Research On Facial Expression Recognition Based On Basic Faces

Facial expression recognition is a comprehensive topics related to many fields and subjects such as computer vision,artificial intelligence and pattern recognition.It also can be widely used in image processing.In this paper,we focus on the technology about facial expression recognition,which contains the image preprocessing,the feature extracting and the method of expression recognition with facial classification.Compared to the traditional method of expression recognition without facial classification,the expression recognition with facial classification takes into account the impact of faces on expression recognition.It constructs different feature libraries for different faces,as a result it can achieve higher recognition rate.For this method,with more faces and libraries,the recognition rate is much higher.But too many libraries bring restrictions to the system.At last we propose the facial expression recognition based on basic faces,which just constructs feature libraries for two basic faces,namely oval and round faces.When an expression is imputed into the system, it is first extracted by Gabor filter and sparse dimensionality reduced.Then we can get the linear fitting expression feature of the test sample from feature libraries of the basic faces,with which and nearest neighbor classifier the process is complete.This paper makes two experimental analysis on JAFFE facial expression database.The first experiment compare the real feature and the fitting feature by drawing curves,and the result shows that the method of fitting feature is effective.The second complete the expression recognition from the basic faces,the final average recognition rate is 65.08%.The low recognition rate proves that the method of basic faces remaind to be improved.