def song_pitch(song, sr, threshold=None, freq_range=None, fft_step=None,
fft_size=None):
"""Return an array of pitch values for the whole song."""
if threshold is None:
threshold = 0.9
windows = get_windows(song, fft_step, fft_size)
nb_windows = windows.shape[0]
win_s = windows.shape[1]
pitch_o = aubio_pitch("yin", 2048, win_s, sr)
pitch_o.set_unit("freq")
pitch_o.set_tolerance(threshold)
pitches = np.zeros(nb_windows)
for i, window in enumerate(windows):
pitches[i] = pitch_o(window.astype(np.float32))
pitches[pitches > sr / 2] = 0
return pitches
def pitch(self, **options):
algorithm = options.get("algorithm") or "yin"
pitchUnit = options.get("pitchUnit") or "freq"
tolerance = options.get("tolerance") or 0.85
sourceBuffer = source(self.audioFilename, self.samplerate, self.hopSize)
samplerate = sourceBuffer.samplerate
pitchSampler = aubio_pitch(algorithm, self.fftWindowSize, self.hopSize, self.samplerate)
pitchSampler.set_unit(pitchUnit)
pitchSampler.set_tolerance(tolerance)
timings = []
frames = []
pitches = []
confidences = []
# total number of frames read
totalFrames = 0
while True:
samples, read = sourceBuffer()
pitchSample = pitchSampler(samples)[0]
confidence = pitchSampler.get_confidence()
timings += [float(totalFrames) / samplerate]
frames += [totalFrames]
pitches += [pitchSample]
confidences += [confidence]
totalFrames += read
if read < self.hopSize: break
return {
"timings": timings,
"frames": frames,
"pitches": pitches,
"confidences": confidences
}
def setup(self, channels=None, samplerate=None,
blocksize=None, totalframes=None):
# Frame parameters setup
if self._blocksize_s:
self.input_blocksize = nextpow2(self._blocksize_s * samplerate)
else:
self.input_blocksize = 2048
if self._stepsize_s:
self.input_stepsize = int(np.round(self._stepsize_s * samplerate))
else:
self.input_stepsize = self.input_blocksize / 2
# Now that frames size metadata are properly set, we can do the set-up
super(AubioPitch, self).setup(channels,
samplerate,
blocksize,
totalframes)
# Aubio Pitch set-up
self.aubio_pitch = aubio_pitch("default", self.input_blocksize,
self.input_stepsize, samplerate)
self.aubio_pitch.set_unit("freq")
def setup(self,
channels=None,
samplerate=None,
blocksize=None,
totalframes=None):
# Frame parameters setup
if self._blocksize_s:
self.input_blocksize = nextpow2(self._blocksize_s * samplerate)
else:
self.input_blocksize = 2048
if self._stepsize_s:
self.input_stepsize = int(np.round(self._stepsize_s * samplerate))
else:
self.input_stepsize = self.input_blocksize / 2
# Now that frames size metadata are properly set, we can do the set-up
super(AubioPitch, self).setup(channels, samplerate, blocksize,
totalframes)
# Aubio Pitch set-up
self.aubio_pitch = aubio_pitch("default", self.input_blocksize,
self.input_stepsize, samplerate)
self.aubio_pitch.set_unit("freq")
def notes(self, **options):
onsetAlgorithm = options.get("onsetAlgorithm") or "default"
algorithm = options.get("algorithm") or "yinfft"
pitchUnit = options.get("pitchUnit") or "freq"
tolerance = options.get("tolerance") or 0.85
sourceBuffer = source(self.audioFilename, self.samplerate, self.hopSize)
onsetSampler = aubio_onset(onsetAlgorithm, self.fftWindowSize, self.hopSize, self.samplerate)
pitchSampler = aubio_pitch(algorithm, self.fftWindowSize * 4, self.hopSize, self.samplerate)
pitchSampler.set_unit(pitchUnit)
pitchSampler.set_tolerance(tolerance)
median = 6
isReady = 0
note_buffer = []
timeStarts = []
timeStops = []
frameStarts = []
frameStops = []
pitches = []
confidences = []
isOn = False
totalFrames = 0
while True:
samples, read = sourceBuffer()
new_pitch = pitchSampler(samples)[0]
curlevel = aubio_level_detection(samples, -90)
note_buffer.insert(0, new_pitch)
if len(note_buffer) > median:
note_buffer.pop()
if onsetSampler(samples):
if curlevel == 1.:
isReady = 0
else:
isReady = 1
else:
if isReady > 0:
isReady += 1
if isReady == median:
new_note = self.pseudo_median(note_buffer)
if isOn:
timeStops += [float(totalFrames) / self.samplerate]
frameStops += [totalFrames]
isOn = False
if new_note > 45:
confidence = pitchSampler.get_confidence()
timeStarts += [float(totalFrames) / self.samplerate]
frameStarts += [totalFrames]
pitches += [new_note]
confidences += [confidence]
isOn = True
totalFrames += read
if read < self.hopSize: break
if len(frameStops) < len(frameStarts): frameStops += [totalFrames]
if len(timeStops) < len(timeStarts): timeStops += [timeStops]
# timeStops += [timeStops]
return {
"timeStarts": timeStarts,
"timeStops": timeStops,
"frameStarts": frameStarts,
"frameStops": frameStops,
"pitches": pitches,
"confidences": confidences
}