基于LMS和神经网络的水下通信信道模型建模matlab仿真,包含仿真操作录像,代码注释

clc; clear; close all; warning off; addpath(genpath(pwd)); rng('default') load('data1'); load('data2'); BerEst2=zeros(10,1); %估计当前信道质量 for channelNumber=1:10 dataIn =x2_all(channelNumber,:)'>0.5; %%第n个通道的输入通道数据 dataOut= y2_all(channelNumber,:)'>0.5; %将通道与输入卷积 BerEst2(channelNumber)=biterr(dataIn,dataOut)/length(dataIn); end figure grid on box on grid minor hold on hold on semilogy(1:10,BerEst2,'-.','LineWidth',2); xlabel('channel通道号'); ylabel('误码率'); %LMS load('data1'); load('data2'); LmsBerEst1=zeros(10,1); LevBerEst1=zeros(10,1); LmsBerEst2=zeros(10,1); LevBerEst2=zeros(10,1); filter_order=100; b1=zeros(filter_order,1); b2=zeros(filter_order,1); %LMS自适应 for channelNumber = 1:10 channelNumber x1 =x_all(channelNumber,:)'>0.5; %第n个通道的输入通道数据 d1 = y_all(channelNumber,:)'>0.5; %将通道与输入卷积 x2 =x2_all(channelNumber,:)'>0.5; %第n个通道的输入通道数据 d2 = y2_all(channelNumber,:)'>0.5; %将通道与输入卷积 lms=dsp.LMSFilter(filter_order,'StepSize',0.01,'WeightsOutputPort',true); [~,e,w1]=lms(double(x1),double(d1)); %LMS滤波器权重 [~,e,w2]=lms(double(x2),double(d2)); %LMS滤波器权重 lms_filter1=w1./w1(1); %估计均衡器系数 lms_filter2=w1./w1(1); %估计均衡器系数 dataOut1=(filter(1,lms_filter1,d1))>0.5; %信息比特通过均衡器 dataOut2=(filter(1,lms_filter2,d2))>0.5; %信息比特通过均衡器 LmsBerEst1(channelNumber)=biterr(x1,dataOut1)/length(x1); %信道BER LmsBerEst2(channelNumber)=biterr(x2,dataOut2)/length(x2); %信道BER %Levinson Durbin [a1 ~]=levinson(autocorr(d1,filter_order),filter_order-1); [a2 ~]=levinson(autocorr(d1,filter_order),filter_order-1); b1=a1; b2=a2; dataOut1=(filter(b1,1,d1))>0.5; %信息比特通过均衡器 dataOut2=(filter(b2,1,d2))>0.5; %信息比特通过均衡器 LevBerEst1(channelNumber)=biterr(x1,dataOut1)/length(x1); LevBerEst2(channelNumber)=biterr(x2,dataOut2)/length(x2); end figure grid on box on grid minor hold on semilogy(1:10,LmsBerEst1,'--','LineWidth',2); hold on semilogy(1:10,LevBerEst1,'-.','LineWidth',2); hold on semilogy(1:10,LmsBerEst2,'--','LineWidth',2); hold on semilogy(1:10,LevBerEst2,'-.','LineWidth',2); hold on xlabel('channel通道号'); ylabel('误码率'); legend( 'LMS 1','Levinson 1','LMS 2','Levinson 2') %Neural Network load data2.mat number_of_layers=10; neurons_per_layer=10; training_fn='trainbfg'; NNBerEst2=zeros(10,1); for channel_num=1:10 number_of_layers=10; neurons_per_layer=10; training_fn='trainbfg'; train_data_input=y2_all(channel_num,1:9000)>0.5; train_data_output=x2_all(channel_num,1:9000)>0.5; hiddensize=neurons_per_layer*ones(1,number_of_layers); net=feedforwardnet(hiddensize,training_fn); % view(net); net=configure(net,train_data_input,train_data_output); net=train(net,train_data_input,train_data_output); %%Test network test_data_input=y2_all(channel_num,9001:10000); test_data_output=sim(net,test_data_input); x_out=test_data_output>0.5; x_ver=x2_all(channel_num,9001:10000)>0.5; NNBerEst2(channel_num)=sum(x_out~=x_ver)/1000; end figure grid on box on grid minor hold on semilogy(1:10,NNBerEst2,'-.','LineWidth',2); pbaspect([1 1 1]) xlabel('channel通道号'); ylabel('误码率'); xlim([1 10])