From: Meng-Ju on
Hi all,

I am trying to compare two signals. One is original and the other has time
distortion. This is an example.
http://picasaweb.google.com/mengjuwu/Image#5455199328995717762.

In the graph, the shapes of C1, C2, C3, and C4 are very similar, but C2~C4
have time distortions in x-axis. I want to compare the similarity just
based on the shape of waveforms, and score the similarity at 0~1 level.

I have thought about using cross-correlation coefficient, but it considers
about time delay, not time distortion. What is the best way to measure
similarity between two signals which have time distortion?

Thanks in advance for your help,

Meng-Ju


From: alos on
>Hi all,
>
>I am trying to compare two signals. One is original and the other has
time
>distortion. This is an example.
>http://picasaweb.google.com/mengjuwu/Image#5455199328995717762.
>
>In the graph, the shapes of C1, C2, C3, and C4 are very similar, but
C2~C4
>have time distortions in x-axis. I want to compare the similarity just
>based on the shape of waveforms, and score the similarity at 0~1 level.
>
>I have thought about using cross-correlation coefficient, but it
considers
>about time delay, not time distortion. What is the best way to measure
>similarity between two signals which have time distortion?
>
>Thanks in advance for your help,
>
>Meng-Ju
>

http://en.wikipedia.org/wiki/Dynamic_time_warping

Regards,
Alek