Voice Driven Animation System Of Face And Mouse

The facial animation is a technique about human-computer interaction, whose core isface controlling. The main driven modes includes voice-driven mode, text-driven mode andvideo-driven mode. It involves a chain of techniques, such as face modeling, speechrecognition, speech signal analysis, texture mapping and so on. At present, it has been widelyused in virtual teaching, animation, online games, virtual reality and other areas.Voice driven animation system of face and mouth belongs to the domain of facialanimation technique. Realistic audio and visual techniques have become hot research invirtual reality, human-computer interaction and artificial intelligence. With the emergence ofMPEG-4, it provides a broader perspective for development of facial animation. It enrichesthe types of facial animation data, reduces the amount of data, lowers the computationalcomplexity, in addition its architecture suits to real-time network transmission.The paper designs and implements a system of facial animation which is driven byspeech. The input is simple, versatile, with low computational complexity and smaller amountof data. The main content of the system includes four aspects which are face modeling,establishing the mapping model between voice characteristic parameters and the facialanimation parameter, the extraction of the voice characteristic parameters and the controllingof face animation. The related techniques are described briefly in the paper, include how tocreate the system and comply with the facial modeling standard which is MPEG-4standard.First of all, in order to get a simple, convenient and efficient way to model human face,the paper rejects the traditional code mode, and uses3D MAX tools to complete3D facialmodeling. The complished3D face model is more precise and detailed.Secondly, with the method of speech analysis and process, this paper developsvoice-driven facial animation based on thought without speech recognition. In this paper, weestablish a mapping model between voice characteristic parameters and facial animationparameters by HMM, it is used to conduct the training and recognition for samples. The mapaims to use the voice characteristic parameters to find the corresponding facial animationparameters which drive the face model. The experimental result shows that, the facialanimation by this method ensures better synchronization between voice and animation of faceand mouse, smooth picture and more realistic animation.Thirdly, the paper analyzes input voice file and extracts the speech feature parameters. Ituses the Mel-frequency cepstral coefficients (MFCC) as the voice characteristic parameters,as the reason of it is proposed based on human hearing mechanism. The experimental resultshows that, it has better noise immunity and higher speech recognition rate. When the MFCC is extracted, with the help of the mapping model, it identifies corresponding facialanimation parameter sequence (FAP), then the face model is driven to complete appropriateaction.Finally, in order to build FAP driven facial model, a face definition table (FDT) is setwhich stores the information of facial controlling points for each FAP. If we want to controlthe movements of the facial model, new3D coordinates are calculated using FDT and somealgorithms as controlling points are affected by FAP. This paper also employes the MPEG-4algorithm.