Recognition and classification of nonlinear chaotic signals.

We propose a method of identifying and classifying signals generated by nonlinear, chaotic processes even when there are other signals and noise present. The method compares probability densities constructed from signals with a library of known densities. We note that there are many different invariant densities that can be constructed and that the characteristic functional of these densities factorizes into a product of characteristic functionals. Each member of the product is the characteristic functional of one of the signals present. This factorization provides a means of identifying the presence of a particular signal. Simple examples are given to demonstrate the method.