Abstract: Faced with the increasingly complex electromagnetic jamming environment, the effectiveness of traditional jamming recognition methods has become increasingly limited, posing challenges to the development of radar anti-jamming technologies. To improve the accuracy and efficiency of jamming recognition, this paper proposes a jamming recognition method based on multi-domain features and ensemble learning. Firstly, theoretical derivation and analysis are conducted on the four transform domains of jamming signals, yielding a high-discriminability feature dataset for different jamming signals. Secondly, a Stacking ensemble learning recognition framework is constructed, where decision trees, support vector machines, and naive Bayes serve as base learners, and logistic regression acts as the meta-learner. A five-fold cross-validation method is adopted for the training and validation of the dataset. Finally, jamming recognition experiments are carried out for 8 types of single jamming signals and 4 types of compound jamming signals under different jamming-to-noise ratios (JNRs). Comparisons are made with traditional machine learning methods and neural network methods. The results demonstrate that the proposed method in this paper exhibits certain advantages in both recognition accuracy and speed, which can effectively enhance the anti-jamming capability of radar.