A Bayesian approach to speaker adaptation for the stochastic segment model

Speaker adaptation is frequently used to achieve good speech recognition performance without the high costs associated with training a speaker-dependent model. The main goal of this study is to investigate speaker adaptation for recognizers using multivariate Gaussian densities, specifically, the stochastic segment model. A Bayesian approach is followed, with estimation of the parameters of a speaker-adapted model based on prior densities obtained from speaker-independent data. Experimental results achieve 16% error reduction using mean adaptation with roughly 3 min of speech, nearly half the difference between speaker-independent and speaker-dependent recognition rates.