python/lib/aubio/cmd.py

   1 #! /usr/bin/env python
   2 # -*- coding: utf-8 -*-
   3
   4 """aubio command line tool
   5
   6 This file was written by Paul Brossier <piem@aubio.org> and is released under
   7 the GNU/GPL v3.
   8
   9 Note: this script is mostly about parsing command line arguments. For more
  10 readable code examples, check out the `python/demos` folder."""
  11
  12 import sys
  13 import argparse
  14 import aubio
  15
  16 def aubio_parser():
  17     epilog = 'use "%(prog)s <command> --help" for more info about each command'
  18     parser = argparse.ArgumentParser(epilog=epilog)
  19     parser.add_argument('-V', '--version', help="show version",
  20             action="store_true", dest="show_version")
  21
  22     subparsers = parser.add_subparsers(title='commands', dest='command',
  23             metavar="")
  24
  25     parser_add_subcommand_onset(subparsers)
  26     parser_add_subcommand_pitch(subparsers)
  27     parser_add_subcommand_beat(subparsers)
  28     parser_add_subcommand_tempo(subparsers)
  29     parser_add_subcommand_notes(subparsers)
  30     parser_add_subcommand_mfcc(subparsers)
  31     parser_add_subcommand_melbands(subparsers)
  32     parser_add_subcommand_quiet(subparsers)
  33
  34     return parser
  35
  36 def parser_add_subcommand_onset(subparsers):
  37     # onset subcommand
  38     subparser = subparsers.add_parser('onset',
  39             help='estimate time of onsets (beginning of sound event)',
  40             formatter_class = argparse.ArgumentDefaultsHelpFormatter)
  41     parser_add_input(subparser)
  42     parser_add_buf_hop_size(subparser)
  43     helpstr = "onset novelty function"
  44     helpstr += " <default|energy|hfc|complex|phase|specdiff|kl|mkl|specflux>"
  45     parser_add_method(subparser, helpstr=helpstr)
  46     parser_add_threshold(subparser)
  47     parser_add_silence(subparser)
  48     parser_add_minioi(subparser)
  49     parser_add_time_format(subparser)
  50     parser_add_verbose_help(subparser)
  51     subparser.set_defaults(process=process_onset)
  52
  53 def parser_add_subcommand_pitch(subparsers):
  54     # pitch subcommand
  55     subparser = subparsers.add_parser('pitch',
  56             help='estimate fundamental frequency (monophonic)')
  57     parser_add_input(subparser)
  58     parser_add_buf_hop_size(subparser, buf_size=2048)
  59     helpstr = "pitch detection method <default|yinfft|yin|mcomb|fcomb|schmitt>"
  60     parser_add_method(subparser, helpstr=helpstr)
  61     parser_add_threshold(subparser)
  62     parser_add_pitch_unit(subparser)
  63     parser_add_silence(subparser)
  64     parser_add_time_format(subparser)
  65     parser_add_verbose_help(subparser)
  66     subparser.set_defaults(process=process_pitch)
  67
  68 def parser_add_subcommand_beat(subparsers):
  69     # beat subcommand
  70     subparser = subparsers.add_parser('beat',
  71             help='estimate location of beats')
  72     parser_add_input(subparser)
  73     parser_add_buf_hop_size(subparser, buf_size=1024, hop_size=512)
  74     parser_add_time_format(subparser)
  75     parser_add_verbose_help(subparser)
  76     subparser.set_defaults(process=process_beat)
  77
  78 def parser_add_subcommand_tempo(subparsers):
  79     # tempo subcommand
  80     subparser = subparsers.add_parser('tempo',
  81             help='estimate overall tempo in bpm')
  82     parser_add_input(subparser)
  83     parser_add_buf_hop_size(subparser, buf_size=1024, hop_size=512)
  84     parser_add_time_format(subparser)
  85     parser_add_verbose_help(subparser)
  86     subparser.set_defaults(process=process_tempo)
  87
  88 def parser_add_subcommand_notes(subparsers):
  89     # notes subcommand
  90     subparser = subparsers.add_parser('notes',
  91             help='estimate midi-like notes (monophonic)')
  92     parser_add_input(subparser)
  93     parser_add_buf_hop_size(subparser)
  94     parser_add_time_format(subparser)
  95     parser_add_verbose_help(subparser)
  96     subparser.set_defaults(process=process_notes)
  97
  98 def parser_add_subcommand_mfcc(subparsers):
  99     # mfcc subcommand
 100     subparser = subparsers.add_parser('mfcc',
 101             help='extract Mel-Frequency Cepstrum Coefficients')
