The Research Of Real Time Facial Expression Recognition Based On Multi-feature Integration

Facial expression and gesture play an important part of our social interaction. The psychological research shows that facial expression contains more emotion information than sounds. Therefore, Human Emotion Computer Automatic Analysis is a key step in realizing natural Human-Computer Interaction. In recent years, facial expression recognition research has become a popular object. Facial expression recognition is a challenging problem in the fields of pattern recognition and psychology. Based on the former research achievements, a Multi-feature extraction is proposed in this paper. The main contributions are as follows:(1) Face detection. Haar-Like feature based, use Adaboost learning algorithm to detect the face area in the picture.(2) Locate facial expression area. Facial expression information is focus on the mouth area and eye area. To get effective feature, this paper adopt the pixel gray project method to locate the mouth area and eye area.(3) Multi-feature extraction. Multi-orientation, multi-frequency gabor filter is applied to extract the texture feature of mouth area and eye area. Because of the large number of texture feature, the principal component analysis is used to reduce the feature number. Get the geometric feature from picture grey level.(4) Construct a multi-classes C-support vector classifier. Use Decision Directed Acyclic Graph C-support vector classifier to classify seven facial expressions. Adjust DDAG-SVM parameters and structure to achieve a better accuracy.(5) Develop the automatic facial expression recognition system, and analyze the experiment result and design the future research.