文献类型：期刊文献

中文题名：基于实值神经网络的FBMC/OQAM系统PAPR降低方法

英文题名：PAPR Reduction Method of FBMC/OQAM System Based on Real Valued Neural Network

作者：何超逸[1];袁伟娜[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2022

卷号：48

期号：6

起止页码：826

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：Scopus;北大核心:【北大核心2020】；CSCD:【CSCD_E2021_2022】；

基金：国家自然科学基金(61501187)。

语种：中文

中文关键词：偏移正交幅度调制滤波器组多载波;峰均功率比;实值神经网络;色散选择性映射

外文关键词：FBMC/OQAM;PAPR;real valued neural network;DSLM

摘要：偏移正交幅度调制滤波器组多载波(FBMC/OQAM)系统是5G多载波通信系统候选方案之一,与正交频分复用(OFDM)等多载波方案一样,存在峰均功率比(PAPR)较高的问题,影响了高功率放大器(HPA)的效率。针对FBMC/OQAM系统PAPR过高的问题,提出了一种基于实值神经网络的方法,该方法在发送端和接收端搭建两个实值神经网络,分别用来降低PAPR和误码率(BER)。仿真结果表明,相较于色散选择性映射(DSLM)、限幅(Clipping)、编码以及PRnet方法,本文方法对PAPR和BER性能都有一定的提升。
Filter bank multicarrier with offset quadrature amplitude modulation(FBMC/OQAM) is one of the candidate schemes for 5G multicarrier communication system. Like orthogonal frequency division multiplexing(OFDM) and other multicarrier schemes, it has the problem of high peak-to-average power ratio(PAPR), which will affect the efficiency of high power amplifier(HPA). Aiming at the problem of too high PAPR in FBMC/OQAM system, this paper proposes a method based on real valued neural network. By establishing two real valued neural networks are established at the transmitter and the receiver, this method can reduce PAPR and bit error ratio(BER). It is shown via simulation results that compared with the dispersive selected mapping(DSLM), clipping, coding, and PRnet, the proposed method can attain better performance in PAPR and BER.