 102     parser_add_input(subparser)
 103     parser_add_buf_hop_size(subparser)
 104     parser_add_time_format(subparser)
 105     parser_add_verbose_help(subparser)
 106     subparser.set_defaults(process=process_mfcc)
 107
 108 def parser_add_subcommand_melbands(subparsers):
 109     # melbands subcommand
 110     subparser = subparsers.add_parser('melbands',
 111             help='extract energies in Mel-frequency bands')
 112     parser_add_input(subparser)
 113     parser_add_buf_hop_size(subparser)
 114     parser_add_time_format(subparser)
 115     parser_add_verbose_help(subparser)
 116     subparser.set_defaults(process=process_melbands)
 117
 118 def parser_add_subcommand_quiet(subparsers):
 119     # quiet subcommand
 120     subparser = subparsers.add_parser('quiet',
 121             help='extract timestamps of quiet and loud regions')
 122     parser_add_input(subparser)
 123     parser_add_hop_size(subparser)
 124     parser_add_silence(subparser)
 125     parser_add_time_format(subparser)
 126     parser_add_verbose_help(subparser)
 127     subparser.set_defaults(process=process_quiet)
 128
 129 def parser_add_input(parser):
 130     parser.add_argument("source_uri", default=None, nargs='?',
 131             help="input sound file to analyse", metavar = "<source_uri>")
 132     parser.add_argument("-i", "--input", dest = "source_uri2",
 133             help="input sound file to analyse", metavar = "<source_uri>")
 134     parser.add_argument("-r", "--samplerate",
 135             metavar = "<freq>", type=int,
 136             action="store", dest="samplerate", default=0,
 137             help="samplerate at which the file should be represented")
 138
 139 def parser_add_verbose_help(parser):
 140     parser.add_argument("-v","--verbose",
 141             action="count", dest="verbose", default=1,
 142             help="make lots of noise [default]")
 143     parser.add_argument("-q","--quiet",
 144             action="store_const", dest="verbose", const=0,
 145             help="be quiet")
 146
 147 def parser_add_buf_hop_size(parser, buf_size=512, hop_size=256):
 148     parser_add_buf_size(parser, buf_size=buf_size)
 149     parser_add_hop_size(parser, hop_size=hop_size)
 150
 151 def parser_add_buf_size(parser, buf_size=512):
 152     parser.add_argument("-B","--bufsize",
 153             action="store", dest="buf_size", default=buf_size,
 154             metavar = "<size>", type=int,
 155             help="buffer size [default=%d]" % buf_size)
 156
 157 def parser_add_hop_size(parser, hop_size=256):
 158     parser.add_argument("-H","--hopsize",
 159             metavar = "<size>", type=int,
 160             action="store", dest="hop_size", default=hop_size,
 161             help="overlap size [default=%d]" % hop_size)
 162
 163 def parser_add_method(parser, method='default', helpstr='method'):
 164     parser.add_argument("-m","--method",
 165             metavar = "<method>", type=str,
 166             action="store", dest="method", default=method,
 167             help="%s [default=%s]" % (helpstr, method))
 168
 169 def parser_add_threshold(parser, default=None):
 170     parser.add_argument("-t","--threshold",
 171             metavar = "<threshold>", type=float,
 172             action="store", dest="threshold", default=default,
 173             help="threshold [default=%s]" % default)
 174
 175 def parser_add_silence(parser):
 176     parser.add_argument("-s", "--silence",
 177             metavar = "<value>", type=float,
 178             action="store", dest="silence", default=-70,
 179             help="silence threshold")
 180
 181 def parser_add_minioi(parser):
 182     parser.add_argument("-M", "--minioi",
 183             metavar = "<value>", type=str,
 184             action="store", dest="minioi", default="12ms",
 185             help="minimum Inter-Onset Interval")
 186
 187 def parser_add_pitch_unit(parser, default="Hz"):
 188     help_str = "frequency unit, should be one of Hz, midi, bin, cent"
 189     help_str += " [default=%s]" % default
 190     parser.add_argument("-u", "--pitch-unit",
 191             metavar = "<value>", type=str,
 192             action="store", dest="pitch_unit", default=default,
 193             help=help_str)
 194
 195 def parser_add_time_format(parser):
