Facial Expression Recognition Based On Geometrical Characteristics And Subspace Learning

In recent years, facial expression recognition becomes a studying focus because of the increase of interest in human-computer interaction. Facial expression recognition plays very important role in human communication, which is important supplement to voice communications. Facial expression recognition is proposed in this background, and the purpose of facial expression recognition is to solve the difficulty in the feature extraction, low recognition rate and low speed in actual application. For the classification effect of facial expression recognition, the present expression classification method still exist many deficiencies, and there are a lot of difficult problems to solve until now there hasn't been a practical system. The research contents of facial expression recognition mainly consist of three following parts: face detection, feature extraction and selection, and expression classification. The research topic mainly researches some key issues in the process of feature extraction and selection, and expression classification. In the recent years, based on the analysis of the algorithms about facial expression recognition, this research topic emphatically studies algorithms about feature extraction and selection and classification methods, and carries on the related tests.The main contents of this paper can be summarized as follows:(1) According to the geometric features of facial expression which exists, a feature table that covers almost all geometric features appeared in the literatures is summed up.(2) Expression image preprocessing is studied. Expression image preprocessing mainly include geometry preprocessing and grayscale preprocessing.(3) Based on a lot of methods of extract features, Gabor wavelet transformation is adopted, and image grey-scale value is substituted for Gabor wavelet transformation coefficients to reduce the sensitivity to variations of lighting and position.(4) Classification algorithms are studied and genetic algorithm is used to optimize expression features.(5) According to the research papers in the feature extraction and face recognition algorithm, a human facial expression recognition system is built. This system can identify happy, sadness, anger, disgust, surprise, fear and neutral expression.Experiment results show the method adopted in this paper can effectively distinguish the seven expressions and can achieve good results.