Study On Key Technology For Multi-view Facial Expression Recognition

The existing facial expression technologies perform really well under ideal conditions,but the vast majority of facial expression recognition performs badly on the occasion of face pose change.In order to solve this problem that the change of face pose cause serious interference to the facial expression recognition result,this paper have been deeply studied multi-view facial expression recognition key technologies,such as multi-view facial expression shape feature extraction,dimension reduction and facial expression recognition.The main study contents are shown as follows:1.Based on multi-view facial expression shape feature extraction method of the sequential cascade of linear regression,we puts forward the regression model of incremental update of multi-view facial expression shape feature extraction method.This method can accurately extract multi-view facial expression shape feature which needs a short time and does meet incremental update.The experiment carried on LFPW database show that the regression model of incremental update for multi-view facial expression shape feature extraction are better than the discriminative response map fitting method.2.Because the support vector machine(SVM)have good classification ability for small sample.The multi-view facial expression recognition method based on support vector machine(SVM)is proposed.This method is adopted view-specific classifier to recognition view-specific sample.The experiment carried on small sample database show that this method is effective for multi-view facial expression recognition.3.A multi-view facial expression recognition method based on regression model of incremental update and Discriminative Shared Gaussian Process Latent Variable Model is proposed.Firstly this method extraction regression model of incremental update feature,then its uses PCA to select the feature,finally adopts Discriminative Shared Gaussian Process Latent Variable Model to recognition multi-view facial expression.The experiment carried on CMU-PIE database show the effectiveness of our method.