def compute(audio, pool, options):
INFO('Computing Tempo extractor...')
use_onset = options['useOnset']
use_bands = options['useBands']
# frameNumber * hopSize ~= about 6 seconds
hopSize = options['hopSize']
frameSize = options['frameSize']
frameNumber = options['frameNumber']
frameHop = options['frameHop']
sampleRate = options['sampleRate']
tolerance = 0.24
period_tol = 2
windowType = options['windowType']
bands_freq = [
40.0, 413.16, 974.51, 1818.94, 3089.19, 5000.0, 7874.4, 12198.29,
17181.13
]
bands_gain = [2.0, 3.0, 2.0, 1.0, 1.2, 2.0, 3.0, 2.5]
maxbpm = 208
minbpm = 40
last_beat_interval = 0.025
frame_time = float(hopSize) / float(sampleRate)
frames = essentia.FrameGenerator(audio=audio,
frameSize=frameSize,
hopSize=hopSize)
window = essentia.Windowing(size=frameSize, zeroPadding=0, type=windowType)
if use_onset:
fft = essentia.FFT(size=frameSize)
cartesian2polar = essentia.CartesianToPolar()
onset_hfc = essentia.OnsetDetection(method='hfc',
sampleRate=sampleRate)
onset_complex = essentia.OnsetDetection(method='complex',
sampleRate=sampleRate)
if use_bands:
espectrum = essentia.Spectrum(size=frameSize)
tempotapbands = essentia.FrequencyBands(frequencyBands=bands_freq)
temposcalebands = essentia.TempoScaleBands(bandsGain=bands_gain)
tempotap = essentia.TempoTap(numberFrames=frameNumber,
sampleRate=sampleRate,
frameHop=frameHop)
tempotapticks = essentia.TempoTapTicks(hopSize=hopSize,
sampleRate=sampleRate,
frameHop=frameHop)
frameTime = float(hopSize) / float(sampleRate)
frameRate = 1. / frameTime
nframes = 0
bpm_estimates_list = []
ticks = []
matchingPeriods = []
oldhfc = 0
fileLength = len(audio) / sampleRate
startSilence = 0
oldSilence = 0
endSilence = round(fileLength * sampleRate / hopSize) + 1
for frame in frames:
windowed_frame = window(frame)
features = []
if use_onset:
complex_fft = fft(windowed_frame)
(spectrum, phase) = cartesian2polar(complex_fft)
hfc = onset_hfc(spectrum, phase)
complexdomain = onset_complex(spectrum, phase)
difhfc = max(hfc - oldhfc, 0)
oldhfc = hfc
features += [hfc, difhfc, complexdomain]
if use_bands:
spectrum_frame = espectrum(windowed_frame)
bands = tempotapbands(spectrum_frame)
(scaled_bands, cumul) = temposcalebands(bands)
features += list(scaled_bands)
features = essentia.array(features)
(periods, phases) = tempotap(features)
(these_ticks, these_matchingPeriods) = tempotapticks(periods, phases)
for period in these_matchingPeriods:
if period != 0:
matchingPeriods += [period]
ticks += list(these_ticks)
if nframes < 5. * sampleRate / hopSize:
if isSilent(frame) and startSilence == nframes - 1:
startSilence = nframes
if nframes > (fileLength - 5.) * sampleRate / hopSize:
if isSilent(frame):
if oldSilence != nframes - 1:
endSilence = nframes
oldSilence = nframes
nframes += 1
# make sure we do not kill beat too close to music
if startSilence > 0: startSilence -= 1
