Skip to Main Content
In this paper, we proposed a method for facial expression recognition. Unlike other existing methods, our method is a simple and fully automatic in which firstly, the face region is detected using local SMQT features, then a Pictorial Structure model is applied to extract key parts of face (consist of left and right eye and mouth) in different facial images with very little time. Then a set of Gabor filters with different angels is applied on each part and whole face to create new representations of them. In the next step, we extract useful features of these parts and their representations using LBP operator. Afterward, Support Vector Machine classifier is adopted for each part as well as whole face and their different representations. Finally Majority voting is applied to combine the votes of adopted SVMs to classify facial expressions. The algorithm is implemented with MATLAB and experimented on JAFFE database. We obtained a facial expression recognition rate of 77.62% for person-independent that shows the effectiveness of the proposed method than other investigated algorithms such as PCA, 2DPCA, LDA and simple LBP.