annotate Code/Classifiers/SVM_Formants_Singing.m @ 4:92ca03a8fa99 tip

Update to ICASSP 2013 benchmark

function [] = SVM_Formants_Singing( varargin )

cd 'C:\Users\dawn\Dropbox\TestResults'

DEBUG = 0;
% output results file name
masterFileOutputID = fopen( 'SVM_paper_SingingFormantsStats.txt', 'a' );
% input results file name
inputFileName = 'paper_singingFormantsStats.txt';

% This function allows the user to stipulate which Singing voice LLD's they
% wish to forward to a k-means classifier and produces a file of
% performance characteristics. Input arguments stipulate the LLD's and
% there is a choice of:-
%
% ---- PRAAT FORMANT MEASUREMENTS ----
%     '_Formant_Burg'
%     '_Formant_all'
%     '_Formant_robust'
%
% A text file entitled kmeans_Singing_LLD1name_LLD2name_ ... LLDNname.txt
% is produced that contains the results of the k-mean classification for
% the LLD's specified and named in the result document title.

fprintf( masterFileOutputID, '\n RESULTS FILE NAME: %s\n', inputFileName);
inputFileID = fopen( inputFileName );

% get the column numbers of the results that we want to classify

% COLUMN NUMBER     :   METRIC
%
% ------------- BURG FORMANTS ---------------
%   11   : Number of BURG formants listed = nBF
%
%   THERE ARE CURRENTLY 24 MEASUREMENTS TAKEN FOR EACH FORMANT
nMetrics = 24;
%
%   12   : mean frequency of the first BURG formant
%   13   : variance of the first BURG formant
%   14   : minimum frequency of the first BURG formant
%   15   : maximum frequency of the first BURG formant
%   16   : mean Frequency Derivative of the first BURG formant
%   17   : varience of the Frequency Derivative of the first BURG formant
%   18   : min of the Frequency Derivative of the first BURG formant
%   19   : max  of the Frequency Derivative of the first BURG formant
%   20   : mean of the Frequency 2nd Derivative of the first BURG formant
%   21   : varience of the Frequency 2nd Derivative of the first BURG formant
%   22   : min of the Frequency 2nd Derivative of the first BURG formant
%   23   :  max of the Frequency 2nd Derivative of the first BURG formant
%   24   : mean of the Bandwidth of the first BURG formant
%   25   :  varience of the Bandwidth of the first BURG formant
%   26   : min of the Bandwidth of the first BURG formant
%   27   : max of the Bandwidth of the first BURG formant
%   28   : mean of the Bandwidth Derivative of the first BURG formant
%   29   : varience of the Bandwidth Derivative of the first BURG formant
%   30   : min of the Bandwidth Derivative of the first BURG formant
%   31   : max of the Bandwidth Derivative of the first BURG formant
%   32   : mean of the Bandwidth 2nd Derivative of the first BURG formant
%   33   : var of the Bandwidth 2nd Derivative of the first BURG formant
%   34   : min of the Bandwidth 2nd Derivative of the first BURG formant
%   35   : max of the Bandwidth 2nd Derivative of the first BURG formant
%
%    ....... there are nMetrics for each formant in nBF formants, so cycle
%    through until the last is reached ......
%
%   36 + ((nBF-1)*nMetrics)       : mean frequency of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 1   : variance of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 2   : minimum frequency of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 3   : maximum frequency of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 4   : mean Frequency Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 5   : varience of the Frequency Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 6   : min of the Frequency Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 7   : max of the Frequency Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 8   : mean of the Frequency 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 9   : varience of the Frequency 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 10  : min of the Frequency 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 11  : max of the Frequency 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 12  : mean of the Bandwidth of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 13  : varience of the Bandwidth of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 14  : min of the Bandwidth of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 15  : max of the Bandwidth of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 16  : mean of the Bandwidth Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 17  : variece of the Bandwidth Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 18  : min of the Bandwidth Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 19  : max of the Bandwidth Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 20  : mean of the Bandwidth 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 21  : var of the Bandwidth 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 22  : min of the Bandwidth 2nd Derivative of the nBF BURG formant
%   36 + ((nBF-1)*nMetrics) + 23  : max of the Bandwidth 2nd Derivative of the nBF BURG formant
%
%   FOR THE MEAN OF ALL BURG FORMANTS
%   36 + (nBF*nMetrics)           : mean of all formants Frequency 
%   36 + (nBF*nMetrics) + 1       : varience of the mean of all formants Frequency 
%   36 + (nBF*nMetrics) + 2       : minimum of the mean of all formants Frequency 
%   36 + (nBF*nMetrics) + 3       : maximum of the mean of all formants Frequency
%   36 + (nBF*nMetrics) + 4       :  mean of all formants mean Frequency Derivative 
%   36 + (nBF*nMetrics) + 5       :  mean of all formants varience Frequency Derivative 
%   36 + (nBF*nMetrics) + 6       :  min of the mean of all formants Frequency Derivative
%   36 + (nBF*nMetrics) + 7       :  max of the mean of all formants Frequency Derivative 
%   36 + (nBF*nMetrics) + 8       :  mean of the mean of all formants Frequency 2nd Derivative 
