Doc/lib/libaudioop.tex

   1 \section{\module{audioop} ---
   2          Manipulate raw audio data}
   3
   4 \declaremodule{builtin}{audioop}
   5 \modulesynopsis{Manipulate raw audio data.}
   6
   7
   8 The \module{audioop} module contains some useful operations on sound
   9 fragments.  It operates on sound fragments consisting of signed
  10 integer samples 8, 16 or 32 bits wide, stored in Python strings.  This
  11 is the same format as used by the \refmodule{al} and \refmodule{sunaudiodev}
  12 modules.  All scalar items are integers, unless specified otherwise.
  13
  14 % This para is mostly here to provide an excuse for the index entries...
  15 This module provides support for u-LAW and Intel/DVI ADPCM encodings.
  16 \index{Intel/DVI ADPCM}
  17 \index{ADPCM, Intel/DVI}
  18 \index{u-LAW}
  19
  20 A few of the more complicated operations only take 16-bit samples,
  21 otherwise the sample size (in bytes) is always a parameter of the
  22 operation.
  23
  24 The module defines the following variables and functions:
  25
  26 \begin{excdesc}{error}
  27 This exception is raised on all errors, such as unknown number of bytes
  28 per sample, etc.
  29 \end{excdesc}
  30
  31 \begin{funcdesc}{add}{fragment1, fragment2, width}
  32 Return a fragment which is the addition of the two samples passed as
  33 parameters.  \var{width} is the sample width in bytes, either
  34 \code{1}, \code{2} or \code{4}.  Both fragments should have the same
  35 length.
  36 \end{funcdesc}
  37
  38 \begin{funcdesc}{adpcm2lin}{adpcmfragment, width, state}
  39 Decode an Intel/DVI ADPCM coded fragment to a linear fragment.  See
  40 the description of \function{lin2adpcm()} for details on ADPCM coding.
  41 Return a tuple \code{(\var{sample}, \var{newstate})} where the sample
  42 has the width specified in \var{width}.
  43 \end{funcdesc}
  44
  45 \begin{funcdesc}{adpcm32lin}{adpcmfragment, width, state}
  46 Decode an alternative 3-bit ADPCM code.  See \function{lin2adpcm3()}
  47 for details.
  48 \end{funcdesc}
  49
  50 \begin{funcdesc}{avg}{fragment, width}
  51 Return the average over all samples in the fragment.
  52 \end{funcdesc}
  53
  54 \begin{funcdesc}{avgpp}{fragment, width}
  55 Return the average peak-peak value over all samples in the fragment.
  56 No filtering is done, so the usefulness of this routine is
  57 questionable.
  58 \end{funcdesc}
  59
  60 \begin{funcdesc}{bias}{fragment, width, bias}
  61 Return a fragment that is the original fragment with a bias added to
  62 each sample.
  63 \end{funcdesc}
  64
  65 \begin{funcdesc}{cross}{fragment, width}
  66 Return the number of zero crossings in the fragment passed as an
  67 argument.
  68 \end{funcdesc}
  69
  70 \begin{funcdesc}{findfactor}{fragment, reference}
  71 Return a factor \var{F} such that
  72 \code{rms(add(\var{fragment}, mul(\var{reference}, -\var{F})))} is
  73 minimal, i.e., return the factor with which you should multiply
  74 \var{reference} to make it match as well as possible to
  75 \var{fragment}.  The fragments should both contain 2-byte samples.
  76
  77 The time taken by this routine is proportional to
  78 \code{len(\var{fragment})}.
  79 \end{funcdesc}
  80
  81 \begin{funcdesc}{findfit}{fragment, reference}
  82 Try to match \var{reference} as well as possible to a portion of
  83 \var{fragment} (which should be the longer fragment).  This is
  84 (conceptually) done by taking slices out of \var{fragment}, using
  85 \function{findfactor()} to compute the best match, and minimizing the
  86 result.  The fragments should both contain 2-byte samples.  Return a
  87 tuple \code{(\var{offset}, \var{factor})} where \var{offset} is the
  88 (integer) offset into \var{fragment} where the optimal match started
  89 and \var{factor} is the (floating-point) factor as per
  90 \function{findfactor()}.
