As one of the important development directions of modern radar technology, radar target recognition is of great significance to the future weapon system. Aiming at the problem of translational sensitivity in high resolution range profile (HRRP) in preprocessing, the global minimum entropy distance alignment algorithm is used to accurately align the range profile and improve the accuracy of distance alignment. To improve the accuracy of radar target recognition, a new method of target recognition based on generalized regression neural network (GRNN) is proposed. The K-fold cross-validation method is used to train the neural network, and the optimal value of the smoothing factor spread of GRNN neural network is found according to the mean square error (MSE). The optimal input and output values of the target recognition training samples are obtained. By contrast, GRNN neural network with optimal smoothing factor spread will greatly improve its convergence speed and generalization ability. The simulation results show that the radar target recognition based on the improved GRNN neural network can obtain higher and more accurate recognition accuracy.