【变分模式分解】基于matlab动态带宽变分模式分解【含Matlab源码 3722期】.zip

%% Add "Method_Scripts" path % Method_Scripts is the folder where DB-VMD and VMD are implemeneted init_pwd = pwd; cd .. addpath('Method_Scripts') cd(init_pwd) %% Preparations and parameters definition clear; clc; close all snr_arr = 20:-2:10; tau_l_arr = [0.1]; max_it = 500; % Number of iterations % Parameters alpha = 1000; % VMD bandwidth factor K = 3; % Components' count tau_ab = 0.1; % Bandwidth rate of change (DB-VMD) DC = 0; % DC impose (0 for none) init = 3; % Central frequencies initialization tol = 1e-7; % Stopping criteria tolerance N = 500; % Signal length n = (1:N)'; % Generating Hanning windows L_arr = [500, 125, 100]; d_arr = [250, 125, 375]; hann_windows = nan(N,K); for i=1:K temp = zeros(N, 1); low = d_arr(i) - L_arr(i)/2 + 1; high = d_arr(i) + L_arr(i)/2; temp(low: high) = hann(L_arr(i)); hann_windows(:, i) = temp; end % Success rate array for (iteration number, SNR, tau_l) combination SR_DB_vmd_arr = nan(max_it, length(snr_arr), length(tau_l_arr)); SR_vmd_arr = nan(max_it, length(snr_arr), length(tau_l_arr)); %% Noise robustness experiment for i_tau_l = 1:length(tau_l_arr) tau_l = tau_l_arr(i_tau_l); for i_snr = 1:length(snr_arr) for it = 1:max_it % Printing progress if mod(it, 50) == 0 fprintf("iteration: %d/%d - SNR: %.2f (%d/%d) - tau_l: %.2f (%d/%d)\n", ... it, max_it, ... snr_arr(i_snr), i_snr, length(snr_arr), ... tau_l_arr(i_tau_l), i_tau_l, length(tau_l_arr)) end % Signal generation omega_arr = unifrnd(0, pi, 3, 1); A_arr = unifrnd(0.5, 1.5, 3, 1); fsub = cell(K, 1); s = zeros(N, 1); for i=1:K fsub{i} = hann_windows(:, i) .* A_arr(i) .* cos(omega_arr(i).*n); s = s + hann_windows(:, i) .* A_arr(i) .* cos(omega_arr(i).*n); end [~, sortIndex] = sort(omega_arr); fsub = fsub(sortIndex); s = awgn(s, snr_arr(i_snr), "measured"); % Add noise % DB-VMD applied [u, ~, omega] = DB_VMD(s, tau_ab, tau_l, K, DC, init, tol); [~, sortIndex] = sort(omega(end, :)); u = u(sortIndex, :); % DB-VMD Success rate corr_arr = nan(K, 1); for k=1:K corr_arr(k) = abs(xcorr(fsub{k}, u(k,:), 0, 'normalized')); end SR_DB_vmd_arr(it, i_snr, i_tau_l) = mean(corr_arr); % VMD applied [u, ~, omega] = VMD(s, alpha, tau_l, K, DC, init, tol); [~, sortIndex] = sort(omega(end, :)); u = u(sortIndex, :); % VMD Success rate corr_arr = nan(K, 1); for k=1:K corr_arr(k) = xcorr(fsub{k}, u(k,:), 0, 'normalized'); end SR_vmd_arr(it, i_snr, i_tau_l) = mean(corr_arr); end end end %% Results % Plotting medians and Median Absolute Deviation (MAD) for i=1:length(tau_l_arr) medians_vec_DB_VMD = median(SR_DB_vmd_arr(:, :, i_tau_l)); mad_vec_DB_VMD = mad(SR_DB_vmd_arr(:, :, i_tau_l), 1); medians_vec_VMD = median(SR_vmd_arr(:, :, i_tau_l)); mad_vec_VMD = mad(SR_vmd_arr(:, :, i_tau_l), 1); figure("Name", sprintf("tau_l = %.2f", tau_l_arr(i_tau_l))) h(1) = semilogx(snr_arr, medians_vec_DB_VMD, 'k-', 'LineWidth', 3); for j_snr = 1:length(snr_arr) hold on plot([snr_arr(j_snr) snr_arr(j_snr)], ... [medians_vec_DB_VMD(j_snr) - mad_vec_DB_VMD(j_snr), ... medians_vec_DB_VMD(j_snr) + mad_vec_DB_VMD(j_snr)], ... 'k:_', 'LineWidth', 2) end hold on h(2) = semilogx(snr_arr, medians_vec_VMD, 'b--', 'LineWidth', 3); for j_snr = 1:length(snr_arr) hold on plot([snr_arr(j_snr) snr_arr(j_snr)], ... [medians_vec_VMD(j_snr) - mad_vec_VMD(j_snr), ... medians_vec_VMD(j_snr) + mad_vec_VMD(j_snr)], ... 'b:_', 'LineWidth', 2) end xlabel("SNR (dB)", 'FontSize', 15) ylabel("Success Rate", 'FontSize', 20) xlim([min(snr_arr) - 0.1, max(snr_arr) + 0.1]) legend(h, "DB-VMD", "VMD", "Location", "Best", 'FontSize', 15) title("Success Rate vs SNR", 'FontSize', 30) set(gca, 'XGrid', 'on', 'XMinorGrid', 'on'); end