%   36 + (nBF*nMetrics) + 9       :  varience of the mean of all formants Frequency 2nd Derivative 
%   36 + (nBF*nMetrics) + 10      :  min of the mean of all formants Frequency 2nd Derivative
%   36 + (nBF*nMetrics) + 11      :  max of the mean of all formants Frequency 2nd Derivative
%
% ------------- ALL FORMANTS ---------------
%
%   36 + (nBF*nMetrics) + 12   : Number of ALL formants listed = nAF
%
%   startOfALLMeasurements = 36 + (nBF*nMetrics) + 13;
%
%   startOfALLMeasurements      : mean frequency of the first ALL formant
%   startOfALLMeasurements + 1     : variance of the first ALL formant
%   startOfALLMeasurements + 2     : minimum frequency of the first ALL formant
%   startOfALLMeasurements + 3     : maximum frequency of the first ALL formant
%   startOfALLMeasurements + 4     : mean Frequency Derivative of the first ALL formant
%   startOfALLMeasurements + 5     : varience of the Frequency Derivative of the first ALL formant
%   startOfALLMeasurements + 6     : min of the Frequency Derivative of the first ALL formant
%   startOfALLMeasurements + 7     : max  of the Frequency Derivative of the first ALL formant
%   startOfALLMeasurements + 8     : mean of the Frequency 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 9     : varience of the Frequency 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 10    : min of the Frequency 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 11    :  max of the Frequency 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 12    : mean of the Bandwidth of the first ALL formant
%   startOfALLMeasurements + 13    :  varience of the Bandwidth of the first ALL formant
%   startOfALLMeasurements + 14    : min of the Bandwidth of the first ALL formant
%   startOfALLMeasurements + 15    : max of the Bandwidth of the first ALL formant
%   startOfALLMeasurements + 16    : mean of the Bandwidth Derivative of the first ALL formant
%   startOfALLMeasurements + 17    : varience of the Bandwidth Derivative of the first ALL formant
%   startOfALLMeasurements + 18    : min of the Bandwidth Derivative of the first ALL formant
%   startOfALLMeasurements + 19    : max of the Bandwidth Derivative of the first ALL formant
%   startOfALLMeasurements + 20    : mean of the Bandwidth 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 21    : var of the Bandwidth 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 22    : min of the Bandwidth 2nd Derivative of the first ALL formant
%   startOfALLMeasurements + 23    : max of the Bandwidth 2nd Derivative of the first ALL formant
%
%    ....... there are nMetrics for each formant in nAF formants, so cycle
%    through until the last is reached ......
%
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean frequency of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : variance of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : minimum frequency of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : maximum frequency of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean Frequency Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : varience of the Frequency Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : min of the Frequency Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : max of the Frequency Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean of the Frequency 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : varience of the Frequency 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : min of the Frequency 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : max of the Frequency 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean of the Bandwidth of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : varience of the Bandwidth of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : min of the Bandwidth of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : max of the Bandwidth of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean of the Bandwidth Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : variece of the Bandwidth Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : min of the Bandwidth Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : max of the Bandwidth Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : mean of the Bandwidth 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : var of the Bandwidth 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : min of the Bandwidth 2nd Derivative of the nAF ALL formant
%   startOfALLMeasurements + ((nAF-1)*nMetrics)       : max of the Bandwidth 2nd Derivative of the nAF ALL formant
%
%   FOR THE MEAN OF ALL ALL FORMANTS
%   startOfALLMeasurements + (nAF*nMetrics)           : mean of all formants Frequency 
%   startOfALLMeasurements + (nAF*nMetrics) + 1       : varience of the mean of all formants Frequency 
%   startOfALLMeasurements + (nAF*nMetrics) + 2       : minimum of the mean of all formants Frequency 
%   startOfALLMeasurements + (nAF*nMetrics) + 3       : maximum of the mean of all formants Frequency
%   startOfALLMeasurements + (nAF*nMetrics) + 4       :  mean of all formants mean Frequency Derivative 
%   startOfALLMeasurements + (nAF*nMetrics) + 5       :  mean of all formants varience Frequency Derivative 
%   startOfALLMeasurements + (nAF*nMetrics) + 6       :  min of the mean of all formants Frequency Derivative
%   startOfALLMeasurements + (nAF*nMetrics) + 7       :  max of the mean of all formants Frequency Derivative 
%   startOfALLMeasurements + (nAF*nMetrics) + 8       :  mean of the mean of all formants Frequency 2nd Derivative 
%   startOfALLMeasurements + (nAF*nMetrics) + 9       :  varience of the mean of all formants Frequency 2nd Derivative 
%   startOfALLMeasurements + (nAF*nMetrics) + 10      :  min of the mean of all formants Frequency 2nd Derivative
