Study Of Virtual Human Speech-Emotion Interaction In Virtual Situation

Emotion Virtual Human technology is a computer virtual technology with the performing ability of multi-modal emotion and behavior coordination, which is essentially an approach to prompt and visual human-computer interaction. Virtual Reality utilizes computer simulation technology to simulate visual, auditory, tactile and other senses in the real world, creating a virtual world for users. Speech Emotion Recognition extracts and distinguishes emotions in human speech so that those emotions can be analyzed by computer. This paper integrates Emotion Virtual Human technology, Virtual Reality technology and Speech Emotion Recognition, with the purpose of developing a simpler, more visual and more user-friendly human-computer interaction technology, and thus bridging human emotion speech in the real world and Emotion Virtual Human in the virtual world.In this paper, the specific work includes:First, the classification of emotion models as well as their distinguished features in the psychology field and Speech Emotion Recognition field is carefully studied and analyzed. Based on the analysis and the features of Mandarin speech signal in Speech Recognition, five emotion models-pleasure, anger, sadness, surprise and neutral emotion-are chosen.Second, in the area of Speech Emotion Recognition, obtaining the material database of Mandarin emotional speech, studying the front-end processing and analysis of speech signal, extracting Emotional eigenvector included in speech signal, using parameters to select extracted eigenvector to conduct a screening for the purpose of dimension reduction, removing the redundant part of speech emotional characteristics parameters, selecting from the sample eigenvector sets which has smaller distance in classes and bigger distance among other classes,increasing recognition rate.Third, using hidden Markov model for emotion recognition. Speech Emotion Recognition and speech semantic recognition are essentially connected and intertwined. Hidden Markov Type has been widely used in speech semantic recognition. Through the feature extraction of obtained speech sample and experiments of model recognition, Hidden Markov Type has been proved to have relatively high recognition rate and good recognition effect in speech emotion recognition.Fourth, building emotion virtual human reality software Virtools, explaining the process of modeling and controlling of virtual human in detail, and through the controlling of module, completing the visual presentation and feedback of the emotion results recognized by virtual human, so as to achieve a good result of human-computer interaction.