哈尔滨工业大学-自适应信号处理-RLS算法实验

close all;clear;clc;
N = 500;        %伯努利序列个数
M = 11;        %滤波器长度
W = 3.2;       n = [1;2;3];
h = 0.5*(1+cos(2*pi*(n-2)/W));       %信道单位冲激响应
mu = 0.05;           %步长参数
for num = 1:200         %独立实验200次
    x = binornd(1,0.5,N,1);
    loc = find(x==0);
    x(loc) = -1;         %1，-1的概率都为0.5的伯努利序列
    deltv1 = 0.1;
    v1 = sqrt(deltv1)*randn(N+2,1);  %均值为0，方差为0.1的高斯白噪声
    u = conv(h(:,1),x)+v1;        %滤波器的输入序列
    %%LMS算法%%
    w = zeros(M,1);         %初始化滤波器系数
     for k = 1:N-M+1
         utemp = u(k+M-1:-1:k);        %存放u(n)的临时向量
         y(k) = w.'*utemp;             %滤波器的输出
         e(k) = x(k+4)-y(k);           %估计误差
         LPe(num,k) = e(k).*e(k);      %估计误差平方
         w = w+mu*utemp*conj(e(k));    %更新滤波器系数
     end
     w = zeros(M,1);          %滤波器初始化
     delt = 0.004;
     P = 1/delt*eye(M);       %初始化逆相关矩阵
     lambda = 1;              %指数加权因子
     for k = 1:N-M+1
         utemp = u(k+M-1:-1:k);        %u(n)的临时向量
         Pai = P*utemp;             
         kn = Pai/(lambda+utemp'*Pai); %增益向量
         xi = x(k+4)-w'*utemp;         %先验估计误差
         w = w+kn*conj(xi);            %更新滤波器系数
         P = 1/lambda*P-1/lambda*kn*utemp'*P;  %更新逆相关矩阵
         RPe(num,k) = xi.*xi;                  %误差平方
     end        
end

figure;           %画图
semilogy(mean(LPe),'b','LineWidth',1);hold on;
semilogy(mean(RPe),'r','LineWidth',1);
xlabel('迭代次数','FontSize',16);ylabel('集平均平方误','FontSize',16);set(gca,'FontSize',16);
ylim([0.1 10]);legend('LMS','RLS');grid on;
title("LMS算法和RLS算法的学习曲线")
text(300,0.7,'LMS','FontSize',16);
text(300,0.2,'RLS','FontSize',16);

close all;clear;clc;
N = 210;           %伯努利序列的个数
M = 11;             %滤波器长度
W0 = [2.8,3.0,3.2,3.4];
W = W0;  W = 1./W;  n = [1;2;3];
h = 0.5*(1+cos(2*pi*(n-2)*W));         %信道的冲激响应函数
delt = 0.004; lambda = 1;           %指数加权因子
for num = 1:200                %独立进行200次实验
    x = binornd(1,0.5,N,1);      %产生1，-1概率都为0.5的伯努利序列
    loc = find(x==0);
    x(loc) = -1;
    deltv1 = 0.001; 
    v1 = sqrt(deltv1)*randn(N+2,1);    %均值为0，方差为0.001的白噪声
    u(:,1) = conv(h(:,1),x)+v1;        %不同参数W下的u(n);
    u(:,2) = conv(h(:,2),x)+v1;
    u(:,3) = conv(h(:,3),x)+v1;
    u(:,4) = conv(h(:,4),x)+v1;
    for wk = 1:4               %4种参数W进行循环计算
        w = zeros(M,1);       %滤波器的初始值
        P = 1/delt*eye(M);      %初始化逆相关矩阵
        for k = 1:N-M+1 
            utemp = u(k+M-1:-1:k,wk);     %u(n)的临时向量
            Pai = P*utemp;              
            kn = Pai/(lambda+utemp'*Pai); %增益向量
            xi = x(k+4)-w'*utemp;         %先验估计误差
            w = w+kn*conj(xi);            %更新滤波器系数
            P = 1/lambda*P-1/lambda*kn*utemp'*P; %更行逆相关矩阵
            Pe(num,k,wk) = xi.*xi;               %误差平方
        end
    end
end

figure;          %画图
semilogy(mean(Pe(:,:,1)),'k','LineWidth',1);hold on;
semilogy(mean(Pe(:,:,2)),'--b','LineWidth',1);hold on;
semilogy(mean(Pe(:,:,3)),':m','LineWidth',1);hold on;
semilogy(mean(Pe(:,:,4)),'-.r','LineWidth',1);hold on;
xlabel('迭代次数','FontSize',16);ylabel('集平均平方误差','FontSize',16);
set(gca,'FontSize',16);
title('RLS算法的学习曲线');
legend(num2str(W0(1)),num2str(W0(2)),num2str(W0(3)),num2str(W0(4)));grid on;
text(200,0.0041,' W=3.4','FontSize',10);
text(200,0.0025,' W=3.2','FontSize',10);
text(200,0.0018,' W=3.0','FontSize',10);
text(200,0.0013,' W=2.8','FontSize',10);