PinpuNet: Towards ISAC-Based Drone Monitoring by Learning Micro-Doppler Spectrum

Monitoring non-cooperative drones and, by extension, other traffic elements is one major use case of future Integrated Sensing and Communication (ISAC) for 6 G. In our attempt to realise an ISAC-based drone monitoring system, we found that conventional micro-Doppler analysis is highly susceptible to what is coined as the smearing effect that the backscattered signal is blurred due to various reasons, rendering drone target detection intractable. To mitigate the smearing effect, we proposed PinpuNet, a data-driven learning network, to extract drone signatures from highly noisy microDoppler spectrograms. We evaluated PinpuNet on a public drone backscattered signal dataset and the drone classification result shows a substantially improved accuracy compared to generic and commonly used learning networks.