| Due to the ever-increasing complexity in modern radar systems it has become apparent that new ways of classifying received signals are required. One such method, proposed by Dr. Pierre Lavoie of Defense Research Establishment Ottawa (DREO), is to model a given radar system as a finite state automaton. In so doing, it is possible to uncover the underlying system processes in a probabilistic fashion using hidden Markov models (HMMs). This thesis investigates the feasibility of this technique for radar identification. Artificial deterministic and pseudo-random signals are used to show that HMMs can provide adequate signal recognition that is far superior to conventional cross-correlation techniques.; Finally, two methods for predicting false recognition rates and devising optimal decision regions between competing HMMs are investigated. The first approach uses only the prior pulse repetition interval (PRI) statistics of a known radar and template matching which can lead to a qualitative understanding into radar correlation. The second uses a posteriori knowledge of HMM recognition probability distributions. (Abstract shortened by UMI.)... |
