Facial expression recognition based on weighted principal component analysis and support vector machines

This paper presents a new method of facial expression recognition using the multi-feature fusion weighted principal component analysis (WPCA) and the improved support vector machines(SVMs). It employed the WPCA with multi-features to extract the facial expression feature and the SVMs to classify human facial expression. A simple way based on the distribution of action units in the different facial area is introduced to determine the weights. The detailed procedures for facial expression training and recognition algorithms are given. Facial expression recognition experimental results on the CKACFEID facial expression database indicate that Radial Basis Function (RBF) SVM performs better than Linear and Polynomial SVMs. We also provide experimental evidence that the proposed method using the WPCA which is convenient to get the training templates, easy to match and recognize has a higher recognition rate for all the basic expressions than other methods using the pure PCA (PPCA). The final recognition rates of the WPCA can achieve 88.25% whereas the PPCA gives 84.75% in our experiments.