Context Navigation

source: python.old/aubio/task/pitch.py @ 493b832

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/pydocstringsfeature/timestretchfix/ffmpeg5pitchshiftsamplertimestretchyinfft+

Last change on this file since 493b832 was b554e41, checked in by Paul Brossier <piem@piem.org>, 12 years ago
moved out old python interface
Property mode set to `100644`
File size: 7.1 KB

Rev	Line
[13c3fba]	1	from aubio.task.task import task
	2	from aubio.task.silence import tasksilence
	3	from aubio.aubioclass import *
	4
	5	class taskpitch(task):
	6	def __init__(self,input,params=None):
	7	task.__init__(self,input,params=params)
	8	self.shortlist = [0. for i in range(self.params.pitchsmooth)]
[3027c6e]	9	if self.params.pitchmode == 'yin':
	10	tolerance = self.params.yinthresh
	11	elif self.params.pitchmode == 'yinfft':
	12	tolerance = self.params.yinfftthresh
[650e39b]	13	else:
[3027c6e]	14	tolerance = 0.
[ca1abdd]	15	self.pitchdet = pitch(mode=self.params.pitchmode,
[13c3fba]	16	bufsize=self.params.bufsize,
	17	hopsize=self.params.hopsize,
	18	samplerate=self.srate,
[83376e9]	19	omode=self.params.omode,
[3027c6e]	20	tolerance = tolerance)
[13c3fba]	21
	22	def __call__(self):
	23	from aubio.median import short_find
	24	task.__call__(self)
	25	if (aubio_silence_detection(self.myvec(),self.params.silence)==1):
	26	freq = -1.
	27	else:
	28	freq = self.pitchdet(self.myvec)
	29	minpitch = self.params.pitchmin
	30	maxpitch = self.params.pitchmax
	31	if maxpitch and freq > maxpitch :
	32	freq = -1.
	33	elif minpitch and freq < minpitch :
	34	freq = -1.
	35	if self.params.pitchsmooth:
	36	self.shortlist.append(freq)
	37	self.shortlist.pop(0)
	38	smoothfreq = short_find(self.shortlist,
	39	len(self.shortlist)/2)
	40	return smoothfreq
	41	else:
	42	return freq
	43
	44	def compute_all(self):
	45	""" Compute data """
	46	mylist = []
	47	while(self.readsize==self.params.hopsize):
	48	freq = self()
	49	mylist.append(freq)
	50	if self.params.verbose:
	51	self.fprint("%s\t%s" % (self.frameread*self.params.step,freq))
	52	return mylist
	53
	54	def gettruth(self):
	55	""" extract ground truth array in frequency """
	56	import os.path
	57	""" from wavfile.txt """
	58	datafile = self.input.replace('.wav','.txt')
	59	if datafile == self.input: datafile = ""
	60	""" from file.<midinote>.wav """
	61	# FIXME very weak check
	62	floatpit = self.input.split('.')[-2]
	63	if not os.path.isfile(datafile) and len(self.input.split('.')) < 3:
	64	print "no ground truth "
	65	return False,False
	66	elif floatpit:
	67	try:
[5d1c070]	68	self.truth = float(floatpit)
	69	#print "ground truth found in filename:", self.truth
	70	tasksil = tasksilence(self.input,params=self.params)
[13c3fba]	71	time,pitch =[],[]
	72	while(tasksil.readsize==tasksil.params.hopsize):
	73	tasksil()
[650e39b]	74	time.append(tasksil.params.step*(tasksil.frameread))
[13c3fba]	75	if not tasksil.issilence:
	76	pitch.append(self.truth)
	77	else:
	78	pitch.append(-1.)
[4798fdf]	79	return time,pitch
[13c3fba]	80	except ValueError:
	81	# FIXME very weak check
	82	if not os.path.isfile(datafile):
	83	print "no ground truth found"
	84	return 0,0
	85	else:
	86	from aubio.txtfile import read_datafile
	87	values = read_datafile(datafile)
	88	time, pitch = [], []
	89	for i in range(len(values)):
	90	time.append(values[i][0])
[83376e9]	91	if values[i][1] == 0.0:
	92	pitch.append(-1.)
	93	else:
	94	pitch.append(aubio_freqtomidi(values[i][1]))
[13c3fba]	95	return time,pitch
	96
[5d1c070]	97	def oldeval(self,results):
[13c3fba]	98	def mmean(l):
	99	return sum(l)/max(float(len(l)),1)
	100
	101	from aubio.median import percental
	102	timet,pitcht = self.gettruth()
	103	res = []
	104	for i in results:
	105	#print i,self.truth
	106	if i <= 0: pass
	107	else: res.append(self.truth-i)
	108	if not res or len(res) < 3:
	109	avg = self.truth; med = self.truth
	110	else:
	111	avg = mmean(res)
	112	med = percental(res,len(res)/2)
	113	return self.truth, self.truth-med, self.truth-avg
	114
[83376e9]	115	def eval(self,pitch,tol=0.5):
[13c3fba]	116	timet,pitcht = self.gettruth()
[5d1c070]	117	pitch = [aubio_freqtomidi(i) for i in pitch]
	118	for i in range(len(pitch)):
	119	if pitch[i] == "nan" or pitch[i] == -1:
	120	pitch[i] = -1
[4798fdf]	121	time = [ (i+self.params.pitchdelay)*self.params.step for i in range(len(pitch)) ]
[83376e9]	122	#print len(timet),len(pitcht)
	123	#print len(time),len(pitch)
