A Novel Automatic Facial Expression Recognition Method Based on AAM

This paper proposes a novel method to recognize facial expression through Active Appearance Model (AAM) to extract facial regions based on Facial Action Coding System (FACS). It is composed of three parts: extraction of facial regions based on AAM, extraction of facial features by Gabor wavelet transformation, and expression recognition through Support Vector Machines (SVMs). AAM has better performance than other methods in eliminations of the influence of different facial region size, head pose and lighting condition and thus can effectively increase the recognition accuracy. Therefore it is used to extract facial regions before extracting features by Gabor wavelet transformation. Finally, SVMs is applied to recognize expression for its advantage of solving the problems of small sample size and overfitting. The feasibility and effectiveness of this method are evaluated and verified by experiments, and satisfactory results are achieved.