% example with the default H matrix. the 1st bit was flipped in the BS
% channel : ldpcdecodersoft([1 0 1 0 1 1])
%for the awgn channel : ldpcdecodersoft([-0.1 0.5 -0.8 1.0 -0.7 0.5]), with
%SBR = 1.25
%main funciton
% the input frame length must match the number of columns of H
% input frame = y
function output = ldpcdecodersoft(input_frame)
%输入帧 是人给定的
success = 0; % decoding stopping condition
max_iterations = 5;
current_iteration = 1;
% H matrix
% It is possible to put another legit matrix here, the code will adapt
H = [1 1 0 1 0 0;
0 1 1 0 1 0;
1 0 0 0 1 1;
0 0 1 1 0 1];
p = 0.2; % p is the crossover probability 误码概率 ，误码0->1.1->0probability that 1 sent and 0 received for example). Used only if the channel is BSC
current_frame = input_frame;%we initialize the frame on which we will work
%当前帧等于输入帧
nb_c_nodes = size(H,1)% the number行 of check nodes (= number of rows行)
%四个列节点
%一个行节点
%检验行列
nb_v_nodes = size(H,2)% the number of v_nodes (= number of columns列)
%6个列节点
%--------- error checking -------------
% if the syndrome is already null, no errors !
areErrorsPresent = check_errors(H, input_frame);
if areErrorsPresent == 0
success = 1;
end
%----------------------------------------------------------------------
%第一次循环
%This is the first iteration of the soft decoding. Since it is not exactly
%the same than for the next loops, it is put outside the loop;
disp("currrent frame = ");
disp(current_frame);
if success == 0
%如果失败
% r is the a priori log like hood. Here you can change the function
% which represents the channel
r_vector = a_priori_log_likehood(input_frame,p);%输入帧，误码概率
% r被p给干扰 已经是含有误码了！
disp("r=");
disp(r_vector);
disp ("------------- ITERATION " + current_iteration + " -----------------------")
% Mji is the matrix "sent" by v_nodes to c_nodes. For the first iteration, the matrix is different.
Mji_matrix_first_iteration = generate_Mji_matrix_first_iteration(H, r_vector, nb_c_nodes, nb_v_nodes);
%开始迭代
disp("M =");
disp(Mji_matrix_first_iteration);
%第一次迭代的结果
% Eji is the matrix of extrinsic probabilities, computed by v_nodes
Eji_matrix = generate_Eji_matrix(Mji_matrix_first_iteration,nb_c_nodes, nb_v_nodes);
disp("E =");%E是概率矩阵
disp(Eji_matrix);
%L is the new log likehood ratio. Computed by v_nodes
Lj_vector = generate_Lj_vector(r_vector, Eji_matrix, nb_v_nodes);
disp("L =");%L新的似然比
disp(Lj_vector);
%finally, updating the frame with the new LLR
current_frame = update_frame(Lj_vector, nb_v_nodes);
disp("current frame =");%更新帧 显示被更新的帧
disp(current_frame);
%We check again if all the errors are corrected
areErrorsPresent = check_errors(H, current_frame);%校验当前帧
if areErrorsPresent == 0
success = 1;
else
disp("still errors");
end
end
%------------end of 1st exchange between c and v
%nodes--------------%迭代第一次结束
%-------- DECODE LOOP --------------- 解码循环
% exactly the same as above, but in a loop and Mji_matrix instead of
% Mji_matrix_first_iteration
while success==0 && current_iteration<max_iterations
%解码迭代次数小于最大规定次数 不断重复第一次的迭代
current_iteration = current_iteration + 1;
%迭代继续 次数+1
disp ("------------- ITERATION " + current_iteration + " -----------------------")
Mji_matrix = generate_Mji_matrix(H, Eji_matrix, r_vector, nb_c_nodes, nb_v_nodes);
disp("M =");
disp(Mji_matrix);
Eji_matrix = generate_Eji_matrix(Mji_matrix,nb_c_nodes, nb_v_nodes);
disp("E =");
disp(Eji_matrix);
Lj_vector = generate_Lj_vector(r_vector, Eji_matrix, nb_v_nodes);
disp("L =");
disp(Lj_vector);
current_frame = update_frame(Lj_vector, nb_v_nodes);
disp("current frame =");
disp(current_frame);
areErrorsPresent = check_errors(H, current_frame);
if areErrorsPresent == 0
success = 1;
else
disp("still errors");
end
end
if success ~= 1
disp("correction failed");
else
disp ("correction success");
end
output = current_frame;
disp ("output = ");
disp(output);
end
%------------- OTHER FUNCTIONS -----------
%return 1 if there's errors in the frame
% return 0 if not
function res = check_errors(H, current_frame)
