auto correlation of an irregularly occurring pulse buried in noise

[PhilipOakley]

I'm looking for techniques to extract the pulse shape of an irregularly occuring pulse in long term data, such as for example, down stream surges in river flow data (UK has irregular rain fall).

The regular autocorrelation isn't really a good predictor of the pulse shape. And cross correlation is a possiblity, but we don't know haw many pulse we are looking for.

I'm not aware of any of the embedded mathcad functions doing this.

Any suggestions?

Philip

[MichaelH]

I've had good luck in using FFTs to pick out pulses from noisy signals.

You could cut up your long term data into smaller chunks and see if anything pops out from the noise level.

You can then use the FFT peak values to reproduce the pulse shape.

[FDS]

Philip, for what it is worth there is a very nice add-on for this type of analysis called "Kornucopia". It is very good in filtering noise and works seemingly with MC 15. You can download a trial version for 30 days at www.bodietech.com. The package has more than 100 additional MC functions. I personally enjoy it very much.

[Fred_Kohlhepp]

Wavelets are supposed to be useful with irregular data . . .