Mark Borgerding wrote:

....
never asked. You made a good suggestion that was certainly
topical, considering more people read a thread than just the few
persons writing it. But the OP had signals "up to 45 seconds
long", far from infinite.

-- Mark

I'm one other reading the thread...
maybe my contribution is far off the original goal of the OP.

Just to satisfy my interest (and enhance my knowledge...), I would
like to put in two more general questions:

1) given two such signals of 45 seconds, which are equal.
now cut 5s from the beginning of the first and paste it to the
end of it .
then correlating this modified signal with the other unmodified.
I reckon that the frequency content is the same except on the
borders where the signal has been cut/pasted.
Therefore I would expect similar FFT results and probably a high
degree of correlation.
My question: does correlation detect the shift of most of the
signal and show good correlation?
Or would the result be a low correlation because signal contents
don't match at any point?

2) Given, I want to compare a signal (endless stream) with a certain
pattern (short piece of signal).
I expect that the signal contains pieces which are similar to the
pattern, but differ either
* in amplitude or
* in width or
* in both.
Which method would be chosen to find the occurrences?
Is cross-correlation the right approach?
Or will it fail in one of the cases?

Remember: it's not a current task which bothers me,
it's only that I'm curious
(maybe I'll need the answer soon, nevertheless ...)
so it's enough to give me a direction.

Bernhard