endSilence += 1
# fill the rest of buffer with zeros
features = essentia.array([0] * len(features))
while nframes % frameNumber != 0:
(periods, phases) = tempotap(features)
(these_ticks, these_matchingPeriods) = tempotapticks(periods, phases)
ticks += list(these_ticks)
matchingPeriods += list(these_matchingPeriods)
nframes += 1
if len(ticks) > 2:
# fill up to end of file
if fileLength > ticks[-1]:
lastPeriod = ticks[-1] - ticks[-2]
while ticks[-1] + lastPeriod < fileLength - last_beat_interval:
if ticks[-1] > fileLength - last_beat_interval:
break
ticks.append(ticks[-1] + lastPeriod)
if len(ticks) > 1:
# remove all negative ticks
i = 0
while i < len(ticks):
if ticks[i] < startSilence / sampleRate * hopSize: ticks.pop(i)
else: i += 1
# kill all ticks from 350ms before the end of the song
i = 0
while i < len(ticks):
if ticks[i] > endSilence / sampleRate * hopSize: ticks.pop(i)
else: i += 1
# prune values closer than tolerance
i = 1
while i < len(ticks):
if ticks[i] - ticks[i - 1] < tolerance: ticks.pop(i)
else: i += 1
# prune all backward offbeat
i = 3
while i < len(ticks):
if abs( (ticks[i] - ticks[i-2]) - 1.5 * (ticks[i] - ticks[i-1]) ) < 0.100 \
and abs( (ticks[i] - ticks[i-1]) - (ticks[i-2] - ticks[i-3]) ) < 0.100 :
ticks.pop(i - 2)
else:
i += 1
for period in matchingPeriods:
if period != 0:
bpm_estimates_list += [lagtobpm(period, sampleRate, hopSize)]
#else:
# bpm_estimates_list += [ 0 ]
# bpm estimates
for bpm_estimate in bpm_estimates_list:
pool.add(namespace + '.' + 'bpm_estimates', bpm_estimate)
# estimate the bpm from the list of candidates
if len(bpm_estimates_list) > 0:
estimates = [bpm / 2. for bpm in bpm_estimates_list]
closestBpm = argmax(bincount(estimates)) * 2.
matching = []
for bpm in bpm_estimates_list:
if abs(closestBpm - bpm) < period_tol:
matching.append(bpm)
if (len(matching) < 1):
# something odd happened
bpm = closestBpm
else:
bpm = mean(matching)
else:
bpm = 0.
# convert to floats, as python bindings yet not support numpy.float32
ticks = [float(tick) for tick in ticks]
pool.add(namespace + '.' + 'bpm', bpm) #, pool.GlobalScope)
pool.add(namespace + '.' + 'beats_position', ticks) #, pool.GlobalScope
bpm_intervals = [ticks[i] - ticks[i - 1] for i in range(1, len(ticks))]
pool.add(namespace + '.' + 'bpm_intervals',
bpm_intervals) #, pool.GlobalScope
from numpy import histogram
tempotap_bpms = [60. / i for i in bpm_intervals]
if len(tempotap_bpms) > 0:
weight, values = histogram(tempotap_bpms,
bins=250,
range=(0, 250),
normed=True)
else:
weight, values = [0.], [0.]
first_peak_weights = [0] * 250
secnd_peak_weights = [0] * 250
for i in range(max(argmax(weight) - 4, 0),
min(argmax(weight) + 5, len(weight))):
first_peak_weights[i] = weight[i]
weight[i] = 0.
for i in range(max(argmax(weight) - 4, 0),
min(argmax(weight) + 5, len(weight))):
secnd_peak_weights[i] = weight[i]
weight[i] = 0.
