Abstract: Using speech data to drive facial animation is an important intellectual technology in areas such as Virtual Reality（VR）. Recently the rapid development of VR stresses an urgent need for natural human-computer communication in immersive environment. The speech-driven facial animation technology can produce vivid animation with emotion, thus it can better help human-computer interaction and improve users’ experience compared with the traditional presupposed facial animation. In order to improve theintelligenceofthis technology and its applications, focusing on the key issue of speech-driven facial animation：audio and video mapping, methods including frame by frame analyzation, multi-frame analyzation and phoneme by phoneme analyzation are reviewed. Meanwhile, ideas of several facial models, means of animation synthesis, emotion fusion andevaluation are summarized, and possible directions for future studies are provided.

Key words: speech driven, facial animation, Virtual Reality（VR）, neural networks

摘要： 利用语音来驱动人脸动画，是虚拟现实（Virtual Reality）等领域重要的智能技术，近年来虚拟现实技术的飞速发展更进一步地突出了在沉浸环境下的人机自然交流的迫切需求。语音驱动的人脸动画技术能够创造出自然生动、带有情感的动画，相对于传统预设的人脸动画而言能够更好地辅助人机交互、提升用户体验。为推进该技术的智能化程度和应用，针对语音驱动人脸动画的关键问题：音视频映射，综述了逐帧分析、多帧分析和逐音素分析的映射方法，同时也梳理了多种脸部模型的思想，动画合成、情感融合、人脸动画评价的方法，及可能的研究发展方向。

关键词: 语音驱动, 人脸动画, 虚拟现实, 神经网络