% MERGESOUNDSFROMSCRIPT     Add noise to all files listed in a script file.
%                           Write output files in the current directory.
%
%   Note: this implementation ensures we obtain exactly the desired SNR
%   in each output file. Thus more noise will be added to lively speakers,
%   whereas less noise will be added to quiet speakers.
%
% MERGESOUNDSFROMSCRIPT(FILESINFO, DESIREDSNRDB, SAMPLERATE, SPEECHTHRESHOLD)
%
% Audio files must be 16-bit.
%
% FILEINFO is a structure containing all needed info about the audio files.
%   (description below)
% DESIREDSNRDB is a float, the target SNR level in decibels.
% SAMPLERATE is a float, the sample frequency common to all audio files.
% SPEECHTHRESHOLD is a flost (optional, default 1e-3). It is a ratio determining 
%   speech/non-speech segment for segmental energy calculation. See mergeSounds().
%
% ---
% FILEINFO.INPUTSCRIPT is a string. This is the name of the script file,
%   that contains all (typically clean speech) audio file names.
%
% FILEINFO.INPUT_HS is an integer, the size of the audio file header 
%   of all the files listed by FILEINFO.INPUTSCRIPT.
%   Typically 0 for raw, 44 for microsoft wave, 1024 for NIST_1A.
%
% FILEINFO.OUTPUTDIR is a string. This is the directory where we will write
%   the noisy files.
%
% FILEINFO.INSERT is the string we use to produce output filenames. 
%   It is inserted in the original audio file names just before the extension.
%
% FILEINFO.INPUT2 is a string, the name of the noise file.
%
% FILEINFO.INPUT2_HS is an integer, the size of the audio file header for FILEINFO.INPUT2.
%   Typically 0 for raw, 44 for microsoft wave, 1024 for NIST_1A.
%
% by Guillaume LATHOUD at IDIAP (lathoud@idiap.ch)

function mergeSoundsFromScript(fileinfo, desiredSNRdB, samplerate, speechthreshold)

    if nargin < 3
        error('I need at least three parameters!')
    end

    if nargin < 4
        speechthreshold = 1e-3;   % Ratio on max energy in a signal (used in mergeSounds())
    end
    
    % 0) read parameters
    
    listfilename1 = fileinfo.inputscript;
    headersize1   = fileinfo.input_hs;
    outputdir     = fileinfo.outputdir;
    insert1       = fileinfo.insert;
    filename2     = fileinfo.input2;
    headersize2   = fileinfo.input2_hs;
    
    % 1) read the list of speech files
    %    for each filename prepare the output filename
    
    fd = fopen(listfilename1, 'r');
    filelist = {};
    while ~feof(fd)
        s = deblank(fgetl(fd));
        if length(s) > 0
            
            % build the output filename
            % first remove path
            aux = findstr(s, '/');
            if length(aux) > 0
                s2 = s(aux(length(aux))+1:length(s));
            end
            
            % second insert string 'insert1' before extension
            aux = findstr(s2, '.');
            if length(aux) < 1
                s2 = strcat(s2, insert1);
            else
                s2 = strcat(s2(1:aux(length(aux))-1), insert1, s2(aux(length(aux)):length(s2)));
            end
            i = length(filelist) + 1;
            filelist{i}.in = s;
            filelist{i}.out = fullfile(outputdir,s2);
        end
    end
    fclose(fd);

    % 2) read the noise file
    
    [sound2 n2] = rawread(filename2, headersize2);
    
    % 3) Loop through the list of files
    
    for i = 1:length(filelist)
        
        disp([filelist{i}.in ' with ' filename2 ' to ' filelist{i}.out])
        
        % read the speech file
        [sound1 n1] = rawread(filelist{i}.in, headersize1);
        
        % merge the speech signal with the noise signal
        [soundMerged a1 a2] = mergeSounds(sound1, sound2, desiredSNRdB, speechthreshold, samplerate, 16);
        
        % write out the result
        nistwrite(soundMerged, samplerate, 16, filelist{i}.out);
        
    end