Group delay-based minimum variance distortion-less response cepstral features for speaker identification in whispered speech
by Vijay M. Sardar; Manisha L. Jadhav; Saurabh H. Deshmukh; Makarand M. Jadhav
International Journal of Computer Applications in Technology (IJCAT), Vol. 73, No. 2, 2023

Abstract: The whispering voice shows a wide difference in characteristics compared to the neutral voice. It makes identification of a person from the whispered sound difficult. The Group Delay Function (GDF) in its spectral form considers the phase information in the short-time FT phase function, which is otherwise ignored in traditional front-end processing. A Minimum Variance Distortion-less Response (MVDR) based on smoothing on the denominator of group delay that enhances speech quality and intelligibility is proposed in this paper. The experiment uses MVDR spectral coefficient features with a multi-class Support Vector Machine (SVM) for classification. The proposed method reported an improvement of 2.41% over the baseline system using the CHAINs database and SVM classifier. The five-fold cross-validation is exercised for accuracy and Speaker Error Rate (SER) to verify the consistency of the results. The proposed system is also evaluated for False Positives (FP) and Precision and reported the enhancement compared to the baseline system.

Online publication date: Thu, 09-Nov-2023

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