Subband-based speech recognition

Abstract
In the framework of Hidden Markov Models (HMM) or hybridHMM/Artificial Neural Network (ANN) systems, wepresent a new approach towards automatic speech recognition(ASR). The general idea is to divide up the full frequencyband (represented in terms of critical bands) intoseveral subbands, compute phone probabilities for each subbandon the basis of subband acoustic features, performdynamic programming independently for each band, andmerge the subband recognizers (recombining the...

This publication has 6 references indexed in Scilit: