(Publisher of Peer Reviewed Open Access Journals)
ICETTR-2013
Full-Text PDF
Paper Title : Speaker Verification Using I-Vectors
Author Name : N.S.Kalkar, S.P.Savarkar, RajaniP.K
Abstract : This paper deals with the study of low-variance multi-taper Mel-Frequency Cepstral Coefficient (MFCC) and Perceptual Linear Prediction (PLP) features in i-vector speaker verification. Hamming windowed periodogram spectrum estimate is a important method to calculate the MFCC and PLP features. Single tapered spectrum estimate has large variance, which can be reduced by averaging spectral estimates obtained using a set of different tapers, leading to a multitaper spectral estimate. The multi-taper spectrum estimation method has proven to be powerful especially when the spectrum of interest has a large dynamic range or varies rapidly. In this study primary goal is to validate those findings using an up-to-date i-vector classifier. Robust Perceptual Linear Prediction (PLP) features using multitapers. Sine – Weighted Cepstrom Estimator based multitaper method provides average relative reductions of 12.3% and 7.5% in Equal Error Rate, respectively. For the Multi-Peak Multi-Taper method, the corresponding reductions are 12.6% and 11.6%, respectively. Finally, the Thomson multitaper method provides error reductions of 9.5% and 5.0% in EER for MFCC and PLP features, respectively. Both the MFCC and PLP features computed via multitapers provide systematic improvements in recognition accuracy.
Keywords : Speaker verification, Multi-taper spectrum, Feature extraction, i-vectors, MFCC, PLP.
Cite this article : N.S.Kalkar, S.P.Savarkar, RajaniP.K " Speaker Verification Using I-Vectors " ,ICETTR-2013 ,Page No : 376-384.