Signal subspace decompositions for perceptual speech and audio processing by