Research And Implementation Of Emotion Recognition System Based On Gabor Transform

With the continuous development of science and technology,computer increasingly becomes an integral part of people's daily life,and intelligent interactive technology has developed rapidly meanwhile.Taking the communication ways of language,body movements and facial expressions among people as consideration,how to establish effective communication with humans by machine under the condition of non-verbal has become a problem we has to be faced.In the non-verbal communication,the person's expression is a true reflection of his or her inner state,so humans can understand his or her inner emotional states by observing each other's facial expressions to achieve the purpose of communication.If the machine has the ability to recognize human facial expressions,it is also possible for machine to establish effective communication with humans.Expression recognition is the technology that captures the current facial expression state of human and analyzes it to identify the category it belongs to through some way.Expression recognition process includes three steps,namely image preprocessing,feature extraction and expression classification.Preprocessing mainly refers to the various normalization process of expression image.The feature extraction and facial expression classification are two more important part of the process,and that the extracted facial features are effective or not directly determines the expression the classification's accuracy,while a good expression classification method plays a decisive role in identifying facial expression correctly.In the existing expression recognition methods,the most of them are used to recognize static image.The relevant technology of this part is relatively mature and can reach a high recognition rate.This paper tries to recognize the facial expression in video.In the feature extraction phase,based on Gabor transform,respectively this paper has proposed two different approaches to image feature extraction,then used a neural network and SVM classification method to classify.By contrast experiments,the result shows the effectiveness of the methods used herein,as well as its advantages in terms of feature extraction.In the face recognition phase,based on the learning and researching a hidden Markov model with time-domain characteristics,this paper has proposed a HMM model which has a three-layer structure,then takes the model as the way of a facial expression recognition method for image sequences,then verifies the effectiveness in expression classification through experiments,and explores its application in the field of other aspects of pattern recognition.