Research On Virtual Human With Ability Of Emotion

Emotion is an important part of human behavior and sensory experience. The most difference between human and machine is human's ability of producing emotion. The developing theory of human emotions supports the research of artificial intelligence in computer science. More and more artificial intelligence researchers realized that in order to actualize the dream of personating computer, let computer have intelligence is not sufficient, we must also make it understand and have similar human emotions, and make it interact with human.This paper mainly research on the theories of artificial emotion and the technologies of affective computing and develop an affective virtual human system with the emotion model. This system includes: (1) The text processing and response module: use the theory of artificial intelligence and natural language understanding to extract the emotion information from the text; (2) The personality module: we adopt the Five Factor Model (FFM) of personality from psychology studies to define the personality space of virtual human. Use the model with the current emotional state, and the rules of Bayesian Belief Network to calculate the next emotional state. (3) The Mood to Emotions module: With the emotion information obtained from the first module, we defined a typical transition probability matrix for each mood to calculate the probabilities of transitions from one emotional state to another. And use the biggest one to be the next emotional state. (4) The synchronization module: To complete the mapping between emotional state and facial expressions. Finally, the experiment results of artificial virtual human system are outlined. The test results of system showed that the emotional reaction calculated by the model is in accord with the human behavior and validate the affective computing theory.