Radar HRRP target recognition in frequency domain based on autoregressive model

In this paper, we adopt the autoregressive (AR) model to characterize the frequency spectrum amplitude of high resolution range profile (HRRP) and extract the AR and partial correlation (PARCOR) coefficients, which are invariant to the initial-phase, translation and scale changes of HRRP, as discriminating features. Moreover, a mixture model based frame partition method is proposed and a Bayesian Ying-Yang (BYY) harmony learning algorithm is adopted to determine the frame number automatically during parameter learning. Experimental results based on measured data demonstrate the proposed features are superior to others in their minor frame number, robustness to sample size and good rejection ability.