Optimized Subspace Weighting for Robust Speech Recognition in Additive Noise Environments

摘要

Signal Subspace (SS) based speech enhancement techniques obtain significnat additive-noise reduction by altering the singular value spectrum of the speech observation matrix.