Font Size: a A A

Facial Expression Recognition Based On Gabor Feature BC Encoding

Posted on:2017-02-04Degree:MasterType:Thesis
Country:ChinaCandidate:X M XuFull Text:PDF
GTID:2348330485462210Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the key methods of computer vision and pattern recognition continuously put forward and improving, facial expression recognition as an important part of affective computing which has become a research focus among many domestic and foreign scholars. Through the study of facial expression recognition, we can make the intelligent machines have the ability of emotional computing, recognition, expression like people do. Detailed analysis and experiments have been done about the feature extraction. So that can effectively increase the friendly and intelligent interaction between humans and machines.Firstly, this paper gives a detailed elaboration about the background and significance of facial expression recognition, and then introduces the overall framework of FER system. Do detail and analyze the advantages and disadvantage of classical algorithms which in the FER field. Following is the main research work and innovations:(1) Expression image preprocessing:the preprocessing effects have a great impact on the late feature extraction and recognition. Through the image scale size, position, and gradation processing can remove the negative factors on expression classification so that can improve the accuracy of face recognition.(2) Expression feature extraction:images do Gabor wavelet transforms after preprocessing. Because Gabor graphs in different direction at the same scale have information redundancy, inspired by the LBP, we propose BC algorithm which has been used in the experiments.(3) Expression recognition:SVM as a classic representative of supervised learning model which has been widely used in pattern recognition, classification, and regression analysis. Experimental results on the Cohn-Kanade face database using the LIBSVM classify have satisfactory results with correct rate 92.5%.
Keywords/Search Tags:Facial expression recognition, Gabor wavelet transform, Histogram of oriented gradients, SVM
PDF Full Text Request
Related items