def run(self, fname):
# load audio and eqLoudness
# Note: MonoLoader resamples the audio signal to 44100 Hz by default
audio = estd.MonoLoader(filename=fname)()
audio = estd.EqualLoudness()(audio)
contours_bins, contours_start_times, contour_saliences, duration = \
self._extract_pitch_contours(audio)
# run the simplified contour selection
[pitch,
pitch_salience] = self.select_contours(contours_bins,
contour_saliences,
contours_start_times, duration)
# cent to Hz conversion
pitch = [
0. if p == 0 else 55. * 2.**(self.bin_resolution * p / 1200.)
for p in pitch
]
pitch = e_array(pitch)
pitch_salience = e_array(pitch_salience)
# pitch filter
if self.filter_pitch:
pitch, pitch_salience = self._post_filter_pitch(
pitch, pitch_salience)
# generate time stamps
time_stamps = self._gen_time_stamps(0, len(pitch))
# [time pitch salience] matrix
out = np.transpose(
np.vstack((time_stamps, pitch.tolist(), pitch_salience.tolist())))
out = out.tolist()
# settings
settings = self.get_settings()
settings.update({'source': fname})
return {'pitch': out, 'settings': settings}
def run(self, fname):
# load audio and eqLoudness
# Note: MonoLoader resamples the audio signal to 44100 Hz by default
audio = estd.MonoLoader(filename=fname)()
audio = estd.EqualLoudness()(audio)
contours_bins, contours_start_times, contour_saliences, duration = \
self._extract_pitch_contours(audio)
# run the simplified contour selection
[pitch, pitch_salience] = self.select_contours(
contours_bins, contour_saliences, contours_start_times, duration)
# cent to Hz conversion
pitch = [0. if p == 0
else 55. * 2. ** (self.bin_resolution * p / 1200.)
for p in pitch]
pitch = e_array(pitch)
pitch_salience = e_array(pitch_salience)
# pitch filter
if self.filter_pitch:
pitch, pitch_salience = self._post_filter_pitch(
pitch, pitch_salience)
# generate time stamps
time_stamps = self._gen_time_stamps(0, len(pitch))
# [time pitch salience] matrix
out = np.transpose(
np.vstack((time_stamps, pitch.tolist(), pitch_salience.tolist())))
out = out.tolist()
# settings
settings = self.get_settings()
settings.update({'source': fname})
return {'pitch': out, 'settings': settings}
def run(self, audio):
contours_bins, contours_start_times, contour_saliences, duration = \
self._extract_pitch_contours(audio)
tmp = []
tmp_times = []
for i, j in zip(contours_bins, contours_start_times):
values = [
0. if p == 0 else 55. * 2.**(self.bin_resolution * p / 1200.)
for p in i
]
time_stamps = self._gen_time_stamps(0, len(i))
time_stamps = [x + j for x in time_stamps]
tmp.append(values)
tmp_times.append(time_stamps)
#print(np.shape(tmp))
#print(np.shape(tmp_times))
#y, sr = librosa.load(fname)
#D = librosa.stft(y) # STFT of y
#S_db = librosa.amplitude_to_db(np.abs(D), ref=np.max)
#plt.figure(figsize=[30, 12])
#fig, ax = plt.subplots()
#for i, j in zip(tmp_times, tmp):
#plt.plot(i, j, linewidth=1.0, c='b', zorder=10)
#img = librosa.display.specshow(S_db, x_axis='time', y_axis='linear', ax=ax)
#ax.set(title='Now with labeled axes!')
#fig.colorbar(img, ax=ax, format="%+2.f dB")
# run the simplified contour selection
[pitch,
pitch_salience] = self.select_contours(contours_bins,
contour_saliences,
contours_start_times, duration)
# cent to Hz conversion
pitch = [
0. if p == 0 else 55. * 2.**(self.bin_resolution * p / 1200.)
for p in pitch
]
pitch = e_array(pitch)
pitch_salience = e_array(pitch_salience)
# pitch filter
if self.filter_pitch:
pitch, pitch_salience = self._post_filter_pitch(
pitch, pitch_salience)
# generate time stamps
time_stamps = self._gen_time_stamps(0, len(pitch))
#plt.plot(time_stamps, pitch.tolist(), c='orange', linewidth=2.5, zorder=0)
#filename_to_plot = fname.split('/')[1].replace('.wav', '')
#plt.savefig('outputs_figs/' + filename_to_plot + '.png')
#plt.show()
# [time pitch salience] matrix
out = np.transpose(
np.vstack((time_stamps, pitch.tolist(), pitch_salience.tolist())))
out = out.tolist()
# settings
settings = self.get_settings()
return {'pitch': out, 'settings': settings}
def run(self, musicbrainzid, fname):
citation = u"""
Atlı, H. S., Uyar, B., Şentürk, S., Bozkurt, B., and Serra, X.
