blob: ce127aa06a61b1d965033c6c300d6bfdbe555d20 [file] [log] [blame]
Georg Brandl116aa622007-08-15 14:28:22 +00001:mod:`audioop` --- Manipulate raw audio data
2============================================
3
4.. module:: audioop
5 :synopsis: Manipulate raw audio data.
6
7
8The :mod:`audioop` module contains some useful operations on sound fragments.
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +03009It operates on sound fragments consisting of signed integer samples 8, 16, 24
Serhiy Storchaka711e91b2013-11-10 21:44:36 +020010or 32 bits wide, stored in :term:`bytes-like object`\ s. All scalar items are
11integers, unless specified otherwise.
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +030012
13.. versionchanged:: 3.4
14 Support for 24-bit samples was added.
Larry Hastings3732ed22014-03-15 21:13:56 -070015 All functions now accept any :term:`bytes-like object`.
16 String input now results in an immediate error.
Serhiy Storchaka711e91b2013-11-10 21:44:36 +020017
Georg Brandl116aa622007-08-15 14:28:22 +000018.. index::
19 single: Intel/DVI ADPCM
20 single: ADPCM, Intel/DVI
21 single: a-LAW
22 single: u-LAW
23
24This module provides support for a-LAW, u-LAW and Intel/DVI ADPCM encodings.
25
Christian Heimes5b5e81c2007-12-31 16:14:33 +000026.. This para is mostly here to provide an excuse for the index entries...
Georg Brandl116aa622007-08-15 14:28:22 +000027
28A few of the more complicated operations only take 16-bit samples, otherwise the
29sample size (in bytes) is always a parameter of the operation.
30
31The module defines the following variables and functions:
32
33
34.. exception:: error
35
36 This exception is raised on all errors, such as unknown number of bytes per
37 sample, etc.
38
39
40.. function:: add(fragment1, fragment2, width)
41
42 Return a fragment which is the addition of the two samples passed as parameters.
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +030043 *width* is the sample width in bytes, either ``1``, ``2``, ``3`` or ``4``. Both
Serhiy Storchaka01ad6222013-02-09 11:10:53 +020044 fragments should have the same length. Samples are truncated in case of overflow.
Georg Brandl116aa622007-08-15 14:28:22 +000045
46
47.. function:: adpcm2lin(adpcmfragment, width, state)
48
49 Decode an Intel/DVI ADPCM coded fragment to a linear fragment. See the
50 description of :func:`lin2adpcm` for details on ADPCM coding. Return a tuple
51 ``(sample, newstate)`` where the sample has the width specified in *width*.
52
53
54.. function:: alaw2lin(fragment, width)
55
56 Convert sound fragments in a-LAW encoding to linearly encoded sound fragments.
57 a-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
58 width of the output fragment here.
59
Georg Brandl116aa622007-08-15 14:28:22 +000060
61.. function:: avg(fragment, width)
62
63 Return the average over all samples in the fragment.
64
65
66.. function:: avgpp(fragment, width)
67
68 Return the average peak-peak value over all samples in the fragment. No
69 filtering is done, so the usefulness of this routine is questionable.
70
71
72.. function:: bias(fragment, width, bias)
73
74 Return a fragment that is the original fragment with a bias added to each
Serhiy Storchaka01ad6222013-02-09 11:10:53 +020075 sample. Samples wrap around in case of overflow.
Georg Brandl116aa622007-08-15 14:28:22 +000076
77
Serhiy Storchaka3062c9a2013-11-23 22:26:01 +020078.. function:: byteswap(fragment, width)
79
80 "Byteswap" all samples in a fragment and returns the modified fragment.
81 Converts big-endian samples to little-endian and vice versa.
82
Larry Hastings3732ed22014-03-15 21:13:56 -070083 .. versionadded:: 3.4
Serhiy Storchaka3062c9a2013-11-23 22:26:01 +020084
85
Georg Brandl116aa622007-08-15 14:28:22 +000086.. function:: cross(fragment, width)
87
88 Return the number of zero crossings in the fragment passed as an argument.
89
90
91.. function:: findfactor(fragment, reference)
92
93 Return a factor *F* such that ``rms(add(fragment, mul(reference, -F)))`` is
94 minimal, i.e., return the factor with which you should multiply *reference* to
95 make it match as well as possible to *fragment*. The fragments should both
96 contain 2-byte samples.
97
98 The time taken by this routine is proportional to ``len(fragment)``.
99
100
101.. function:: findfit(fragment, reference)
102
103 Try to match *reference* as well as possible to a portion of *fragment* (which
104 should be the longer fragment). This is (conceptually) done by taking slices
105 out of *fragment*, using :func:`findfactor` to compute the best match, and
106 minimizing the result. The fragments should both contain 2-byte samples.
