Pattern recognition in speech perception research

Auteurs-es

  • Terrance M. Nearey Dept. of Linguistics, University of Alberta, Edmonton, AB T6G 2E7, Canada

Mots-clés :

Feature extraction, Pattern recognition, Speech, Automatic speech recognitions, Clustering, Flexible patterns, Logistic models, Logistic regressions, Multinomial logistic regressions, Response factors, Rms errors, Speech perceptions

Résumé

The use of more flexible pattern recognition methods in speech perception research was investigated. The logistic regression and methods imported from automatic speech recognition (ASR) technology was included. The APP scores generated by a NAPP model was expressed in the form of a multinomial logistic regression (MNLR). An initial MNLR analysis was conducted and the response factor comprised the 6 response categories. The clustering patterns of consonant responses suggested a factored solution, whereby judgment log-odds were tuned continuously by only the three factors that include closing place, opening place, and gap duration. a reduced logistic model enforcing the decomposition shows RMS error and modal agreement that are nearly indistinguishable from the full CC model.

Fichiers supplémentaires

Publié-e

2008-09-01

Comment citer

1.
Nearey TM. Pattern recognition in speech perception research. Canadian Acoustics [Internet]. 1 sept. 2008 [cité 16 févr. 2025];36(3):150-1. Disponible à: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/2082

Numéro

Rubrique

Actes du congrès de la Semaine canadienne d'acoustique

Articles les plus lus du,de la,des même-s auteur-e-s