Facial Expression Recognition Based On The Improved LLE Algorithm And The Improved Adaboost Algorithm

Facial expression is the external manifestation of human thought and emotion.It is one of the basic ways to express emotions.Facial recognition is a prerequisite for computer to understand human emotions.There are wide application prospects in real life,such as human-computer interaction,intelligent game,video surveillance,distance education and so on.In the facial expression recognition,the work done in this paper is as follows:(1)The process of expression image preprocessing is as follows.Firstly,we use gray histogram equalization and binarization to preprocess expression images.And then we use the gray-scale integral projection to determine the location of the human eye to cut the facial expression area.Finally,the scale of the expression image normalized to unified all the size of the expression image.(2)The process of facial image feature extraction is as follows.The use of 40 Gabor filter to extract facial expression characteristics.(3)The process of expression image preprocessing is as follows.We use the PCA algorithm,the LLE algorithm,the SLLE algorithm and the improved LLE algorithm to reduce facial expression features dimensionality.(4)The process of expression classification is as follows.We use the SVM algorithm,the Adaboost algorithm and the improved Adaboost algorithm to classify facial expressions.In this paper,the study of facial expression recognition focuses on data reduction dimension and facial expression classification.The LLE algorithm is a kind of nonlinear dimension reduction algorithm,it can make the data after dimension reduction maintain the original topology and there is a wide application in facial expression recognition.Because the LLE algorithm does not take the sample category information into account,so the SLLE algorithm is put forward.However,the SLLE algorithm only considers the category information of samples,without considering the relationship between the various expressions.So this paper proposes an improved LLE algorithm,this algorithm considers the neutral expression as the center of the other kinds of expressions.The Adaboost algorithm is an excellent classification algorithm,it hardly produces over-fitting.The Adaboost algorithm is usually used to deal with the binary classification problems,the accuracy rate needs to be improved when dealing with multi-classification problems,and facial expression classification is a multi-classification problem.Therefore,this paper presents an improved Adaboost algorithm to deal with facial expression classification.In this paper,through a large number of experimental comparative analysis,we found that the improved LLE algorithm has better experimental results than the PCA algorithm,the LLE algorithm and the SLLE algorithm.Similarly,the improved Adaboost algorithm has better experimental results than the SVM algorithm and the Adaboost algorithm.Thus,the improved LLE algorithm and the improved Adaboost algorithm are valid algorithms.