[ file synthmultisource/README ]

3 recordings based on the 5 french voyels 'a', 'e', 'i', 'o', 'u' ( resp. /ah/ /eh/ /i/ /oh/ /y/ ).

3 loudspeakers play synchronously the 3 channels (16 kHz files [ synth-channel-*.wav ] in this directory).
loudspeaker 1 = channel 1, pitch 100 Hz
loudspeaker 2 = channel 2, pitch 120 Hz
loudspeaker 3 = channel 3, pitch 180 Hz
Meanwhile an 8-microphone array was recording (16 kHz files [ setup*-mic*.wav ] in this directory).
Each setup corresponds to a different position for loudspeaker 1.
Each channel is a loop : 4 seconds of ( silence or one voyel )
                         2 seconds of silence

Below is the list of contents for each channel ( '-' means silence ).
Basically all possible combinations of 2 and 3 channels and 5 voyels (3*5*5+5*5*5 = 200) are played. Total duration for each setup therefore is 200 * ( 4 + 2 ) = 1200 seconds.

Note that in fact, a small drift in the loudspeaker playing/microphone recording synchronization system lead to a slight advance of the emission time (50 ms after 20 minutes). This drift was observed to be linear with time, so the ground-truth correction was simple to do.
=> the "*gt.mat" file have been corrected accordingly, based on the recording of microphone 1.

Final note: in the recorded files, some obscure Windows OS-related problem replaced the beginning (less than 1 second) with zeroes.



Contents:

synth-channel-1.wav:

aeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiou-------------------------aeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiouaeiou

synth-channel-2.wav:

aaaaaeeeeeiiiiiooooouuuuu-------------------------aeiouaeiouaeiouaeiouaeiouaaaaaeeeeeiiiiiooooouuuuuaaaaaeeeeeiiiiiooooouuuuuaaaaaeeeeeiiiiiooooouuuuuaaaaaeeeeeiiiiiooooouuuuuaaaaaeeeeeiiiiiooooouuuuu

synth-channel-3.wav:

-------------------------aaaaaeeeeeiiiiiooooouuuuuaaaaaeeeeeiiiiiooooouuuuuaaaaaaaaaaaaaaaaaaaaaaaaaeeeeeeeeeeeeeeeeeeeeeeeeeiiiiiiiiiiiiiiiiiiiiiiiiiooooooooooooooooooooooooouuuuuuuuuuuuuuuuuuuuuuuuu


Geometry (MATLAB lines: rows 1,2,3 are X,Y,Z respectively):

a = (1 + (0:7)/4) * pi;
r = 0.1;
array_geometry = [ r * cos( a ); r * sin( a ); 0 * a ];

setup = [];
setup( 1 ).loudspeaker_geometry = [  0.0   0.7   0.7;      % precise
                                     0.8   0.35 -0.35;     % precise
                                     0.30  0.30  0.30 ];   % approximative

setup( 2 ).loudspeaker_geometry = [  0.0   0.7   0.7;      % precise
                                     1.8   0.35 -0.35;     % precise
                                     0.30  0.30  0.30 ];   % approximative

setup( 3 ).loudspeaker_geometry = [  1.45  0.7   0.7;      % precise
                                     0.0   0.35 -0.35;     % precise
                                     0.30  0.30  0.30 ];   % approximative


for i_setup = 1:length( setup )
  xyz = array_geometry;
  figure; plot3( xyz(1,:), xyz(2,:), xyz(3,:), 'bo');
  hold on;
  xyz = setup( i_setup ).loudspeaker_geometry;
  plot3( xyz(1,:), xyz(2,:), xyz(3,:), 'r*');
  axis equal; grid on;
  title( sprintf( 'setup %d', i_setup ) );
  xlabel( 'x' );
  ylabel( 'y' );
  zlabel( 'z' );
end