Deep Learning Based Detection for Communications Systems With Radar Interference

Due to the increasing demand for spectrum resources, co-existence of communications and radar systems has been proposed that allows operate in same frequency band. On other hand, deep learning shown great potential revolutionizing systems. In this work, we investigate use subject interference from Specifically, consider a single-carrier system. Linear frequency-modulated (LFM) continuous-wave (FMCW) are considered radar. Several important system parameters, including level noise interference, coverage, symbol duration, feature extraction methods number hidden layers investigated performance detector. Fully connected neural network (FCDNN) long short-term memory (LSTM) detectors implemented, where principle component analysis (PCA) is applied preprocess observed signals FCDNN Numerical results show learning-based detector achieves comparable radar-communication traditional but without cancellation. Preprocessing received with PCA can improve when strong. Also, LSTM shows more robust than channel time-related distortion.