Asian Science Citation Index is committed to provide an authoritative, trusted and significant information by the coverage of the most important and influential journals to meet the needs of the global scientific community.  
ASCI Database
308-Lasani Town,
Sargodha Road,
Faisalabad, Pakistan
Fax: +92-41-8815544
Contact Via Web
Suggest a Journal
Journal of Applied Sciences
Year: 2015  |  Volume: 15  |  Issue: 2  |  Page No.: 295 - 300

FLVQ Based GMM in Speaker Verification

P. Shanmugapriya and Y. Venkataramani    

Abstract: In order to improve the verification rate of automatic speaker verification system, a novel training algorithm for Gaussian Mixture Model is proposed in this study. A novel feature extraction method for automatic speaker verification system is also presented. This system includes extraction of discrete wavelet transform based Mel frequency cepstral coefficients from speech and Fuzzy Learning Vector Quantization based gaussian mixture model training. This feature extraction approach utilizes the dynamic spectral features which are useful for recognizing the speaker. The proposed training method for speaker model not only reduces the number of features vectors used to train the model but also increases the verification rate than the conventional GMM-expectation maximization algorithm. The proposed method of speaker verification is evaluated using TIMIT database. Experiments are also conducted with other vector quantization algorithms: (1) Learning vector quantization, (2) K-means, (3) Fuzzy C-means and (4) Linde-buzo-grey algorithm as training algorithms for GMM. Experimental results demonstrate that the performance of the proposed system is better when compared to conventional systems in terms of verification rate.

Cited References   |    Fulltext    |   Related Articles   |   Back
   
 
 
 
  Related Articles

 
 
 
 
 
 
 
 
 
Copyright   |   Desclaimer   |    Privacy Policy   |   Browsers   |   Accessibility