shape similarity between two signals

How can I compare the shape similarity of two signals? corrcoef does not work, because two signals having exactly the same shape of different length tend to have a small coefficient.

Answer 1

[Matt Tearle]

If the signals have different time bases, you'll have to align them first, then use corrcoef. To do that you could use either set operations, like intersect, or interpolation of some kind, in which case use interp1. Something like this:

ti = linspace(tmin,tmax,npts);
y1i = interp1(t1,y1,ti,'cubic');
y2i = interp1(t2,y2,ti,'cubic');
corrcoef(y1i,y2i);

[Bo Wu]

Is there any other method for the comparison, such as in the frequency domain after some transformation?