Ghost Target Suppression in PMCW-Based Sensing Systems Under IQ Imbalance

초록

This study proposes a deep learning-based compensation method to effectively correct in-phase/quadrature (I/Q) imbalance and direct current (DC) offset in phase-modulated continuous-wave (PMCW) radar systems. In practical PMCW radar systems, I/Q imbalance and DC offset generate ghost targets and increase sidelobe levels in the range-velocity (RV) map, respectively, thereby degrading the accuracy of target detection. Conventional signal-processing-based methods for compensating hardware impairments are effective in suppressing ghost targets in single-target scenarios. However, their robustness can degrade under Doppler shifts induced by target velocity, and they are not sufficiently addressed for multitarget scenarios. To overcome these limitations, we propose a U-Net-based compensation method that operates with conditional inputs. The proposed method constructs reference signals from target information estimated from the RV map generated using the received signal. The network uses the distorted received signal and the reference signals to suppress ghost targets caused by hardware impairments. Simulation results demonstrate that the proposed method effectively suppresses ghost targets and sidelobes in both single-target and multitarget scenarios under various signal-to-noise ratio (SNR) levels. Across the SNR range from −5 to 20 dB, the proposed method consistently improves both SNR and improvement factor (IF) compared with the signal without compensation. At an input SNR of 0 dB, the SNR improvement and IF are 5.76 and 25.03 dB, respectively.

키워드