Maximum likelihood estimation of Micro-Doppler parameters based on MCMC
Yihua Hu, Liren Guo, Xiao Dong, Shilong Xu · 2016
In recent years, target remote sensing based on the Micro-Doppler effect opens up a new way to target recognition and classification. One of the critical works in this research is parameter estimation. With maximum likelihood estimator (MLE) method, we can achieve the optimal unbiased result comparing with those suboptimal estimations, but the highly nonlinear and multimodal cost function of the Micro-Doppler signal detected by lidar yields the direct MLE method impractical because of the huge computational burden. For this reason, the paper proposes a new method utilizing a mean likelihood estimator based on Markov Chain Monte Carlo (MCMC) sampling methods. Firstly, the closed expression for estimation and Cramer-Rao bound(CRB) are derived, then the effect of initialization and the proposal distribution of the algorithm on estimation accuracy is analyzed. Lastly the parameter is estimated through a simulation. The simulating results present that with the increase in data length, Markov chain converges to the target distribution, resulting in that the estimating performance can achieve the Cramer-Rao lower bound. Moreover the computational complexity is acceptable than the other algorithms with the same performance.