palmar
Posts:
40
Registered:
11/16/07
|
|
Re: wavelet packet decomposition
Posted:
Mar 14, 2012 5:49 AM
|
|
"dypiet " <vidyapurohit88@gmail.com> wrote in message <jjo7je$fca$1@newscl01ah.mathworks.com>...
> hi...
> i have many basic doubts about wave packet decomposition(wpd).
> is it right that the wpd divides both freq n length of signal?.
> the fs of my signal found by command
> [s,fs]= wavread('G:\4.wav'); is 11025.
>
> then i wanted to convert the signal in 8khz bandwidth i.e(0-8khz) so i resampled it at 16khz(sampling freq must be twice of the max freq. here i want it 8khz so my new fs must be 16khz)
> [p,q] = rat(16000/11025);
> y = resample(s,p,q);
>
> so now if i do the wpd at level 3
> t = wpdec(y,3,'db6');
> the 8 nodes that i get will have freq 0-1khz,1-2khz,2-3khz,3-4khz,4-5khz,5-6khz,6-7khz,7-8khz?..
> is what i am saying right?..
> please guide me..
>
> -vidya
Vidya:
Do you really need to have the distribution of 0-1khz,1-2khz,2-3khz,3-4khz,4-5khz,5-6khz,6-7khz,7-8khz in the 3rd level. Because this seems to be your motivation do the resampling prossecing right?
If you decide to keep this procedure you must not resample (this will introduce aliasing). You should interpolate because I believe (may be wrong) that resample,m does not filter the data, and you must do it with resampling.
You can adjust the level of your wavelet packet decomposition to reach a frequency distribition close to the one you want anr discard the data interpolation
|
|
