An efficient voice driven face animation method for cyber telepresence applications

In on-line cyber spaces there are artificial bodies which imitate realistic behavior controlled by remote users. An important aspect is the realistic facial motion of human like characters according to the actual speech sounds. This paper describes a memory and CPU efficient method for visual speech synthesis for on-line applications using voice connection over network. The method is real-time, can be activated on the receiver client without server support. It is needed only to send coded speech signal and the visual speech synthesis is the task of the receiving client. This way deaf or hard of hearing people can lipread the transmitted audio speech. The animation rendering is supported by graphical accelerator devices, so CPU load of the conversion is insignificant.