%   startOfALLMeasurements + (nAF*nMetrics) + 11      :  max of the mean of all formants Frequency 2nd Derivative
%
% ------------- ROBUST FORMANTS ---------------
%
%   startOfALLMeasurements + (nAF*nMetrics) + 12    : Number of ROBUST formants listed = nRF
%
%   startOfROBUSTMeasurements = startOfALLMeasurements + (nAF*nMetrics) + 13;
%
%   startOfROBUSTMeasurements         : mean frequency of the first ROBUST formant
%   startOfROBUSTMeasurements + 1     : variance of the first ROBUST formant
%   startOfROBUSTMeasurements + 2     : minimum frequency of the first ROBUST formant
%   startOfROBUSTMeasurements + 3     : maximum frequency of the first ROBUST formant
%   startOfROBUSTMeasurements + 4     : mean Frequency Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 5     : varience of the Frequency Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 6     : min of the Frequency Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 7     : max  of the Frequency Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 8     : mean of the Frequency 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 9     : varience of the Frequency 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 10    : min of the Frequency 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 11    :  max of the Frequency 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 12    : mean of the Bandwidth of the first ROBUST formant
%   startOfROBUSTMeasurements + 13    :  varience of the Bandwidth of the first ROBUST formant
%   startOfROBUSTMeasurements + 14    : min of the Bandwidth of the first ROBUST formant
%   startOfROBUSTMeasurements + 15    : max of the Bandwidth of the first ROBUST formant
%   startOfROBUSTMeasurements + 16    : mean of the Bandwidth Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 17    : varience of the Bandwidth Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 18    : min of the Bandwidth Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 19    : max of the Bandwidth Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 20    : mean of the Bandwidth 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 21    : var of the Bandwidth 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 22    : min of the Bandwidth 2nd Derivative of the first ROBUST formant
%   startOfROBUSTMeasurements + 23    : max of the Bandwidth 2nd Derivative of the first ROBUST formant
%
%    ....... there are nMetrics for each formant in nRF formants, so cycle
%    through until the last is reached ......
%
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean frequency of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : variance of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : minimum frequency of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : maximum frequency of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean Frequency Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : varience of the Frequency Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : min of the Frequency Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : max of the Frequency Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean of the Frequency 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : varience of the Frequency 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : min of the Frequency 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : max of the Frequency 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean of the Bandwidth of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : varience of the Bandwidth of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : min of the Bandwidth of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : max of the Bandwidth of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean of the Bandwidth Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : variece of the Bandwidth Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : min of the Bandwidth Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : max of the Bandwidth Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : mean of the Bandwidth 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : var of the Bandwidth 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : min of the Bandwidth 2nd Derivative of the nRF ROBUST formant
%   startOfROBUSTMeasurements + ((nRF-1)*nMetrics)       : max of the Bandwidth 2nd Derivative of the nRF ROBUST formant
%
%   FOR THE MEAN OF ALL ROBUST FORMANTS
%   startOfROBUSTMeasurements + (nRF*nMetrics)           : mean of all formants Frequency 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 1       : varience of the mean of all formants Frequency 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 2       : minimum of the mean of all formants Frequency 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 3       : maximum of the mean of all formants Frequency
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 4       :  mean of all formants mean Frequency Derivative 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 5       :  mean of all formants varience Frequency Derivative 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 6       :  min of the mean of all formants Frequency Derivative
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 7       :  max of the mean of all formants Frequency Derivative 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 8       :  mean of the mean of all formants Frequency 2nd Derivative 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 9       :  varience of the mean of all formants Frequency 2nd Derivative 
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 10      :  min of the mean of all formants Frequency 2nd Derivative
%   startOfROBUSTMeasurements + (nRF*nMetrics) + 11      :  max of the mean of all formants Frequency 2nd Derivative
%

