|
wolffd@0
|
1 %% MP3 reading and writing
|
|
wolffd@0
|
2 %
|
|
wolffd@0
|
3 % These function, mp3read and mp3write, aim to exactly duplicate
|
|
wolffd@0
|
4 % the operation of wavread and wavwrite for accessing soundfiles,
|
|
wolffd@0
|
5 % except the soundfiles are in Mpeg-Audio layer 3 (MP3) compressed
|
|
wolffd@0
|
6 % format. All the hard work is done by external binaries written
|
|
wolffd@0
|
7 % by others: mp3info to query the format of existing mp3 files,
|
|
wolffd@0
|
8 % mpg123 to decode mp3 files, and lame to encode audio files.
|
|
wolffd@0
|
9 % Binaries for these files are widely available (and may be
|
|
wolffd@0
|
10 % included in this distribution).
|
|
wolffd@0
|
11 %
|
|
wolffd@0
|
12 % These functions were originally developed for access to very
|
|
wolffd@0
|
13 % large mp3 files (i.e. many hours long), and so avoid creating
|
|
wolffd@0
|
14 % the entire uncompressed audio stream if possible. mp3read
|
|
wolffd@0
|
15 % allows you to specify the range of frames you want to read
|
|
wolffd@0
|
16 % (as a second argument), and mp3read will construct an mpg123
|
|
wolffd@0
|
17 % command that skips blocks to decode only the part of the file
|
|
wolffd@0
|
18 % that is required. This can be much quicker (and require less
|
|
wolffd@0
|
19 % memory/temporary disk) than decoding the whole file.
|
|
wolffd@0
|
20 %
|
|
wolffd@0
|
21 % mpg123 also provides for "on the fly" downsampling at conversion
|
|
wolffd@0
|
22 % to mono, which are supported as extra options in mp3read.
|
|
wolffd@0
|
23 %
|
|
wolffd@0
|
24 % mpg123 can read MP3s across the network. This is supported
|
|
wolffd@0
|
25 % if the FILE argument is a URL (e.g. beginning 'http://...').
|
|
wolffd@0
|
26 %
|
|
wolffd@0
|
27 % mp3info sometimes gets the file size wrong (as returned by the
|
|
wolffd@0
|
28 % mp3read(...'size') syntax). I'm not sure when this happens
|
|
wolffd@0
|
29 % exactly, but it's probably a result of VBR files. In the worst
|
|
wolffd@0
|
30 % case, figuring the number of samples in such a file requires
|
|
wolffd@0
|
31 % scanning through the whole file, and mp3info doesn't usually do
|
|
wolffd@0
|
32 % this.
|
|
wolffd@0
|
33 %
|
|
wolffd@0
|
34 % For more information, including advice on handling MP4 files,
|
|
wolffd@0
|
35 % see http://labrosa.ee.columbia.edu/matlab/mp3read.html
|
|
wolffd@0
|
36
|
|
wolffd@0
|
37 %% Example usage
|
|
wolffd@0
|
38 % Here, we read a wav file in, then write it out as an MP3, then
|
|
wolffd@0
|
39 % read the resulting MP3 back in, and compare it to the original
|
|
wolffd@0
|
40 % file.
|
|
wolffd@0
|
41
|
|
wolffd@0
|
42 % Read an audio waveform
|
|
wolffd@0
|
43 [d,sr] = wavread('piano.wav');
|
|
wolffd@0
|
44 % Save to mp3 (default settings)
|
|
wolffd@0
|
45 mp3write(d,sr,'piano.mp3');
|
|
wolffd@0
|
46 % Read it back again
|
|
wolffd@0
|
47 [d2,sr] = mp3read('piano.mp3');
|
|
wolffd@0
|
48 % mp3 encoding involves some extra padding at each end; we attempt
|
|
wolffd@0
|
49 % to cut it off at the start, but can't do that at the end, because
|
|
wolffd@0
|
50 % mp3read doesn't know how long the original was. But we do, so..
|
|
wolffd@0
|
51 % Chop it down to be the same length as the original
|
|
wolffd@0
|
52 d2 = d2(1:length(d),:);
|
|
wolffd@0
|
53 % What is the SNR (distortion)?
|
|
wolffd@0
|
54 ddiff = d - d2;
|
|
wolffd@0
|
55 disp(['SNR is ',num2str(10*log10(sum(d(:).^2)/sum(ddiff(:).^2))),' dB']);
|
|
wolffd@0
|
56 % Do they look similar?
|
|
wolffd@0
|
57 subplot(211)
|
|
wolffd@0
|
58 specgram(d(:,1),1024,sr);
|
|
wolffd@0
|
59 subplot(212)
|
|
wolffd@0
|
60 plot(1:5000,d(10000+(1:5000),1),1:5000,d2(10000+(1:5000)));
|
|
wolffd@0
|
61 % Yes, pretty close
|
|
wolffd@0
|
62 %
|
|
wolffd@0
|
63 % NB: lame followed by mpg123 causes a little attenuation; you
|
|
wolffd@0
|
64 % can get a better match by scaling up the read-back waveform:
|
|
wolffd@0
|
65 ddiff = d - 1.052*d2;
|
|
wolffd@0
|
66 disp(['SNR is ',num2str(10*log10(sum(d(:).^2)/sum(ddiff(:).^2))),' dB']);
|
|
wolffd@0
|
67
|
|
wolffd@0
|
68 %% Delay, size, and alignment
|
|
wolffd@0
|
69 %
|
|
wolffd@0
|
70 % In mid-2006 I noticed that mp3read followed by mp3write followed by
|
|
wolffd@0
|
71 % mp3read effectively delayed the waveform by 2257 samples (at 44
|
|
wolffd@0
|
72 % kHz). So I introduced code to discard the first 2257 samples to ensure
|
|
wolffd@0
|
73 % that the waveforms remained time aligned. As best I could understand,
|
|
wolffd@0
|
74 % mpg123 (v 0.5.9) was including the "warm-up" samples from the
|
|
wolffd@0
|
75 % synthesis filterbank which are more properly discarded.
