Echo Cancellation

Started by Jacky_42 July 26, 2005
Hi,

I have a problem where I send a signal and listen for reflections of
different surfaces but I dont want to listen for signals that are
reflected of two or more surfaces.

How can I use an adaptive filter to remove these kind of reflections? Any
help on this topic will be greatly appreciated.

Thanks,
Jack



		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com
Jacky_42 wrote:
> Hi, > > I have a problem where I send a signal and listen for reflections of > different surfaces but I dont want to listen for signals that are > reflected of two or more surfaces. > > How can I use an adaptive filter to remove these kind of reflections? Any > help on this topic will be greatly appreciated. > > Thanks, > Jack > > > > > This message was sent using the Comp.DSP web interface on > www.DSPRelated.com
Jacky, Hopefully your first double reflection arrives back after your last single reflection. In this case you could discriminate on the basis of time delay. I suggest you use the usual adaptive filter configuration for suppressing echoes, and make the number of taps large enough so that the signal sample delay from the first tap to the last is longer than your longest multiple-reflection time. (There may be a lot of taps involved.) When you do the FIR filter MACs and when you do the coefficient updates, only do it for the taps corresponding to the delay times of the range of delayed signals that you want to suppress, i.e the multiple-reflection signals, (and probably also the immediate, undelayed signal, if you want to get rid of that as well). I can't confirm that this works, but offer it as something to think about. Regards, John
Jacky_42 wrote:
> Hi, > > I have a problem where I send a signal and listen for reflections of > different surfaces but I dont want to listen for signals that are > reflected of two or more surfaces. > > How can I use an adaptive filter to remove these kind of reflections? Any > help on this topic will be greatly appreciated.
Jacky, that's a toughie. All you have to go on is time delay. (With light, there could be handedness, which distinguishes between even and odd numbers of reflections.) Is there some special character of the space in which reflections occur that might help? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
>Jacky_42 wrote: >> Hi, >> >> I have a problem where I send a signal and listen for reflections of >> different surfaces but I dont want to listen for signals that are >> reflected of two or more surfaces. >> >> How can I use an adaptive filter to remove these kind of reflections?
Any
>> help on this topic will be greatly appreciated. > >Jacky, that's a toughie. All you have to go on is time delay. (With >light, there could be handedness, which distinguishes between even and >odd numbers of reflections.) Is there some special character of the >space in which reflections occur that might help? > >Jerry >-- >Engineering is the art of making what you want from things you can get. >����������������������������������������������������������������������� >
Jerry/John, Firstly, thank you for responding to my query. The time delay approach is not suitable as I am always expecting a reflection from one surface or another. The internal reflections at some point in time are overlapping with the surface reflected signals. I don't know how to remove the echo if its correlated with your expected signal. The signals that I am sending are sound signals. I am not familiar with handedness. I dont know if it applies to sound waves. -Jack This message was sent using the Comp.DSP web interface on www.DSPRelated.com
Jacky_42 wrote:

   ...

> The signals that I am sending are sound signals. I am not familiar with > handedness. I dont know if it applies to sound waves.
"Mirror images" don't happen with sound, as I wrote. I was thinking aloud; a bad habit. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Jacky_42 wrote:
>>Jacky_42 wrote: >> >>>Hi, >>> >>>I have a problem where I send a signal and listen for reflections of >>>different surfaces but I dont want to listen for signals that are >>>reflected of two or more surfaces. >>> >>>How can I use an adaptive filter to remove these kind of reflections? > > Any > >>>help on this topic will be greatly appreciated. >> >>Jacky, that's a toughie. All you have to go on is time delay. (With >>light, there could be handedness, which distinguishes between even and >>odd numbers of reflections.) Is there some special character of the >>space in which reflections occur that might help? >> >>Jerry >>-- >>Engineering is the art of making what you want from things you can get. >>����������������������������������������������������������������������� >> > > > Jerry/John, > > Firstly, thank you for responding to my query. The time delay approach is > not suitable as I am always expecting a reflection from one surface or > another. The internal reflections at some point in time are overlapping > with the surface reflected signals. > > I don't know how to remove the echo if its correlated with your expected > signal. > > The signals that I am sending are sound signals. I am not familiar with > handedness. I dont know if it applies to sound waves. > > -Jack > > > > This message was sent using the Comp.DSP web interface on > www.DSPRelated.com
Jack, I'm guessing at your application, but if your problem is similar to the removal of 'clutter' in RADAR, a lot has been written on this. If the desired reflections are from a moving object and the undesired ones are from stationary objects, then an adaptive filter with a long adaption time could be useful as newly-introduced reflections from the moving object will be only partially canceled. Regards, John
John Monro wrote:

   ...

> Jack, > I'm guessing at your application, but if your problem is similar to the > removal of 'clutter' in RADAR, a lot has been written on this. > > If the desired reflections are from a moving object and the undesired > ones are from stationary objects, then an adaptive filter with a long > adaption time could be useful as newly-introduced reflections from the > moving object will be only partially canceled.
With adequate frequency discrimination, reflectors moving at different velocities can be distinguished by Doppler shifts. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������