0

I'm trying to solve the following problem and will appreciate your help. In the example I have two signals. The 1st is the original signal and the 2nd is a some slightly stretched version of it. The 1st signal I segment into N overlapping windows and for each window I want to find its new length in the 2nd signal in order to know eventually the difference.

As I see, using cross-correlation is not suitable here, because the segmented waveform in the 2nd signal is not the same length as the original.

Using peak detection works only for part of the signal since close to the end the peaks become more and more distorted.

Will be happy for any help.

Overlaid signals

Window shift

Alex Z
  • 11
  • 1
  • First idea would be to use DTW, Dynamic time warping – Irreducible Aug 06 '19 at 10:37
  • Thanks for the idea! In this method both time series are stretched to get the best match, however in my case I just want to keep the length, say, of the 1st timeseries to be constant and stretch only the 2nd until it matches the first. Is there any modification to DTW to do so? – Alex Z Aug 06 '19 at 12:05
  • Right now I am not sure if both timeseries are being manipulated, in my memory DTW searches for an alignment of both timeseries and quantifies the "distance" between them. – Irreducible Aug 06 '19 at 12:32
  • As what I see it manipulates both making them the same length and with minimal distance – Alex Z Aug 06 '19 at 14:23

0 Answers0