107 Return a tuple ``(offset, factor)`` where *offset* is the (integer) offset into
108 *fragment* where the optimal match started and *factor* is the (floating-point)
109 factor as per :func:`findfactor`.
110
111
112.. function:: findmax(fragment, length)
113
114 Search *fragment* for a slice of length *length* samples (not bytes!) with
115 maximum energy, i.e., return *i* for which ``rms(fragment[i*2:(i+length)*2])``
116 is maximal. The fragments should both contain 2-byte samples.
117
118 The routine takes time proportional to ``len(fragment)``.
119
120
121.. function:: getsample(fragment, width, index)
122
123 Return the value of sample *index* from the fragment.
124
125
126.. function:: lin2adpcm(fragment, width, state)
127
128 Convert samples to 4 bit Intel/DVI ADPCM encoding. ADPCM coding is an adaptive
129 coding scheme, whereby each 4 bit number is the difference between one sample
130 and the next, divided by a (varying) step. The Intel/DVI ADPCM algorithm has
131 been selected for use by the IMA, so it may well become a standard.
132
133 *state* is a tuple containing the state of the coder. The coder returns a tuple
134 ``(adpcmfrag, newstate)``, and the *newstate* should be passed to the next call
135 of :func:`lin2adpcm`. In the initial call, ``None`` can be passed as the state.
136 *adpcmfrag* is the ADPCM coded fragment packed 2 4-bit values per byte.
137
138
139.. function:: lin2alaw(fragment, width)
140
141 Convert samples in the audio fragment to a-LAW encoding and return this as a
Serhiy Storchakac8bd74d2012-12-27 20:43:36 +0200142 bytes object. a-LAW is an audio encoding format whereby you get a dynamic
Georg Brandl116aa622007-08-15 14:28:22 +0000143 range of about 13 bits using only 8 bit samples. It is used by the Sun audio
144 hardware, among others.
145
Georg Brandl116aa622007-08-15 14:28:22 +0000146
147.. function:: lin2lin(fragment, width, newwidth)
148
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +0300149 Convert samples between 1-, 2-, 3- and 4-byte formats.
Georg Brandl116aa622007-08-15 14:28:22 +0000150
Christian Heimescc47b052008-03-25 14:56:36 +0000151 .. note::
152
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +0300153 In some audio formats, such as .WAV files, 16, 24 and 32 bit samples are
Christian Heimescc47b052008-03-25 14:56:36 +0000154 signed, but 8 bit samples are unsigned. So when converting to 8 bit wide
155 samples for these formats, you need to also add 128 to the result::
156
157 new_frames = audioop.lin2lin(frames, old_width, 1)
158 new_frames = audioop.bias(new_frames, 1, 128)
159
Serhiy Storchakaeaea5e92013-10-19 21:10:46 +0300160 The same, in reverse, has to be applied when converting from 8 to 16, 24
161 or 32 bit width samples.
Christian Heimescc47b052008-03-25 14:56:36 +0000162
Georg Brandl116aa622007-08-15 14:28:22 +0000163
164.. function:: lin2ulaw(fragment, width)
165
166 Convert samples in the audio fragment to u-LAW encoding and return this as a
Serhiy Storchakac8bd74d2012-12-27 20:43:36 +0200167 bytes object. u-LAW is an audio encoding format whereby you get a dynamic
Georg Brandl116aa622007-08-15 14:28:22 +0000168 range of about 14 bits using only 8 bit samples. It is used by the Sun audio
169 hardware, among others.
170
171
Georg Brandl116aa622007-08-15 14:28:22 +0000172.. function:: max(fragment, width)
173
174 Return the maximum of the *absolute value* of all samples in a fragment.
175
176
177.. function:: maxpp(fragment, width)
178
179 Return the maximum peak-peak value in the sound fragment.
180
181
Ezio Melottie0035a22012-12-14 20:18:46 +0200182.. function:: minmax(fragment, width)
183
184 Return a tuple consisting of the minimum and maximum values of all samples in
185 the sound fragment.
186
187
Georg Brandl116aa622007-08-15 14:28:22 +0000188.. function:: mul(fragment, width, factor)
189
190 Return a fragment that has all samples in the original fragment multiplied by
Serhiy Storchaka01ad6222013-02-09 11:10:53 +0200191 the floating-point value *factor*. Samples are truncated in case of overflow.
