Speech Enhancement Using Kernel Recursive Least-Squares Method
Keywords:
Speech Enhancement, Linear AdaptiveFiltering, Reproducing Kernel Hilbert Spaces, KernelMethods, Kernel Adaptive Filtering
Abstract:
In this paper, we propose a new speech enhancementstructure based on kernel recursive least squares adaptivefiltering. The combination of the famed kernel trick andrecursive least squares (RLS) algorithm yields powerfulnonlinear extensions, named collectively here as KRLS. Thismethod improves the adaptive filtering performance innonlinear adaptive filtering scenarios. We compare theperformance of this kernel based algorithm in the area of dualchannelspeech enhancement with other linear adaptive filteringtechniques. Experimental results show that the proposedenhancement structure has better performance in a sense ofmean-squares error (MSE) and speech quality improvementthan the those based on standard LMS, Normalized LMS, Affineprojection, and conventional RLS algorithms.