Parallel recognizer algorithm for automatic speech recognition

Authors

  • Akakpo Agbago Institute for Information Technology, National Research Council of Canada, Pavillon Lucien Brault, 101 rue St-Jean-Bosco, Gatineau, Que. J8Y 3G4, Canada
  • Caroline Barrière Institute for Information Technology, National Research Council of Canada, Pavillon Lucien Brault, 101 rue St-Jean-Bosco, Gatineau, Que. J8Y 3G4, Canada

Keywords:

Computational methods, Dynamic programming, Heuristic methods, Matrix algebra, Parallel algorithms, Parameter estimation, Signal distortion, Speech processing, Vectors, Automatic speech recognition (ASR), Parallel recognizer algorithms, Phonemes, Two-level dynamic programming (TLDP)

Abstract

A parallel recognizer algorithm for automatic speech recognition (ASR) was presented. This algorithm uses the principle of best-fit in an open competition for all phonemes in the Knowledge Base of Reference Phonemes (KBRP). The phonemes were concurrently matched in KBRP to an input T using a set of defined segment entities. This made it possible to convert the matrix computation of two-level dynamic programming (TLDP) to a vector space.

Downloads

Published

2004-09-01

How to Cite

1.
Agbago A, Barrière C. Parallel recognizer algorithm for automatic speech recognition. Canadian Acoustics [Internet]. 2004 Sep. 1 [cited 2021 Oct. 25];32(3):148-50. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/1671

Issue

Section

Proceedings of the Acoustics Week in Canada