Quantitative Perceptual Separation of Two Kinds of Degradation in Speech Denoising Applications