摘要：This paper proposes anovel method to recognize facial expression through ActiveAppearance Model (AAM)to extract facial regions based on Facial Action CodingSystem (FACS). Itis composed of three parts: extractionof facial regions based on AAM,extraction of facial featuresby Gabor wavelettransformation, and expressionrecognition through Support Vector Machines (SVMs).AAM has better performance thanother methodsin eliminations of the influenceof different facialregion size, head pose and lighting condition and thus can effectively increase the recognitionaccuracy. Therefore it is usedto extract facial regions before extracting features by Gabor wavelettransformation. Finally, SVMsis appliedto recognize expression for its advantage of solvingthe problems of small sample size and overfitting. The feasibility and effectiveness of this method are evaluated and verified by experiments, and satisfactoryresults are achieved.
