cross correlation improvement to detect delay between two mikes

[Alberto Novell]

Hi there I'm trying to measure time delay of one percussive event recorded by two contact mikes on a glass surface. I decided to check the cross correlation (via FFT) and detect the index of the max in the function. The method is really jittery as the medium affects the transfer function of the signal. I wonder if there is a better method than cross correlation or some signal cleaning that can help before the calculation, or in the cross-correlation function. Also there might be a way to reconstruct the distortion transfer through the medium (as in phase cancellation)..any help?

 

[weetabixharry]

How jittery is jittery? I would expect the cross-correlation to perform pretty well. Can you upload your input data?

Do you have any control over the percussive events? (i.e. can you design what shapes these waveforms take?).

 

[Alberto Novell]

HI weetabixharry, thanks for your answer,,,
well in certain cases it jitters about 100-200 samples, while I'd like a resolution of 20 at least. It's in materials like glass where sound travels fast and the material distorts the wave. I uploaded some recordings using a stick and a finger, in each recording we tapped around the center, right and left (you can easily see which side comes first) creating three impulses.

download here:
**broken link removed**

I can't really predict the form of the impulse, but I could claibrate the system with some previous signals. So now I'm trying to mark the signal in the calibration.

Even just by looking at the signals is difficult to associate the peaks and the valleys..

If you have any idea please shoot

some images here