Context Navigation

source: python/aubio/task/pitch.py @ fe163ad

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

Last change on this file since fe163ad was fe163ad, checked in by Paul Brossier <piem@piem.org>, 15 years ago
src/pitch: use a string to set pitch method, add a new function to set pitch unit, keep pitch enums private, update pitch methods where they are used
Property mode set to `100644`
File size: 7.2 KB

Rev	Line
[13c3fba]	1	from aubio.task.task import task
	2	from aubio.task.silence import tasksilence
	3	from aubio.task.utils import *
	4	from aubio.aubioclass import *
	5
	6	class taskpitch(task):
	7	def __init__(self,input,params=None):
	8	task.__init__(self,input,params=params)
	9	self.shortlist = [0. for i in range(self.params.pitchsmooth)]
[3027c6e]	10	if self.params.pitchmode == 'yin':
	11	tolerance = self.params.yinthresh
	12	elif self.params.pitchmode == 'yinfft':
	13	tolerance = self.params.yinfftthresh
[650e39b]	14	else:
[3027c6e]	15	tolerance = 0.
[fe163ad]	16	self.pitchdet = pitchdetection(mode=self.params.pitchmode,
[13c3fba]	17	bufsize=self.params.bufsize,
	18	hopsize=self.params.hopsize,
	19	channels=self.channels,
	20	samplerate=self.srate,
[83376e9]	21	omode=self.params.omode,
[3027c6e]	22	tolerance = tolerance)
[13c3fba]	23
	24	def __call__(self):
	25	from aubio.median import short_find
	26	task.__call__(self)
	27	if (aubio_silence_detection(self.myvec(),self.params.silence)==1):
	28	freq = -1.
	29	else:
	30	freq = self.pitchdet(self.myvec)
	31	minpitch = self.params.pitchmin
	32	maxpitch = self.params.pitchmax
	33	if maxpitch and freq > maxpitch :
	34	freq = -1.
	35	elif minpitch and freq < minpitch :
	36	freq = -1.
	37	if self.params.pitchsmooth:
	38	self.shortlist.append(freq)
	39	self.shortlist.pop(0)
	40	smoothfreq = short_find(self.shortlist,
	41	len(self.shortlist)/2)
	42	return smoothfreq
	43	else:
	44	return freq
	45
	46	def compute_all(self):
	47	""" Compute data """
	48	mylist = []
	49	while(self.readsize==self.params.hopsize):
	50	freq = self()
	51	mylist.append(freq)
	52	if self.params.verbose:
	53	self.fprint("%s\t%s" % (self.frameread*self.params.step,freq))
	54	return mylist
	55
	56	def gettruth(self):
	57	""" extract ground truth array in frequency """
	58	import os.path
	59	""" from wavfile.txt """
	60	datafile = self.input.replace('.wav','.txt')
	61	if datafile == self.input: datafile = ""
	62	""" from file.<midinote>.wav """
	63	# FIXME very weak check
	64	floatpit = self.input.split('.')[-2]
	65	if not os.path.isfile(datafile) and len(self.input.split('.')) < 3:
	66	print "no ground truth "
	67	return False,False
	68	elif floatpit:
	69	try:
[5d1c070]	70	self.truth = float(floatpit)
	71	#print "ground truth found in filename:", self.truth
	72	tasksil = tasksilence(self.input,params=self.params)
[13c3fba]	73	time,pitch =[],[]
	74	while(tasksil.readsize==tasksil.params.hopsize):
	75	tasksil()
[650e39b]	76	time.append(tasksil.params.step*(tasksil.frameread))
[13c3fba]	77	if not tasksil.issilence:
	78	pitch.append(self.truth)
	79	else:
	80	pitch.append(-1.)
[4798fdf]	81	return time,pitch
[13c3fba]	82	except ValueError:
	83	# FIXME very weak check
	84	if not os.path.isfile(datafile):
	85	print "no ground truth found"
	86	return 0,0
	87	else:
	88	from aubio.txtfile import read_datafile
	89	values = read_datafile(datafile)
	90	time, pitch = [], []
	91	for i in range(len(values)):
	92	time.append(values[i][0])
[83376e9]	93	if values[i][1] == 0.0:
	94	pitch.append(-1.)
	95	else:
	96	pitch.append(aubio_freqtomidi(values[i][1]))
[13c3fba]	97	return time,pitch
	98
[5d1c070]	99	def oldeval(self,results):
[13c3fba]	100	def mmean(l):
	101	return sum(l)/max(float(len(l)),1)
	102
	103	from aubio.median import percental
	104	timet,pitcht = self.gettruth()
	105	res = []
	106	for i in results:
	107	#print i,self.truth
	108	if i <= 0: pass
	109	else: res.append(self.truth-i)
	110	if not res or len(res) < 3:
	111	avg = self.truth; med = self.truth
	112	else:
	113	avg = mmean(res)
	114	med = percental(res,len(res)/2)
	115	return self.truth, self.truth-med, self.truth-avg
	116
[83376e9]	117	def eval(self,pitch,tol=0.5):
[13c3fba]	118	timet,pitcht = self.gettruth()
[5d1c070]	119	pitch = [aubio_freqtomidi(i) for i in pitch]
	120	for i in range(len(pitch)):
	121	if pitch[i] == "nan" or pitch[i] == -1:
	122	pitch[i] = -1
[4798fdf]	123	time = [ (i+self.params.pitchdelay)*self.params.step for i in range(len(pitch)) ]
