DSPRelated.com
Forums

Adaptive filtering

Started by thom June 20, 2006
Hi everyone

I'm trying to do adaptive filtering with Matlab (LMS adaptive filtering).
But with this method, it seems that I need to know the original signal
form (without noise). And I don't have information about the original
signal (it's a large band signal, ofdm modulation). Do you know if it's
possible to reduce noise with this method?

Thanks a lot.
thom wrote:
> Hi everyone > > I'm trying to do adaptive filtering with Matlab (LMS adaptive filtering). > But with this method, it seems that I need to know the original signal > form (without noise). And I don't have information about the original > signal (it's a large band signal, ofdm modulation). Do you know if it's > possible to reduce noise with this method? > > Thanks a lot.
Don't you think "I don't have information about the original signal" is rather incompatible with "it's a large band signal, ofdm modulation"? Steve
Maybe I'm not clear... Yes I know for example the signal bandwidth. But I
don't have any reference which could be the "desired signal" used in
adaptive filtering algorithms. So I don't really how to use this technique
in my case...

Thanks for your help.



>Don't you think "I don't have information about the original signal" is >rather incompatible with "it's a large band signal, ofdm modulation"? > >Steve >
thom wrote:
(top posting fixed)
> >>Don't you think "I don't have information about the original signal" is >>rather incompatible with "it's a large band signal, ofdm modulation"? >> >>Steve >> > > >
> Maybe I'm not clear... Yes I know for example the signal bandwidth. > But I don't have any reference which could be the "desired signal" > used in adaptive filtering algorithms. So I don't really how to use > this technique in my case... > > Thanks for your help. > With adaptive equalization it is common to use the demodulated data to generate the prototype signal. If your SNR is good enough to get mostly good data out of a demodulator then this should work for you -- and if it's not you may not be able to recover the signal with any kind of filtering. My understanding of OFDM is that it is inherently resistant to the kinds of effects that demand adaptive filtering or equalization -- what effects are you trying to mitigate, and do you really need to? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Posting from Google? See http://cfaj.freeshell.org/google/ "Applied Control Theory for Embedded Systems" came out in April. See details at http://www.wescottdesign.com/actfes/actfes.html
Actually I have to work on signals that are good enough to be
demodulated... Although I would like to reduce the noise part in the
received signal to improve the transmission sytem. And I don't know what
signal I can use as a reference, since I only have the received signal
with white noise.

Thom




>With adaptive equalization it is common to use the demodulated data to >generate the prototype signal. If your SNR is good enough to get mostly
>good data out of a demodulator then this should work for you -- and if >it's not you may not be able to recover the signal with any kind of >filtering. > >My understanding of OFDM is that it is inherently resistant to the kinds
>of effects that demand adaptive filtering or equalization -- what >effects are you trying to mitigate, and do you really need to? > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Posting from Google? See http://cfaj.freeshell.org/google/ > >"Applied Control Theory for Embedded Systems" came out in April. >See details at http://www.wescottdesign.com/actfes/actfes.html >
thom wrote:
(top posting fixed)
> >>With adaptive equalization it is common to use the demodulated data to >>generate the prototype signal. If your SNR is good enough to get mostly > > >>good data out of a demodulator then this should work for you -- and if >>it's not you may not be able to recover the signal with any kind of >>filtering. >> >>My understanding of OFDM is that it is inherently resistant to the kinds > > >>of effects that demand adaptive filtering or equalization -- what >>effects are you trying to mitigate, and do you really need to? >>
> Actually I have to work on signals that are good enough to be > demodulated... Although I would like to reduce the noise part in the > received signal to improve the transmission system. And I don't know > what signal I can use as a reference, since I only have the received > signal with white noise. > > Thom * If the noise characteristics are known you don't need an adaptive filter. * If your channel doesn't suffer from fading or dispersion then you don't need an equalizer, adaptive or not -- and OFDM is designed to eliminate the need for fancy equalization in any case. * If the noise is white and the channel is purely attenuating then the classical demodulation technique of using a matched filter is optimal: any extra filtering, adaptive or otherwise, will only degrade your system performance. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Posting from Google? See http://cfaj.freeshell.org/google/ "Applied Control Theory for Embedded Systems" came out in April. See details at http://www.wescottdesign.com/actfes/actfes.html