DC-Former Network Empowered UAV and Bird Recognition Based on Integrated Sensing and Communication System

The integrated sensing and communication (ISAC) technology is considered as one of the key features of 6 G networks, with potential far surpassing that of traditional singlefunction systems. At present, there are serious difficulties in the identification of birds and small unmanned aerial vehicles (UAVs) in urban low-altitude scenes. How to use existing base stations and ISAC technology to identify UAVs and birds becomes a new challenge. In order to classify the radar echo images of UAV and bird more accurately, we collected the echo data of rotor UAV and bird through simulation and actual measurement using 5G signals, and performed short-time Fourier transform (STFT) on the data to obtain the spectrograms of the micro-Doppler signatures. We propose a classification method, DC-Former network, which can accurately capture the channel information of the image and periodically extract effective features according to the object's micro-motion. By putting the spectrograms into the DC-Former network, we can extract the effective features and classification results of the target. The experimental results show that compared with the existing classification methods, the DCFormer network can obtain more accurate classification accuracy, so as to effectively distinguish the radar echo images of UAVs and birds.