%0 Conference Paper %1 Wang2011 %A Wang, Lijuan %A Wu, Yi-Jian %A Zhuang, Xiaodan %A Soong, F.K. %B Proceedings of the 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) %C Prague, Czech Republic %D 2011 %K HMM;heuristic Markov algorithm;traditional alignment;probabilistic descent error error;photo-real;talking estimation;visual generation head;trajectory-guided;visual hidden likelihood maximum method;mean method;optimal methods;speech models;Speech;Speech models;mean reduction;minimum speech square state synthesis synthesis;Acoustics;Hidden synthesis;Training;Trajectory;Visualization;minimum synthesis;audiovisual training trajectory %P 4580-4583 %R 10.1109/ICASSP.2011.5947374 %T Synthesizing visual speech trajectory with minimum generation error %X In this paper, we propose a minimum generation error (MGE) training method to refine the audio-visual HMM to improve visual speech trajectory synthesis. Compared with the traditional maximum likelihood (ML) estimation, the proposed MGE training explicitly optimizes the quality of generated visual speech trajectory, where the audio-visual HMM modeling is jointly refined by using a heuristic method to find the optimal state alignment and a probabilistic descent algorithm to optimize the model parameters under the MGE criterion. In objective evaluation, compared with the ML-based method, the proposed MGE-based method achieves consistent improvement in the mean square error reduction, correlation increase, and recovery of global variance. It also improves the naturalness and audio-visual consistency perceptually in the subjective test.

@inproceedings{Wang2011, abstract = {In this paper, we propose a minimum generation error (MGE) training method to refine the audio-visual HMM to improve visual speech trajectory synthesis. Compared with the traditional maximum likelihood (ML) estimation, the proposed MGE training explicitly optimizes the quality of generated visual speech trajectory, where the audio-visual HMM modeling is jointly refined by using a heuristic method to find the optimal state alignment and a probabilistic descent algorithm to optimize the model parameters under the MGE criterion. In objective evaluation, compared with the ML-based method, the proposed MGE-based method achieves consistent improvement in the mean square error reduction, correlation increase, and recovery of global variance. It also improves the naturalness and audio-visual consistency perceptually in the subjective test.}, added-at = {2021-02-01T10:51:23.000+0100}, address = {Prague, Czech Republic}, author = {Wang, Lijuan and Wu, Yi-Jian and Zhuang, Xiaodan and Soong, F.K.}, biburl = {https://www.bibsonomy.org/bibtex/2eb518745eaa39f4079c7bd708523e3f7/m-toman}, booktitle = {Proceedings of the 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, doi = {10.1109/ICASSP.2011.5947374}, file = {:pdfs/wang_icassp_2011.pdf:PDF}, interhash = {486adda00892403e13b5a862eb1b8bc0}, intrahash = {eb518745eaa39f4079c7bd708523e3f7}, keywords = {HMM;heuristic Markov algorithm;traditional alignment;probabilistic descent error error;photo-real;talking estimation;visual generation head;trajectory-guided;visual hidden likelihood maximum method;mean method;optimal methods;speech models;Speech;Speech models;mean reduction;minimum speech square state synthesis synthesis;Acoustics;Hidden synthesis;Training;Trajectory;Visualization;minimum synthesis;audiovisual training trajectory}, month = may, owner = {schabus}, pages = {4580-4583}, timestamp = {2021-02-01T10:51:23.000+0100}, title = {Synthesizing visual speech trajectory with minimum generation error}, year = 2011 }