[music-dsp] Noise Cancellation

Larry Trammell (aka RidgeRat) ltramme1476 at earthlink.net
Tue Jun 8 03:10:12 EDT 2010

Sorry for chiming in late on this thread, but the adaptive filtering 
approach to this problem is described nicely in "Adaptive Inverse 
Control" by Bernard Widrow and Eugene Walach, section 7.6. Basically, 
this approach uses a transversal (delay line) feedforward filter tuned 
by the LMS (gradient) algorithm. One microphone senses the external 
noise pressure you want to suppress, and one senses the sound pressure 
within the chamber that you want to protect. The objective of the LMS 
tuning algorithm is to adapt the coefficient of the filter in such a way 
that given the current and past levels of sound pressure measured from 
the outside noise, the filter predicts (i.e. feedforward) the signal to 
apply to the cancellation driver (with its own response dynamics) to 
reduce the correlation to the outside noise as close to zero as possible 
within the protected chamber. You don't want to cancel the program 
source, so you can either calibrate the dynamic response to the intended 
program signal within the chamber, or you can use a second LMS process 
to adaptively learn these response dynamics too. Removing the program 
sound pressure reasonably well from the measured internal pressure helps 
to isolate the external noise component and improve the adaptive tuning 

This process is much easier to implement than it is to describe.

Thomas Strathmann wrote:
> Am 6/2/10 13:03 , schrieb Victor Lazzarini:
>> Also, I'm not sure how the filter is supposed to work, since the noise
>> you are trying to block is bypassing any of the DSP
>> (ie. is made up of the environment sounds). I can see how some sort of
>> phase cancellation works, so I guess what you
>> are trying to say is that the system would generate some signal based on
>> an adaptive process that would provide the
>> desired effect. Maybe you mean that the filters will be used in an
>> analysis stage?
> I do not know what is used in practical applications (Robert mentioned 
> headphones), but in theory you would use an adaptive filter such as a 
> Wiener filter which is based on linear prediction of an additive noise
> source. If I recall correctly this approach only works if the stochastic 
> processes that "describe" the signal and noise source do not have time 
> varying covariances.
>     Thomas
> -- 
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews, 
> dsp links
> http://music.columbia.edu/cmc/music-dsp
> http://music.columbia.edu/mailman/listinfo/music-dsp

More information about the music-dsp mailing list