Would you plz tell me how did u extract noise frame from silence periods. Did
you use Zero-Crossing-Rate approach or Energy-Based approach. I have tried both
of them and they work pretty well in clean speech but not with speech
contaminated with background noise.
Your comments plz...
Hi all,
>Has anyone here done Spectral Subtraction in Wavelet
Domain? How is this different from the one in FFT domain?
>I have estimated the noise frame from the silence periods in the Noisy speech
signal. Now the next step is applying DWT. My question is
>1) Should i apply DWT to Noise Estimate frame and Noisy speech frame by frame
or at once i.e. I have one frame (length = 512) as Noise Estimate and my size of
noisy speech is 20,000. Should I divide my noisy speech in to frames of 512 and
then apply DWT to both Noise estimate and Noisy speech or apply DWT to the whole
of Noisy Speech(without Framing).
>2) Now once both Noise Estimate and Noisy signal are in Wavelet domain how do i
perform the Spectral subtraction.
>Can you give me the necessary equations/ Titles of the Related Papers, I would
be very thankful for any help/ Suggestion on this issue.
>Thank you,
>Khalandar Pasha N