Better analysis for automatic speech recognition
Douglas D. O’Shaughnessy · Canadian acoustics · 2002
An efficient automatic speech recognition (ASR) analysis method was studied. An ideal robust ASR should decode speech from all speakers of a language and with different microphones and transmission channels. Various advantages of this method are: it is an automatic and effcient method; actual ASR results that appear to be better than with some other methods extensively examined in the past like basic linear predictive coding (LPC) or a filter bank.