  91 \end{funcdesc}
  92
  93 \begin{funcdesc}{findmax}{fragment, length}
  94 Search \var{fragment} for a slice of length \var{length} samples (not
  95 bytes!)\ with maximum energy, i.e., return \var{i} for which
  96 \code{rms(fragment[i*2:(i+length)*2])} is maximal.  The fragments
  97 should both contain 2-byte samples.
  98
  99 The routine takes time proportional to \code{len(\var{fragment})}.
 100 \end{funcdesc}
 101
 102 \begin{funcdesc}{getsample}{fragment, width, index}
 103 Return the value of sample \var{index} from the fragment.
 104 \end{funcdesc}
 105
 106 \begin{funcdesc}{lin2lin}{fragment, width, newwidth}
 107 Convert samples between 1-, 2- and 4-byte formats.
 108 \end{funcdesc}
 109
 110 \begin{funcdesc}{lin2adpcm}{fragment, width, state}
 111 Convert samples to 4 bit Intel/DVI ADPCM encoding.  ADPCM coding is an
 112 adaptive coding scheme, whereby each 4 bit number is the difference
 113 between one sample and the next, divided by a (varying) step.  The
 114 Intel/DVI ADPCM algorithm has been selected for use by the IMA, so it
 115 may well become a standard.
 116
 117 \var{state} is a tuple containing the state of the coder.  The coder
 118 returns a tuple \code{(\var{adpcmfrag}, \var{newstate})}, and the
 119 \var{newstate} should be passed to the next call of
 120 \function{lin2adpcm()}.  In the initial call, \code{None} can be
 121 passed as the state.  \var{adpcmfrag} is the ADPCM coded fragment
 122 packed 2 4-bit values per byte.
 123 \end{funcdesc}
 124
 125 \begin{funcdesc}{lin2adpcm3}{fragment, width, state}
 126 This is an alternative ADPCM coder that uses only 3 bits per sample.
 127 It is not compatible with the Intel/DVI ADPCM coder and its output is
 128 not packed (due to laziness on the side of the author).  Its use is
 129 discouraged.
 130 \end{funcdesc}
 131
 132 \begin{funcdesc}{lin2ulaw}{fragment, width}
 133 Convert samples in the audio fragment to u-LAW encoding and return
 134 this as a Python string.  u-LAW is an audio encoding format whereby
 135 you get a dynamic range of about 14 bits using only 8 bit samples.  It
 136 is used by the Sun audio hardware, among others.
 137 \end{funcdesc}
 138
 139 \begin{funcdesc}{minmax}{fragment, width}
 140 Return a tuple consisting of the minimum and maximum values of all
 141 samples in the sound fragment.
 142 \end{funcdesc}
 143
 144 \begin{funcdesc}{max}{fragment, width}
 145 Return the maximum of the \emph{absolute value} of all samples in a
 146 fragment.
 147 \end{funcdesc}
 148
 149 \begin{funcdesc}{maxpp}{fragment, width}
 150 Return the maximum peak-peak value in the sound fragment.
 151 \end{funcdesc}
 152
 153 \begin{funcdesc}{mul}{fragment, width, factor}
 154 Return a fragment that has all samples in the original fragment
 155 multiplied by the floating-point value \var{factor}.  Overflow is
 156 silently ignored.
 157 \end{funcdesc}
 158
 159 \begin{funcdesc}{ratecv}{fragment, width, nchannels, inrate, outrate,
 160                          state\optional{, weightA\optional{, weightB}}}
 161 Convert the frame rate of the input fragment.
 162
 163 \var{state} is a tuple containing the state of the converter.  The
 164 converter returns a tuple \code{(\var{newfragment}, \var{newstate})},
 165 and \var{newstate} should be passed to the next call of
 166 \function{ratecv()}.  The initial call should pass \code{None}
 167 as the state.
 168
 169 The \var{weightA} and \var{weightB} arguments are parameters for a
 170 simple digital filter and default to \code{1} and \code{0} respectively.
