Adaptation of male acoustic models on the basis of language-independent female speaking data

摘要

A method for producing proxy - acoustic models for the use in the automatic speech recognition, which comprises the steps of that:(a) starting from speech material, which via microphone of male speakers a first language has been obtained, acoustic models be trained, and(b) in step (a) trained acoustic models in response to languages independent speaking data of female speakers be adapted to a second language, in order to proxy - acoustic models for use during the transit time of the speech recognition of an expression of a female speaker to produce the first language.