Speaker-dependent 100 word recognition using CombNET and dynamic spectral features of speech

Author

Kitamura, Tadashi ; Nishioka, Ken ; Iwata, A. ; Hayahara, Etsuro

Author_Institution

Nagoya Inst. of Technol., Japan

fYear

1991

fDate

14-17 May 1991

Firstpage

83

Abstract

Present speaker-dependent 100-word recognition using CombNET, which consists of a four-layered neural network with a comb structure, and dynamic spectral features of speech based on a two-dimensional mel-cepstrum. CombNET consists of two types of neural network. The first one is a stem network which utilizes a self-organizing algorithm and roughly classifies an input pattern. The second one consists of many branch networks using a back-propagation algorithm and precisely classifies the pattern. Experimental results on speaker-dependent word recognition for 100 Japanese city names uttered by nine male speakers show that the recognition accuracy is 97.3%

Keywords

backpropagation; feedforward neural nets; speech recognition; CombNET; Japanese city names; back-propagation algorithm; branch networks; comb structure; dynamic spectral features; four-layered neural network; male speakers; recognition accuracy; self-organizing algorithm; speaker-dependent recognition; stem network; two-dimensional mel-cepstrum; word recognition; Backpropagation algorithms; Feedforward neural networks; Fourier transforms; Frequency domain analysis; Large-scale systems; Neural networks; Organizing; Speaker recognition; Speech recognition; Time domain analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1991., Proceedings of the 34th Midwest Symposium on

Conference_Location

Monterey, CA

Print_ISBN

0-7803-0620-1

Type

conf

DOI

10.1109/MWSCAS.1991.252132

Filename

252132