comparison Code/General/readFormantValueFromFile.m @ 4:92ca03a8fa99 tip

Update to ICASSP 2013 benchmark

+function [] = readFormantValueFromFile( statsFileID, metricFileName, metricName, vuv )
+
+DEBUG=0;
+
+FileID = fopen( metricFileName );
+
+if( FileID <= 0 ) %does the file exist?
+    % if not
+    
+    message = ['WARNING: MISSING ' metricName ' FILE'];
+    disp( message );
+    fprintf( statsFileID, '\t \s missing', metricName );
+   
+ else
+    % file format is not straight forward
+    noOfValues = 0;
+    formants = [];
+    while( ~(feof(FileID)) )
+
+        % search for numberOfFormants
+        thestr = fgetl(FileID);
+
+        if( strfind( thestr , 'numberOfFormants' ) > 0 )
+            noOfValues = noOfValues + 1;
+            %numberOfFormants found
+            pos = find( thestr == '=' );
+            numberOfFormants = str2num(thestr(pos+2:end));
+            formants( noOfValues, 1 ) = numberOfFormants;
+            % discard the 'formant []' line
+            thestr = fgetl(FileID);
+            % now read the formant positions
+            for (i=0:numberOfFormants-1)
+                thestr = fgetl(FileID);
+                pos = find( thestr == '=' );
+                formants( noOfValues, i+2 ) =  str2num(thestr( pos+2 : end ));
+            end
+
+            % discard the 'bandwidth []' line
+            thestr = fgetl(FileID);
+            % now read the formant bandwidths
+            for (i=0:numberOfFormants-1)
+                thestr = fgetl(FileID);
+                pos = find( thestr == '=' );
+                formants( noOfValues, i+2+numberOfFormants ) =  str2num(thestr( pos+2 : end ));
+            end
+
+        end
+    end
+    fclose(FileID);    
+
+
+    if(DEBUG)
+        timeShift = length(formants) / length(vuv);
+        subplot(211); hold off;
+        subplot(212); hold off;
+        for( idx = 1: length( formants )/10 )
+            frameIdx = floor( idx / timeShift ) + 1;
+            if( frameIdx <= length(vuv) )
+                if( vuv( frameIdx ) == 0 )
+                    % frame is silent
+                    subplot(211); plot( idx, formants( idx,2 ), 'w.' ); hold on;
+                    subplot(211); plot( idx, formants( idx,3 ), 'w.' );
+                    subplot(211); plot( idx, formants( idx,4 ), 'w.' );
+                    % plot bandwidths
+                    subplot(212); plot( idx, formants( idx,5 ), 'w.' ); hold on;
+                    subplot(212); plot( idx, formants( idx,6 ), 'w.' );
+                    subplot(212); plot( idx, formants( idx,7 ), 'w.' );
+                elseif( vuv( frameIdx ) == 1 )
+                    % frame is voiced
+                    subplot(211); plot( idx, formants( idx,2 ), 'y.' ); hold on;
+                    subplot(211); plot( idx, formants( idx,3 ), 'c.' );
+                    subplot(211); plot( idx, formants( idx,4 ), 'g.' );
+                    % plot bandwidths
+                    subplot(212); plot( idx, formants( idx,5 ), 'y.' ); hold on;
+                    subplot(212); plot( idx, formants( idx,6 ), 'c.' );
+                    subplot(212); plot( idx, formants( idx,7 ), 'g.' );
+                else
+                    %frame is unvoiced
+                    subplot(211); plot( idx, formants( idx,2 ), 'r.' ); hold on;
+                    subplot(211); plot( idx, formants( idx,3 ), 'r.' );
+                    subplot(211); plot( idx, formants( idx,4 ), 'r.' );
+                    % plot bandwidths
+                    subplot(212); plot( idx, formants( idx,5 ), 'r.' ); hold on;
+                    subplot(212); plot( idx, formants( idx,6 ), 'r.' );
+                    subplot(212); plot( idx, formants( idx,7 ), 'r.' );
+                end
+            else
+                frameIdx
+            end
+        end
+    end
+
+    [ noOfSamples formantDataSize ] = size( formants );
+    maxNoOfFormants = (formantDataSize - 1)/2; % the array contains the number of formants, then the position of each formant, then the bandwidth of each formant.
+    voicedFormant = [];
+    voicedBandwidth = [];
+    % find the basic metric set for the voiced frames only
+    validSamples = min( noOfSamples, length(vuv) );
+    for( idx = 1: validSamples )
+        if( vuv( idx ) == 1 )
+            % frame is voiced
+            formantsThisFrame = formants(idx,1);
+            theFormants = formants( idx, 2 : formantsThisFrame+1 );
+            theBandwidths = formants( idx, formantsThisFrame+2 : (formantsThisFrame*2)+1 );
+            % we want to ignore any results that contain less than the
+            % max number of formants as then one of the formants is set to
+            % zero and the detection is not reliable
+            % DON'T THINK THIS IS STILL VALID
+            % DO WE WANT TO DISCARD FORMANT SETS THAT CONTAIN '0' FORMANT
+            % POSITIONS FOR VOICED SEGMENTS?
+            if( formantsThisFrame == maxNoOfFormants )
+                voicedFormant = [ voicedFormant ; theFormants ];
+                voicedBandwidth = [ voicedBandwidth; theBandwidths ];
+            else
+                % pad with zeros
+                diff = maxNoOfFormants - formantsThisFrame;
+                theFormants = [ theFormants zeros( 1, diff )];
+                voicedFormant = [ voicedFormant ; theFormants ];
+                theBandwidths = [ theBandwidths zeros( 1, diff )];
+                voicedBandwidth = [ voicedBandwidth; theBandwidths ];
+            end
+
+        end
+    end
+
+    % get the basic formant metrics
+    formantFrequencyMeans = [];
+    formantGradientMean = [];
+    formantBandwidthMean = [];
+
+    fprintf( statsFileID, '\t %s %d ', 'maxNoOfFormants', maxNoOfFormants );
+    
+    for( formantIdx = 1:maxNoOfFormants )
+        % formant position
+        [ metrics ] = basicMetricSet( voicedFormant(:,formantIdx), statsFileID );
+        formantFrequencyMeans = [ formantFrequencyMeans metrics(1) ];
+        % formant gradient
+        formantGradient = gradient( voicedFormant(:,formantIdx) );
+        [ metrics ] = basicMetricSet( formantGradient, statsFileID );
+        formantGradientMean = [formantGradientMean metrics(1)];
+        % formant second derivative
+        formantGradient = gradient( formantGradient );
+        [ metrics ] = basicMetricSet( formantGradient, statsFileID );
+        % formant bandwidth
+        [ metrics ] = basicMetricSet( voicedBandwidth(:,formantIdx), statsFileID );
+        formantBandwidthMean = [formantBandwidthMean metrics(1)];
+        % formant bandwidth gradient
+        formantGradient = gradient( voicedBandwidth(:,formantIdx));
+        [ metrics ] = basicMetricSet( formantGradient, statsFileID );
+        % formant bandwidth second derivative
+        formantGradient = gradient( formantGradient );
+        [ metrics ] = basicMetricSet( formantGradient, statsFileID );
+    end
+
+    % calculate the basic metrics set for the ...
+    % ... mean of all formant frequencies
+    meanMetrics = basicMetricSet( formantFrequencyMeans, statsFileID );
+    % ... mean of all formant frequency gradients
+    meanMetrics = basicMetricSet( formantGradientMean, statsFileID );
+    % ... mean of all formant bandwidths
+    meanMetrics = basicMetricSet( formantBandwidthMean, statsFileID );
+end
+
+    
+