FeatureSpectralSkewness.m

%computes the spectral skewness from the magnitude spectrum
%> called by ::ComputeFeature
%>
%> @param X: spectrogram (dimension FFTLength X Observations)
%> @param f_s: sample rate of audio data (unused)
%>
%> @retval v spectral skewness
% ======================================================================
function [vssk] = FeatureSpectralSkewness (X, f_s)

    % compute 'mean' and 'standard deviation'
    mu_X = FeatureSpectralCentroid(X, f_s) * 2 / f_s * (size(X, 1)-1);
    std_X = FeatureSpectralSpread(X, f_s) * 2 / f_s * (size(X, 1)-1);
    tmp = repmat(0:size(X,1)-1, size(X,2), 1) - repmat(mu_X, size(X,1), 1)';

    vssk = sum((tmp.^3)'.*X)' ./ (std_X'.^3 .* sum(X, 1)');

    vssk = vssk';
       
    % avoid NaN for silence frames
    vssk (sum(X, 1) == 0) = 0;
    
end