wolffd@0: function [Y,FS,NBITS,OPTS] = mp3read(FILE,N,MONO,DOWNSAMP,DELAY)
wolffd@0: % MP3READ   Read MP3 audio file via use of external binaries.
wolffd@0: %   Y = MP3READ(FILE) reads an mp3-encoded audio file into the
wolffd@0: %     vector Y just like wavread reads a wav-encoded file (one channel 
wolffd@0: %     per column).  Extension ".mp3" is added if FILE has none.
wolffd@0: %     Also accepts other formats of wavread, such as
wolffd@0: %   Y = MP3READ(FILE,N) to read just the first N sample frames (N
wolffd@0: %     scalar), or the frames from N(1) to N(2) if N is a two-element vector.  
wolffd@0: %   Y = MP3READ(FILE,FMT) or Y = mp3read(FILE,N,FMT) 
wolffd@0: %     with FMT as 'native' returns int16 samples instead of doubles; 
wolffd@0: %     FMT can be 'double' for default behavior (to exactly mirror the
wolffd@0: %     syntax of wavread).
wolffd@0: %
wolffd@0: %   [Y,FS,NBITS,OPTS] = MP3READ(FILE...) returns extra information:
wolffd@0: %     FS is the sampling rate,  NBITS is the bit depth (always 16), 
wolffd@0: %     OPTS.fmt is a format info string; OPTS has multiple other
wolffd@0: %     fields, see WAVREAD.
wolffd@0: %
wolffd@0: %   SIZ = MP3READ(FILE,'size') returns the size of the audio data contained
wolffd@0: %     in the file in place of the actual audio data, returning the
wolffd@0: %     2-element vector SIZ=[samples channels].
wolffd@0: %
wolffd@0: %   [Y...] = MP3READ(FILE,N,MONO,DOWNSAMP,DELAY) extends the
wolffd@0: %     WAVREAD syntax to allow access to special features of the
wolffd@0: %     mpg123 engine:  MONO = 1 forces output to be mono (by
wolffd@0: %     averaging stereo channels); DOWNSAMP = 2 or 4 downsamples by 
wolffd@0: %     a factor of 2 or 4 (thus FS returns as 22050 or 11025
wolffd@0: %     respectively for a 44 kHz mp3 file); 
wolffd@0: %     To accommodate a bug in mpg123-0.59, DELAY controls how many
wolffd@0: %     "warm up" samples to drop at the start of the file; the
wolffd@0: %     default value of 2257 makes an mp3write/mp3read loop for a 44
wolffd@0: %     kHz mp3 file be as close as possible to being temporally
wolffd@0: %     aligned; specify as 0 to prevent discard of initial samples.
wolffd@0: %     For later versions of mpg123 (e.g. 1.9.0) this is not needed; 
wolffd@0: %     a flag in mp3read.m makes the default DELAY zero in this case.
wolffd@0: %
wolffd@0: %   [Y...] = MP3READ(URL...)  uses the built-in network
wolffd@0: %     functionality of mpg123 to read an MP3 file across the
wolffd@0: %     network.  URL must be of the form 'http://...' or
wolffd@0: %     'ftp://...'.  'size' and OPTS are not available in this mode.
wolffd@0: %
wolffd@0: %   Example:
wolffd@0: %   To read an mp3 file as doubles at its original width and sampling rate:
wolffd@0: %     [Y,FS] = mp3read('piano.mp3');
wolffd@0: %   To read the first 1 second of the same file, downsampled by a
wolffd@0: %   factor of 4, cast to mono, using the default filename
wolffd@0: %   extension:
wolffd@0: %     [Y,FS4] = mp3read('piano', FS/4, 1, 4);
wolffd@0: %
wolffd@0: %   Note: Because the mp3 format encodes samples in blocks of 26 ms (at
wolffd@0: %   44 kHz), and because of the "warm up" period of the encoder,
wolffd@0: %   the file length may not be exactly what you expect, depending 
wolffd@0: %   on your version of mpg123 (recent versions fix warmup).
wolffd@0: %
wolffd@0: %   Note: requires external binaries mpg123 and mp3info; you
wolffd@0: %   can find binaries for several platforms at:
wolffd@0: %     http://labrosa.ee.columbia.edu/matlab/mp3read.html
wolffd@0: %
wolffd@0: %   See also mp3write, wavread.
