Facial Expression Recognition Based On Haar-like Feature And Expression Validity Analysis

Facial expression recognition is the process of analyze facial expression and itschanges by computer and then verify its inner feelings and ideological activities. Acomplete facial expression recognition process consists of three important technicalaspects, which are the detection and process of face region, the extraction of facialexpression feature, recognition of facial expression. This paper mainly studies facedetection and process, facial expression feature extraction, facial expressionrecognition, and then analyze the valid period of each facial expression.Haar-like feature and Adaboost are used to detect face region in a given picture.Each Haar-like feature is trained as threshold-based weak classifier, and Adaboost isused to extract some Haar-like features. Then a strong classifier is constructed bythose selected features. After face detection, pre-processing operations of Histogramequalization on face region is used in this paper.On the basis of face detection, Haar-like feature is used to represent facialexpression. Because facial expression classification is a multi-classification problem,facial expression features are extracted by multi-class real Adaboost and featuretemplate is constructed on the basis of extracted facial expression features. Facialexpression feature extraction and facial expression recognition are constructed indifferent face database, and the experimental result indicates the feasibility of theproposed method and it has higher pervasiveness.Because the state and strength of each facial expression are not immutable, theanalyzed results of facial expression are effective within a certain time, so a model isproposed in this paper to analyze the valid period of facial expression states.Facial expression recognition based on Haar-like feature is conducted after facedetection, the proposed model improves pervasive and maintains high recognition rate,an analysis model of facial expression valid period is proposed at last.