Research On Facial Feature Extraction And Recognition In Automatic Facial Expression Analysis System

Facial expression contains numerous information of human behavior. In human society, facial expression is the very important communication method besides words. As the carrier of information, expressions can convey lots of information that words can not transport. Automatic Facial Expression Analysis(AFEA) extracts the features of human facial expressions information, then classify and synthesis them according to the human cognization. It usually contains three basic parts: pretreatment of facial images, facial expression features extraction, and facial expression recognition. In this paper, we reviewed the history and the recent development of facial expression, studied the methods of facial expression recognition in the way of pattern recognition and statistic study. We discussed the model-matching algorithm in the feature extraction and proposed a different model-matching algorithm, which efficiently reduced the computation with almost the same right rate as the gabor wavelet. For the purpose of automatic facial expression recognition, we also discussed Active Shape Model(ASM), and used it into the real-time system.We proposed a facial expression recognition method based on support vector machine. Multi-classifiers were constructed and the different ways to build them are compared. We also used Kernel Canonical Correlation Analysis to recognize the facial expressions. It extend the traditional CCA, which is used in linear feature analysis, to the non-linear analysis field. We also compared the different policies in recognition. The experiment result shows the efficiency in facial emotion recognition.