Facial Expression Recognition Based On Gabor And Neural Network

Facial expression is a method that depicts the mood, cognition and main body state. It is also the method which depicts their effect and role in social interaction. Mehrabian pointed out, if we analyzed the information of transferring between people, language accounted for 7%, tone accounted for 38%, and expression accounted for 55%. These data indicate the important effect of face expression in exchanges between people. And the Intelligence robot is the target that people seek all the time. Face expression recognition is able to provide the technology help for this target, and change the human-computer interaction way completely.Over the past few years, facial expression recognition has already become a very popular field, which is alternating in psychology, computer vision, principles of pattern recognition. Therefore, many problems need to study and solve in expression recognition field. The target of this thesis is to study the mothed of extracting and classifying facial expressional features.Firstly, the thesis introduces the history of facial expression recognition, summarizes its main method, and analyses its characteristic and difficulty. Base on the shortcoming of current method, we propose the method of extracting facial expression features which bases on Gabor and twice Discrete Cosine Tranform. Effective texture information is retained while reducing the dimension of expression characteristic vector.Secondly, we classify the expression characteristic, study the theoretics of BP nerual network and ensemble neural networks. Based on upwards, we design a three layers BP network, and build ensemble neural networks by bagging theory. Then apply them to classfy expression characteristic vector.Finally, we experiment with the CUM's Cohn-Kanade facial expression database to validate the validity of the method which proposes in this thesis...