function SPEC = computeSpectrum (fftLength,winShift,s)
% --------------------------------------------
% compute spectrum from time signal
%
% Returns power spectrum (|X(f)|^2) in
% matrix SPEC.X(coefficientIndex,frameIndex) .
% No energy normalization is performed.
% The signal energy (sum of power spectrum coefficients)
% is returned in vector SPEC.e(frameIndex)
%
% parameters:
% fftLength: length of FFT
% winShift: window shift [number of samples]
% s: vector of time samples
%
% last update 18.1.04
% --------------------------------------------

% compute local variables
nofSamples = size(s);
maxFFTIdx = fftLength/2;
% compute time window
win = hamming(fftLength);

% compute matrix X(fftIndex,timeFrameIndex) short term spectra

k = 1;
for m = 1:winShift:nofSamples-fftLength
 
    spec = fft( (win.*s(m:m+fftLength-1)) ,fftLength);
    %use only lower half of fft coefficients
    SPEC.X(:,k) = ( abs( spec(1:maxFFTIdx) ) ).^2;    
    %compute energy
    SPEC.e(k) = sum(SPEC.X(:,k));
 
    k = k+1;
end %m