Georg Brandl116aa622007-08-15 14:28:22 +0000192
193
194.. function:: ratecv(fragment, width, nchannels, inrate, outrate, state[, weightA[, weightB]])
195
196 Convert the frame rate of the input fragment.
197
198 *state* is a tuple containing the state of the converter. The converter returns
199 a tuple ``(newfragment, newstate)``, and *newstate* should be passed to the next
200 call of :func:`ratecv`. The initial call should pass ``None`` as the state.
201
202 The *weightA* and *weightB* arguments are parameters for a simple digital filter
203 and default to ``1`` and ``0`` respectively.
204
205
206.. function:: reverse(fragment, width)
207
208 Reverse the samples in a fragment and returns the modified fragment.
209
210
211.. function:: rms(fragment, width)
212
213 Return the root-mean-square of the fragment, i.e. ``sqrt(sum(S_i^2)/n)``.
214
215 This is a measure of the power in an audio signal.
216
217
218.. function:: tomono(fragment, width, lfactor, rfactor)
219
220 Convert a stereo fragment to a mono fragment. The left channel is multiplied by
221 *lfactor* and the right channel by *rfactor* before adding the two channels to
222 give a mono signal.
223
224
225.. function:: tostereo(fragment, width, lfactor, rfactor)
226
227 Generate a stereo fragment from a mono fragment. Each pair of samples in the
228 stereo fragment are computed from the mono sample, whereby left channel samples
229 are multiplied by *lfactor* and right channel samples by *rfactor*.
230
231
232.. function:: ulaw2lin(fragment, width)
233
234 Convert sound fragments in u-LAW encoding to linearly encoded sound fragments.
235 u-LAW encoding always uses 8 bits samples, so *width* refers only to the sample
236 width of the output fragment here.
237
Georg Brandl502d9a52009-07-26 15:02:41 +0000238Note that operations such as :func:`.mul` or :func:`.max` make no distinction
Georg Brandl116aa622007-08-15 14:28:22 +0000239between mono and stereo fragments, i.e. all samples are treated equal. If this
240is a problem the stereo fragment should be split into two mono fragments first
241and recombined later. Here is an example of how to do that::
242
243 def mul_stereo(sample, width, lfactor, rfactor):
244 lsample = audioop.tomono(sample, width, 1, 0)
245 rsample = audioop.tomono(sample, width, 0, 1)
Georg Brandlf3d00872010-10-17 10:07:29 +0000246 lsample = audioop.mul(lsample, width, lfactor)
247 rsample = audioop.mul(rsample, width, rfactor)
Georg Brandl116aa622007-08-15 14:28:22 +0000248 lsample = audioop.tostereo(lsample, width, 1, 0)
249 rsample = audioop.tostereo(rsample, width, 0, 1)
250 return audioop.add(lsample, rsample, width)
251
252If you use the ADPCM coder to build network packets and you want your protocol
253to be stateless (i.e. to be able to tolerate packet loss) you should not only
254transmit the data but also the state. Note that you should send the *initial*
255state (the one you passed to :func:`lin2adpcm`) along to the decoder, not the
256final state (as returned by the coder). If you want to use
Serhiy Storchakabfdcd432013-10-13 23:09:14 +0300257:class:`struct.Struct` to store the state in binary you can code the first
Georg Brandl116aa622007-08-15 14:28:22 +0000258element (the predicted value) in 16 bits and the second (the delta index) in 8.
259
260The ADPCM coders have never been tried against other ADPCM coders, only against
261themselves. It could well be that I misinterpreted the standards in which case
262they will not be interoperable with the respective standards.
263
264The :func:`find\*` routines might look a bit funny at first sight. They are
265primarily meant to do echo cancellation. A reasonably fast way to do this is to
266pick the most energetic piece of the output sample, locate that in the input
267sample and subtract the whole output sample from the input sample::
268
269 def echocancel(outputdata, inputdata):
270 pos = audioop.findmax(outputdata, 800) # one tenth second
271 out_test = outputdata[pos*2:]
272 in_test = inputdata[pos*2:]
273 ipos, factor = audioop.findfit(in_test, out_test)
274 # Optional (for better cancellation):
Georg Brandl48310cd2009-01-03 21:18:54 +0000275 # factor = audioop.findfactor(in_test[ipos*2:ipos*2+len(out_test)],
Georg Brandl116aa622007-08-15 14:28:22 +0000276 # out_test)
277 prefill = '\0'*(pos+ipos)*2
278 postfill = '\0'*(len(inputdata)-len(prefill)-len(outputdata))
279 outputdata = prefill + audioop.mul(outputdata,2,-factor) + postfill
280 return audioop.add(inputdata, outputdata, 2)
281