def get_file_bpm(path, params=None):
""" Calculate the beats per minute (bpm) of a given file.
path: path to the file
param: dictionary of parameters
"""
if params is None:
params = {}
# default:
samplerate, win_s, hop_s = 44100, 1024, 512
if 'mode' in params:
if params.mode in ['super-fast']:
# super fast
samplerate, win_s, hop_s = 4000, 128, 64
elif params.mode in ['fast']:
# fast
samplerate, win_s, hop_s = 8000, 512, 128
elif params.mode in ['default']:
pass
else:
print("unknown mode {:s}".format(params.mode))
# manual settings
if 'samplerate' in params:
samplerate = params.samplerate
if 'win_s' in params:
win_s = params.win_s
if 'hop_s' in params:
hop_s = params.hop_s
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
o = tempo("specdiff", win_s, hop_s, samplerate)
# List of beats, in samples
beats = []
# Total number of frames read
total_frames = 0
while True:
samples, read = s()
is_beat = o(samples)
if is_beat:
this_beat = o.get_last_s()
beats.append(this_beat)
#if o.get_confidence() > .2 and len(beats) > 2.:
# break
total_frames += read
if read < hop_s:
break
def beats_to_bpm(beats, path):
# if enough beats are found, convert to periods then to bpm
if len(beats) > 1:
if len(beats) < 4:
print("few beats found in {:s}".format(path))
bpms = 60./diff(beats)
return median(bpms)
else:
print("not enough beats found in {:s}".format(path))
return 0
return beats_to_bpm(beats, path)
def __init__(self, filename, samplerate=44100, win_s=1024, hop_s=512):
'''
'''
self.source = source(filename, samplerate, hop_s)
self.win_s = win_s
self.hop_s = hop_s
def onset(self, **options):
onsetAlgorithm = options.get("onsetAlgorithm") or "default"
onsetThreshold = options.get("onsetThreshold") or -90
sourceBuffer = source(self.audioFilename, self.samplerate, self.hopSize)
onsetSampler = aubio_onset(onsetAlgorithm, self.fftWindowSize, self.hopSize, self.samplerate)
# onsetSampler = aubio_level_detection(-70)
# list of onsets, in samples
onsets = []
timings = []
frames = []
# total number of frames read
totalFrames = 0
while True:
samples, read = sourceBuffer()
if onsetSampler(samples):
# print "%f" % onsetSampler.get_last_s()
timings.append(onsetSampler.get_last_s())
frames += [totalFrames]
onsets.append(onsetSampler.get_last())
totalFrames += read
if read < self.hopSize: break
return {
"timings": timings,
"frames": frames,
"onsets": onsets
}
def pitches(filename, samplerate=None, downsample=1):
if samplerate is None:
samplerate = 44100 / downsample
win_s = 4096 / downsample # fft size
hop_s = 512 / downsample # hop size
s = aubio.source(filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = aubio.pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("midi")
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# total number of frames read
total_frames = 0
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
#pitch = int(round(pitch))
confidence = pitch_o.get_confidence()
#if confidence < 0.8: pitch = 0.
#print "%f %f %f" % (total_frames / float(samplerate), pitch, confidence)
pitches += [pitch]
confidences += [confidence]
total_frames += read
if read < hop_s: break
return samples, pitches, confidences, total_frames
def mfcc(self, **options):
nFilters = options.get("nFilters") or 40 # must be 40 for mfcc
nCoeffs = options.get("nCoefs") or 13
sourceBuffer = source(self.audioFilename, self.samplerate, self.hopSize)
pvocBuffer = pvoc(self.fftWindowSize, self.hopSize)
mfccBuffer = mfcc(self.fftWindowSize, nFilters, nCoeffs, self.samplerate)
mfccs = np.zeros([nCoeffs, ])
timings = []
frames = []
totalFrames = 0
while True:
samples, read = sourceBuffer()
spec = pvocBuffer(samples)
mfcc_out = mfccBuffer(spec)
mfccs = np.vstack((mfccs, mfcc_out))
totalFrames += read
timings += [float(totalFrames) / self.samplerate]
frames += [totalFrames]
if read < self.hopSize: break
return mfccs
def getPitch(filename):
from aubio import pitch
downsample = 1
samplerate = 44100 / downsample
win_s = 1024 / downsample # fft size
hop_s = 256/ downsample # hop size
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("midi")
pitch_o.set_tolerance(tolerance)
total_frames = 0
pitches = []
confidences=[]
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
confidence = pitch_o.get_confidence()
confidences+=[confidence]
pitches += [pitch]
total_frames += read
if read < hop_s: break
pitches = array(pitches[1:])
confidences = array(confidences[1:])
cleaned_pitches = pitches
cleaned_pitches = ma.masked_where(confidences < tolerance, cleaned_pitches,copy=False)
cleaned_pitches = cleaned_pitches[~cleaned_pitches.mask]
return max(cleaned_pitches)
def extract_beats (self):
samplerate, win_s, hop_s = 44100, 1024, 512
s = aubio.source('good_test.wav', 44100, 512)
o = aubio.tempo("specdiff", self.win_s, self.hop_s, self.rate)
self._beats = []
total_frames = 0
i = 0
print "Starting extraction ..."
