UAV Jamming Recognition Algorithm Based on Multi-step Singular Spectrum Analysis

Due to the widespread use of electronic devices, unmanned aerial vehicle (UAV) wireless communication systems are inevitably affected by complex electromagnetic environments. In UAV anti-jamming systems, jamming recognition is a critical component. The multi-step singular spectrum analysis (MS-SSA) algorithm is proposed to address the issue of low jamming recognition accuracy caused by the ambiguity of UAV jamming signals under low jamming-to-noise ratio (JNR) conditions. MS-SSA effectively performs denoising on jamming signals through advanced filtering techniques. This approach reduces the computational complexity of the singular spectrum analysis (SSA) algorithm while enhancing recognition accuracy. By increasing the delay step, MS-SSA improves the construction of the trajectory matrix and sequence reconstruction method. Noise is separated from UAV jamming signals by the MS-SSA filtering algorithm in high-noise environments. By using the MS-SSA filtering algorithm to decrease noise from UAV jamming signals in high-noise environments and combining the multi-domain feature parameters extracted from the signals, the channel space attention convolutional neural network (CSA-CNN) is proposed for UAV jamming recognition. Experimental results show that compared to traditional filtering methods such as wavelet and Kalman algorithms, the average root mean square error (RMSE) of MS-SSA is reduced by 55.8% and 37.94%, respectively. Particularly under low JNR conditions, MS-SSA demonstrates superior filtering performance. After processing with MS-SSA, the effectiveness of MS-SSA filtering is further validated through UAV jamming recognition using CSA-CNN.