基于深度学习的NOMA系统符号检测算法研究

Symbol Detection Algorithm of NOMA System Based on Deep Learning

非正交多址(NOMA)是5G通信领域的一项关键技术.它通过在时间、频率和空间上为使用相同资源的多个用户提供服务,来提高系统容量和频谱效率.提出一种基于深度学习的NOMA接收机,基于信道统计的仿真数据进行离线神经网络训练,在部署阶段直接用于恢复传输的符号.仿真结果表明,与传统的信道估计方法相比,深度学习检测算法能提高NOMA系统符号检测的性能.最后分析了不同参数对所提方案性能的影响.

The non-orthogonal multiple access(NOMA)is a key technology in the field of 5G communication.Its basic idea is to provide services for multiple users using the same resources in time,frequency and space,to improve the system capacity and spectral efficiency.The receiver based on deep learning was proposed,which performs off-line neural network training based on the simulation data of channel statistics.The simulation results show that the deep learning detection algorithm can improve the symbol detection performance of NOMA system compared with the traditional channel estimation method.The effects of different parameters on the performance of the proposed scheme were analyzed.