This paper presents a new frequency-domain approach to implement an ad
aptive postfilter for enhancement of noisy speech. The postfilter is d
escribed by a set of DFT coefficients which suppress noise in the spec
tral valleys and allow for more noise in formant regions which is mask
ed by the speech signal. First, we perform an LPC analysis of the nois
y speech and calculate the log magnitude spectrum of the input speech.
After identifying the formants and valleys (by a new method), the log
magnitude spectrum is modified to obtain the postfilter coefficients.
The filtering operation is also done in the frequency domain through
an FFT and an overlap-add strategy to get the postfiltered speech. Exp
erimental results on 8-kHz-sampled speech show that this new frequency
-domain approach results in enhanced speech of better perceptual quali
ty than obtained by a time-domain method. This new method is especiall
y efficient in eliminating high frequency noise and in preserving the
weaker, high frequency formants in sonorant sounds.