 196     helpstr = "select time values output format (samples, ms, seconds)"
 197     helpstr += " [default=seconds]"
 198     parser.add_argument("-T", "--time-format",
 199              metavar='format',
 200              dest="time_format",
 201              default=None,
 202              help=helpstr)
 203
 204 # some utilities
 205
 206 def samples2seconds(n_frames, samplerate):
 207     return "%f\t" % (n_frames / float(samplerate))
 208
 209 def samples2milliseconds(n_frames, samplerate):
 210     return "%f\t" % (1000. * n_frames / float(samplerate))
 211
 212 def samples2samples(n_frames, _samplerate):
 213     return "%d\t" % n_frames
 214
 215 def timefunc(mode):
 216     if mode is None or mode == 'seconds' or mode == 's':
 217         return samples2seconds
 218     elif mode == 'ms' or mode == 'milliseconds':
 219         return samples2milliseconds
 220     elif mode == 'samples':
 221         return samples2samples
 222     else:
 223         raise ValueError("invalid time format '%s'" % mode)
 224
 225 # definition of processing classes
 226
 227 class default_process(object):
 228     def __init__(self, args):
 229         if 'time_format' in args:
 230             self.time2string = timefunc(args.time_format)
 231         if args.verbose > 2 and hasattr(self, 'options'):
 232             name = type(self).__name__.split('_')[1]
 233             optstr = ' '.join(['running', name, 'with options', repr(self.options), '\n'])
 234             sys.stderr.write(optstr)
 235     def flush(self, frames_read, samplerate):
 236         # optionally called at the end of process
 237         pass
 238
 239     def parse_options(self, args, valid_opts):
 240         # get any valid options found in a dictionnary of arguments
 241         options = {k :v for k,v in vars(args).items() if k in valid_opts}
 242         self.options = options
 243
 244     def remap_pvoc_options(self, options):
 245         # FIXME: we need to remap buf_size to win_s, hop_size to hop_s
 246         # adjust python/ext/py-phasevoc.c to understand buf_size/hop_size
 247         if 'buf_size' in options:
 248             options['win_s'] = options['buf_size']
 249             del options['buf_size']
 250         if 'hop_size' in options:
 251             options['hop_s'] = options['hop_size']
 252             del options['hop_size']
 253         self.options = options
 254
 255 class process_onset(default_process):
 256     valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
 257     def __init__(self, args):
 258         self.parse_options(args, self.valid_opts)
 259         self.onset = aubio.onset(**self.options)
 260         if args.threshold is not None:
 261             self.onset.set_threshold(args.threshold)
 262         if args.minioi:
 263             if args.minioi.endswith('ms'):
 264                 self.onset.set_minioi_ms(float(args.minioi[:-2]))
 265             elif args.minioi.endswith('s'):
 266                 self.onset.set_minioi_s(float(args.minioi[:-1]))
 267             else:
 268                 self.onset.set_minioi(int(args.minioi))
 269         if args.silence:
 270             self.onset.set_silence(args.silence)
 271         super(process_onset, self).__init__(args)
 272     def __call__(self, block):
 273         return self.onset(block)
 274     def repr_res(self, res, _frames_read, samplerate):
 275         if res[0] != 0:
 276             outstr = self.time2string(self.onset.get_last(), samplerate)
 277             sys.stdout.write(outstr + '\n')
 278
 279 class process_pitch(default_process):
 280     valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
 281     def __init__(self, args):
 282         self.parse_options(args, self.valid_opts)
 283         self.pitch = aubio.pitch(**self.options)
 284         if args.pitch_unit is not None:
 285             self.pitch.set_unit(args.pitch_unit)
 286         if args.threshold is not None:
 287             self.pitch.set_tolerance(args.threshold)
 288         if args.silence is not None:
 289             self.pitch.set_silence(args.silence)
 290         super(process_pitch, self).__init__(args)
 291     def __call__(self, block):
 292         return self.pitch(block)
 293     def repr_res(self, res, frames_read, samplerate):