[83376e9]	124	#print len(timet),len(pitcht)
	125	#print len(time),len(pitch)
	126	if len(timet) != len(time):
	127	time = time[1:len(timet)+1]
	128	pitch = pitch[1:len(pitcht)+1]
	129	#pitcht = [aubio_freqtomidi(i) for i in pitcht]
	130	for i in range(len(pitcht)):
	131	if pitcht[i] == "nan" or pitcht[i] == "-inf" or pitcht[i] == -1:
	132	pitcht[i] = -1
	133	assert len(timet) == len(time)
[5d1c070]	134	assert len(pitcht) == len(pitch)
	135	osil, esil, opit, epit, echr = 0, 0, 0, 0, 0
	136	for i in range(len(pitcht)):
	137	if pitcht[i] == -1: # currently silent
	138	osil += 1 # count a silence
[83376e9]	139	if pitch[i] <= 0. or pitch[i] == "nan":
[5d1c070]	140	esil += 1 # found a silence
	141	else:
	142	opit +=1
	143	if abs(pitcht[i] - pitch[i]) < tol:
	144	epit += 1
	145	echr += 1
	146	elif abs(pitcht[i] - pitch[i]) % 12. < tol:
	147	echr += 1
	148	#else:
	149	# print timet[i], pitcht[i], time[i], pitch[i]
	150	#print "origsilence", "foundsilence", "origpitch", "foundpitch", "orig pitchroma", "found pitchchroma"
	151	#print 100.esil/float(osil), 100.epit/float(opit), 100.*echr/float(opit)
	152	return osil, esil, opit, epit, echr
[13c3fba]	153
[fb8bde7]	154	def plot(self,pitch,wplot,oplots,titles,outplot=None):
[13c3fba]	155	import Gnuplot
	156
[4798fdf]	157	time = [ (i+self.params.pitchdelay)*self.params.step for i in range(len(pitch)) ]
[83376e9]	158	pitch = [aubio_freqtomidi(i) for i in pitch]
[5f23f66]	159	oplots.append(Gnuplot.Data(time,pitch,with_='lines',
[13c3fba]	160	title=self.params.pitchmode))
[fb8bde7]	161	titles.append(self.params.pitchmode)
[13c3fba]	162
	163
[470a772]	164	def plotplot(self,wplot,oplots,titles,outplot=None,extension=None,xsize=1.,ysize=1.,multiplot = 1, midi = 1, truth = 1):
	165	from aubio.gnuplot import gnuplot_create , audio_to_array, make_audio_plot
[13c3fba]	166	import re
	167	import Gnuplot
	168
	169	# check if ground truth exists
[fb8bde7]	170	if truth:
	171	timet,pitcht = self.gettruth()
	172	if timet and pitcht:
[5f23f66]	173	oplots = [Gnuplot.Data(timet,pitcht,with_='lines',
[fb8bde7]	174	title='ground truth')] + oplots
[13c3fba]	175
[470a772]	176	g = gnuplot_create(outplot=outplot, extension=extension)
[13c3fba]	177	g('set title \'%s\'' % (re.sub('.*/','',self.input)))
[470a772]	178	g('set size %f,%f' % (xsize,ysize) )
[13c3fba]	179	g('set multiplot')
	180	# hack to align left axis
[fb8bde7]	181	g('set lmargin 4')
	182	g('set rmargin 4')
[470a772]	183	# plot waveform
	184	time,data = audio_to_array(self.input)
	185	wplot = [make_audio_plot(time,data)]
	186	g('set origin 0,%f' % (0.7*ysize) )
	187	g('set size %f,%f' % (xsize,0.3*ysize))
	188	#g('set format y "%1f"')
[13c3fba]	189	g('set xrange [0:%f]' % max(time))
	190	g('set yrange [-1:1]')
[fb8bde7]	191	g('set noytics')
	192	g('set y2tics -1,1')
[470a772]	193	g.xlabel('time (s)',offset=(0,0.7))
[13c3fba]	194	g.ylabel('amplitude')
[470a772]	195	g.plot(*wplot)
[13c3fba]	196
[470a772]	197	# default settings for next plots
	198	g('unset title')
	199	g('set format x ""')
	200	g('set format y "%3e"')
	201	g('set tmargin 0')
	202	g.xlabel('')
	203	g('set noclip one')
[13c3fba]	204
[83376e9]	205	if not midi:
	206	g('set log y')
	207	#g.xlabel('time (s)')
	208	g.ylabel('f0 (Hz)')
	209	g('set yrange [100:%f]' % self.params.pitchmax)
	210	else:
[4798fdf]	211	g.ylabel('midi')
[650e39b]	212	g('set yrange [%f:%f]' % (aubio_freqtomidi(self.params.pitchmin), aubio_freqtomidi(self.params.pitchmax)))
[fb8bde7]	213	g('set y2tics %f,%f' % (round(aubio_freqtomidi(self.params.pitchmin)+.5),12))
[470a772]	214
[13c3fba]	215	if multiplot:
[470a772]	216	N = len(oplots)
	217	y = 0.7*ysize # the vertical proportion of the plot taken by onset functions
	218	delta = 0.035 # the constant part of y taken by last plot label and data
	219	for i in range(N):
	220	# plot pitch detection functions
	221	g('set size %f,%f' % ( xsize, (y-delta)/N))
	222	g('set origin 0,%f' % ((N-i-1)*(y-delta)/N + delta ))
[fb8bde7]	223	g('set nokey')
[470a772]	224	g('set xrange [0:%f]' % max(time))
[fb8bde7]	225	g.ylabel(titles[i])
[470a772]	226	if i == N-1:
	227	g('set size %f,%f' % (xsize, (y-delta)/N + delta ) )
	228	g('set origin 0,0')
	229	g.xlabel('time (s)', offset=(0,0.7))
	230	g('set format x')
[13c3fba]	231	g.plot(oplots[i])
	232	else:
[470a772]	233	g('set key right top')
[13c3fba]	234	g.plot(*oplots)
	235	g('unset multiplot')
	236

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

Download in other formats: