Decision Threshold Setting for Radar System Identification Using HMM

The principal objective of applying the Hidden Markov Model (HMM) in Radar System Identification (RSI) is to show that the recognition performance of a HMM exceeds that of the conventional methods such as cross-correlation. An important offshoot of this research is to provide a method for choosing the optimal model parameters for an actual radar signal so that a library of HMMs can be created and used for practical EW tasks. Therefore, optimal threshold settings between competing HMMs should be investigated to improve overall recognition performance. In this paper, a new method for predicting the false recognition rates and deriving optimal decision regions between competing HMMs that model the dynamic behavior of the radars stored in the threat library of the Electronic Warfare (EW) is proposed. The proposed method 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. Moreover, the paper investigates the idea of a threshold setting
so that the receiver can have a reject option and decide that an observation sequence does not belong to any of the HMMs in its library. Computer simulations are performed through the paper to validate the obtained theoretical analysis.