3D Facial Expression Recognition Algorithm Using Local Threshold Binary Pattern And Histogram Of Oriented Gradient

Facial expression which carries rich information of body behavior is the leading carrier of human affective and the symbol of intelligence.The main research object is 3D facial expression data,and our goal is to recognize 3D human facial expression.The research in this paper includes the expression feature extraction algorithm and integration with different kinds of feature.The main tasks are concluded as follows.(1)To contain additional local texture feature information,we put forward the concept Local threshold binary pattern(LTBP)features which based on Local binary pattern(LBP).The proposed algorithm calculate the difference of gray value standard between neighboring pixels and the center pixel as a threshold to binary instead of the traditional Local Binary Pattern(LBP)operation which only comparison of size between neighboring pixels and the center pixel.Compared with LBP features,LTBP features can describe the local areas that the dramatic changes of grayscale amplitude are different better and more effectively.After we get the LTBP feature,HOG(Histogram of Oriented Gradient)feature is used to describe facial local direction and gradient feature.(2)According to canonical correlation analysis theory,this dissertation put forward a feature fusion algorithm based on feature correlation criterion,the proposed algorithm uses the correlation between two sets of feature vectors as the criterion of feature fusion to solve the canonical projective vectors of LTBP features and HOG features,and then we integrate the LTBP and HOG(Histogram of Oriented Gradient)features to get multi-feature integration for 3D facial expression recognition.(3)In the experimental simulation,this dissertation uses the international mainstream 3D facial expression database BU-3DFE,and in the processing of 3D facial expression data,thinking of the different human face size,rather than take the current common method of using the fixed radius to cut the face with the nose as the center,we put forward a new method that calculate the position of center point of eyes based on the corner eyes inside and outside,then cut the face according to the distance between the center of the two eyes.This ensures that the proportion of facial features can be same roughly,in particular the eyes will be located in 1/4 height of the cut face.Then according to the algorithm,we take depth image to feature transform,establish the criterion function to describe the correlation between the two groups of feature vectors,then the fusion features of LTBP and HOG are obtained,At last,we use KNN(k-nearest neighbor)or SVM(support vector machine)to recognize the expression classes.The experimental results on BU-3DFE database show that the proposed method can recognize six kinds of basic expressions and has a high recognition rate.