wolffd@0: 
wolffd@0: % $Header: /Users/dpwe/matlab/columbiafns/RCS/mp3read.m,v 1.7 2010/04/09 18:13:00 dpwe Exp dpwe $
wolffd@0: 
wolffd@0: % 2003-07-20 dpwe@ee.columbia.edu  This version calls mpg123.
wolffd@0: % 2004-08-31 Fixed to read whole files correctly
wolffd@0: % 2004-09-08 Uses mp3info to get info about mp3 files too
wolffd@0: % 2004-09-18 Reports all mp3info fields in OPTS.fmt; handles MPG2LSF sizes
wolffd@0: %            + added MONO, DOWNSAMP flags, changed default behavior.
wolffd@0: % 2005-09-28 Fixed bug reading full-rate stereo as 1ch (thx bjoerns@vjk.dk)
wolffd@0: % 2006-09-17 Chop off initial 2257 sample delay (for 44.1 kHz mp3)
wolffd@0: %            so read-write loop doesn't get progressively delayed.
wolffd@0: %            You can suppress this with a 5th argument of 0.
wolffd@0: % 2007-02-04 Added support for FMT argument to match wavread
wolffd@0: %            Added automatic selection of binary etc. to allow it
wolffd@0: %            to work cross-platform without editing prior to
wolffd@0: %            submitting to Matlab File Exchange
wolffd@0: % 2007-07-23 Tweaks to 'size' mode so it exactly agrees with read data.
wolffd@0: % 2009-03-15 Added fixes so 'http://...' file URLs will work.
wolffd@0: % 2009-03-26 Added filename length check to http: test (thx fabricio guzman)
wolffd@0: 
wolffd@0: % find our baseline directory
wolffd@0: path = fileparts(which('mp3read'));
wolffd@0: 
wolffd@0: % %%%%% Directory for temporary file (if needed)
wolffd@0: % % Try to read from environment, or use /tmp if it exists, or use CWD
wolffd@0: tmpdir = getenv('TMPDIR');
wolffd@0: if isempty(tmpdir) || exist(tmpdir,'file')==0
wolffd@0:   tmpdir = '/tmp';
wolffd@0: end
wolffd@0: if exist(tmpdir,'file')==0
wolffd@0:   tmpdir = '';
wolffd@0: end
wolffd@0: % ensure it exists
wolffd@0: %if length(tmpdir) > 0 && exist(tmpdir,'file')==0
wolffd@0: %  mkdir(tmpdir);
wolffd@0: %end
wolffd@0: 
wolffd@0: %%%%%% Command to delete temporary file (if needed)
wolffd@0: rmcmd = 'rm';
wolffd@0: 
wolffd@0: %%%%%% Location of the binaries - attempt to choose automatically
wolffd@0: %%%%%% (or edit to be hard-coded for your installation)
wolffd@0: ext = lower(computer);
wolffd@0: if ispc
wolffd@0:   ext = 'exe';
wolffd@0:   rmcmd = 'del';
wolffd@0: end
wolffd@0: % mpg123-0.59 inserts silence at the start of decoded files, which
wolffd@0: % we compensate.  However, this is fixed in mpg123-1.9.0, so 
wolffd@0: % make this flag 1 only if you have mpg123-0.5.9
wolffd@0: MPG123059 = 0;
wolffd@0: mpg123 = fullfile(path,['mpg123.',ext]);
wolffd@0: mp3info = fullfile(path,['mp3info.',ext]);
wolffd@0: 
wolffd@0: %%%%% Check for network mode
wolffd@0: if length(FILE) > 6 && (strcmp(lower(FILE(1:7)),'http://') == 1 ...
