NAISTAR
Advanced Search
Japanese | English

naistar (NAIST Academic Repository) >
学術リポジトリ naistar / NAIST Academic Repository naistar >
国際会議発表論文 / Proceedings >
情報科学研究科 / Graduate School of Information Science >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10061/8040

Title: Gaussian Mixture Selection Using Context-Independent HMM
Authors: Akinobu Lee
Tatsuya Kawahara
Kiyohiro Shikano
Issue Date: May-2001
Publisher: IEEE
Start page: 69
End page: 72
Abstract: We address a method to efficiently select Gaussian mixtures for fast acoustic likelihood computation. It makes use of context-independent models for selection and back-off of corresponding triphone models. Specifically, for the k-best phone models by the preliminary evaluation, triphone models of higher resolution are applied, and others are assigned likelihoods with the monophone models. This selection scheme assigns more reliable back-off likelihoods to the un-selected states than the conventional Gaussian selection based on a VQ codebook. It can also incorporate efficient Gaussian pruning at the preliminary evaluation, which offsets the increased size of the pre-selection model. Experimental results show that the proposed method achieves comparable performance as the standard Gaussian selection, and performs much better under aggressive pruning condition. Together with the phonetic tied-mixture modeling, acoustic matching cost is reduced to almost 14% with little loss of accuracy
Description: ICASSP2001: IEEE International Conference on Acoustics, Speech and Signal Processing, May 7-11, 2001, Salt Lake City, Utah, US.
URI: http://hdl.handle.net/10061/8040
ISBN: 0780370414
ISSN: 1520-6149
Rights: Copyright 2001 IEEE
Text Version: Publisher
Publisher DOI: 10.1109/ICASSP.2001.940769
Appears in Collections:情報科学研究科 / Graduate School of Information Science

Files in This Item:

File SizeFormat
ICASSP_2001_69.pdf359.33 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Copyright (c) 2007-2012 Nara Institute of Science and Technology All Rights Reserved.
DSpace Software Copyright © 2002-2010  Duraspace - Feedback