Study On Related Technology Of Artifical Psychology

Facial expression recognition and affective computing are very useful to harmonious human computer interactions. The key for harmonious human computer interactions is to endow computer with the ability to identify, understand,express human feeling. Based on the theory of affective computing and artificial psychology, this paper discusses in detail facial expression recognition and affective modeling and affords technique supports for harmonious human computer interactions. The main contributions of the thesis include:A new facial expression recognition method based on image algebraic characters is presented. First, eyes and mouths are segmented from the facial expression image and variant moments and singular value feature vectors of eyes and mouth are extracted. Then Fisher linear discriminant analysis is used to find a set of optimal feature vectors. Finally, we have trained a SVM Classifier.A new facial expression recognition method based on Gabor Transform and Adaboost Algorithm is put forward. We discuss facial feature extraction from static images using Gabor transform. To reduce the dimension, we first downsample Gabor tansform results by a factorρ, then PCA is used to reduce the dimension of the Gabor feature. Finally adaboost algorithm is used to resolve facial expression recognition.Aiming at the shortcomings of optical flow technique, two improved optical flow algorithms are presented. One way is to introduce forward and backword constraint eqution and Hessian matrix of optical flow equation is computed. The well-posedness of each point of local neighbourhood is examined and the weight of Lucas-Kanade's method is defined as the reciprocal of the conditioning number of its Hessian Matrix. This can eliminate those uncertainty constrains and improve the stability of the solution of the gradient constraint equation. Another way is based on the extended optical flow constraint equation, a novel approach for estimating facial expression motion based first-order and second-order div-curl splines constraint is presented. The experiment results show the before-mentioned methods can compute effectively facial expression optical flow fields.A new approach for HMM-training which is based on the improved maximum mutual information crtiterion is presented. HMM parameter adajustment rules are induced. By adopting a more realistic MMI definition, discriminative information contained in the training data could be used to improve the performance of HMM. A facial expression hybrid classificer based improved HMM and BP neural network is designed.Maximum model distance crtiterion is discussed and some reasonable modification is presented. A new method for HMM-training which is based on the improved maximum model distance crtiterion is described and the HMM parameter adajustment rules are induced. The trained HMM has strong recognition ability. A facial expression hybrid classificer based improved HMM-training algorithm and BP neural network is designed.According to the basic emotions theory, the paper presents personality, mood, and emotion space. The mapping relationship among personality, mood and emotion is built. An equation for updating the affective and mood states is induced and a generic computing model for personality, mood and emotion simulation for virtual humans is constructed. The simulation results demonstrate that the affective model can better simulate the dynamic process of emotion and mood change under various environment stimul.