Daniel@0: function [Y,FS,NBITS,OPTS] = mp3read(FILE,N,MONO,DOWNSAMP,DELAY)
Daniel@0: % MP3READ   Read MP3 audio file via use of external binaries.
Daniel@0: %   Y = MP3READ(FILE) reads an mp3-encoded audio file into the
Daniel@0: %     vector Y just like wavread reads a wav-encoded file (one channel 
Daniel@0: %     per column).  Extension ".mp3" is added if FILE has none.
Daniel@0: %     Also accepts other formats of wavread, such as
Daniel@0: %   Y = MP3READ(FILE,N) to read just the first N sample frames (N
Daniel@0: %     scalar), or the frames from N(1) to N(2) if N is a two-element vector.  
Daniel@0: %   Y = MP3READ(FILE,FMT) or Y = mp3read(FILE,N,FMT) 
Daniel@0: %     with FMT as 'native' returns int16 samples instead of doubles; 
Daniel@0: %     FMT can be 'double' for default behavior (to exactly mirror the
Daniel@0: %     syntax of wavread).
Daniel@0: %
Daniel@0: %   [Y,FS,NBITS,OPTS] = MP3READ(FILE...) returns extra information:
Daniel@0: %     FS is the sampling rate,  NBITS is the bit depth (always 16), 
Daniel@0: %     OPTS.fmt is a format info string; OPTS has multiple other
Daniel@0: %     fields, see WAVREAD.
Daniel@0: %
Daniel@0: %   SIZ = MP3READ(FILE,'size') returns the size of the audio data contained
Daniel@0: %     in the file in place of the actual audio data, returning the
Daniel@0: %     2-element vector SIZ=[samples channels].
Daniel@0: %
Daniel@0: %   [Y...] = MP3READ(FILE,N,MONO,DOWNSAMP,DELAY) extends the
Daniel@0: %     WAVREAD syntax to allow access to special features of the
Daniel@0: %     mpg123 engine:  MONO = 1 forces output to be mono (by
Daniel@0: %     averaging stereo channels); DOWNSAMP = 2 or 4 downsamples by 
Daniel@0: %     a factor of 2 or 4 (thus FS returns as 22050 or 11025
Daniel@0: %     respectively for a 44 kHz mp3 file); DELAY controls how many
Daniel@0: %     "warm up" samples to drop at the start of the file; the
Daniel@0: %     default value of 2257 makes an mp3write/mp3read loop for a 44
Daniel@0: %     kHz mp3 file be as close as possible to being temporally
Daniel@0: %     aligned; specify as 0 to prevent discard of initial samples.
Daniel@0: %
Daniel@0: %   Example:
Daniel@0: %   To read an mp3 file as doubles at its original width and sampling rate:
Daniel@0: %     [Y,FS] = mp3read('piano.mp3');
Daniel@0: %   To read the first 1 second of the same file, downsampled by a
Daniel@0: %   factor of 4, cast to mono, using the default filename
Daniel@0: %   extension:
Daniel@0: %     [Y,FS4] = mp3read('piano', FS/4, 1, 4);
Daniel@0: %
Daniel@0: %   Note: Because the mp3 format encodes samples in blocks of 26 ms (at
Daniel@0: %   44 kHz), and because of the "warm up" period of the encoder,
Daniel@0: %   the file length may not be exactly what you expect.
Daniel@0: %
Daniel@0: %   Note: requires external binaries mpg123 and mp3info; you
Daniel@0: %   can find binaries for several platforms at:
Daniel@0: %     http://labrosa.ee.columbia.edu/matlab/mp3read.html
Daniel@0: %
Daniel@0: %   See also mp3write, wavread.
Daniel@0: 
Daniel@0: % 2003-07-20 dpwe@ee.columbia.edu  This version calls mpg123.
Daniel@0: % 2004-08-31 Fixed to read whole files correctly
Daniel@0: % 2004-09-08 Uses mp3info to get info about mp3 files too
Daniel@0: % 2004-09-18 Reports all mp3info fields in OPTS.fmt; handles MPG2LSF sizes
Daniel@0: %            + added MONO, DOWNSAMP flags, changed default behavior.
Daniel@0: % 2005-09-28 Fixed bug reading full-rate stereo as 1ch (thx bjoerns@vjk.dk)
Daniel@0: % 2006-09-17 Chop off initial 2257 sample delay (for 44.1 kHz mp3)
Daniel@0: %            so read-write loop doesn't get progressively delayed.
Daniel@0: %            You can suppress this with a 5th argument of 0.
