Abstract: Translation sensitivity is one of the important limits of machine learning methods applied on radar target recognition based on high resolution range profiles. In the real radar situations, the targets are located on the different positions in the high resolution range profiles, leading to the result that the information represented by the same dimension of different high resolution range profile samples is not coherent with each other if the high resolution range profiles are regarded as the input of a machine learning system directly, which reduces the performance of the machine learning system. To solve the problem, a novel project function is proposed for high resolution range profile alignment, and solved by an approximation algorithm. Benefit from the alignment process, with machine learning methods, the performance of radar target recognition based on high resolution range profiles is improved significantly. The performance of the proposed method is validated based on measured data, and experimental results show the effectiveness of the proposed method.