基于神经网络与RLS算法的短波信道均衡研究  

Research on High Frequency Channel Equalization Based on Neural Network and RLS Algorithm

作　　者：蒋洪晖 吴小东 朱庆昊 王若琨 JIANG Honghui;WU Xiaodong;ZHU Qinghao;WANG Ruokun(Navy Petty Office Academy,Bengbu 233000;Wuhan Maritime Communication Research Institute,Wuhan 430000)

机构地区：[1]海军士官学校,蚌埠233000 [2]武汉船舶通信研究所,武汉430000

出　　处：《舰船电子工程》2024年第9期65-69,共5页Ship Electronic Engineering

摘　　要：针对短波信道中信号传输产生的码间串扰问题,文中结合人工神经网络(Artificial Neural Network,ANN)与递归最小二乘法(Recursive Least Square,RLS)算法,实现改进的RLS-ANN短波信道均衡算法。通过有监督与无监督分段均衡处理,实时校准经由短波信道传播之后产生的信号幅度和相位失真。对实测的高频数据链(High Frequency Data Link,HFDL)短波信号进行均衡处理,结果表明RLS-ANN算法相比传统的最小均方算法(Least Mean Square,LMS)和RLS算法在星座图收敛速度、平均误差及误码率等方面效果更优,该算法通过降低误码率,可有效改善信号经由短波信道传输的通信质量。Combined artificial neural network(ANN)and recursive least square(RLS)algorithm,an improved RLS-ANN equalization algorithm is proposed to eliminate the inter-symbol interference(ISI)in high frequency signal transmission.In the use of supervised and unsupervised equalization adapted to different data segments,the amplitude and phase distortion of the high frequency signals can be calibrated in real time.High frequency data link(HFDL)signals are tested in data equalization,the results reveal that compared with LMS and RLS algorithms,the proposed RLS-ANN algorithm is better at improving the convergence rate and reducing the average error of the signal constellation,as well as reducing the bit error rate of the tested signals,showing that the RLS-ANN algorithm can effectively improve the signal communication quality in in high frequency channel.

关 键 词：神经网络 递归最小二乘法 短波信道均衡 HFDL信号 

分 类 号：TN918.4[电子电信—通信与信息系统]