Forums

Methods for digital signal comparison?

Started by Shafik December 13, 2004
Hello wizards,

Is there a "best" known way for comparing two finite-same-sized
digital signals? I want a good way to measure the degree of similarity
of the *shape* of the two signals. This is for comparing magnetic
signatures of vehicles.

Is the best way just to do a scaled cross correlation between the two
signals?

--Shafik

Shafik wrote:
> Hello wizards, > > Is there a "best" known way for comparing two finite-same-sized > digital signals? I want a good way to measure the degree of
similarity
> of the *shape* of the two signals.
For comparing two signals, use correlation or coherence, or something like that. Check out a book on statistics.
> This is for comparing magnetic > signatures of vehicles. > > Is the best way just to do a scaled cross correlation between the two > signals?
No. There are ways of doing scaled correlation measurements, but whether they are "best" or not, depends on the application. Any method works very well in some situations and not at all in others. You essentially have to check out different approaches and see what works in your particular application. Rune
Shafik wrote:
> Hello wizards, > > Is there a "best" known way for comparing two finite-same-sized > digital signals? I want a good way to measure the degree of
similarity
> of the *shape* of the two signals. This is for comparing magnetic > signatures of vehicles. > > Is the best way just to do a scaled cross correlation between the two > signals? > > --Shafik
Well that is one of the best methods. If you like to extract frequency domain info and like to compare their spectrum then take FFT of both signals . As correlation being a second order statistics is phase blind you need to apply higher order statistics to get higher order moments and observe degree of similarity. If you are not interested in these infos then what you did is ok to my opinion. santosh