noOfArguments = length(varargin);
columnIndices = [];

getBURGFormants = 0;
getAllFormants=0;
getRobustFormants=0;

for i=1 : noOfArguments
    if( strcmp( varargin{i}, 'formant_Burg' ))
        getBURGFormants = 1;
    elseif( strcmp( varargin{i}, 'formant_all' ))
        getAllFormants=1;
    elseif( strcmp( varargin{i}, 'formant_robust' ))
        getRobustFormants=1;
    end
end

titleName = '';
for i=1 : noOfArguments
    titleName = [ titleName varargin{i} '_'];
    fprintf( masterFileOutputID, '%s_', varargin{i} );
end

fprintf( masterFileOutputID, '\t' );

% -------------------- get the data from the results file ---------------
lineCount = 0;
fileCount = 0;
data = [];
groups = [];

while( ~(feof(inputFileID)) )
    
    outputValues = [];

    thestr = fgetl(inputFileID);
    if( lineCount > 10 )    % skip the file header
        fileCount = fileCount + 1;
        
        % determine whether we have a positive or negative sample
        sampleEmotion( fileCount ) = 'U';
        if( ~(isempty(strfind(thestr,'pos'))))
            % sample is positive
            sampleEmotion( fileCount ) = 'P';
            groups( fileCount ) = 1;
        elseif( ~(isempty(strfind(thestr,'neg'))))
            % sample is negative
            sampleEmotion( fileCount ) = 'N';
            groups( fileCount ) = 0;
        else
            disp('EEEK!');
            pause;
        end
        
%         % determine whether we have a male, female or trans sample
%         gender( fileCount ) = '?';
%         if( ~(isempty(strfind(thestr,'fem'))))
%             % gender is female
%             gender( fileCount ) = 'F';
%         elseif( ~(isempty(strfind(thestr,'male'))))
%             % gender is male
%             gender( fileCount ) = 'M';
%         elseif( ~(isempty(strfind(thestr,'trans'))))
%             % gender is trans
%             gender( fileCount ) = 'T';
%         else
%             disp('EEEK!');
%             pause;
%         end

        %how many values are in the string?
        spaces = strfind( thestr, ' ' );
        numberstr = thestr( spaces(1) : end ); % chop off the file name
        frmtpos = strfind( numberstr, 'maxNoOfFormants'); % find the position of the label for number of formants
        
