function X=extractDWT(x_train,startS,endS,wStep,wRange)

%  x_train  = input signal

%  startS   = from second

%  endS     = end second

%  wStep   = overlapping

%  wRange = window size

FS=128;

N=size(x_train,3);

sz=floor((endS-(startS+wRange))/wStep)+1;

X=zeros(sz*140,2);

cn=0;

for i=1:N

    for sig=startS:wStep:endS-wRange

        sW=sig*FS+1;

        eW=(sig+wRange)*FS;

        C3Sig=x_train(sW:eW,1,i);

        C4Sig=x_train(sW:eW,3,i);

        waveletFunction = 'db4';

        waveletLevel=3;

        [wCoe,L] = wavedec(C3Sig,waveletLevel,waveletFunction);

        C3D3 = detcoef(wCoe,L,3);   % Mu

        [wCoe,L] = wavedec(C4Sig,waveletLevel,waveletFunction);

        C4D3 = detcoef(wCoe,L,3);  % Mu

        cn=cn+1;

        % Mean of the absolute values

        X(cn,1)=sum(C3D3.^2)/numel(C3D3);

        X(cn,2)=sum(C4D3.^2)/numel(C4D3);

    end

end

end