通信与网络 | Communication and Network ntizer, the power consumption of the system will be greatly increased, which will hinder the widespread application of ultra - large - scale MIMO systems. Therefore, this article assumes that each antenna of the base station is equipped with a pair of 1 - bit analog - to - digital converters (ADC ), and uses the mapping relationship between the sub - array and the user to describe the non - stationary channel characteristics. Based on the powerful generalization ability of neural network(DNN), this paper designs a new generative supervised DNN model that can be trained with a reasonable number of pilots. The simulation results show that the proposed network can achieve better estimation performance with less pilots and achieve a good balance between performance and complexity. Key words :channel estimation ; deep learning ; spatial non-stationary ; 1-bit ADC 0引言 超高的数据速率和系统吞吐量。超大规模MIMO技术也 大规模多输入多输出(Multiple Input Multiple Output, 因此成为第六代(Sixth Generation,6G)移动通信关键技术 MIMO)技术是第五代(Fifth Generation,5G)移动通信的关 的候选[1] $然而,大口径阵列的使用会造成不同的信道 键技术之一。随着天线阵列尺寸数量级的增加,形成了 条件$当整个阵列的孔径有限并且服务相同的用户时