mat_emd.m

fastest algorithm to find EMD.

function [IMF,cumIMF] = mat_emd(x,numIMF,MAXITERATIONS);
% [IMF,CUM_IMF] = MAT_EMD(X,nIMF,nITS) loops over all
%   columns in X, performing the empirical mode decomposition
%   on each one.  The outputs are two matrices each having
%   the same number of rows as X but with X*nIMF columns.  The 
%   EMD is performed with nITS as the sifting limit. IMF
%   contains the intrinsic mode functions while CUM_IMF
%   consists of the cumulative sums for each IMF such that the
%   X = cumIMF(nIMF:nIMF:end) 
% 
% author: Bradley M. Battista
%   University of South Carolina
%   Department of Geological Sciences
%   701 Sumter Street, EWS 617
%   Columbia, SC. 29208
%
% COPYRIGHT: see the associated COPYRIGHT.txt file, and also
% http://software.seg.org/disclaimer2.txt
% This source code may be found online at:
% http://software.seg.org/2007/0003
%

% Pre-allocate memory for IMFs
IMF = zeros(size(x,1),size(x,2)*numIMF);
cumIMF = IMF;

% Loop the EMD over all columns
h = waitbar(0,'Working on data');
hh = get(get(h,'children'),'title');

sz = size(x,2);
for n = 1:sz;
    set(hh,'String',...
        ['Working: ',num2str(n),' of ',num2str(sz),' traces.']);
    waitbar(n/sz,h);
    % Perform EMD on the n^th trace
    imf = semd(x(:,n),numIMF-1,MAXITERATIONS);
    
    % Pad with zeros to get 'nm_IMF' IMFs...
    if size(imf,2) < numIMF
        imf(:,size(imf,2)+1:numIMF) = 0;
    end
    
    % Cumulative sums of IMFs one thru sz
    cumimf = cumsum(imf,2);
    
    % Determine where the IMFs should be written
    idx = (n-1)*numIMF+1;
    
    % Write the IMFs into the big bank of IMFs
    IMF(:,idx:idx+numIMF-1) = imf;
    cumIMF(:,idx:idx+numIMF-1) = cumimf;
end
close(h);