 171 \end{funcdesc}
 172
 173 \begin{funcdesc}{reverse}{fragment, width}
 174 Reverse the samples in a fragment and returns the modified fragment.
 175 \end{funcdesc}
 176
 177 \begin{funcdesc}{rms}{fragment, width}
 178 Return the root-mean-square of the fragment, i.e.
 179 \begin{displaymath}
 180 \catcode`_=8
 181 \sqrt{\frac{\sum{{S_{i}}^{2}}}{n}}
 182 \end{displaymath}
 183 This is a measure of the power in an audio signal.
 184 \end{funcdesc}
 185
 186 \begin{funcdesc}{tomono}{fragment, width, lfactor, rfactor}
 187 Convert a stereo fragment to a mono fragment.  The left channel is
 188 multiplied by \var{lfactor} and the right channel by \var{rfactor}
 189 before adding the two channels to give a mono signal.
 190 \end{funcdesc}
 191
 192 \begin{funcdesc}{tostereo}{fragment, width, lfactor, rfactor}
 193 Generate a stereo fragment from a mono fragment.  Each pair of samples
 194 in the stereo fragment are computed from the mono sample, whereby left
 195 channel samples are multiplied by \var{lfactor} and right channel
 196 samples by \var{rfactor}.
 197 \end{funcdesc}
 198
 199 \begin{funcdesc}{ulaw2lin}{fragment, width}
 200 Convert sound fragments in u-LAW encoding to linearly encoded sound
 201 fragments.  u-LAW encoding always uses 8 bits samples, so \var{width}
 202 refers only to the sample width of the output fragment here.
 203 \end{funcdesc}
 204
 205 Note that operations such as \function{mul()} or \function{max()} make
 206 no distinction between mono and stereo fragments, i.e.\ all samples
 207 are treated equal.  If this is a problem the stereo fragment should be
 208 split into two mono fragments first and recombined later.  Here is an
 209 example of how to do that:
 210
 211 \begin{verbatim}
 212 def mul_stereo(sample, width, lfactor, rfactor):
 213     lsample = audioop.tomono(sample, width, 1, 0)
 214     rsample = audioop.tomono(sample, width, 0, 1)
 215     lsample = audioop.mul(sample, width, lfactor)
 216     rsample = audioop.mul(sample, width, rfactor)
 217     lsample = audioop.tostereo(lsample, width, 1, 0)
 218     rsample = audioop.tostereo(rsample, width, 0, 1)
 219     return audioop.add(lsample, rsample, width)
 220 \end{verbatim}
 221
 222 If you use the ADPCM coder to build network packets and you want your
 223 protocol to be stateless (i.e.\ to be able to tolerate packet loss)
 224 you should not only transmit the data but also the state.  Note that
 225 you should send the \var{initial} state (the one you passed to
 226 \function{lin2adpcm()}) along to the decoder, not the final state (as
 227 returned by the coder).  If you want to use \function{struct.struct()}
 228 to store the state in binary you can code the first element (the
 229 predicted value) in 16 bits and the second (the delta index) in 8.
 230
 231 The ADPCM coders have never been tried against other ADPCM coders,
 232 only against themselves.  It could well be that I misinterpreted the
 233 standards in which case they will not be interoperable with the
 234 respective standards.
 235
 236 The \function{find*()} routines might look a bit funny at first sight.
 237 They are primarily meant to do echo cancellation.  A reasonably
 238 fast way to do this is to pick the most energetic piece of the output
 239 sample, locate that in the input sample and subtract the whole output
 240 sample from the input sample:
 241
 242 \begin{verbatim}
 243 def echocancel(outputdata, inputdata):
 244     pos = audioop.findmax(outputdata, 800)    # one tenth second
 245     out_test = outputdata[pos*2:]
 246     in_test = inputdata[pos*2:]
 247     ipos, factor = audioop.findfit(in_test, out_test)
 248     # Optional (for better cancellation):
 249     # factor = audioop.findfactor(in_test[ipos*2:ipos*2+len(out_test)],
 250     #              out_test)
 251     prefill = '\0'*(pos+ipos)*2
 252     postfill = '\0'*(len(inputdata)-len(prefill)-len(outputdata))
 253     outputdata = prefill + audioop.mul(outputdata,2,-factor) + postfill
 254     return audioop.add(inputdata, outputdata, 2)
 255 \end{verbatim}