%         str1 = numberstr( 1 : frmtpos(1)-1 ); % string contains jitter and shimmer values
        str2 = numberstr( frmtpos(1) : frmtpos(2)-1 ); % string contains all BURG formant information
        str3 = numberstr( frmtpos(2) : frmtpos(3)-1 ); % string contains all ALL formant information
        str4 = numberstr( frmtpos(3) : end ); % string contains all ROBUST formant information
        
        
%         vars = sscanf( str1, '%f', inf );
%         % extract the shimmer and jitter values
%         outputValues = [ outputValues vars( columnIndices )'];
        
        if( getBURGFormants )
            spaces = strfind( str2, ' ' ); % remove the string 'maxNoOfFormants'
            vars = sscanf( str2( spaces(1) : end ), '%f', inf );
            outputValues = stripOutFormantValues( vars, outputValues );
        end
        
        if( getAllFormants )
            spaces = strfind( str3, ' ' ); % remove the string 'maxNoOfFormants'
            vars = sscanf( str3( spaces(1) : end ), '%f', inf );
            outputValues = stripOutFormantValues( vars, outputValues );
        end
        
        if( getRobustFormants )
            spaces = strfind( str4, ' ' ); % remove the string 'maxNoOfFormants'
            vars = sscanf( str4( spaces(1) : end ), '%f', inf );
            outputValues = stripOutFormantValues( vars, outputValues );
        end
        
        [m n] = size( data );
        % sometimes the 'all' formants command gives us fewer formants than
        % usual. If this is the case,then we will have to pad with zeros
        % for now.
        if( n > length( outputValues ) )
            lenDiff = n - length( outputValues );
            outputValues = [ outputValues zeros( 1, lenDiff ) ];
        elseif( n < length( outputValues ) )
            lenDiff = length( outputValues ) - n;
            outputValues = [ outputValues zeros( 1, lenDiff ) ];
        end
        
        data( fileCount, : ) = outputValues;
        
    end
    lineCount = lineCount + 1;
    
end
fclose(inputFileID);



% ------------  apply the SVM classifier  ------------------------

resultMatrix = [];

noOfIterations = 10;

for n = 1:noOfIterations
    % Randomly select training and test sets, perhaps we should try all and
    % choose the best?
    [train, test] = crossvalind('holdOut',groups);
    cp = classperf(groups);
    
    % Use a linear support vector machine classifier
    svmStruct = svmtrain(data(train,:),groups(train));
    classes = svmclassify(svmStruct,data(test,:));
    % See how well the classifier performed
    classperf(cp,classes,test);
    numbers = cp.CountingMatrix;

    resultMatrix (n,:,:) = cp.DiagnosticTable;
%     
end




% for emotion detection give the confusion matrix as 
    % -----------------------------------------------------------------
    % positive correctly identified         | positive incorrectly identified (1,2)
    % negative incorrectly identified (2,1) | negative correctly identified
    % ------------------------------------------------------------------
    
% takes the average of 10 iterations - do we want to take the best?  

meanResults(1,1) = mean( resultMatrix(:,1,1) );
meanResults(1,2) = mean( resultMatrix(:,2,1) );
meanResults(2,1) = mean( resultMatrix(:,1,2) );
meanResults(2,2) = mean( resultMatrix(:,2,2) );

meanResults(3,:)=0;
meanResults(:,3)=0;

meanResults(3,3) = (meanResults(1,1) + meanResults(2,2));% / sum(sum(meanResults));