	124	if len(timet) != len(time):
	125	time = time[1:len(timet)+1]
	126	pitch = pitch[1:len(pitcht)+1]
	127	#pitcht = [aubio_freqtomidi(i) for i in pitcht]
	128	for i in range(len(pitcht)):
	129	if pitcht[i] == "nan" or pitcht[i] == "-inf" or pitcht[i] == -1:
	130	pitcht[i] = -1
	131	assert len(timet) == len(time)
[5d1c070]	132	assert len(pitcht) == len(pitch)
	133	osil, esil, opit, epit, echr = 0, 0, 0, 0, 0
	134	for i in range(len(pitcht)):
	135	if pitcht[i] == -1: # currently silent
	136	osil += 1 # count a silence
[83376e9]	137	if pitch[i] <= 0. or pitch[i] == "nan":
[5d1c070]	138	esil += 1 # found a silence
	139	else:
	140	opit +=1
	141	if abs(pitcht[i] - pitch[i]) < tol:
	142	epit += 1
	143	echr += 1
	144	elif abs(pitcht[i] - pitch[i]) % 12. < tol:
	145	echr += 1
	146	#else:
	147	# print timet[i], pitcht[i], time[i], pitch[i]
	148	#print "origsilence", "foundsilence", "origpitch", "foundpitch", "orig pitchroma", "found pitchchroma"
	149	#print 100.esil/float(osil), 100.epit/float(opit), 100.*echr/float(opit)
	150	return osil, esil, opit, epit, echr
[13c3fba]	151
[fb8bde7]	152	def plot(self,pitch,wplot,oplots,titles,outplot=None):
[13c3fba]	153	import Gnuplot
	154
[4798fdf]	155	time = [ (i+self.params.pitchdelay)*self.params.step for i in range(len(pitch)) ]
[83376e9]	156	pitch = [aubio_freqtomidi(i) for i in pitch]
[5f23f66]	157	oplots.append(Gnuplot.Data(time,pitch,with_='lines',
[13c3fba]	158	title=self.params.pitchmode))
[fb8bde7]	159	titles.append(self.params.pitchmode)
[13c3fba]	160
	161
[470a772]	162	def plotplot(self,wplot,oplots,titles,outplot=None,extension=None,xsize=1.,ysize=1.,multiplot = 1, midi = 1, truth = 1):
	163	from aubio.gnuplot import gnuplot_create , audio_to_array, make_audio_plot
[13c3fba]	164	import re
	165	import Gnuplot
	166
	167	# check if ground truth exists
[fb8bde7]	168	if truth:
	169	timet,pitcht = self.gettruth()
	170	if timet and pitcht:
[5f23f66]	171	oplots = [Gnuplot.Data(timet,pitcht,with_='lines',
[fb8bde7]	172	title='ground truth')] + oplots
[13c3fba]	173
[470a772]	174	g = gnuplot_create(outplot=outplot, extension=extension)
[13c3fba]	175	g('set title \'%s\'' % (re.sub('.*/','',self.input)))
[470a772]	176	g('set size %f,%f' % (xsize,ysize) )
[13c3fba]	177	g('set multiplot')
	178	# hack to align left axis
[fb8bde7]	179	g('set lmargin 4')
	180	g('set rmargin 4')
[470a772]	181	# plot waveform
	182	time,data = audio_to_array(self.input)
	183	wplot = [make_audio_plot(time,data)]
	184	g('set origin 0,%f' % (0.7*ysize) )
	185	g('set size %f,%f' % (xsize,0.3*ysize))
	186	#g('set format y "%1f"')
[13c3fba]	187	g('set xrange [0:%f]' % max(time))
	188	g('set yrange [-1:1]')
[fb8bde7]	189	g('set noytics')
	190	g('set y2tics -1,1')
[470a772]	191	g.xlabel('time (s)',offset=(0,0.7))
[13c3fba]	192	g.ylabel('amplitude')
[470a772]	193	g.plot(*wplot)
[13c3fba]	194
[470a772]	195	# default settings for next plots
	196	g('unset title')
	197	g('set format x ""')
	198	g('set format y "%3e"')
	199	g('set tmargin 0')
	200	g.xlabel('')
	201	g('set noclip one')
[13c3fba]	202
[83376e9]	203	if not midi:
	204	g('set log y')
	205	#g.xlabel('time (s)')
	206	g.ylabel('f0 (Hz)')
	207	g('set yrange [100:%f]' % self.params.pitchmax)
	208	else:
[4798fdf]	209	g.ylabel('midi')
[650e39b]	210	g('set yrange [%f:%f]' % (aubio_freqtomidi(self.params.pitchmin), aubio_freqtomidi(self.params.pitchmax)))
[fb8bde7]	211	g('set y2tics %f,%f' % (round(aubio_freqtomidi(self.params.pitchmin)+.5),12))
[470a772]	212
[13c3fba]	213	if multiplot:
[470a772]	214	N = len(oplots)
	215	y = 0.7*ysize # the vertical proportion of the plot taken by onset functions
	216	delta = 0.035 # the constant part of y taken by last plot label and data
	217	for i in range(N):
	218	# plot pitch detection functions
	219	g('set size %f,%f' % ( xsize, (y-delta)/N))
	220	g('set origin 0,%f' % ((N-i-1)*(y-delta)/N + delta ))
[fb8bde7]	221	g('set nokey')
[470a772]	222	g('set xrange [0:%f]' % max(time))
[fb8bde7]	223	g.ylabel(titles[i])
[470a772]	224	if i == N-1:
	225	g('set size %f,%f' % (xsize, (y-delta)/N + delta ) )
	226	g('set origin 0,0')
	227	g.xlabel('time (s)', offset=(0,0.7))
	228	g('set format x')
[13c3fba]	229	g.plot(oplots[i])
	230	else:
[470a772]	231	g('set key right top')
[13c3fba]	232	g.plot(*oplots)
	233	g('unset multiplot')
	234

Note: See TracBrowser for help on using the repository browser.

Download in other formats: