source: python/aubiocut @ 588a09f

#! /usr/bin/python

""" this file was written by Paul Brossier 
  it is released under the GNU/GPL license.
"""

import sys
import numarray
from aubio.aubioclass import *

usage = "usage: %s [options] -i soundfile" % sys.argv[0]

def check_mode(option, opt, value, parser):
        nvalue = parser.rargs[0]
        if   nvalue == 'complexdomain' :
                 setattr(parser.values, option.dest, complexdomain)
        elif nvalue == 'hfc'           :
                 setattr(parser.values, option.dest, hfc)
        elif nvalue == 'phase'         :
                 setattr(parser.values, option.dest, phase)
        elif nvalue == 'specdiff'      :
                 setattr(parser.values, option.dest, specdiff)
        elif nvalue == 'energy'        :
                 setattr(parser.values, option.dest, energy)
        elif nvalue == 'dual'          :
                 setattr(parser.values, option.dest, 'dual')

def parse_args():
        from optparse import OptionParser
        parser = OptionParser(usage=usage)
        parser.add_option("-i","--input",
                          action="store", dest="filename", 
                          help="input sound file")
        parser.add_option("-m","--mode", action="callback", 
                          callback=check_mode, dest="mode", default='dual', 
                          help="onsetdetection mode [default=dual] \
                          complexdomain|hfc|phase|specdiff|energy|dual")
        parser.add_option("-B","--bufsize",
                          action="store", dest="bufsize", default=1024, 
                          help="buffer size [default=1024]")
        parser.add_option("-H","--hopsize",
                          action="store", dest="hopsize", default=512, 
                          help="overlap size [default=512]")
        parser.add_option("-t","--threshold",
                          action="store", dest="threshold", default=0.3, 
                          help="onset peak picking threshold [default=0.3]")
        parser.add_option("-s","--silence",
                          action="store", dest="silence", default=-70, 
                          help="silence threshold [default=-70]")
        parser.add_option("-M","--mintol",
                          action="store", dest="mintol", default=0.048, 
                          help="minimum inter onset interval [default=0.048]")
        parser.add_option("-D","--delay",
                          action="store", dest="delay", default=0.022, 
                          help="number of seconds to take back [default=0.022]")
        parser.add_option("-L","--localmin",
                          action="store_true", dest="localmin", default=False, 
                          help="use local minima after peak detection")
        parser.add_option("-c","--cut",
                          action="store_true", dest="cut", default=False,
                          help="cut input sound file at detected labels \
                          best used with option -L")
        # to be implemented
        # plotting functions
        parser.add_option("-d","--derivative",
                          action="store_true", dest="derivative", default=False, 
                          help="NOT IMPLEMENTED derivate onset detection function")
        parser.add_option("-p","--plot",
                          action="store_true", dest="doplot", default=False, 
                          help="NOT IMPLEMENTED draw plot")
        parser.add_option("-O","--outplot",
                          action="store", dest="output-plot", default=None, 
                          help="NOT IMPLEMENTED save plot to output.{ps,png}")
        parser.add_option("-z","--zerocross",
                          action="store_true", dest="zerocross", default=False, 
                          help="NOT IMPLEMENTED zero crossing matching")
        parser.add_option("-b","--beat",
                          action="store_true", dest="beat", default=False,
                          help="NOT IMPLEMENTED output beat locations")
        parser.add_option("-v","--verbose",
                          action="store_true", dest="verbose", default=False,
                          help="make lots of noise [default]")
        parser.add_option("-q","--quiet",
                          action="store_false", dest="verbose", default=False, 
                          help="be quiet")
        (options, args) = parser.parse_args()
        if not options.filename: 
                 print "no file name given\n", usage
                 sys.exit(1)
        return options, args

options, args = parse_args()

filename   = options.filename
samplerate = float(sndfile(filename).samplerate())
hopsize    = int(options.hopsize)
bufsize    = int(options.bufsize)
step       = float(samplerate)/float(hopsize)
threshold  = float(options.threshold)
silence    = float(options.silence)
mintol     = float(options.mintol)*step
delay      = float(options.delay)

if options.beat:
        #onsets = getbeats(filename,threshold,silence,mode=options.mode)
        exit("not implemented yet")
else:
        onsets = getonsets(filename,threshold,silence,
                mode=options.mode,localmin=options.localmin,
                bufsize=bufsize,hopsize=hopsize)

# take back system delay
if delay != 0:
        for i in range(len(onsets)):
                onsets[i] -= delay*step

# prune doubled 
if mintol > 0:
        last = -2*mintol
        newonsets = []
        for new in onsets:
                if (new - last > mintol): 
                        newonsets.append(new)
                last = new
        onsets = newonsets

# print times in second
if options.verbose:
        for i in onsets: print "%f" % (i/step)

if options.cut:
        cutfile(filename,onsets,bufsize=bufsize,hopsize=hopsize)