Thread: cross correlation or similar approach to find signals with exact/ similar amplitude

Hey everyone,

I am currently trying to find signals which are exactly like or at least similar to my reference / test signal.
The way I approach this problem right now is a simple cross correlation. This, however, turns out to be insufficient so far, as results show that it will find signals which are similar in shape, curve and durations, but the amplitudes can vary a lot. The problem is that I need to separate these signals due to their amplitudes as well and I just have no clue how to do so.

I am using the matlab function xcorr to approach this issue (if anyone is familiar with that).

If anyone can help? Theoretical and/ or matlab help is very much appreciated.

Thank you very much!

Originally Posted by Burrita

Hey everyone,

I am currently trying to find signals which are exactly like or at least similar to my reference / test signal.
The way I approach this problem right now is a simple cross correlation. This, however, turns out to be insufficient so far, as results show that it will find signals which are similar in shape, curve and durations, but the amplitudes can vary a lot. The problem is that I need to separate these signals due to their amplitudes as well and I just have no clue how to do so.

I am using the matlab function xcorr to approach this issue (if anyone is familiar with that).

If anyone can help? Theoretical and/ or matlab help is very much appreciated.

Thank you very much!

Look at the normalised cross correlation xcorr(x,y)/xcorr(y,y) where x is the test signal and y the replica. Now you are interested signals with a normalised cross correlation close to 1.

CB