Reduction of musical noise in spectral subtraction method using subframe phase randomisation

A subframe phase randomisation method is proposed and applied to the enhanc ed speech with spectral subtraction method to reduce musical noise in the n onvoicing region. The musical noise in the spectral subtraction method is l argely due to narrowband tonal components appearing somewhat periodically i n the spectrogram of unvoiced and silence regions. Thus, each synthesis fra me in a nonvoicing region is divided into several subframes to broaden the narrowband spectrum, and then phases of silence and unvoiced regions are ra ndomised to disrupt the tonal spectrum structure.