Robust Speech Recognition Using Perceptual Wavelet Denoising and Mel-Frequency Product Spectrum Cepstral Coefficient Features. INF 32 (3).