On improvements of CI-based GMM selection (2005)
| Venue: | in Proceedings of Interspeech |
| Citations: | 1 - 1 self |
BibTeX
@INPROCEEDINGS{Chan05onimprovements,
author = {Arthur Chan and Mosur Ravishankar and Er I. Rudnicky},
title = {On improvements of CI-based GMM selection},
booktitle = {in Proceedings of Interspeech},
year = {2005}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Gaussian Mixture Model (GMM) computation is known to be one of the most computation-intensive components of speech recognition. In our previous work, context-independent model based GMM selection (CIGMMS) was found to be an effective way to reduce the cost of GMM computation without significant loss in recognition accuracy. In this work, we propose three methods to further improve the performance of CIGMMS. Each method brings an additional 5-10 % relative speed improvement, with a cumulative improvement up to 37 % on some tasks. Detailed analysis and experimental results on three corpora are presented. 1.
Citations
| 35 | Subphonetic Acoustic Modeling for Speaker-Independent Continuous Speech Recognition - Hwang - 1993 |
| 5 | The Design for the Wall Street - Paul, Baker - 1992 |
| 4 | Gaussian mixture selection using context-independent HMM - Lee, Kawahara, et al. - 2001 |
| 2 | The 1999 CMU 10x real time broadcast news transcription system - Ravishankar - 2000 |
| 1 | et al, “Four-Layer Categorization of Fast GMM - Chan |
| 1 | et al, “Fast Likelihood Computation in NearestNeighbor Based Search for Continuous Speech Recognition - Pellom - 2001 |
| 1 | et al, “Use of Gaussian Selection in Large Vocabulary Speech Recognition Using HMMs” ICSLP - Gales - 1996 |
| 1 | et al, “Sub-Vector Clustering to Improve Memory and Speed Performance of Acoustic Likelihood Computation.”, Eurospeech - Mosur - 1997 |
