Abstract: Feature extraction is the key technique for Radar Automatic Target Recognition (RATR) based on High Resolution Range Profile (HRRP). Traditional feature extraction algorithms usually use shallow models. When applying such models, the inherent structure of the target is always ignored, which is disadvantageous for learning effective features. To address this issue, a deep framework for radar HRRP target recognition is proposed, which adopts multi-layered nonlinear networks for feature learning. Ground on the stable physical properties of the average profile in each HRRP frame without migration through resolution cell, Stacked Robust Auto-Encoders (SRAEs) are further developed, which are stacked by a series of RAEs. SRAEs can not only reconstruct the original HRRP samples, but also constrain the HRRPs in one frame close to the average profile. Then the top-level output of the networks is used as the input to the classifier. Experimental results on measured radar HRRP dataset validate the effectiveness of the proposed method.