 294         fmt_out = self.time2string(frames_read, samplerate)
 295         sys.stdout.write(fmt_out + "%.6f\n" % res[0])
 296
 297 class process_beat(default_process):
 298     valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
 299     def __init__(self, args):
 300         self.parse_options(args, self.valid_opts)
 301         self.tempo = aubio.tempo(**self.options)
 302         super(process_beat, self).__init__(args)
 303     def __call__(self, block):
 304         return self.tempo(block)
 305     def repr_res(self, res, _frames_read, samplerate):
 306         if res[0] != 0:
 307             outstr = self.time2string(self.tempo.get_last(), samplerate)
 308             sys.stdout.write(outstr + '\n')
 309
 310 class process_tempo(process_beat):
 311     def __init__(self, args):
 312         super(process_tempo, self).__init__(args)
 313         self.beat_locations = []
 314     def repr_res(self, res, _frames_read, samplerate):
 315         if res[0] != 0:
 316             self.beat_locations.append(self.tempo.get_last_s())
 317     def flush(self, frames_read, samplerate):
 318         import numpy as np
 319         if len(self.beat_locations) < 2:
 320             outstr = "unknown bpm"
 321         else:
 322             bpms = 60./ np.diff(self.beat_locations)
 323             median_bpm = np.mean(bpms)
 324             if len(self.beat_locations) < 10:
 325                 outstr = "%.2f bpm (uncertain)" % median_bpm
 326             else:
 327                 outstr = "%.2f bpm" % median_bpm
 328         sys.stdout.write(outstr + '\n')
 329
 330 class process_notes(default_process):
 331     valid_opts = ['method', 'hop_size', 'buf_size', 'samplerate']
 332     def __init__(self, args):
 333         self.parse_options(args, self.valid_opts)
 334         self.notes = aubio.notes(**self.options)
 335         super(process_notes, self).__init__(args)
 336     def __call__(self, block):
 337         return self.notes(block)
 338     def repr_res(self, res, frames_read, samplerate):
 339         if res[2] != 0: # note off
 340             fmt_out = self.time2string(frames_read, samplerate)
 341             sys.stdout.write(fmt_out + '\n')
 342         if res[0] != 0: # note on
 343             lastmidi = res[0]
 344             fmt_out = "%f\t" % lastmidi
 345             fmt_out += self.time2string(frames_read, samplerate)
 346             sys.stdout.write(fmt_out) # + '\t')
 347     def flush(self, frames_read, samplerate):
 348         eof = self.time2string(frames_read, samplerate)
 349         sys.stdout.write(eof + '\n')
 350
 351 class process_mfcc(default_process):
 352     def __init__(self, args):
 353         valid_opts1 = ['hop_size', 'buf_size']
 354         self.parse_options(args, valid_opts1)
 355         self.remap_pvoc_options(self.options)
 356         self.pv = aubio.pvoc(**self.options)
 357
 358         valid_opts2 = ['buf_size', 'n_filters', 'n_coeffs', 'samplerate']
 359         self.parse_options(args, valid_opts2)
 360         self.mfcc = aubio.mfcc(**self.options)
 361
 362         # remember all options
 363         self.parse_options(args, list(set(valid_opts1 + valid_opts2)))
 364
 365         super(process_mfcc, self).__init__(args)
 366
 367     def __call__(self, block):
 368         fftgrain = self.pv(block)
 369         return self.mfcc(fftgrain)
 370     def repr_res(self, res, frames_read, samplerate):
 371         fmt_out = self.time2string(frames_read, samplerate)
 372         fmt_out += ' '.join(["% 9.7f" % f for f in res.tolist()])
 373         sys.stdout.write(fmt_out + '\n')
 374
 375 class process_melbands(default_process):
 376     def __init__(self, args):
 377         self.args = args
 378         valid_opts = ['hop_size', 'buf_size']
 379         self.parse_options(args, valid_opts)
 380         self.remap_pvoc_options(self.options)
 381         self.pv = aubio.pvoc(**self.options)
 382
 383         valid_opts = ['buf_size', 'n_filters']
 384         self.parse_options(args, valid_opts)
 385         self.remap_pvoc_options(self.options)
 386         self.filterbank = aubio.filterbank(**self.options)
 387         self.filterbank.set_mel_coeffs_slaney(args.samplerate)
 388
 389         super(process_melbands, self).__init__(args)
 390     def __call__(self, block):
