Co-Channel Interference and Colored Noise Mitigation: An Iterative Structure with Convolutional Neural Network

Abstract

Reliable information transmission in complex electromagnetic interference environments is an essential proposition for wireless communication systems. This paper proposes an iterative structure for single antenna receiver, which combines both blind signal extraction (BSE) and convolutional neural network (CNN) to mitigate potential co-channel interference (CCI) as well as colored noise. Firstly, the single channel received signal is transformed into a multi-channel observations so that the proposed structure can use BSE to extract the target signal. Then, the belief propagation (BP) algorithm is employed to decode low-density parity-check (LDPC) codes. After that, the residual interferences and colored noise are iteratively learned by a one-dimensional CNN model and fed back to be gradually canceled out from the original input. Finally, a valid estimation of the desired sequence is obtained from the output of the BP decoder. The simulation results show that, in contrast, the proposed structure can effectively reduce the bit error rate (BER) of the system, which indicates that it has a strong ability to mitigate the co-channel interference and colored noise simultaneously.