Deep Learning-Based Interference Mitigation of SF-Modulated Signals for FMCW Systems

We present a method for mitigating interference in frequency-modulated continuous-wave (FMCW) systems caused by stepped-frequency FMCW (SF-FMCW) signals. Mutual interference between radar systems can lead to a raised noise floor and obscured targets, resulting in degraded target detection performance. While conventional FMCW-to-FMCW interference is well-documented, the emergence of varied radar waveforms, such as SF-FMCW, introduces new challenges. Because these characteristics are different from those of typical FMCW-to-FMCW interference, a novel mitigation method is required. To address this issue, we propose a deep learning-based method to mitigate SF-FMCW-induced interference in FMCW systems. The proposed SF-FMCW interference mitigation network (SF-IMNet) is designed to predict the two-dimensional interference component. Then, by subtracting this predicted interference from the received signal, it is possible to eliminate the interference while minimizing the loss of critical target information. Simulation results show that the proposed method achieves a signal-to-interference-plus-noise ratio (SINR) of 22.94 dB and a correlation coefficient of 0.997 under an input SINR of -10 dB. In addition, the method provides consistent performance in both equal-bandwidth and different-bandwidth interference scenarios. Compared with existing methods, the proposed SF-IMNet improves the SINR by up to 24.1 dB.