wolffd@0:       || strcmp(lower(FILE(1:6)),'ftp://'))
wolffd@0:   % mp3info not available over network
wolffd@0:   OVERNET = 1;
wolffd@0: else
wolffd@0:   OVERNET = 0;
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%% Process input arguments
wolffd@0: if nargin < 2
wolffd@0:   N = 0;
wolffd@0: end
wolffd@0: 
wolffd@0: % Check for FMT spec (per wavread)
wolffd@0: FMT = 'double';
wolffd@0: if ischar(N)
wolffd@0:   FMT = lower(N);
wolffd@0:   N = 0;
wolffd@0: end
wolffd@0: 
wolffd@0: if length(N) == 1
wolffd@0:   % Specified N was upper limit
wolffd@0:   N = [1 N];
wolffd@0: end
wolffd@0: if nargin < 3
wolffd@0:   forcemono = 0;
wolffd@0: else
wolffd@0:   % Check for 3rd arg as FMT
wolffd@0:   if ischar(MONO)
wolffd@0:     FMT = lower(MONO);
wolffd@0:     MONO = 0;
wolffd@0:   end
wolffd@0:   forcemono = (MONO ~= 0);
wolffd@0: end
wolffd@0: if nargin < 4
wolffd@0:   downsamp = 1;
wolffd@0: else
wolffd@0:   downsamp = DOWNSAMP;
wolffd@0: end
wolffd@0: if downsamp ~= 1 && downsamp ~= 2 && downsamp ~= 4
wolffd@0:   error('DOWNSAMP can only be 1, 2, or 4');
wolffd@0: end
wolffd@0: 
wolffd@0: % process DELAY option (nargin 5) after we've read the SR
wolffd@0: 
wolffd@0: if strcmp(FMT,'native') == 0 && strcmp(FMT,'double') == 0 && ...
wolffd@0:       strcmp(FMT,'size') == 0
wolffd@0:   error(['FMT must be ''native'' or ''double'' (or ''size''), not ''',FMT,'''']);
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%% Constants
wolffd@0: NBITS=16;
wolffd@0: 
wolffd@0: %%%%% add extension if none (like wavread)
wolffd@0: [path,file,ext] = fileparts(FILE);
wolffd@0: if isempty(ext)
wolffd@0:   FILE = [FILE, '.mp3'];
wolffd@0: end
wolffd@0: 
wolffd@0: %%%%% maybe expand ~ %%%%%%
wolffd@0: if FILE(1) == '~'
wolffd@0:   FILE = [getenv('HOME'),FILE(2:end)];
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: if ~OVERNET
wolffd@0:   %%%%%% Probe file to find format, size, etc. using "mp3info" utility
wolffd@0:   cmd = ['"',mp3info, '" -r m -p "%Q %u %b %r %v * %C %e %E %L %O %o %p" "', FILE,'"'];
wolffd@0:   % Q = samprate, u = #frames, b = #badframes (needed to get right answer from %u) 
wolffd@0:   % r = bitrate, v = mpeg version (1/2/2.5)
wolffd@0:   % C = Copyright, e = emph, E = CRC, L = layer, O = orig, o = mono, p = pad
wolffd@0:   w = mysystem(cmd);
wolffd@0:   % Break into numerical and ascii parts by finding the delimiter we put in
wolffd@0:   starpos = findstr(w,'*');
wolffd@0:   nums = str2num(w(1:(starpos - 2)));
wolffd@0:   strs = tokenize(w((starpos+2):end));
wolffd@0: 
wolffd@0:   SR = nums(1);
wolffd@0:   nframes = nums(2);
wolffd@0:   nchans = 2 - strcmp(strs{6}, 'mono');
wolffd@0:   layer = length(strs{4});
wolffd@0:   bitrate = nums(4)*1000;
wolffd@0:   mpgv = nums(5);
wolffd@0:   % Figure samples per frame, after
wolffd@0:   % http://board.mp3-tech.org/view.php3?bn=agora_mp3techorg&key=1019510889
wolffd@0:   if layer == 1
wolffd@0:     smpspfrm = 384;
wolffd@0:   elseif SR < 32000 && layer ==3
wolffd@0:     smpspfrm = 576;
wolffd@0:     if mpgv == 1
wolffd@0:       error('SR < 32000 but mpeg version = 1');
wolffd@0:     end
wolffd@0:   else
wolffd@0:     smpspfrm = 1152;
wolffd@0:   end
wolffd@0: 
wolffd@0:   OPTS.fmt.mpgBitrate = bitrate;
wolffd@0:   OPTS.fmt.mpgVersion = mpgv;
wolffd@0:   % fields from wavread's OPTS
wolffd@0:   OPTS.fmt.nAvgBytesPerSec = bitrate/8;
wolffd@0:   OPTS.fmt.nSamplesPerSec = SR;
wolffd@0:   OPTS.fmt.nChannels = nchans;
wolffd@0:   OPTS.fmt.nBlockAlign = smpspfrm/SR*bitrate/8;
wolffd@0:   OPTS.fmt.nBitsPerSample = NBITS;
wolffd@0:   OPTS.fmt.mpgNFrames = nframes;
