The Denoise and Reconstruction Method for Radar HRRP Using Enhanced Sparse Auto-Encoder

The HRRP is a summation vector of sub-echoes of the target scattering points acquired by the wide-band radar. Generally, the HRRP obtained under the non-cooperative complex electromagnetic environment is contaminated by strong noise, which result in a significant reduction in recognition accuracy. Effective pre-processing of the HRRP can solve the above problem. A denoise and reconstruction method for HRRP is proposed based on Enhanced Sparse Auto-Encoder. Auto-Encoder is a representative non-linear model. To better sparse reconstruct the HRRP, an enhanced sparsity term is proposed in the objective function and the sparse coefficient is computed according to the intrinsic dimension of HRRP. Simulation results show that the proposed method can both reconstruct the signal and suppress noise effectively, which significantly outperforms comparison methods, especially at low SNR conditions.