% EX_ONLINE.M%% P. Flandrin, Mar. 13, 2003%% computes on-line EMD for the periodized version of the% sum of 2 sinusoidal FM's + 1 Gaussian logon used in EMD_FMSIN.M%% displays reassigned spectrograms of the sum signal and of the 3 first% modes extracted%% illustrates Section 3.4 in%% G. Rilling, P. Flandrin and P. Gon鏰lv鑣% "On Empirical Mode Decomposition and its algorithms"% IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing% NSIP-03, Grado (I), June 2003% elementary signalN = 2000;% # of data samplesT = 1:4:N;t = 1:N;p = N/2;% period of the 2 sinusoidal FM's% sinusoidal FM 1fmin1 = 1/64;% min frequencyfmax1 = 1.5*1/8;% max frequencyx1 = fmsin(N,fmin1,fmax1,p,N/2,fmax1);% sinusoidal FM 2fmin2 = 1/32;% min frequencyfmax2 = 1.5*1/4;% max frequencyx2 = fmsin(N,fmin2,fmax2,p,N/2,fmax2);% logonf0 = 1.5*1/16;% center frequencyx3 = amgauss(N,N/2,N/8).*fmconst(N,f0);a1 = 1;a2 = 1;a3 = 1;x = real(a1*x1+a2*x2+a3*x3);x = x/max(abs(x));% periodized signaly = [x' x' x' x' x' x' x' x'];[imf_el,ort_el,nbits_el] = emd_online(y,1:length(y),[0.05,0.5,20,100],4,1,1);T = 1:32:length(y);emd_visu(y,1:length(y),imf_el,1);figure(1)% time-frequency distributionsNf = 256;% # of frequency binsNh = 127;% short-time window lengthw = window(Nh,'Kaiser');[s,rs] = tfrrsp(y',T,Nf,w,1);[s,rs1] = tfrrsp(imf_el(1,:)',T,Nf,w,1);[s,rs2] = tfrrsp(imf_el(2,:)',T,Nf,w,1);[s,rs3] = tfrrsp(imf_el(3,:)',T,Nf,w,1);figure(4)subplot(221)imagesc(flipud(rs(1:128,:)))set(gca,'YTick',[]);set(gca,'XTick',[])xlabel('time')ylabel('frequency')title('signal')subplot(222)imagesc(flipud(rs1(1:128,:)))set(gca,'YTick',[]);set(gca,'XTick',[])xlabel('time')ylabel('frequency')title('mode #1')subplot(223)imagesc(flipud(rs2(1:128,:)))set(gca,'YTick',[]);set(gca,'XTick',[])xlabel('time')ylabel('frequency')title('mode #2')subplot(224)imagesc(flipud(rs3(1:128,:)))set(gca,'YTick',[]);set(gca,'XTick',[])xlabel('time')ylabel('frequency')title('mode #3')%colormap(flipud(gray))colormap(jet)