pool.add(namespace + '.' + 'first_peak_bpm',
values[argmax(first_peak_weights)]) #, pool.GlobalScope
pool.add(
namespace + '.' + 'first_peak_weight',
first_peak_weights[argmax(first_peak_weights)]) #, pool.GlobalScope
if sum(first_peak_weights) != 0.:
pool.add(namespace + '.' + 'first_peak_spread',
1. - first_peak_weights[argmax(first_peak_weights)] /
sum(first_peak_weights)) #, pool.GlobalScope
else:
pool.add(namespace + '.' + 'first_peak_spread',
0.) #, pool.GlobalScope
pool.add(namespace + '.' + 'second_peak_bpm',
values[argmax(secnd_peak_weights)]) #, pool.GlobalScope
pool.add(
namespace + '.' + 'second_peak_weight',
secnd_peak_weights[argmax(secnd_peak_weights)]) #, pool.GlobalScope
if sum(secnd_peak_weights) != 0.:
pool.add(namespace + '.' + 'second_peak_spread',
1. - secnd_peak_weights[argmax(secnd_peak_weights)] /
sum(secnd_peak_weights)) #, pool.GlobalScope
else:
pool.add(namespace + '.' + 'second_peak_spread',
0.) #, pool.GlobalScope
'''
def rubato(ticks):
bpm_rubato_python = []
tolerance = 0.08
i = 5
tmp1 = 60./ float(ticks[i ] - ticks[i-1])
tmp2 = 60./ float(ticks[i-1] - ticks[i-2])
tmp3 = 60./ float(ticks[i-2] - ticks[i-3])
tmp4 = 60./ float(ticks[i-3] - ticks[i-4])
tmp5 = 60./ float(ticks[i-4] - ticks[i-5])
for i in range(6, len(ticks)):
if ( abs(1. - tmp1 / tmp4) >= tolerance
and abs(1. - tmp2 / tmp5) >= tolerance
and abs(1. - tmp2 / tmp4) >= tolerance
and abs(1. - tmp1 / tmp5) >= tolerance
and abs(1. - tmp1 / tmp2) <= tolerance
and abs(1. - tmp4 / tmp5) <= tolerance ):
bpm_rubato_python.append(ticks[i-2])
tmp5 = tmp4; tmp4 = tmp3; tmp3 = tmp2; tmp2 = tmp1
tmp1 = 60./ (ticks[i] - ticks[i-1])
print bpm_rubato_python
return bpm_rubato_python
'''
# FIXME we need better rubato algorithm
#rubato = essentia.BpmRubato()
#bpm_rubato_start, bpm_rubato_stop = rubato(ticks)
#pool.add(namespace + '.' + 'rubato_start', bpm_rubato_start)#, pool.GlobalScope
#pool.add(namespace + '.' + 'rubato_stop', bpm_rubato_stop)#, pool.GlobalScope)
INFO('100% done...')
def compute(audio, pool, options):
sampleRate = options['sampleRate']
frameSize = options['frameSize']
hopSize = options['hopSize']
zeroPadding = options['zeroPadding']
windowType = options['windowType']
frameRate = float(sampleRate) / float(frameSize - hopSize)
INFO('Computing Onset Detection...')
frames = essentia.FrameGenerator(audio=audio,
frameSize=frameSize,
hopSize=hopSize)
window = essentia.Windowing(size=frameSize,
zeroPadding=zeroPadding,
type=windowType)
fft = essentia.FFT()
cartesian2polar = essentia.CartesianToPolar()
onsetdetectionHFC = essentia.OnsetDetection(method="hfc",
sampleRate=sampleRate)
onsetdetectionComplex = essentia.OnsetDetection(method="complex",
sampleRate=sampleRate)
onsets = essentia.Onsets(frameRate=frameRate)
total_frames = frames.num_frames()
n_frames = 0
start_of_frame = -frameSize * 0.5
hfc = []
complex = []
progress = Progress(total=total_frames)
for frame in frames:
if essentia.instantPower(frame) < 1.e-4:
total_frames -= 1
start_of_frame += hopSize
hfc.append(0.)
complex.append(0.)
continue
windowed_frame = window(frame)
complex_fft = fft(windowed_frame)
(spectrum, phase) = cartesian2polar(complex_fft)
hfc.append(onsetdetectionHFC(spectrum, phase))
complex.append(onsetdetectionComplex(spectrum, phase))
# display of progress report
progress.update(n_frames)
n_frames += 1
start_of_frame += hopSize
# The onset rate is defined as the number of onsets per seconds
detections = numpy.concatenate(
[essentia.array([hfc]),
essentia.array([complex])])
# prune all 'doubled' detections
time_onsets = list(onsets(detections, essentia.array([1, 1])))
t = 1
while t < len(time_onsets):
if time_onsets[t] - time_onsets[t - 1] < 0.080: time_onsets.pop(t)
else: t += 1
onsetrate = len(time_onsets) / (len(audio) / sampleRate)
pool.add(namespace + '.' + "onset_times",
essentia.array(time_onsets)) #, pool.GlobalScope)
pool.add(namespace + '.' + "onset_rate", onsetrate) #, pool.GlobalScope)
progress.finish()