Speech Recognition Based Upon a Segment Classification and Labelling Technique and Hidden Markov Model

Authors

  • W. A. Mahmoud
  • L. A. M. Bennett

Abstract

A new structure for isolated-word speech recognition via vector quantisation (VQ) is described, namely the segment classification and labelling technique (SCLT). The proposed recognizer requires the generation of separate codebooks for the acoustically dissimilar events and then the merging of them to produce a single reference codebook. Three major acoustic events were considered, namely voiced, unvoiced and silence (V/U/S). The results show that the proposed structure has the capability of reducing the degradation of VQ in speech recognition and provides a better set of observations for the hidden Markov model (HMM).

Additional Files

Published

2022-12-03

How to Cite

1.
Mahmoud WA, Bennett LAM. Speech Recognition Based Upon a Segment Classification and Labelling Technique and Hidden Markov Model. Canadian Acoustics [Internet]. 2022 Dec. 3 [cited 2024 Mar. 28];14(3 bis):91-2. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/3535

Issue

Section

Proceedings of the Acoustics Week in Canada