function [data_estimation,data_decision,C]=dferls_equation(K1,K2,signal_in,signal_original)
%基于數字通信書的P495頁的算法
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                          訓練模式                                   %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
N=K1+K2+1;
%均衡濾波器的階數
w=0.999;
%加權因子
C=zeros(N,1);
%均衡濾波器各抽頭的系數初始化為零
Num=length(signal_in)-N+1;
%由于反饋的引入,實際對Num個數據進行了均衡
I=zeros(Num,1);
%均衡后的結果，尚未判決
error=zeros(Num,1); 
%誤差矩陣
Yall=zeros(N,Num);
%每一列代表每次均衡所用的數據，其中前K1+1個是接收到的值，后K2個是判決反饋后的值,在訓練模式下,即為已知值
Call=zeros(N,Num);
%每列代表每次迭代所得到的濾波器的系數的值
P=1000*eye(N) ; 
%R矩陣的逆,R矩陣初值取deltaI，取delta為0.001
train_length=length(signal_original);
for t=K2+1:train_length
    Yall(:,t)=[signal_in(K1+t:-1:t);signal_original(t-1:-1:t-K2)];
    %第t次迭代對應Y矩陣的第t列，每列元素的前K1+1個為接收到的信號，后K2個為已知序列，
    Y=Yall(:,t);
    %代表第t次迭代用到的數據
    Y_star=conj(Y);
    %取共軛
    Y_trans=Y.';
    %取轉置
    I(t)=Y_trans*C; 
    %計算均衡輸出
    error(t)=signal_original(t)-I(t);
    %計算誤差
    K=(P*Y_star)/(w+(Y_trans*P*Y_star));
    %計算卡爾曼增益
    P=(1/w)*(P-(K*Y_trans*P));
    %更新矩陣P
    C=C+(error(t)*K);
    %更新均衡器的系數矩陣
    Call(:,t)=C;
end
semilogy(power(abs(error),2))
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                          傳輸模式                                   %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Num=length(signal_in)-train_length-K1-1;%所傳輸的信息序列的長度
data_decision=zeros(Num,1);
data_estimation=zeros(Num,1);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%前K2
data(:,1)=[signal_in(train_length+K1+1:-1:train_length+1);signal_original(train_length:-1:train_length-K2+1)];
data_estimation(1)= data(:,1).'*C;
data_decision(1)=decision(data_estimation(1));
for t=2:K2
    data(:,t)=[signal_in(train_length+K1+t:-1:train_length+t);data_decision(t-1:-1:1);signal_original(train_length:-1:train_length-K2+t)];
    data_estimation(t)= data(:,t).'*C;
    data_decision(t)=decision(data_estimation(t));
end
Y=zeros(N,Num);
for t=(K2+1):Num
     for n=1:K1+1
        data(n,t)=signal_in(train_length+t+K1-n+1);
     end;
     for n=(K1+2):(K1+K2+1)
        data(n,t)=data_decision(t+K1-n+1);
     end 
     data_estimation(t)=C.'*data(:,t);
     data_decision(t)=decision(data_estimation(t));
end