Research On Chinese Speech Driven Virtual Human

With the development of virtual reality, the research on virtual human becomes a hot spot in the virtual reality field. In the television animations and computer games, people can obtain good visual effect from a lifelike virtual human model. In recent years, more and more researchers have been interested in the research of virtual human model, as well as the ways to drive virtual human, they put forward their own methods respectively. The study of 3D virtual human driven methods not only reduces the investment of developers, but also extends the choices of users. Therefore, the commercial value and practical value is immeasurable.As there is obvious mapping relationship between the voice and mouth shape, we can use the speech signal to drive the virtual human mouth through this mapping relation. On the contrary, there is no direct corresponding relationship between voice and the action of virtual human, it may be difficult to drive the virtual human using speech signal directly. At present, researchers have proposed a lot of methods to drive the virtual human. However, most of these methods have high complexity, so it is necessary to improve it. According to the above problems, this thesis proposes a method based on Chinese speech recognition to drive the virtual human, this method can drive the virtual human’s body movements and facial animation as well. First of all, a speech recognition module is established so that the speech signal can be converted into text information; then, the speech recognition module is added into the virtual human chat system, achieving the goal that the virtual human action and virtual human chat system are controlled by speech signal; next, the speech recognition module is added into the 3D virtual human face animation system, so the virtual human mouth animation and expression details animation can be driven by speech signal. Moreover, a 3D facial animation, which is synchronous with the input speech signal, can also be gotten through the TTS technology.In this thesis, we use the Microsoft speech development kit to realize above process. The speech recognition module is established from the class of speech recognition, which converts the speech signal into text information. Then, the speech recognition module is added into a virtual human chat system and a 3D facial animation system respectively, as a result, both the action and facial animation of the virtual human are driven by Chinese speech signal. The experiment results show that this method is simple and practical, recognition accuracy can also be accepted by users.