Recent advances in automatic speech recognition are mainly accomplished by designing a plug-in maximum a posteriori decision rule such that the forms of the acoustic and language model distributions are specified and the parameters of the assumed distributions are estimated from a collection of speech and language training examples. Maximum likelihood point estimation of decision parameters is by far the most prevailing training method. However, due to the problems of unknown speech and language distributions, sparse training data, high spectral and temporal variabilities in speech, and possible mismatch between training and testing conditions, a dynamic learning strategy is needed. In this talk, the mathematical foundations of Bayesian adaptation of decision parameters is first described. Maximum a posteriori point estimation is then developed. Finally robust decision rules for compensating mismatch are formulated and discusseds.

Material from talks    pdf (207KB)    postscript (256KB)

Mathematical Foundations of Speech Processing and Recognition

2000-2001 Program: Mathematics in Multimedia