%检测所有的syndrome的元素 不是0 的就是错误
H
syndrome = H * transpose(current_frame) % syndrome = H * v^t
% syndrome 结果向量形式展示
areErrors = any(mod(syndrome,2)) % if there's something else than 0 in the syndrome, ERROR detected
%不是0 就是错误
res = areErrors;
end
%returns a vector r containig a priori log likehood for each received bit
% This function represents the channel and can be modified. This one is to
% get LLR of of a BSC (binary symetric channel), where only p is needed
function r = a_priori_log_likehood(input_frame,p)
r = zeros(1,size(input_frame,2));
for i = 1: size(input_frame,2)
if input_frame(i) == 1
r(i) = log(p/(1-p));
else
r(i) = log((1-p)/p);
end
end
end
% example of function to get the LLR of AWGN channel. Signal to Noise Ratio
% needed
function r = a_priori_log_likehood_AWGN(input_frame,SBR)
r = zeros(1,size(input_frame,2));
for i = 1: size(input_frame,2)
r(i) = 4 * input_frame(i) * SBR;
end
end
% creates the Mji matrix which will be "sent" to check nodes. In the first
% iteration, Mji=ri
function Mji_matrix_first_iteration = generate_Mji_matrix_first_iteration(H, r, nb_c_nodes, nb_v_nodes)
Mji_matrix_first_iteration = zeros(nb_c_nodes, nb_v_nodes); %先创一个初始化的0矩阵
for i = 1:nb_c_nodes
for j=1:nb_v_nodes
if H(i,j) == 1
Mji_matrix_first_iteration(i,j) = r(j);
end
end
end
end
% creates the Mji matrix which will be "sent" to check nodes
function Mji_matrix = generate_Mji_matrix(H, Eji_matrix, r_vector, nb_c_nodes, nb_v_nodes)
Mji_matrix = zeros(nb_c_nodes, nb_v_nodes);
for i = 1:nb_c_nodes%行与列都要循环 确保每一个元素都能遍历
for j=1:nb_v_nodes
if H(i,j) == 1
Mji = generate_Mji(Eji_matrix, r_vector, nb_c_nodes, i, j)
%
Mji_matrix(i,j) = Mji
end
end
end
end
function Mji = generate_Mji(Eji_matrix, r_vector, nb_c_nodes, i, j)
tmp = 0;
forbidden_index = i;
for i= 1:1:nb_c_nodes
if (Eji_matrix(i,j) ~= 0) && (i ~= forbidden_index)
tmp = tmp + Eji_matrix(i,j);
end
end
Mji = tmp + r_vector(j);
end
% creates the Eji matrix, aka matrix of extrinsic probabilities. for each
% probability of Mji, the extrinic probability is the """sum""" of the
% other probabilities except the current one
function Eji_matrix = generate_Eji_matrix(Mji_matrix, nb_c_nodes, nb_v_nodes)
Eji_matrix = zeros(nb_c_nodes, nb_v_nodes);
for i = 1:1:nb_c_nodes %iteative parsing of Mji
for j = 1:1:nb_v_nodes
if Mji_matrix(i,j) ~= 0 % we compute the extrinsic probability only if there's a probability in these coordinates
Eji = generate_Eji(i,j,Mji_matrix,nb_v_nodes)
Eji_matrix(i,j) = Eji
end
end
end
end
% for a given probability, generates its extrinsic probability. This
% functions is performed by check nodes
function Eji = generate_Eji(i,j,Mji_matrix,nb_v_nodes)
tmp = 1;
forbidden_index = j;
for j = 1:1:nb_v_nodes % iteration on each proability get by the i-th c_node
if (j ~= forbidden_index) && (Mji_matrix(i,j) ~= 0)
not_current_Mji = Mji_matrix(i,j);
tmp = tmp * tanh(not_current_Mji / 2);
end
end
Eji = log((1+tmp)/(1-tmp)); % the said """sum"""
end
% creates the new log likehood ratio vector after 1 iteration
function Lj_vector = generate_Lj_vector(r_vector, Eji_matrix, nb_v_nodes)
Lj_vector = zeros(1,nb_v_nodes);
for j = 1:1:nb_v_nodes % iteration of Eji column by column (ie for each v_node)
Lj = generate_Lj(j, r_vector, Eji_matrix(:,j));
Lj_vector(j) = Lj;
end
end
% creates the j-th new log likehood ratio after 1 iteration. It is the sum
% of all extrinsic probabilities received by the v_node + initial log
% likehood ratio
function Lj = generate_Lj(j, r_vector, extrinsic_probabilities)
Lj = sum(extrinsic_probabilities) + r_vector(j);
end
% updating the current frame with the inforation of the new LLR
% the sign gives the bit ( minus means it should be 1) and the |value|
% is the reliability
function current_frame = update_frame(Lj_vector, nb_v_nodes)
current_frame = zeros(1, nb_v_nodes);
for j=1:1:nb_v_nodes
if Lj_vector(j) < 0
current_frame(j) = 1;
end
end
end
命令行输入 ldpcdecodersoft([1 0 1 0 1 1]).