% convert to percentages
% how many of each sample do we have?
groupNumbers = unique( groups( test ));
groupNames = unique( sampleEmotion( test ));
sampleEmotionTest = sampleEmotion( test );
% which group is which emotion?
thisGroupNumber = groupNumbers(1);
thisGroup = find( groups( test ) == thisGroupNumber );
thisGroupName = unique( sampleEmotionTest( thisGroup ));

thatGroupNumber = groupNumbers(2);
thatGroup = find( groups( test ) == thatGroupNumber );
thatGroupName = unique( sampleEmotionTest( thatGroup ));

if(length( thisGroupName ) ~= 1 )
    disp('ARGH!');
    pause;
end

thisGroupNumberOfSamples = length( thisGroup );
thatGroupNumberOfSamples = length( thatGroup );

if( thisGroupName == 'P' )
    %swap all the variables ready for checking
    temp = thisGroupNumberOfSamples;
    thisGroupNumberOfSamples = thatGroupNumberOfSamples;
    thatGroupNumberOfSamples = temp;
    
    temp = thisGroupName;
    thisGroupName = thatGroupName
    thatGroupName = temp;
    disp('CHECK ME!');
end

if( thisGroupName == 'N' )
     % group 0 is negative
    if( sum( meanResults(1,:) ) == thisGroupNumberOfSamples )
        %if the elements in the first row add up to the number of negative
        %samples, then swap the rows because we want the top row to be the
        %results for the positive samples
        temp(:,1) = meanResults(1:2,2);
        temp(:,2) = meanResults(1:2,1);
        temp2(1,:) = temp(2,:);
        temp2(2,:) = temp(1,:);
        
        meanResults(1:2,1) = temp2(:,1);
        meanResults(1:2,2) = temp2(:,2);

        % check the number of positive samples
        if(( sum( meanResults(1,:) ) == thatGroupNumberOfSamples ) ...
            && ( thatGroupName == 'P' ) )
            % row 1 is positive
            disp('matrix correct');
        else
            disp('ARGH!');
            pause;
        end

    elseif( sum( meanResults(2,:) ) == thisGroupNumberOfSamples  )

        % the elements in the second row add up to the number of negative
        % samples, so the matrix is the correct way around

        % check the number of positive samples
        if(( sum( meanResults(1,:) ) == thatGroupNumberOfSamples ) ...
            && ( thatGroupName == 'P' ) )
            % row 0 is positive
            disp('matrix correct');
        else
            disp('ARGH!');
            pause;
        end
    end
end

% calculate the percentages
numberOfSamples = sum(sum( meanResults(1:2,1:2)));
percentageResults = meanResults;
percentageResults(1,1) = meanResults(1,1) / numberOfSamples;
percentageResults(1,2) = meanResults(1,2) / numberOfSamples;
percentageResults(2,1) = meanResults(2,1) / numberOfSamples;
percentageResults(2,2) = meanResults(2,2) / numberOfSamples;
percentageResults(3,3) = meanResults(3,3) / numberOfSamples;

percentageResults = percentageResults * 100

confusionMatrix = percentageResults;
 fprintf( masterFileOutputID, '\n %f \t %f \n %f \t %f \n %f \t %f \t %f \n', confusionMatrix(1,1), confusionMatrix(1,2), confusionMatrix(2,1), confusionMatrix(2,2), 0, 0, confusionMatrix(3,3));

 % print latex results to the screen
str1 = sprintf(' & %2.2f & %2.2f &  \\\\', percentageResults(1,1), percentageResults(1,2) );
disp(str1);
str1 = sprintf(' & %2.2f & %2.2f &  \\\\', percentageResults(2,1), percentageResults(2,2) );
disp(str1);
str1 = sprintf(' & & & %2.2f \\\\',percentageResults(3,3) );
disp(str1);

fprintf( masterFileOutputID, '\n' );
fclose( masterFileOutputID );

end

%------------------------------------------------------------------

function [ outputValues ] = stripOutFormantValues( vars, outputValues )
    
    noOfFormantValues = length( vars ) - 1; % gives the number of formant arguments only
    noOfFormants = vars(1);
    % there are 12 measurements for the mean of all formants (so the number
    % of formants is not important) for each formant measurement.
    if( noOfFormants ~= (noOfFormantValues-12)/24 )
        disp('EEK!');
        pause;
    else
        outputValues = [ outputValues vars( 2:end )' ];
    end

end

%-------------------------------------------------------------------