Jamming Recognition Methods Based on Spherical-Coordinate Mapping

Accurate recognition of active jamming signals is critical for electromagnetic measurement and anti-jamming systems. However, existing studies usually assume bandwidth matching between the signal and the receiver. In practical instrumentation scenarios, wideband electromagnetic reconnaissance receivers often monitor narrowband jamming signals, resulting in severely compressed and indistinct time–frequency representations that degrade recognition performance. In addition, the high computational cost of conventional deep neural networks limits their deployment on resource-constrained maneuverable platforms. To address these issues, this paper proposes a feature enhancement method based on Spherical Coordinate Mapping (SCM), which reconstructs compressed time–frequency data in a spherical coordinate space to improve feature discriminability under severe bandwidth mismatch. Complementing this, a lightweight recognition network, termed Jamming Recognition Network-LiteX (JRN-LiteX), is further developed for real-time edge processing. Extensive experiments on both simulated and real-world measured datasets show that SCM consistently improves recognition accuracy, with a maximum gain of 25.00 percentage points on measured data. Moreover, the proposed framework provides reliable predictive uncertainty estimates, shows reliable rejection capability for unknown jamming signals under open-set conditions, and achieves competitive recognition performance with low computational overhead. These results demonstrate that the proposed SCM and JRN-LiteX offer an effective and deployable solution for real-time jamming recognition in practical electromagnetic measurement systems.