BLMS.m

format compact;
clc;
close all;
clear all;

channel_taps = 16; % number of channel taps present in the FIR filter
desired_noise_SNR = 0; % gaussian noise present in the desired output data
filter_weights = rand(channel_taps,1); % initializing normalized random values for the channel taps of FIR filter
weight_update = zeros(channel_taps,1); % inital guess of the filter weights choosen to be all zero vector
step_size = 0.01; % step size or step length


wait_bar = waitbar(0,'Starting processing');
mu_BLMS = step_size; % step length of the BLMS update method
experiment= 100; % ensemble-average independent runs
iteration = 5000; % total number of iterations done
block_length = 20; %block length in which all the inputs are divided

% selected parameters
mean_square_deviation_main = zeros(iteration,1); % Mean Square Deviation
mean_square_error_main = zeros(iteration,1); % Mean Square Error
excess_mean_square_error_main = zeros(iteration,1); % Excess Mean Square Error


for dummy_var_2 = 1:experiment
    wait_bar_percentage = dummy_var_2/experiment *100;
    wait_bar = waitbar(dummy_var_2/experiment, wait_bar, strcat('Percentage complete.....',string(floor(wait_bar_percentage)),'%'));
    u_i = zeros(1,channel_taps); % input vector
    mean_square_deviation = zeros(iteration,1); % Mean Square Deviation
    mean_square_error = zeros(iteration,1); % Mean Square Error
    excess_mean_square_error = zeros(iteration,1); % Excess Mean Square Error
    w_BLMS = weight_update; % setting the weight update vector equal to the initial guess which is all zero vector
    for dummy_var = 1:iteration/block_length
        for dummy_var_3 = 1:block_length
            new_tx_symbol = abs(normrnd(0,1)); % Gaussian random numbers with mean 0 and variance 1
            tx_symbol(dummy_var) = new_tx_symbol;
            u_i = [new_tx_symbol u_i(1:end-1)]; % generate regressor/input signal (u_i - a row vector of size 1xM)
            u_i_block(dummy_var_2,:) = u_i; % generating input block matrix
            [d_i,desired_noise_variance] = awgn(u_i*filter_weights, desired_noise_SNR); % generate noisy version of channel output as received symbol
            d_i_block(dummy_var_2) = d_i; % generating output block vector
            e_i_BLMS_block(dummy_var_2) = d_i - u_i*w_BLMS; % generating error block vector
        end
        % Block LMS update
        e_i_BLMS = d_i - u_i*w_BLMS; % finding error between desired output and filter output to update adaptive filter at end of block
        w_BLMS = w_BLMS + (mu_BLMS/block_length)*(sum(e_i_BLMS .* u_i_block))'; % updating the adaptive filter after finding the error using BLMS algorithm

        %calculation of the parameter
        mean_square_deviation(dummy_var) = norm(w_BLMS-filter_weights)^2; % mean square deviation calculation
        excess_mean_square_error(dummy_var) = norm(e_i_BLMS)^2; % Excess Mean Square Error calculation
        mean_square_error(dummy_var) = excess_mean_square_error(dummy_var)+ desired_noise_variance; % Mean Square Error Calculation
    end
    mean_square_deviation = mean_square_deviation/max(mean_square_deviation);
    excess_mean_square_error = excess_mean_square_error/max(excess_mean_square_error);
    mean_square_error = mean_square_error/max(mean_square_error);
% 
    mean_square_deviation_main = mean_square_deviation_main + mean_square_deviation;
    mean_square_error_main = mean_square_error_main + mean_square_error;
    excess_mean_square_error_main = excess_mean_square_error_main + excess_mean_square_error;
    

end
mean_square_deviation = mean_square_deviation_main/experiment;
mean_square_error = mean_square_error_main/experiment;
excess_mean_square_error = excess_mean_square_error_main/experiment;
 
% Plot for Mean Square Devivation Curve
figure;
plot(10*log10(mean_square_deviation),'linewidth',1);
xlabel('block')
ylabel('Mean Square Deviation (dB)');
legend('BLMS')


% Plot for Excess Mean Square Error Curve
figure;
plot(10*log10(excess_mean_square_error),'linewidth',1);
xlabel('block')
ylabel('Excess Mean Square Error (dB)');
legend('BLMS')


% Plot for Mean Square Error Curve
figure;
plot(10*log10(mean_square_error),'linewidth',1);
xlabel('block')
ylabel('Mean Square Error (dB)');
legend('BLMS')



close(wait_bar);