基于深度神经网络降低OFDM系统立方度量方法  

A Deep Neural Network Based Algorithm for Cubic Metric Reduction in OFDM Systems

作　　者：张毅 朱红亮 周娟 ZHANG Yi;ZHU Hong-liang;ZHOU Juan(The 10th Research Institute of China Electronic Technology Croup Corporation,Chengdu Sichuan 610036,China;School of Information and Communication Engineering,University of Electronic Scienceand Technology of China,Chengdu Sichuan 611731,China;College of Communication Engineering,Chengdu University of Information Technology,Chengdu Sichuan 610225,China)

机构地区：[1]中国电子科技集团公司第十研究所,四川成都610036 [2]电子科技大学信息与通信工程学院,四川成都611731 [3]成都信息工程大学通信工程学院,四川成都610103

出　　处：《计算机仿真》2021年第5期199-203,共5页Computer Simulation

基　　金：国家自然科学基金(61601064);四川省教育厅重点项目(18ZA0225)。

摘　　要：针对正交频分复用信号包络波动过大的缺点,提出了利用深度神经网络降低正交频分复用系统立方度量从而减少信号包络波动的方法。仿真表明,提出的方法能够灵活控制信号包络波动大小和误比特率性能;与已有算法相比,深度神经网络算法具有更好的立方度量抑制性能,并且算法执行时间大幅降低。因此,提出的深度神经网络算法更有希望应用于实际通信系统,满足系统对于实时性的要求。Aim at the drawback of large envelope fluctuations of orthogonal frequency division multiplexing(OFDM) signals, this paper proposes a deep neural network(DNN) based algorithm to reduce the cubic metric(CM) and thus the OFDM signals. Simulation results show that the DNN based algorithm exhibits the flexible performance in CM reduction and bit error rate;compared with the existing schemes, the DNN based algorithm reduces not only the CM but also the execution time. Therefore, the proposed algorithm is more promising to be applied in practical systems to satisfy the requirement of systems for real-time communication.

关 键 词：立方度量 峰均功率比 深度神经网络 正交频分复用 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]