Radar HRRP adaptive denoising via sparse and redundant representations

We address the radar high resolution range profile (HRRP) denoising problem for improving the recognition rate of HRRP at low signal-to-noise ratio (SNR). Gaussian white noise in HRRP return is suppressed by an approach based on sparse representation. A Fourier redundant dictionary is established for sparsely representing HRRP returns. An adaptive signal recovering algorithm, Orthogonal Matching Pursuit-Modified Cross Validation (OMP-MCV), is proposed for obtaining denoised HRRP without requiring any knowledge about the noise statistics. As a modification to OMP-CV, OMP-MCV modifies the cross validation iteration condition, which can prevent the iteration procedure from terminating at local minimum impacted by noise. Simulation results show that OMP-MCV achieves better performance than OMP-CV and some other traditional denoising method, like discrete wavelet transform, for HRRP returns denoising.