Fitering noisy signals that should have opposite signs

[rforoozan]

Hi there,

I have 2 signals without time lag. What I know is that because of the physics of my problem, these two signals should always have opposite signs, if there be no noises. I can clearly see it when I filter the low frequencies. But my results are highly dependent on the filtering I apply. Can you suggest me a criteria for designing the filter? What I've done is to minimize the dot product of the two signals, varying the cut off frequency of the filter which works but not as good as I want.

Thanks...

I'm a beginner, the band pass filter I've wrote just omits the out of range frequencies.

 

[ring0]

Why not simply calculate a correlation function of the two signals? For truly opposite signals it should be negative and have maximum absoulte value.

C(x,y) = ∫x(t)y(t)dt

 

[rforoozan]

Thanks, but I think it will be the same as the dot product I've done.