Speaker identification using ensembles of feature enhancement methods

Abstract

In this paper, we propose a classifier ensemble of various channel compensation and feature enhancement methods for robust speaker identification on various environments. The proposed ensemble system is constructed with 15 classifiers including three channel compensation methods (including CMS and variance normalization, and without compensation) and five feature enhancement methods (including PCA, kernel PCA, greedy kernel PCA, kernel multimodal discriminant analysis, and without enhancement). Experimental results show that the proposed ensemble system gives the highest average speaker identification rate in various environments (channels, noises, and sessions). © 2011 Springer-Verlag.