# Random signal modeling with Matlab

I want to build a detector of sorts. Say I have a bunch of signals and they all share some patterns, like some peaks in frequency or something more complicated that I can't bother to calculate by hand. I do however have plenty of these signals, so maybe I could "model" statistically, and more importantly, numerically through Matlab, their behavior. I'm familiar with the concepts of auto correlation and spectral density, but I'm not sure how to infer them from of the samples. I know Matlab has a function called autocorr that calculates the autocorrelation of the sample, but I'm not sure I understand how it works. In any case, I don't know how to get the best autocorrelation model out of all the autocorrelations calculated for each sample. Taking an average seems a bit too easy and I'm not sure how it would work.

Thank you!

• There is so many approaches to design such models, it is mostly impossible to give advice without knowing what (more exact than "build a detector or sorts") you want to do and why you think such a model might be beneficial. – jan Nov 18 '13 at 12:59