wolffd@0:   OPTS.fmt.mpgCopyright = strs{1};
wolffd@0:   OPTS.fmt.mpgEmphasis = strs{2};
wolffd@0:   OPTS.fmt.mpgCRC = strs{3};
wolffd@0:   OPTS.fmt.mpgLayer = strs{4};
wolffd@0:   OPTS.fmt.mpgOriginal = strs{5};
wolffd@0:   OPTS.fmt.mpgChanmode = strs{6};
wolffd@0:   OPTS.fmt.mpgPad = strs{7};
wolffd@0:   OPTS.fmt.mpgSampsPerFrame = smpspfrm;
wolffd@0: else
wolffd@0:   % OVERNET mode
wolffd@0:   OPTS = [];
wolffd@0:   % guesses
wolffd@0:   smpspfrm = 1152;
wolffd@0:   SR = 44100;
wolffd@0:   nframes = 0;
wolffd@0: end
wolffd@0:   
wolffd@0: if SR == 16000 && downsamp == 4
wolffd@0:   error('mpg123 will not downsample 16 kHz files by 4 (only 2)');
wolffd@0: end
wolffd@0: 
wolffd@0: % from libmpg123/frame.h
wolffd@0: GAPLESS_DELAY = 529;
wolffd@0: 
wolffd@0: % process or set delay
wolffd@0: if nargin < 5
wolffd@0: 
wolffd@0:   if MPG123059
wolffd@0:     mpg123delay44kHz = 2257;  % empirical delay of lame/mpg123 loop
wolffd@0:     mpg123delay16kHz = 1105;  % empirical delay of lame/mpg123 loop
wolffd@0:                               % for 16 kHz sampling - one 1152
wolffd@0:                               % sample frame less??
wolffd@0:     if SR == 16000
wolffd@0:       rawdelay = mpg123delay16kHz;
wolffd@0:     else
wolffd@0:       rawdelay = mpg123delay44kHz;  % until we know better
wolffd@0:     end
wolffd@0:     delay = round(rawdelay/downsamp);
wolffd@0:   else
wolffd@0:     % seems like predelay is fixed in mpg123-1.9.0
wolffd@0:     delay = 0;
wolffd@0:   end
wolffd@0: else
wolffd@0:   delay = DELAY;
wolffd@0: end
wolffd@0: 
wolffd@0: if downsamp == 1
wolffd@0:   downsampstr = '';
wolffd@0: else
wolffd@0:   downsampstr = [' -',num2str(downsamp)];
wolffd@0: end
wolffd@0: FS = SR/downsamp;
wolffd@0: 
wolffd@0: if forcemono == 1
wolffd@0:   nchans = 1;
wolffd@0:   chansstr = ' -m';
wolffd@0: else
wolffd@0:   chansstr = '';
wolffd@0: end
wolffd@0: 
wolffd@0: % Size-reading version
wolffd@0: if strcmp(FMT,'size') == 1
wolffd@0:   if MPG123059
wolffd@0:     Y = [floor(smpspfrm*nframes/downsamp)-delay, nchans];
wolffd@0:   else
wolffd@0:     Y = [floor(smpspfrm*nframes/downsamp)-GAPLESS_DELAY, nchans];
wolffd@0:   end    
wolffd@0: else
wolffd@0: 
wolffd@0:   % Temporary file to use
wolffd@0:   tmpfile = fullfile(tmpdir, ['tmp',num2str(round(1000*rand(1))),'.wav']);
wolffd@0: 
wolffd@0:   skipx = 0;
wolffd@0:   skipblks = 0;
wolffd@0:   skipstr = '';
wolffd@0:   sttfrm = N(1)-1;
wolffd@0: 
wolffd@0:   % chop off transcoding delay?
wolffd@0:   %sttfrm = sttfrm + delay;  % empirically measured
wolffd@0:   % no, we want to *decode* those samples, then drop them
wolffd@0:   % so delay gets added to skipx instead
wolffd@0:   
wolffd@0:   if sttfrm > 0
wolffd@0:     skipblks = floor(sttfrm*downsamp/smpspfrm);
wolffd@0:     skipx = sttfrm - (skipblks*smpspfrm/downsamp);
wolffd@0:     skipstr = [' -k ', num2str(skipblks)];
wolffd@0:   end
wolffd@0:   skipx = skipx + delay;
wolffd@0:   
wolffd@0:   lenstr = '';
wolffd@0:   endfrm = -1;
wolffd@0:   decblk = 0;
wolffd@0:   if length(N) > 1
wolffd@0:     endfrm = N(2);
wolffd@0:     if endfrm > sttfrm
wolffd@0:       decblk = ceil((endfrm+delay)*downsamp/smpspfrm) - skipblks + 10;   
wolffd@0:       % we read 10 extra blks (+10) to cover the case where up to 10 bad 
wolffd@0:       % blocks are included in the part we are trying to read (it happened)
wolffd@0:       lenstr = [' -n ', num2str(decblk)];
wolffd@0:       % This generates a spurious "Warn: requested..." if reading right 
wolffd@0:       % to the last sample by index (or bad blks), but no matter.
