Channel compensation for SVM speaker recognition.

Abstract

Cross-channel degradation is one of the significant chal-lenges facing speaker recognition systems. We study the problem for speaker recognition using support vector ma-chines (SVMs). We perform channel compensation in SVM modeling by removing non-speaker nuisance dimensions in the SVM expansion space via projections. Training to re-move these dimensions is accomplished via an eigenvalue problem. The eigenvalue problem attempts to reduce mul-tisession variation for the same speaker, reduce different channel effects, and increase “distance ” between different speakers. We apply our methods to a subset of the Switch-board 2 corpus. Experiments show dramatic improvement in performance for the cross-channel case. 1.