Facial Expression Recognition Research Based On Embedded System

Facial expression characteristic is an external form of the inner world. In recent years,face recognition technology has become a very popular research topic. Computer facialexpression recognition contains two main aspects, one is the human face recognitiontechnology, and another is the classification of facial expression which need to be classified tothe six basic expression delight, fear, surprise, anger, hatred, grief.In this paper, the face recognition algorithm is proposed based on skin color, in which thecore algorithm is to calculate the color likelihood figure of a picture. After the imagebinaryzation processing, the extraction of the face is completed by combining with the priorknowledge of the facial features.The facial expression is thoroughly analyzed and the method of half-face detection isproposed for facial expression recognition with the problems of face shade, slope, andunbalanced exposure on both sides. Firstly, the image is preprocessed to finish the geometricnormalization in expression regions and histogram equalization. Secondly, Gaborwavelet-based feature extraction is used in the expression area. Then with the elastictemplates matching, the Euclidean distance between the measured image and the standardimage is calculated. Finally facial expression is classified with the K neighbor expressionclassification method to get the recognition results.The experiments show that the face recognition algorithm based on skin color canaccurately extract facial region. Using Gabor wavelet-based feature extraction and Elastictemplates matching can recognize facial expression and resolve the problems of face shade,slope, and unbalanced exposure on both sides. Comparing with traditional methods, half-facemethod is faster in recognition speed. The recognition effect is better after normalization.In addition, an embedded platform based on the Android system framework is built inorder to port the facial expression recognition algorithm.