A modified MCMC approach for classifying target and decoy

In towed radar active decoy (TRAD) scenario, the target and decoy, locating in same radar half-power beam, make object tracking more challenging in today's electronic warfare. Since the DOAs (direction-of-arrival) of target and decoy are the parameters of the likelihood of the observation data, the categorization of their becomes a sampling problem of machine learning field. Therefore, we, in this paper, propose a modified Markov Chain Monte Carlo (M-MCMC) approach towards classifying the target and decoy. First, we construct the observation signal model. Then, we find out that the parameters of the localization of target and decoy can be achieved by computing the covariance matrix of the observation vector. Moreover, rather than conventional numerical computation, our approach, intrinsically, combines the advantages of random walk and simulation annealing. The simulational results demonstrate the effectiveness of our approach.