A Cross Polarization Interference Cancellation Scheme Based on CNN-LSTM for 6G Satellite Communication

Satellite communication has become a hot topic in the development of the sixth-generation (6G) mobile communication due to its advantage of wide coverage and broadband transmission. Dual polarization frequency reuse (DPFR) is often used to improve the spectrum utilization in satellite communication system. However, cross-polarization interference (XPI) often occurs owing to the imperfect devices and atmosphere conditions. This paper proposes a joint convolutional neural network (CNN) and long short-term memory (LSTM) network to cancel the XPI for satellite communication. Since the cross-polarization interference cancellation (XPIC) is normally implemented at baseband, it is inevitable that nonlinear effects will be introduced by the high-power amplifier (HPA) on satellite. Conventional XPIC schemes based on transversal filter can not deal with the nonlinear impairments of the received signal, resulting in poor bit error ratio (BER) performance. We combine CNN and LSTM to extract the nonlinear feature and temporal correlation of the received signal, which can effectively eliminate the XPI and nonlinear effects. Simulation results show that, compared with conventional XPIC schemes, the BER performance of proposed scheme is improved by two orders of magnitude when the signalto-noise ratio (SNR) is 20 dB. Besides, the proposed scheme can achieve fast convergence with small mean square error (MSE).