Title :
A novel static parameter calculation method for model compensation
Author :
Suliang Bu ; Yunxin Zhao ; Yanmin Qian ; Kai Yu
Author_Institution :
Dept. of Comput. Sci., Univ. of Missouri - Columbia, Columbia, MO, USA
Abstract :
Vector Taylor Series (VTS) based model compensation approach has been successfully applied to various robust speech recognition tasks. In this paper, we propose a novel method of variable transformation to calculate the static statistics. In addition, we provide a detailed explanation of VTS and random variable transformations adopted in some recent papers. Experiments on Aurora 4 showed that the proposed approach obtained 22.8% relative WER reduction over the traditional first-order VTS methods.
Keywords :
compensation; speech recognition; vectors; Aurora 4; VTS based model compensation approach; WER reduction; first-order VTS methods; random variable transformations; robust speech recognition tasks; static statistics; vector Taylor Series based model compensation approach; Approximation methods; Hidden Markov models; Noise; Noise measurement; Speech; Speech recognition; Taylor series; Vector Taylor Series; model compensation approach; robust speech recognition;
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on
Conference_Location :
South Brisbane, QLD
DOI :
10.1109/ICASSP.2015.7178824
