Enhancement of speech by adaptive filtering

In a variety of situations the problem of enhancing speech degraded by the presence of a competing speaker or background noise arises. A possible key to such enhancement lies in the quasi—periodic nature of the speech waveform which corresponds to narrow harmonically spaced bands of energy in the frequency domain. One approach to speech enhancement has been to utilize a time— variant digital comb filter for which the frequency spacing of the filter passbands varies with the fundamental frequency of the speech signal that is to be enhanced1. When the fundamental frequency varies sufficiently slowly, the use of a comb filter leads to significant enhancement of the desired speaker, but it degrades when the fundamental frequency varies rapidly. The procedure discussed here involves the use of an adaptive filter. When the fundamental frequency is constant, this adaptive filter reduces to a comb filter but more generally takes into account the variation of fundamental fre-