Exploiting Chaos to Improve Signal Detection

We will use our existing commercial grade chaotic data analysis tool kit to characterize ocean clutter and noise in hostile environments. This knowledge will be used to demonstrate how Receiver Operating Curves can be improved by reducing the mismatch between real noise and the noise model in traditional filters. We will also study the feasibility of signal detection with chaotic filters. For this task, we will also evaluate new algorithms by using real data to generate ROC curves. Finally, a new detection method for specific signals of interest will be investigated. Randle, Inc. and The Institute for Nonlinear Science (UCSD) have been conducting applied research since 1989 on signal-processing methods derived from the study of chaotic systems. Chaotic signal processing in state space is effective for many types of signals, including those that are nonperiodic or chaotic. The remarkable aspect of these methods is that they usually operate directly on data. No prior knowledge or model of the system is required. We have shown that real signals of interest are chaotic and can be detected and classified using these methods. In many cases, the nonlinear characteristics of the signals of interest are known or can easily be derived.