Clean speech reconstruction from noisy mel-frequency cepstral coefficients using a sinusoidal model | IEEE Conference Publication | IEEE Xplore