Research And Implementation Of Facial Expression Recognition Algorithm

With the rapid development of computer technology,more and more people pay attention to the artificial intelligence.Moreover,with the increasing demand of the facial expression recognition technology in the field of artificial intelligence,facial expression recognition has become the researching focus of experts and scholars.They mainly divide the basic human expressions into 7 categories,and make feature extraction and classification by a variety of algorithms,among which feature extraction is the focus of the study.The main contributions of this paper are shown as follow:1)Through the operation of histogram equalization and gray integral projection on the expression image,the position of the human eye is located,furthermore,the angle and size level of the image are normalized,which make the image unified and the effect of facial expression recognition could be obtained in the next process.2)For the problem of the higher computational complexity and dimensionality of the multi-scaled and multi-directional 2D-Gabor conversion,this paper adopts the feature extraction algorithm based on the enhanced multi-directional 2D-Gabor filter feature extraction,improves the 2D-Gabor filter and carries out the feature extraction combining with Local binary pattern(LBP).First,it extracts the amplitude images with 5 scales and 8directions from the expression images by using 2D-Gabor filter and fuses the amplitude images with the same direction and different scales.Then,it makes a quadratic dimension reduction by the coding in uniform pattern of LBP.Finally,it makes a statistics of weighted partitioned histogram as the eigenvector.The proposed method in this paper can reduce the dimensionality and also retain more feature information of the facial expression image.3)In order to design and implement the facial expression recognition system,this paper selects JAFFE facial expression database on the MATLAB experimental simulation platform as the test sample.On the basis of the above methods,it uses the support vector machine classification method at the classification and recognition stage,and finally gets the classification result.Conduced from experimental results,a recognition rate of 90.24%can be obtained by JAFFE expression database,which verifies the effectiveness of the scheme.