wolffd@0:     end
wolffd@0:  end
wolffd@0: 
wolffd@0:   % Run the decode
wolffd@0:   cmd=['"',mpg123,'"', downsampstr, chansstr, skipstr, lenstr, ...
wolffd@0:        ' -q -w "', tmpfile,'"  "',FILE,'"'];
wolffd@0:   %w = 
wolffd@0:   mysystem(cmd);
wolffd@0: 
wolffd@0:   % Load the data (may update FS if it was based on a guess previously)
wolffd@0:   [Y,FS] = wavread(tmpfile);
wolffd@0: 
wolffd@0: %  % pad delay on to end, just in case
wolffd@0: %  Y = [Y; zeros(delay,size(Y,2))];
wolffd@0: %  % no, the saved file is just longer
wolffd@0:   
wolffd@0:   if decblk > 0 && length(Y) < decblk*smpspfrm/downsamp
wolffd@0:     % This will happen if the selected block range includes >1 bad block
wolffd@0:     disp(['Warn: requested ', num2str(decblk*smpspfrm/downsamp),' frames, returned ',num2str(length(Y))]);
wolffd@0:   end
wolffd@0:   
wolffd@0:   % Delete tmp file
wolffd@0:   mysystem([rmcmd,' "', tmpfile,'"']);
wolffd@0:   
wolffd@0:   % debug
wolffd@0: %  disp(['sttfrm=',num2str(sttfrm),' endfrm=',num2str(endfrm),' skipx=',num2str(skipx),' delay=',num2str(delay),' len=',num2str(length(Y))]);
wolffd@0:   
wolffd@0:   % Select the desired part
wolffd@0:   if skipx+endfrm-sttfrm > length(Y)
wolffd@0:       endfrm = length(Y)+sttfrm-skipx;
wolffd@0:   end
wolffd@0:   
wolffd@0:   if endfrm > sttfrm
wolffd@0:     Y = Y(skipx+(1:(endfrm-sttfrm)),:);
wolffd@0:   elseif skipx > 0
wolffd@0:     Y = Y((skipx+1):end,:);
wolffd@0:   end
wolffd@0:   
wolffd@0:   % Convert to int if format = 'native'
wolffd@0:   if strcmp(FMT,'native')
wolffd@0:     Y = int16((2^15)*Y);
wolffd@0:   end
wolffd@0: 
wolffd@0: end
wolffd@0: 
wolffd@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0: function w = mysystem(cmd)
wolffd@0: % Run system command; report error; strip all but last line
wolffd@0: [s,w] = system(cmd);
wolffd@0: if s ~= 0 
wolffd@0:   error(['unable to execute ',cmd,' (',w,')']);
wolffd@0: end
wolffd@0: % Keep just final line
wolffd@0: w = w((1+max([0,findstr(w,10)])):end);
wolffd@0: % Debug
wolffd@0: %disp([cmd,' -> ','*',w,'*']);
wolffd@0: 
wolffd@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0: function a = tokenize(s,t)
wolffd@0: % Break space-separated string into cell array of strings.
wolffd@0: % Optional second arg gives alternate separator (default ' ')
wolffd@0: % 2004-09-18 dpwe@ee.columbia.edu
wolffd@0: if nargin < 2;  t = ' '; end
wolffd@0: a = [];
wolffd@0: p = 1;
wolffd@0: n = 1;
wolffd@0: l = length(s);
wolffd@0: nss = findstr([s(p:end),t],t);
wolffd@0: for ns = nss
wolffd@0:   % Skip initial spaces (separators)
wolffd@0:   if ns == p
wolffd@0:     p = p+1;
wolffd@0:   else
wolffd@0:     if p <= l
wolffd@0:       a{n} = s(p:(ns-1));
wolffd@0:       n = n+1;
wolffd@0:       p = ns+1;
wolffd@0:     end
wolffd@0:   end
wolffd@0: end
wolffd@0: