and analysis for the test result.