 391         fftgrain = self.pv(block)
 392         return self.filterbank(fftgrain)
 393     def repr_res(self, res, frames_read, samplerate):
 394         fmt_out = self.time2string(frames_read, samplerate)
 395         fmt_out += ' '.join(["% 9.7f" % f for f in res.tolist()])
 396         sys.stdout.write(fmt_out + '\n')
 397
 398 class process_quiet(default_process):
 399     def __init__(self, args):
 400         self.args = args
 401         valid_opts = ['hop_size', 'silence']
 402         self.parse_options(args, valid_opts)
 403         self.wassilence = 1
 404
 405         if args.silence is not None:
 406             self.silence = args.silence
 407         super(process_quiet, self).__init__(args)
 408
 409     def __call__(self, block):
 410         if aubio.silence_detection(block, self.silence) == 1:
 411             if self.wassilence != 1:
 412                 self.wassilence = 1
 413                 return 2 # newly found silence
 414             return 1 # silence again
 415         else:
 416             if self.wassilence != 0:
 417                 self.wassilence = 0
 418                 return -1 # newly found noise
 419             return 0 # noise again
 420
 421     def repr_res(self, res, frames_read, samplerate):
 422         fmt_out = None
 423         if res == -1:
 424             fmt_out = "NOISY: "
 425         if res == 2:
 426             fmt_out = "QUIET: "
 427         if fmt_out is not None:
 428             fmt_out += self.time2string(frames_read, samplerate)
 429             sys.stdout.write(fmt_out + '\n')
 430
 431 def main():
 432     parser = aubio_parser()
 433     args = parser.parse_args()
 434     if 'show_version' in args and args.show_version:
 435         sys.stdout.write('aubio version ' + aubio.version + '\n')
 436         sys.exit(0)
 437     elif 'verbose' in args and args.verbose > 3:
 438         sys.stderr.write('aubio version ' + aubio.version + '\n')
 439     if 'command' not in args or args.command is None:
 440         # no command given, print help and return 1
 441         parser.print_help()
 442         sys.exit(1)
 443     elif not args.source_uri and not args.source_uri2:
 444         sys.stderr.write("Error: a source is required\n")
 445         parser.print_help()
 446         sys.exit(1)
 447     elif args.source_uri2 is not None:
 448         args.source_uri = args.source_uri2
 449     try:
 450         # open source_uri
 451         with aubio.source(args.source_uri, hop_size=args.hop_size,
 452                 samplerate=args.samplerate) as a_source:
 453             # always update args.samplerate to native samplerate, in case
 454             # source was opened with args.samplerate=0
 455             args.samplerate = a_source.samplerate
 456             # create the processor for this subcommand
 457             processor = args.process(args)
 458             frames_read = 0
 459             while True:
 460                 # read new block from source
 461                 block, read = a_source()
 462                 # execute processor on this block
 463                 res = processor(block)
 464                 # print results for this block
 465                 if args.verbose > 0:
 466                     processor.repr_res(res, frames_read, a_source.samplerate)
 467                 # increment total number of frames read
 468                 frames_read += read
 469                 # exit loop at end of file
 470                 if read < a_source.hop_size: break
 471             # flush the processor if needed
 472             processor.flush(frames_read, a_source.samplerate)
 473             if args.verbose > 1:
 474                 fmt_string = "read {:.2f}s"
 475                 fmt_string += " ({:d} samples in {:d} blocks of {:d})"
 476                 fmt_string += " from {:s} at {:d}Hz\n"
 477                 sys.stderr.write(fmt_string.format(
 478                         frames_read/float(a_source.samplerate),
 479                         frames_read,
 480                         frames_read // a_source.hop_size + 1,
 481                         a_source.hop_size,
 482                         a_source.uri,
 483                         a_source.samplerate))
 484     except KeyboardInterrupt:
 485         sys.exit(1)