Gabor And Local Binary Pattern-based Human Face Expression Recognition

Facial expression recognition is one of the most challenging problems in the fields of affective computing, image processing, machine vision, movement tracking, pattern recognition, biometric identification, physiology and psychology. Facial expression recognition is an important part of affective computing and intelligent human-machine interactive.Based on the former research achievements, some improved algorithms were proposed to overcome the difficulties in the steps of feature extraction and expression classification. The major contributions of this paper are as follows:(1) Because the Gabor filter which is insensitive to the illumination, can tolerate a certain degree of image rotation and deformation and improve the robustness of the system. This paper uses Gabor filter to extract the features of facial expression to get the Gabor Coefficient Maps (GCMs).(2) Because the Local Binary Pattern (LBP) operator is extremely invariant to gray and rotation, it can solve the problems of rotation, displacement and uneven illumination which come out in the period of image pretreatment. Thus this paper uses the LBP operator encoding on each GCM to extract the Local Gabor Binary Pattern (LGBP) of facial expression. Then we use LBP operator to code the Gabor coefficients in the neighborhood region, and use histogram to analyze the partial variation attribute after LBP coding. Through analyzing GCM directly, we can get the Histogram Sequence of Local Gabor Binary Patterns (HSLGBP), and reduce dimension simultaneously.(3) Then we use multi-class Support Vector Machine (SVM) to do the feature classification for its good performance in the case of small samples. To reduce the computation and improve the recognition rate, through training samples we can get the transcendental knowledge of the recognition result of every scale and every orientation HSLGBP, and then we get the final recognition result by combining the transcendental knowledge with the estimative value of the test sample.(4) Facial Expression Recognition System (FERS) is developed. It consists of such modules as video reading, eye location, face detection, image pretreatment, feature extraction and expression classification. Users can choose the video frequency or static images to do the facial expression recognition.