Neural network classifier architectures for phoneme recognition / by William Treurniet. : Co24-3/8-1992-1E-PDF

"Automatic recognition of words in continuous speech is difficult to do with whole-word template models, especially when the vocabulary is reasonably large. Instead, the currently preferred approaches for this task hypothesize the presence of words or sub-words, such as syllables or phonemes, on the basis of likelihood estimates obtained from comparisons of the acoustic data with statistical models derived from training data. This paper is concerned with the application of artificial neural networks, trained with the back-propagation learning algorithm, to modelling phonemes extracted from the DARPA TIMIT multi-speaker, continuous speech data base"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.890766&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs