A Study For Facial Expression Recognition

Facial expression are the facial changes in response to a person's emotional states, intentions, or social communications. Facial expressions, and other gestures, convey non-verbal communication cues in face-to-face interactions. These cues may also complement speech by helping the listener to elicit the intended meaning of spoken words. As a consequence of the information that they carry, facial expressions can play an important role wherever humans interact with machines.Automatic facial expressions recognition(FER) could be traced back to the preliminary work of Suwa et al.[1] in 1978 and gained much popularity starting with the pioneering work of Mase and Pentland[2] in the nineties. More recently, facial expression analysis has become a very hot topic in computer vision and pattern recognition, and various approaches have been proposed to this goal. Because facial expression analysis is a research topic relating with multi-subject, therefore there are many problems relating with it to deal with. In this paper, we conduct the facial expression recognition using pattern recognition and machine learning methods, and has gained some accomplishments.In this paper, Support Vector Machine are utilized for facial expression recognition. The procedure contains two part: (1) Extracting 34 fiducial points from each facial image and using the coordinates of these points to form a vector as the input data; (2) The SVM classifier is used to classify the facial expression. The performance of the proposed method is confirmed by using the Japanese Female Facial Expression Database (JAFFE).Facial expression recognition based on kernel discriminant plane is proposed in this paper. This method is a nonlinear extension of the Sammon's optimal discriminant plane via the kernel trick. When we construct the recognition, we also use the coordinates of the fiducial points as the input data. The procedures are as follows: (1) Converting a multi-class classification problem into a multiple binary classification problem; (2) The SVM classifier is used to classify the facial expression in every binary classifier. The better performance of the proposed method is confirmed by the Japanese Female Facial Expression Database (JAFFE).In this paper, facial expression recognition based on kernel canonical correlation analysis isproposed. The procedure are as follows: (1) Locating 34 fiducial points from each facial image as the landmark locations and converting these geometric locations into a labeled graph(LG)[3] vector using Gabor wavelet transformation to represent the image.; (2)Using a semantic expression vector consisting of the semantic rating of each facial image as the semantic expression representation; (3)Learning the correlation between the LG vector and the semantic expression vector is performed by kernel canonical correlation analysis; (4)Estimating the associating semantic expression vector of a given test image and performing the classification according to this semantic expression vector. The better performance of the proposed method is confirmed by the Japanese Female Facial Expression Database (JAFFE) and the Ekman's Expression Database.