The traditional guidance radar is facing increasingly complex new types of active interference patterns, and the radar must discriminate against various types of interference. The traditional interference recognition method is only effective for a specific single pattern, which has the disadvantages of poor generality and weak generalization ability, resulting in its inability to cope with the complex and changing interference countermeasures. Therefore, it is necessary to propose a more intelligent and robust universal interference recognition method to enhance the anti-interference ability of guided weapons. In order to improve the accuracy of interference signal recognition, this paper studied the multi-mode feature fusion algorithm, and finally fused time-domain, time-frequency domain, and information theory features to achieve classification. This paper introduces for the first time the concepts of entropy, relative entropy, and relative distance in information theory into the application scenario of interference signal classification. Through simulation experiments, it is shown that this method can effectively identify common interference and has good recognition accuracy even at low jam-to-noise ratios.