> thanks for the reply
> but, do we have to go about it recursively?
> can't we implement it in 1 step as in digital BPFs
> eg. an LPF for 0 to 2KHz and a BPF for 2 to 4KHz
>
> the pt is it takes up a lot of time to obtain many such filterbanks
>
> and, btw i am not taking abt wavelet compression,
> where the highpassed signals are ignored
> rather, they should be retained in order to get other filterbanks in
> the higher frequency region
Perhaps you're intrested in wavelet packets (google them up)? And yes, you
don't have to perform wavelet transform recursively.
--
Jani Huhtanen
Tampere University of Technology, Pori
Reply by Sukrut●May 11, 20062006-05-11
thanks for the reply
but, do we have to go about it recursively?
can't we implement it in 1 step as in digital BPFs
eg. an LPF for 0 to 2KHz and a BPF for 2 to 4KHz
the pt is it takes up a lot of time to obtain many such filterbanks
and, btw i am not taking abt wavelet compression,
where the highpassed signals are ignored
rather, they should be retained in order to get other filterbanks in
the higher frequency region
Reply by Andor●May 11, 20062006-05-11
Sukrut wrote:
> Is there a way to take a 1d wavelet trans to get 2 bands (lpf and hpf)
> over a fixed BW (say 4KHz) if the sampling freq is greater (eg 20KHz)
That's the point of wavelet filter banks. You apply them recursively to
the lowpassed signals (approximations), leaving the highpassed signals
(details).
Reply by Sukrut●May 10, 20062006-05-10
Is there a way to take a 1d wavelet trans to get 2 bands (lpf and hpf)
over a fixed BW (say 4KHz) if the sampling freq is greater (eg 20KHz)