Daniel@0: % 2007-02-04 Added support for FMT argument to match wavread
Daniel@0: %            Added automatic selection of binary etc. to allow it
Daniel@0: %            to work cross-platform without editing prior to
Daniel@0: %            submitting to Matlab File Exchange
Daniel@0: % 2007-07-23 Tweaks to 'size' mode so it exactly agrees with read data.
Daniel@0: 
Daniel@0: % find our baseline directory
Daniel@0: path = fileparts(which('mp3read'));
Daniel@0: 
Daniel@0: % %%%%% Directory for temporary file (if needed)
Daniel@0: % % Try to read from environment, or use /tmp if it exists, or use CWD
Daniel@0: tmpdir = getenv('TMPDIR');
Daniel@0: if isempty(tmpdir) || exist(tmpdir,'file')==0
Daniel@0:   tmpdir = '/tmp';
Daniel@0: end
Daniel@0: if exist(tmpdir,'file')==0
Daniel@0:   tmpdir = '';
Daniel@0: end
Daniel@0: % ensure it exists
Daniel@0: %if length(tmpdir) > 0 && exist(tmpdir,'file')==0
Daniel@0: %  mkdir(tmpdir);
Daniel@0: %end
Daniel@0: 
Daniel@0: %%%%%% Command to delete temporary file (if needed)
Daniel@0: rmcmd = 'rm';
Daniel@0: 
Daniel@0: %%%%%% Location of the binaries - attempt to choose automatically
Daniel@0: %%%%%% (or edit to be hard-coded for your installation)
Daniel@0: ext = lower(computer);
Daniel@0: if ispc
Daniel@0:   ext = 'exe';
Daniel@0:   rmcmd = 'del';
Daniel@0: end
Daniel@0: mpg123 = fullfile(path,['mpg123.',ext]);
Daniel@0: mp3info = fullfile(path,['mp3info.',ext]);
Daniel@0: 
Daniel@0: %%%%% Process input arguments
Daniel@0: if nargin < 2
Daniel@0:   N = 0;
Daniel@0: end
Daniel@0: 
Daniel@0: % Check for FMT spec (per wavread)
Daniel@0: FMT = 'double';
Daniel@0: if ischar(N)
Daniel@0:   FMT = lower(N);
Daniel@0:   N = 0;
Daniel@0: end
Daniel@0: 
Daniel@0: if length(N) == 1
Daniel@0:   % Specified N was upper limit
Daniel@0:   N = [1 N];
Daniel@0: end
Daniel@0: if nargin < 3
Daniel@0:   forcemono = 0;
Daniel@0: else
Daniel@0:   % Check for 3rd arg as FMT
Daniel@0:   if ischar(MONO)
Daniel@0:     FMT = lower(MONO);
Daniel@0:     MONO = 0;
Daniel@0:   end
Daniel@0:   forcemono = (MONO ~= 0);
Daniel@0: end
Daniel@0: if nargin < 4
Daniel@0:   downsamp = 1;
Daniel@0: else
Daniel@0:   downsamp = DOWNSAMP;
Daniel@0: end
Daniel@0: if downsamp ~= 1 && downsamp ~= 2 && downsamp ~= 4
Daniel@0:   error('DOWNSAMP can only be 1, 2, or 4');
Daniel@0: end
Daniel@0: if nargin < 5
Daniel@0:   mpg123delay44kHz = 2257;  % empirical delay of lame/mpg123 loop
Daniel@0:   delay = round(mpg123delay44kHz/downsamp);
Daniel@0: else
Daniel@0:   delay = DELAY;
Daniel@0: end
Daniel@0: 
Daniel@0: if strcmp(FMT,'native') == 0 && strcmp(FMT,'double') == 0 && ...
