Arakawa Kaoru
Department Undergraduate School , School of Interdisciplinary Mathematical Sciences Position Professor |
|
Date | 2012/06/14 |
Presentation Theme | Feature Analyses of Japanese Language Spoken by Chinese and a Method for Improving Its Automatic Recognition |
Conference | IEICE Technical Report |
Promoters | IEICE |
Presentation Type | Speech (General) |
Contribution Type | Collaborative |
Publisher and common publisher | Xin LI and Kaoru ARAKAWA |
Details | A method to improve the recognition rate of Japanese language spoken by Chinese people is proposed. When automatic speech recognition is applied to Japanese language spoken by Chinese people, the recognition rate is low compared with that by Japanese. Feature analyses show that the duration of vowels is longer in Chinese speech than in Japanese, which causes degradation in recognition rate. In this paper, a method to improve the recognition rate is proposed by shortening the length of vowels on the time axis as preprocessing. The high performance of the new speech recognition system is verified for actual Chinese speeches. |