Paper Abstract and Keywords |
Presentation |
2007-06-29 10:00
Voice Activity Detection Applied to Hands-Free Speech Recognition based on Decoding using Acoustic and Language Models Hiroyuki Sakai, Tobias Cincarek, Hiromichi Kawanami, Hiroshi Saruwatari, Kiyohiro Shikano (NAIST), Akinobu Lee (Nagoya Inst. of Tech.) SP2007-17 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
This paper proposes a new Voice Activity Detection (VAD) method by decoding using Acoustic Model(AM) and Language Model (AM) applied to Hands-Free speech recognition. SNR degrades in real-environment Hands-Free because of the various background noise. This means that the common VAD based on Ampulitude Level (AL) becomes difficult. We focuses on the premise that non-speech segments exist before and after utterance.
VAD by comparing phoneme to silence segment while decoding using AM and LM. Effective VAD and real-time decoding is realized without using AL. We implemented the method on the speech recognition decoder Julius. Experimental results using various SNRs show that our method attains the higher VAD accuracy and the recognition rate than the conventional method. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Voice Activity Detection (VAD) by decoding based on Acoustic Model and Language Model / Hands-Free speech recognition / Real-environment / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 107, no. 116, SP2007-17, pp. 55-60, June 2007. |
Paper # |
SP2007-17 |
Date of Issue |
2007-06-21 (SP) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
SP2007-17 |
|