Daniel@0:       strcmp(FMT,'size') == 0
Daniel@0:   error(['FMT must be ''native'' or ''double'' (or ''size''), not ''',FMT,'''']);
Daniel@0: end
Daniel@0: 
Daniel@0: 
Daniel@0: %%%%%% Constants
Daniel@0: NBITS=16;
Daniel@0: 
Daniel@0: %%%%% add extension if none (like wavread)
Daniel@0: [path,file,ext] = fileparts(FILE);
Daniel@0: if isempty(ext)
Daniel@0:   FILE = [FILE, '.mp3'];
Daniel@0: end
Daniel@0: 
Daniel@0: %%%%%% Probe file to find format, size, etc. using "mp3info" utility
Daniel@0: cmd = ['"',mp3info, '" -r m -p "%Q %u %b %r %v * %C %e %E %L %O %o %p" "', FILE,'"'];
Daniel@0: % Q = samprate, u = #frames, b = #badframes (needed to get right answer from %u) 
Daniel@0: % r = bitrate, v = mpeg version (1/2/2.5)
Daniel@0: % C = Copyright, e = emph, E = CRC, L = layer, O = orig, o = mono, p = pad
Daniel@0: w = mysystem(cmd);
Daniel@0: % Break into numerical and ascii parts by finding the delimiter we put in
Daniel@0: starpos = findstr(w,'*');
Daniel@0: nums = str2num(w(1:(starpos - 2)));
Daniel@0: strs = tokenize(w((starpos+2):end));
Daniel@0: 
Daniel@0: SR = nums(1);
Daniel@0: nframes = nums(2);
Daniel@0: nchans = 2 - strcmp(strs{6}, 'mono');
Daniel@0: layer = length(strs{4});
Daniel@0: bitrate = nums(4)*1000;
Daniel@0: mpgv = nums(5);
Daniel@0: % Figure samples per frame, after
Daniel@0: % http://board.mp3-tech.org/view.php3?bn=agora_mp3techorg&key=1019510889
Daniel@0: if layer == 1
Daniel@0:   smpspfrm = 384;
Daniel@0: elseif SR < 32000 && layer ==3
Daniel@0:   smpspfrm = 576;
Daniel@0:   if mpgv == 1
Daniel@0:     error('SR < 32000 but mpeg version = 1');
Daniel@0:   end
Daniel@0: else
Daniel@0:   smpspfrm = 1152;
Daniel@0: end
Daniel@0: 
Daniel@0: OPTS.fmt.mpgBitrate = bitrate;
Daniel@0: OPTS.fmt.mpgVersion = mpgv;
Daniel@0: % fields from wavread's OPTS
Daniel@0: OPTS.fmt.nAvgBytesPerSec = bitrate/8;
Daniel@0: OPTS.fmt.nSamplesPerSec = SR;
Daniel@0: OPTS.fmt.nChannels = nchans;
Daniel@0: OPTS.fmt.nBlockAlign = smpspfrm/SR*bitrate/8;
Daniel@0: OPTS.fmt.nBitsPerSample = NBITS;
Daniel@0: OPTS.fmt.mpgNFrames = nframes;
Daniel@0: OPTS.fmt.mpgCopyright = strs{1};
Daniel@0: OPTS.fmt.mpgEmphasis = strs{2};
Daniel@0: OPTS.fmt.mpgCRC = strs{3};
Daniel@0: OPTS.fmt.mpgLayer = strs{4};
Daniel@0: OPTS.fmt.mpgOriginal = strs{5};
Daniel@0: OPTS.fmt.mpgChanmode = strs{6};
Daniel@0: OPTS.fmt.mpgPad = strs{7};
Daniel@0: OPTS.fmt.mpgSampsPerFrame = smpspfrm;
Daniel@0: 
Daniel@0: if SR == 16000 && downsamp == 4
Daniel@0:   error('mpg123 will not downsample 16 kHz files by 4 (only 2)');
Daniel@0: end
Daniel@0: 
Daniel@0: if downsamp == 1
Daniel@0:   downsampstr = '';
Daniel@0: else
Daniel@0:   downsampstr = [' -',num2str(downsamp)];
Daniel@0: end
Daniel@0: FS = SR/downsamp;
Daniel@0: 
Daniel@0: if forcemono == 1
Daniel@0:   nchans = 1;
Daniel@0:   chansstr = ' -m';
Daniel@0: else
Daniel@0:   chansstr = '';
Daniel@0: end
Daniel@0: 
Daniel@0: % Size-reading version
Daniel@0: if strcmp(FMT,'size') == 1
Daniel@0:    Y = [floor(smpspfrm*nframes/downsamp)-delay, nchans];
Daniel@0: else
Daniel@0: 
Daniel@0:   % Temporary file to use
Daniel@0:   tmpfile = fullfile(tmpdir, ['tmp',num2str(round(1000*rand(1))),'.wav']);
Daniel@0: 
Daniel@0:   skipx = 0;
Daniel@0:   skipblks = 0;
Daniel@0:   skipstr = '';
Daniel@0:   sttfrm = N(1)-1;
Daniel@0: 
Daniel@0:   % chop off transcoding delay?
