Article Information
Text-Independent Speaker Verification Based on Information Theoretic Learning

Keywords: Speaker Verification, Kmeans, LBG, Information Theoretic Learning, Vector Quantization.

Mehran University Research Journal of Engineering & Technology

Volume 30 ,  Issue 3

Sheeraz   Memon,Tariq Jameel  Saifullah  Khanzada,Sania   Bhatti

Abstract

In this paper VQ (Vector Quantization) based on information theoretic learning is investigated for the task of text-independent speaker verification. A novel VQ method based on the IT (Information Theoretic) principles is used for the task of speaker verification and compared with two classical VQ approaches: the K-means algorithm and the LBG (Linde Buzo Gray) algorithm. The paper provides a theoretical background of the vector quantization techniques, which is followed by experimental results illustrating their performance. The results demonstrated that the ITVQ (Information Theoretic Vector Quantization) provided the best performance in terms of classification rates, EER (Equal Error Rates) and the MSE (Mean Squared Error) compare to Kmeans and the LBG algorithms. The outstanding performance of the ITVQ algorithm can be attributed to the fact that the IT criteria used by this algorithm provide superior matching between distribution of the original data vectors and the codewords.