Signal similarity analysis after cross correlation

I have a time series signal and using this as reference, from the main data set extracted possible matching sections using cross correlation technique.(xcorr ).Based on the threshold xcorr value used, it may contain some not-matching signals as well(slight difference).

I am looking for a way to further analyse the output and remove the not-matching signals. Any better way to do that ? Since I used cross correlation method(mathematical) would it be beneficial to use image analysis of the two signals ?

Reference picture attached.

Answer 1

[Daniel M]

Wouldn't it work just to set the threshold higher?

It depends on what you mean by matching. If you want the exact same signal, you could just take the residuals of reference signal and the candidate signal and check if the max(abs()) value is less than some tolerance level.

For example:

res = abs(refWave-candidateWave); 
if max(res) > 10*eps
    fprintf('These two waves do not match according to the criteria.\n')
end

[Sreeraj Arole]

Thank you Danier for the answer.

Wouldn't it work just to set the threshold higher? I am using an automated script to ensure all the matching signal shapes are captured so the script is dynamically adjusting the threshold based on expted number of captured signals

Matching signals some times means the amplitude differ by an offset. The idea is two signal should have similar shape

Thanks

[Daniel M]

Similar shape but not exact shape. Well whenever you determine your criteria, people can help you code it.

[Sreeraj Arole]

Sorry, bit confusing isn't it !! So is my requirment :) ! ! Needed a bit more work around on the input data before the cross correlation. fiugred it out. Thanks for your inputs.