def computeNoveltyCurve(self,
filename,
frameSize=1024,
hopSize=512,
windowType='hann',
weightCurveType='inverse_quadratic',
sampleRate=44100.0,
startTime=0,
endTime=2000):
loader = EasyLoader(filename=filename,
startTime=startTime,
endTime=endTime,
sampleRate=sampleRate,
downmix='left')
fc = FrameCutter(frameSize=frameSize,
hopSize=hopSize,
silentFrames="keep",
startFromZero=False,
lastFrameToEndOfFile=True)
window = Windowing(type=windowType,
zeroPhase=True,
zeroPadding=1024 - frameSize)
freqBands = FrequencyBands(
sampleRate=sampleRate) # using barkbands by default
pool = Pool()
spec = Spectrum()
loader.audio >> fc.signal
fc.frame >> window.frame >> spec.frame
spec.spectrum >> freqBands.spectrum
freqBands.bands >> (pool, 'frequency_bands')
essentia.run(loader)
noveltyCurve = std.NoveltyCurve(frameRate=sampleRate / float(hopSize),
weightCurveType=weightCurveType)(
pool['frequency_bands'])
return noveltyCurve
frameSize = 2048
hopSize = 128
weight = 'hybrid'
print("Frame size: %d" % frameSize)
print("Hop size: %d" % hopSize)
print("weight: %s" % weight)
audio = es.MonoLoader(filename=input_file)()
w = es.Windowing(type='hann')
s = es.Spectrum()
freq_bands = es.FrequencyBands()
bands_energies = []
for frame in es.FrameGenerator(audio, frameSize=frameSize, hopSize=hopSize):
bands_energies.append(freq_bands(s(w(frame))))
novelty = es.NoveltyCurve(frameRate=44100. / hopSize,
weightCurveType=weight)(numpy.array(bands_energies))
bpm, candidates, magnitudes, tempogram, _, ticks, ticks_strength, sinusoid = es.BpmHistogram(
frameRate=44100. / hopSize)(novelty)
print("BPM = %0.1f" % bpm)
#pylab.plot(novelty)
#pylab.show()
pylab.matshow(tempogram.transpose(), origin='lower', aspect='auto')
pylab.show()
def computeNoveltyCurve(filename, pool):
loader = EasyLoader(filename=filename,
sampleRate=pool['samplerate'],
startTime=STARTTIME,
endTime=ENDTIME,
downmix=pool['downmix'])
fc = FrameCutter(frameSize=int(pool['framesize']),
silentFrames='noise',
hopSize=int(pool['hopsize']),
startFromZero=False)
window = Windowing(type=pool['window'], zeroPhase=False)
#freqBands = FrequencyBands(frequencyBands=EqBands, sampleRate=pool['samplerate'])
freqBands = FrequencyBands(sampleRate=pool['samplerate'])
spec = Spectrum()
hfc = HFC()
loader.audio >> fc.signal
fc.frame >> window.frame >> spec.frame
spec.spectrum >> freqBands.spectrum
spec.spectrum >> hfc.spectrum
freqBands.bands >> (pool, 'frequency_bands')
hfc.hfc >> (pool, 'hfc')
essentia.run(loader)
pool.set('size', loader.audio.totalProduced())
pool.set('length', pool['size'] / pool['samplerate'])
# compute a weighting curve that is according to frequency bands:
frequencyBands = pool['frequency_bands']
nFrames = len(frequencyBands)
weightCurve = np.sum(frequencyBands, axis=0)
weightCurve = [val / float(nFrames) for val in weightCurve]
weightCurve = essentia.normalize(weightCurve)
#pyplot.plot(weightCurve)
#pyplot.show()
noveltyCurve = std.NoveltyCurve(frameRate=pool['framerate'],
weightCurveType=pool['weight'],
weightCurve=weightCurve)(frequencyBands)
#for x in noveltyCurve: pool.add('novelty_curve', x)
#return
# derivative of hfc seems to help in finding more precise beats...
hfc = essentia.normalize(pool['hfc'])
dhfc = essentia.derivative(hfc)
for i, val in enumerate(dhfc):
if val < 0: continue
noveltyCurve[i] += val
# low pass filter novelty curve:
env = std.Envelope(attackTime=2. / pool['framerate'],
releaseTime=2. / pool['framerate'])(noveltyCurve)
# apply median filter:
windowSize = 8 #samples
size = len(env)
filtered = zeros(size)
for i in range(size):
start = i - windowSize
if start < 0: start = 0
end = start + windowSize
if end > size:
end = size
start = size - windowSize
filtered[i] = env[i] - np.median(env[start:end])
if filtered[i] < 0: filtered[i] = 0
#pyplot.subplot(311)
#pyplot.plot(noveltyCurve)
#pyplot.subplot(312)
#pyplot.plot(env, 'r')
#pyplot.subplot(313)
#pyplot.plot(filtered, 'g')
#pyplot.show()
#for x in noveltyCurve: pool.add('novelty_curve', x)
for x in filtered:
pool.add('novelty_curve', x)