Processing and feature analysis of stressed speech in speech recognition
by Yonglian Wang, Nazeih M. Botros, Ismail Shahin
International Journal of Functional Informatics and Personalised Medicine (IJFIPM), Vol. 2, No. 4, 2009

Abstract: In this paper, stressed speech signals are processed, analysed in three feature domains, and then recognised in Hidden Markov Models (HMMs) in Matlab. The Linear Predictive Coding (LPC) cepstral feature analysis is used to obtain the observation vector and the input vector for HMM. Feature analysis explored the characteristics of stress-induced speech compared with normal talking speech. Speech recognition performance degrades greatly owing to the occurrence of stress. A stress compensation technique is utilised to compensate for stress distortion so as to improve speech recognition performance. Four stress styles (neutral, angry, question and Soft) from SUSAS database are used to test recognition performance of speaker-independent isolated word system. Our results show that speech spoken under angry and question stress contains more stress components generating extremely wide fluctuations with average higher pitch, higher intensity and more energy compared with neutral. The recognition rate has increased up to 10% for stressed speech with stress compensation.

Online publication date: Fri, 08-Jan-2010

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 Functional Informatics and Personalised Medicine (IJFIPM):
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