一种基于共振峰分析的语音驱动人脸动画方法

An Approach of Speech-driven Facial Animation Based on Formants Analysis

快速、高效地实现语音驱动下的唇形自动合成,以及优化语音与唇动的同步是语音驱动人脸动画的重点。提出了一种基于共振峰分析的语音驱动人脸动画的方法。对语音信号进行加窗分帧,DFT变换,再对短时音频信号的频谱进行第一、第二共振峰分析,将分析结果映射为一组控制序列,并对控制序列进行去奇异点等后处理。设定三维人脸模型的动态基本口形,以定时方式将控制序列导入模型,完成人脸动画驱动。实验结果表明,该方法简单快速,有效实现了语音和唇形的同步,动画效果连贯自然,可广泛用于各类虚拟角色的配音,缩短虚拟人物的制作周期。

Automatic synthesis of lip animation driven by speech and lip synchronization is the key issues in speech driven facial animation system. A new approach of speech-driven facial animation based on formants analysis is presented. The input audio signal is divided into partly overlapped frames and multiplied by a Hamming window, and then a DFT （Discrete Fourier Transformation） is used. In the frequency domain of the short time signal, the 1st and 2nd formant are analyzed in order to form a control sequence. Several basic dynamic mouth shapes of the 3D facial model are defined, and the control sequence is used to drive the facial movements. The results show that the input speech and facial lip animation is synchronized precisely in this way, and the effect of the animation is fluent and looks real.