关键词: 干扰效果在线评估, 迁移学习, DS证据理论, 马尔可夫模型

Abstract:

Online jamming effect evaluation is a key component of closed-loop decision-making in electronic countermeasures, aiming to assess the impact of jamming on radar operation in real time and provide a basis for jamming strategies. To address the problems of non-cooperative radar target evaluation, including the absence of labeled data in the target domain, conflicts among multi-source evidences, and the dynamic evolution of jamming states, an evaluation method integrating transfer learning with spatio-temporal feature optimization is proposed. A domain adversarial neural network-based transfer learning framework is constructed to transfer jamming-response discrimination experience from known radars to unknown ones. Dempster–Shafer （DS） evidence theory is introduced to fuse the outputs of multiple support vector machine classifiers as multi-source heterogeneous evidences and solve the conflict problem of discrimination results from different feature perspectives. A dual-chain Markov model is designed to characterize the coupled evolution law between jamming states and radar operating modes, enabling temporal tracking of jamming effects. Experimental results show that the proposed method achieves an overall recognition accuracy of 90.33% for effective, partially effective, and ineffective jamming states on unknown radar targets, with an average single-sample inference time of 15.6 ms, satisfying the requirements of realtime decision-making in electronic countermeasure.

Key words: jamming effect online evaluation, transfer learning, Dempster-Shafer (DS) evidence theory, Markov model