The development of analysis methods for speech recognition

摘要: This paper is intended to give an outline of the development and state art in acoustical analysis speech for recognition purposes. Starting with a short view on pattern problem general, shows how spectral primary will be more refined by including knowledge production perception processes. Recognizers high performance need secondary which produces feature vectors components are phonetically or articulatorically meaningful. Because it complicated elaborate less complete set detectors, universal procedures required producing this description. Several approaches discussed.