Adaptive weighting of subband-classifier responses for robust text-independent speaker recognition.

A new strategy for robust speaker recognition using multiple classifiers in the subband domain is presented. This approach relies on the computation of a basis for the null space to implement an adaptive combination of subband-classifier responses. The simulation results using speech corrupted by different types of environmental noises show that the proposed method significantly outperforms other techniques reported in the literature. [ABSTRACT FROM AUTHOR]