Radar HRRP statistical recognition based on hypersphere model

The theoretical analysis and experimental results in this paper show that the independence assumption regarding elements in a radar high-resolution range profile (HRRP) sample, under which some statistical recognition methods were proposed, is not true. In addition, geometrically speaking, target HRRP samples spread on a unit hypersphere, since L2-normalized HRRP samples are applied to radar automatic target recognition (RATR) to deal with the amplitude-scale sensitivity problem. Therefore, this paper considers the two issues and proposes a novel statistical recognition method based on hypersphere model for power transformed HRRP samples under the jointly multivariate Gaussian distribution hypothesis. Compared with the conventional principal components analysis (PCA)-based subspace statistical recognition method, the hyperspherical spread of HRRP samples and the effectively discriminating information contained in the noise subspace can be fully utilized in this method without increasing computation complexity. The experimental results based on measured data show that our proposed method can greatly improve the recognition performance.