Radar signal recognition is an important step of radar countermeasure processing. The classical recognition method is called weight distance, in which the feature parameter weights are obtained by expert and then the weight distance of unknown radar signal and signal template in database is computed. For it existing subjectivity in setting of feature parameter weights with classical recognition method, and the computing method of recognition is too simple, all of which make recognition result can't reflect the true fact objectively. Considering this point, a fuzzy pattern recognition method based on neural network getting weights to radar signal recognition is studied in this paper, the feature parameter weights in this method are fixed on by neural network and then the unknown radar signal is recognized by fuzzy pattern recognition method. Simulation experiment and its result show the method in this paper is practicable and more reliable compared with classical method.