(2014). Audio feature extraction for exploring Turkish makam music.
In Proceedings of 3rd International Conference on Audio Technologies
for Music and Media, Ankara, Turkey.
"""
run_windowing = estd.Windowing(zeroPadding = 3 * self.settings.frameSize) # Hann window with x4 zero padding
run_spectrum = estd.Spectrum(size=self.settings.frameSize * 4)
run_spectral_peaks = estd.SpectralPeaks(minFrequency=self.settings.minFrequency,
maxFrequency = self.settings.maxFrequency,
sampleRate = self.settings.sampleRate,
magnitudeThreshold = self.settings.magnitudeThreshold,
orderBy = 'magnitude')
run_pitch_salience_function = estd.PitchSalienceFunction(binResolution=self.settings.binResolution) # converts unit to cents, 55 Hz is taken as the default reference
run_pitch_salience_function_peaks = estd.PitchSalienceFunctionPeaks(binResolution=self.settings.binResolution,
minFrequency=self.settings.minFrequency,
maxFrequency = self.settings.maxFrequency)
run_pitch_contours = estd.PitchContours(hopSize=self.settings.hopSize,
binResolution=self.settings.binResolution,
peakDistributionThreshold = self.settings.peakDistributionThreshold)
run_pitch_filter = estd.PitchFilter(confidenceThreshold=self.settings.confidenceThreshold,
minChunkSize=self.settings.minChunkSize)
pool = Pool()
# load audio and eqLoudness
audio = estd.MonoLoader(filename = fname)() # MonoLoader resamples the audio signal to 44100 Hz by default
audio = estd.EqualLoudness()(audio)
for frame in estd.FrameGenerator(audio,frameSize=self.settings.frameSize, hopSize=self.settings.hopSize):
frame = run_windowing(frame)
spectrum = run_spectrum(frame)
peak_frequencies, peak_magnitudes = run_spectral_peaks(spectrum)
salience = run_pitch_salience_function(peak_frequencies, peak_magnitudes)
salience_peaks_bins, salience_peaks_contourSaliences = run_pitch_salience_function_peaks(salience)
if not size(salience_peaks_bins):
salience_peaks_bins = array([0])
if not size(salience_peaks_contourSaliences):
salience_peaks_contourSaliences = array([0])
pool.add('allframes_salience_peaks_bins', salience_peaks_bins)
pool.add('allframes_salience_peaks_contourSaliences', salience_peaks_contourSaliences)
# post-processing: contour tracking
contours_bins, contours_contourSaliences, contours_start_times, duration = run_pitch_contours(
pool['allframes_salience_peaks_bins'],
pool['allframes_salience_peaks_contourSaliences'])
# run the simplified contour selection
[pitch, pitch_salience] = self.ContourSelection(contours_bins,contours_contourSaliences,contours_start_times,duration)
# cent to Hz conversion
pitch = e_array([0. if p == 0 else 55.*(2.**(((self.settings.binResolution*(p)))/1200)) for p in pitch])
pitch_salience = e_array(pitch_salience)
# pitch filter
if self.settings.filterPitch:
pitch = run_pitch_filter(pitch, pitch_salience)
# generate time stamps
time_stamps = [s*self.settings.hopSize/float(self.settings.sampleRate) for s in xrange(0,len(pitch))]
# [time pitch salience] matrix
out = transpose(vstack((time_stamps, pitch.tolist(), pitch_salience.tolist())))
out = out.tolist()
# settings
settings = self.settings
settings.update({'version':self._version,
'slug':self._slug,
'source': fname,
'essentiaVersion': essentia_version,
'pitchUnit': 'Hz',
'citation': citation})
# matlab
matout = cStringIO.StringIO()
matob = {'pitch': out}
matob.update(settings)
scipy.io.savemat(matout, matob)
return {'pitch': out,
'matlab': matout.getvalue(),
'settings': settings}