|
|
wolffd@0
|
76 %
|
|
wolffd@0
|
77 % Then in late 2009 I noticed that some chord recognition code, which
|
|
wolffd@0
|
78 % used mp3read to read files which were then segmented on the basis of
|
|
wolffd@0
|
79 % some hand-marked timings, suddenly started getting much poorer
|
|
wolffd@0
|
80 % results. It turned out that I had upgraded my version of mpg123 to v
|
|
wolffd@0
|
81 % 1.9.0, and the warm-up samples had been fixed in this version. So my
|
|
wolffd@0
|
82 % code was discarding 2257 *good* samples, and the data was skewed 51ms
|
|
wolffd@0
|
83 % early relative to the hand labels.
|
|
wolffd@0
|
84 %
|
|
wolffd@0
|
85 % Hence, the current version of mp3read does not
|
|
wolffd@0
|
86 % discard any samples by default -- appropriate for the recent versions
|
|
wolffd@0
|
87 % of mpg123 included here. But if you know you're running an old, v
|
|
wolffd@0
|
88 % 0.5.9, mpg123, you should edit the mp3read.m source to set the flag
|
|
wolffd@0
|
89 % MPG123059 = 1.
|
|
wolffd@0
|
90 %
|
|
wolffd@0
|
91 % Note also that the 'size' function relies on the number of
|
|
wolffd@0
|
92 % blocks reported by mp3info. However, many mp3 files include
|
|
wolffd@0
|
93 % additional information about the size of the file in the
|
|
wolffd@0
|
94 % so-called Xing header, embedded in the first frame, which can
|
|
wolffd@0
|
95 % specify that a certain number of samples from start and end
|
|
wolffd@0
|
96 % should additionally be dropped. mp3info doesn't read that,
|
|
wolffd@0
|
97 % and there's no way for my code to probe it except by running
|
|
wolffd@0
|
98 % mpg123. Hence, the results of mp3read(fn,'size') may sometimes
|
|
wolffd@0
|
99 % overestimate the length of the actual vector you'll get if
|
|
wolffd@0
|
100 % you read the whole file.
|
|
wolffd@0
|
101
|
|
wolffd@0
|
102 %% External binaries
|
|
wolffd@0
|
103 % The m files rely on three external binaries, each of which is
|
|
wolffd@0
|
104 % available for Linux, Mac OS X, or Windows:
|
|
wolffd@0
|
105 %
|
|
wolffd@0
|
106 % *mpg123* is a high-performance mp3 decoder. Its home page is
|
|
wolffd@0
|
107 % http://www.mpg123.de/ .
|
|
wolffd@0
|
108 %
|
|
wolffd@0
|
109 % *mp3info* is a utility to read technical information on an mp3
|
|
wolffd@0
|
110 % file. Its home page is http://www.ibiblio.org/mp3info/ .
|
|
wolffd@0
|
111 %
|
|
wolffd@0
|
112 % *lame* is an open-source MP3 encoder. Its homepage is
|
|
wolffd@0
|
113 % http://lame.sourceforge.net/ .
|
|
wolffd@0
|
114 %
|
|
wolffd@0
|
115 % The various authors of these packages are gratefully acknowledged
|
|
wolffd@0
|
116 % for doing all the hard work to make these Matlab functions possible.
|
|
wolffd@0
|
117
|
|
wolffd@0
|
118 %% Installation
|
|
wolffd@0
|
119 % The two routines, mp3read.m and mp3write.m, will look for their
|
|
wolffd@0
|
120 % binaries (mpg123 and mp3info for mp3read; lame for mp3write) in
|
|
wolffd@0
|
121 % the same directory where they are installed. Binaries for
|
|
wolffd@0
|
122 % different architectures are distinguished by their extension,
|
|
wolffd@0
|
123 % which is the standard Matlab computer code e.g. ".mac" for Mac
|
|
wolffd@0
|
124 % PPC OS X, ".glnx86" for i386-linux. The exception is Windows,
|
|
wolffd@0
|
125 % where the binaries have the extension ".exe".
|
|
wolffd@0
|
126 %
|
|
wolffd@0
|
127 % Temporary files
|
|
wolffd@0
|
128 % will be written to (a) a directory taken from the environment
|
|
wolffd@0
|
129 % variable TMPDIR (b) /tmp if it exists, or (c) the current
|
|
wolffd@0
|
130 % directory. This can easily be changed by editing the m files.
|
|
wolffd@0
|
131
|
|
wolffd@0
|
132 % Last updated: $Date: 2009/03/15 18:29:58 $
|
|
wolffd@0
|
133 % Dan Ellis <dpwe@ee.columbia.edu>
|
