Reply by Vladimir Vassilevsky February 21, 20072007-02-21

minfitlike@yahoo.co.uk wrote:

> I know how to estimate time-delays using FFTs and generalised cross- > correlation but I was wondering if LMS offers any advantages. eg if I > have say two acoustic sensors I would know the delay by assuming there > are 'zeros' in the leading coefficients of the LMS (ie the weights). > What mechanism would be used to say what is near to zero? Practically > you might have something obvious like > > 0.01,0.02 0.003,1.5,0.0, etc and the 1.5 stands out but is there a > more obvious way? Do we just find the peak of the weights?
Yes, it can be done as simple as that. Actually, there is not going to be just one delta peak. The result will be more or less fuzzy. For that purpose, you can perform the LMS on the decimated data for the reduced computation. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by February 21, 20072007-02-21
I know how to estimate time-delays using FFTs and generalised cross-
correlation but I was wondering if LMS offers any advantages. eg if I
have say two acoustic sensors I would know the delay by assuming there
are 'zeros' in the leading coefficients of the LMS (ie the weights).
What mechanism would be used to say what is near to zero? Practically
you might have something obvious like

0.01,0.02 0.003,1.5,0.0, etc and the 1.5 stands out but is there a
more obvious way? Do we just find the peak of the weights?

Thanks

F.