Daniel@0:   %sttfrm = sttfrm + delay;  % empirically measured
Daniel@0:   % no, we want to *decode* those samples, then drop them
Daniel@0:   % so delay gets added to skipx instead
Daniel@0:   
Daniel@0:   if sttfrm > 0
Daniel@0:     skipblks = floor(sttfrm*downsamp/smpspfrm);
Daniel@0:     skipx = sttfrm - (skipblks*smpspfrm/downsamp);
Daniel@0:     skipstr = [' -k ', num2str(skipblks)];
Daniel@0:   end
Daniel@0:   skipx = skipx + delay;
Daniel@0:   
Daniel@0:   lenstr = '';
Daniel@0:   endfrm = -1;
Daniel@0:   decblk = 0;
Daniel@0:   if length(N) > 1
Daniel@0:     endfrm = N(2);
Daniel@0:     if endfrm > sttfrm
Daniel@0:       decblk = ceil((endfrm+delay)*downsamp/smpspfrm) - skipblks + 10;   
Daniel@0:       % we read 10 extra blks (+10) to cover the case where up to 10 bad 
Daniel@0:       % blocks are included in the part we are trying to read (it happened)
Daniel@0:       lenstr = [' -n ', num2str(decblk)];
Daniel@0:       % This generates a spurious "Warn: requested..." if reading right 
Daniel@0:       % to the last sample by index (or bad blks), but no matter.
Daniel@0:     end
Daniel@0:  end
Daniel@0: 
Daniel@0:   % Run the decode
Daniel@0:   cmd=['"',mpg123,'"', downsampstr, chansstr, skipstr, lenstr, ...
Daniel@0:        ' -q -w "', tmpfile,'"  "',FILE,'"'];
Daniel@0:   %w = 
Daniel@0:   mysystem(cmd);
Daniel@0: 
Daniel@0:   % Load the data
Daniel@0:   Y = wavread(tmpfile);
Daniel@0: 
Daniel@0: %  % pad delay on to end, just in case
Daniel@0: %  Y = [Y; zeros(delay,size(Y,2))];
Daniel@0: %  % no, the saved file is just longer
Daniel@0:   
Daniel@0:   if decblk > 0 && length(Y) < decblk*smpspfrm/downsamp
Daniel@0:     % This will happen if the selected block range includes >1 bad block
Daniel@0:     disp(['Warn: requested ', num2str(decblk*smpspfrm/downsamp),' frames, returned ',num2str(length(Y))]);
Daniel@0:   end
Daniel@0:   
Daniel@0:   % Delete tmp file
Daniel@0:   mysystem([rmcmd,' "', tmpfile,'"']);
Daniel@0:   
Daniel@0:   % debug
Daniel@0: %  disp(['sttfrm=',num2str(sttfrm),' endfrm=',num2str(endfrm),' skipx=',num2str(skipx),' delay=',num2str(delay),' len=',num2str(length(Y))]);
Daniel@0:   
Daniel@0:   % Select the desired part
Daniel@0:   if skipx+endfrm-sttfrm > length(Y)
Daniel@0:       endfrm = length(Y)+sttfrm-skipx;
Daniel@0:   end
Daniel@0:   
Daniel@0:   if endfrm > sttfrm
Daniel@0:     Y = Y(skipx+(1:(endfrm-sttfrm)),:);
Daniel@0:   elseif skipx > 0
Daniel@0:     Y = Y((skipx+1):end,:);
Daniel@0:   end
Daniel@0:   
Daniel@0:   % Convert to int if format = 'native'
Daniel@0:   if strcmp(FMT,'native')
Daniel@0:     Y = int16((2^15)*Y);
Daniel@0:   end
Daniel@0: 
Daniel@0: end
Daniel@0: 
Daniel@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Daniel@0: function w = mysystem(cmd)
Daniel@0: % Run system command; report error; strip all but last line
Daniel@0: [s,w] = system(cmd);
Daniel@0: if s ~= 0 
Daniel@0:   error(['unable to execute ',cmd,' (',w,')']);
Daniel@0: end
Daniel@0: % Keep just final line
Daniel@0: w = w((1+max([0,findstr(w,10)])):end);
Daniel@0: % Debug
Daniel@0: %disp([cmd,' -> ','*',w,'*']);
Daniel@0: 
Daniel@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Daniel@0: function a = tokenize(s)
Daniel@0: % Break space-separated string into cell array of strings
Daniel@0: % 2004-09-18 dpwe@ee.columbia.edu
Daniel@0: a = [];
Daniel@0: p = 1;
Daniel@0: n = 1;
Daniel@0: l = length(s);
Daniel@0: nss = findstr([s(p:end),' '],' ');
Daniel@0: for ns = nss
Daniel@0:   % Skip initial spaces
Daniel@0:   if ns == p
Daniel@0:     p = p+1;
Daniel@0:   else
Daniel@0:     if p <= l
Daniel@0:       a{n} = s(p:(ns-1));
Daniel@0:       n = n+1;
Daniel@0:       p = ns+1;
Daniel@0:     end
Daniel@0:   end
Daniel@0: end
Daniel@0: