Facial Expression Recognition And Application Research Based On Affective Model

As a new way of teaching by network, distance education has realized the multimedia teaching for many people in many places at any time. But due to the separation of the time and the space in teaching process, teachers and students are unable to have real-time affective communication. In order to improve this situation,this dissertation study a method of facial expression recognition based on the affective model, including affective modeling, face detection and facial expression recognition.Based on the method, accurate and detailed affective information can be obtain.This dissertation establishes the improved three dimensional affective model based on fuzzy theory. In the model, the Ekman’s six class emotions are choosed as the basic emotions, and the different emotions can be as the different combinations of the three coordinates of the three dimensional state space. At the same time, the strength of every emotion is defined by the fuzzy rules. Compared with the traditional model, the improved model realizes facial expression to the corresponding emotion by the Ekman’s six class emotion, and the locations of the emotion can be shown quantitatively and intuitively, also the affective information is more detailed through defining the strength of every emotion, at the same time, the model decides the process of using facial expression recognition to obtain detailed affective information.Every collected image from the distance education almost contains a single face,and the angles of faces are indefinite. If the angle of face is large, the traditional face detection method based on Adaboost algorithm will not detect face accurately. In order to solve the problem, this dissertation firstly excludes some non-face regions by color characteristics, later calculating the angle of the face by means of the method of eye location and rotating the image. At last, the face of the image can be detected using the Adaboost algorithm. In the end, the experimental results show that the modified face detection method has higher detection accuracy compared with the traditional face detection method.During the process of recognizing facial expression, the usage of the support vector machine in facial expression recognition can solve practical problems such as small samples, high dimension and nonlinear. But the parameters of the algorithm have great influences on the classification results, and may reduce the recognition accuracies. Tosettle the problem, using bacteria foraging algorithm to choose the best parameters of the support vector machine used in facial expression recognition is presented. The method uses bacteria foraging algorithm to optimize the parameters of the support vector machine by chemotaxis, reproduction, elimination and dispersal as bacteria forage for food and the optimal parameters are obtained finally. At last, the results of experiments show that the support vector machine optimized by the bacteria foraging algorithm can obtain the best parameters in short time and the average accuracy is the highest compared with the support vector machine optimized by the bacteria foraging algorithm, the unimproved support vector machine, the support vector machine optimized by other optimization algorithms in the simulation experiments.