Radar Active Jamming Recognition Based on Wavelet Packet Decomposition and Random Forest Feature Selection

Jamming recognition is an important part of radar anti-jamming technology research. In order to improve the accuracy of jamming signal classification and recognition, a jamming signal classification and recognition method based on wavelet packet decomposition, time-frequency domain features and random forest is proposed. The wavelet packet transform is used to decompose and reconstruct jamming signals, and the multi-dimensional time-frequency domain characteristic parameters of each reconstructed signal are calculated separately, then the signal recognition ultra-high dimensional features set is constructed. For the redundant features in the original feature set, random forest is used to analyze the feature importance, feature components with higher importance are labeled from the original high-dimensional feature set, and sensitive features are selected to input into the support vector machine classifier for jamming classification and recognition. Experimental analysis of 8 types of radar jamming simulation data shows that the method proposed in this paper can effectively improve the dimensionality of signal feature extraction and screen out low dimensional sensitive feature sets with high importance, thereby in strong noise environments, it can achieve high jamming recognition accuracy and significantly improve recognition efficiency.