A method of facial expression recognition based on LBP fusion of key expressions areas

For facial expression recognition, the LBP feature is an important way of texture feature, but usually the whole of image is taken as extracting area, ignoring to extract the key areas of facial expression. In order to solve this problem, based on previous LBP feature extraction method, as well as the division of facial motion unit, we put forward a kind of expression recognition method using the fusion feature of key facial areas expression based on LBP, by dividing into several parts: eyes, eyebrows, between-eyebrow, nose, mouth, then we get the key areas of expression to extracted features independently, at the meaning time to hold global facial features, features of the whole facial is also extracted. After that we combine this two different features together and get a new feature which is called combine feature fused key expression ares. The features combined then classified by SVM and NN to recognize different expressions. This article carries on the experiment in JAFFE database, the results show that the method of facial expression recognition rate obtained obvious ascension.