Audio-to-Visual Conversion Via HMM Inversion for Speech-Driven Facial Animation

LUCAS D. TERISSI, JUAN CARLOS GÓMEZ

LECTURE NOTES IN COMPUTER SCIENCE

Springer

Año: 2008 vol. 5249 p. 33 - 42

0302-9743

In this paper, the inversion of a joint Audio-Visual Hidden Markov Model is proposed to estimate the visual information from speech data in a speech driven MPEG-4 compliant facial animation system. The inversion algorithm is derived for the general case of considering full covariance matrices for the audio-visual observations. The system performance is evaluated for the cases of full and diagonal covariance matrices. Experimental results show that full covariance matrices are preferable since similar, to the case of using diagonal matrices, performance can be achieved using a less complex model. The experiments are carried out using audio-visual databases compiled by the authors.