Accuracy improvement for a voice recognition using field association knowledge
by Yasuhiko Fujita, El-Sayed Atlam, Atsushi Sakakibara, Masao Fuketa
International Journal of Computer Applications in Technology (IJCAT), Vol. 33, No. 2/3, 2008

Abstract: Voice recognition is an important technique for man machine interfaces, but the accuracy degrades if a variety of communications is required. One of practical solution is to restrict the general communications into the domain specific topics dynamically. However, it is difficult to detect the specific domain from the general conversation and to adopt the language model for the specified domain effectively. This paper presents a domain specific voice recognition technique using field association knowledge that can find only the segment of utterances. Therefore, it enables us to change the domain of the language models in the real time processing. In this method, field association knowledge is used as the preprocessor of the first voice recognition stage and determines the appropriate domain in the conversation. The voice recognition of the second stage just changes the language model suggested by the first stage. The proposed approach is evaluated for 600,000 field association terms related to <Baseball>, <Sumo>, <Soccer>, etc. According to experimental results, it turns out that the average of recognition rate improved by 15.6%.

Online publication date: Wed, 10-Dec-2008

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Computer Applications in Technology (IJCAT):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com