Prev: image processing
Next: Problems initializing structure of structure.
From: Peng Liang on 11 Jun 2010 19:52
I have a time domain signal, which has two mixed frequency sine wave. 350Hz and 700Hz. I did FFT of this signal and be able to see the spectrum in frequency domain. however, I made an array which contains same length as that signal. I want to this array will be the high pass filter which ones everywhere except the zeros where I want to zeroing the spectrum that I dont want to keep. for example, zeroing from 0 to 400Hz will filter out the 350Hz sine wave but keep the 700Hz sine wave. Since convolving in time-domain will have the same result in multiplication in frequency domain. therefore, I multiply the filter(the zeros and ones array) with the signal after FFT. and then using IFFT reconstruct the signal back to time domain. Theoretically, the reconstruct signal should only have 700Hz sine wave signal and 350Hz filtered out. However, the reconstruct signal is not the one I expect. Is
there any thing wrong with my frequency domain multiplication?

Thanks in advance.
From: dbd on 11 Jun 2010 20:00
On Jun 11, 4:52 pm, "Peng Liang" <madokaauk...(a)hotmail.com> wrote:
>... However, the reconstruct signal is not the one I expect. Is
> there any thing wrong with my frequency domain multiplication?
>
> Thanks in advance.

Take a look at:
Steven W. Smith: Digital Signal Processing - A Practical Guide for
Engineers and Scientists, Newnes, ISBN 0-7506-7444-X

available on the net at:
http://www.dspguide.com/pdfbook.htm

Particularly Chapter 18 on FFT Convolution
http://www.dspguide.com/ch18.htm
for the requirements for achieving correct frequency domain filtering.

Dale B. Dalrymple
From: Peng Liang on 14 Jun 2010 22:04
dbd, thank you for your replying.

I looked at the spectrum after the multiplication in frequency domain. from the spectrum, I can tell one frequency has been filtered. however, when I convert back to time domain using IFFT, and plot it in time domain, the plot is not the sine wave anymore, it looks weird. and it wont be played correctly. Then, I tried x = abs(ifft(X)), and plot this x, it gives sine wave and with correct magnitude. and can be able to play out but with some distortion. I dont know where is wrong. hope get helps.

Thank you~

dbd <dbd(a)ieee.org> wrote in message <a7b183ff-19dd-435a-9102-c649470bf355(a)v12g2000prb.googlegroups.com>...
> On Jun 11, 4:52 pm, "Peng Liang" <madokaauk...(a)hotmail.com> wrote:
> >... However, the reconstruct signal is not the one I expect. Is
> > there any thing wrong with my frequency domain multiplication?
> >
> > Thanks in advance.
>
> Take a look at:
> Steven W. Smith: Digital Signal Processing - A Practical Guide for
> Engineers and Scientists, Newnes, ISBN 0-7506-7444-X
>
> available on the net at:
> http://www.dspguide.com/pdfbook.htm
>
> Particularly Chapter 18 on FFT Convolution
> http://www.dspguide.com/ch18.htm
> for the requirements for achieving correct frequency domain filtering.
>
> Dale B. Dalrymple
 | 
Pages: 1
Prev: image processing
Next: Problems initializing structure of structure.