Research Of Some Problems On Emotional Virtual Human

In recent years, with multimedia Human-Computer Interaction (HCI), humanized of computers has become a hotspot, virtual human, virtual tour guider, virtual broadcasters have now published. Emotional virtual human has been widely used in HCI (voice, sign language, facial expressions etc), movement show, man-machine efficiency, video compression, games, military training, digital library, a new life in areas such as new product development and so on. For the research of emotional virtual human system, not only provide a platform for individual robot character and emotional, but also lay the foundation for studying robot group. This paper includes mainly four aspects of work on the following:First, based on the picture of facial expression recognition, we adopt Carnegie Mellon University (CMU) by the United States for facial expression database, to pretreat the pictures, to use 2DPCA for feature extraction, in the end, to category expressions for the face of the pictures with quantum neural network classifier.Second, we build up emotional modeling for emotional virtual human with Hidden Markov Model (HMM), because HMM is a probability model which is described random process and statistical characteristic, it is a double random process, the corresponding human mood and expression, mood is sightless state. Because forward-backward algorithm and Baum-Welch algorithm bring easily underflow in the course of training parameters, and that do not get the optimal solution. We revaluate parameters with Baum-Welch algorithm of scale factor, testing results show that algorithm satisfies the design requirements.Third, we make up and estimatein probability transfer matrix with emotional entropy in the emotional space, then wo expand the probability transfer matrix of emotional space from the simple three-dimension and three-graduation to arbitrary dimension and graduation, we educe emotional volatility factor with reasoning method.Fourth, we discuss some expression synthesis technologies for drawing facial model about combining 3DS MAX 7 and OpenGL, to complete the construction of virtual human systems, to test the entire system, to demonstrate expression subtle changes along with different encouraging circumstances of the outside. The outside encouraging signal is provided numeric signal.The results showed that established emotional virtual system can accomplish the facial expression recognition, we are able to fulfill the abundant varieties of expression details through the psychological model.