while True:
samples = self._waveform [i*self.hop_s:(i+1)*self.hop_s]
samples = np.float32(samples)
is_beat = o (samples)
if is_beat:
this_beat = o.get_last_s()
print this_beat
self._beats.append(this_beat)
i += 1
if (i+1)*self.hop_s > len(self._waveform): break
#bpms = 60./np.diff(self._beats)
print "Beats:"
print self._beats
print "--- BMP:"
b = 60./np.diff(self._beats)
self._bpm = np.median(b)
print self._bpm
def tempo_ext(filename):
win_s = 512 # fft size
hop_s = win_s / 2 # hop size
samplerate = 11000
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
o = tempo("default", win_s, hop_s, samplerate)
# tempo detection delay, in samples
# default to 4 blocks delay to catch up with
delay = 4.0 * hop_s
# list of beats, in samples
beats = []
# total number of frames read
total_frames = 0
while True:
samples, read = s()
is_beat = o(samples)
if is_beat:
this_beat = int(total_frames - delay + is_beat[0] * hop_s)
# print "%f" % (this_beat / float(samplerate))
beats.append(this_beat)
total_frames += read
if read < hop_s:
break
# convert samples to seconds
beats = map(lambda x: x / float(samplerate), beats)
bpms = [60.0 / (b - a) for a, b in zip(beats[:-1], beats[1:])]
print "BPM: ", median(bpms) * 4
def getBmp(path, params={}):
try:
win_s = params['win_s']
samplerate = params['samplerate']
hop_s = params['hop_s']
except:
#default:
samplerate, win_s, hop_s = 44100, 1024, 512
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
o = tempo("specdiff", win_s, hop_s, samplerate)
#list of beats, in samples
beats = []
#Total number of frames read
total_frames = 0
while True:
samples, read = s()
is_beat = o(samples)
if is_beat:
this_beat = o.get_last_s()
beats.append(this_beat)
total_frames += read
if read < hop_s:
break
bpms = 60./np.diff(beats)
b = np.median(bpms)
return b
def test_samplerate_hopsize(self, hop_size, samplerate, soundfile):
try:
f = source(soundfile, samplerate, hop_size)
except RuntimeError as e:
self.skipTest('failed opening with hop_s = {:d}, samplerate = {:d} ({:s})'.format(hop_size, samplerate, str(e)))
assert f.samplerate != 0
self.read_from_source(f)
def test_close_file_twice(self):
samplerate = 0 # use native samplerate
hop_size = 256
for p in list_of_sounds:
f = source(p, samplerate, hop_size)
f.close()
f.close()
def aubiodata(filename, onsetType, threshold, silence): win_s = 512 # fft size hop_s = win_s / 2 # hop size samplerate = 0 # Just an initial setting # threshold = 0.8 # silence = -10 s = source(filename, samplerate, hop_s) samplerate = s.samplerate o = onset(onsetType, win_s, hop_s, samplerate) o.set_threshold(threshold) o.set_silence(silence) # list of onsets, in seconds onsets = [] # total number of frames read total_frames = 0 while True: samples, read = s() if o(samples): # print "%f" % o.get_last_s() onsets.append(o.get_last_s()) total_frames += read if read < hop_s: break # 'onsets' holds, at each entry, the time where an onset is detected (in seconds). return onsets
def apply_filter(path):
from aubio import source, sink, digital_filter
from os.path import basename, splitext
# open input file, get its samplerate
s = source(path)
samplerate = s.samplerate
# create an A-weighting filter
f = digital_filter(7)
f.set_a_weighting(samplerate)
# alternatively, apply another filter
# create output file
o = sink("filtered_" + splitext(basename(path))[0] + ".wav", samplerate)
total_frames = 0
while True:
samples, read = s()
filtered_samples = f(samples)
o(filtered_samples, read)
total_frames += read
if read < s.hop_size: break
duration = total_frames / float(samplerate)
print ("read {:s}".format(s.uri))
print ("applied A-weighting filtered ({:d} Hz)".format(samplerate))
print ("wrote {:s} ({:.2f} s)".format(o.uri, duration))
def getPitchDiff(path):
downsample = 1
samplerate = 44100 / downsample
win_s = 4096 / downsample # fft size
hop_s = 512 / downsample # hop size
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = aubio.pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("midi")
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# total number of frames read
total_frames = 0
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
#pitch = int(round(pitch))
confidence = pitch_o.get_confidence()
#if confidence < 0.8: pitch = 0.
#print "%f %f %f" % (total_frames / float(samplerate), pitch, confidence)
pitches += [pitch]
confidences += [confidence]
total_frames += read
if read < hop_s: break
return np.median(np.fabs(np.diff(pitches)))
def vocoder(file_id):
in_file = '/tmp/don_robot/output/' + file_id + ".mp3"
out_file = '/tmp/don_robot/output/' + file_id + "-voc.mp3"
samplerate = 44100
f = source(in_file, samplerate, 256)
g = sink(out_file, samplerate)
total_frames, read = 0, 256
win_s = 512 # fft size
hop_s = win_s // 2 # hop size
pv = pvoc(win_s, hop_s) # phase vocoder
while read:
samples, read = f()
spectrum = pv(samples) # compute spectrum
spectrum.norm *= .5 # reduce amplitude a bit .8
spectrum.phas[:] = 0. # zero phase
new_samples = pv.rdo(spectrum) # compute modified samples
g(new_samples, read) # write to output
total_frames += read
format_str = "read {:d} samples from {:s}, written to {:s}"
print(format_str.format(total_frames, f.uri, g.uri))
return out_file
def getMelEnergy(path):
win_s = 512 #fft size
hop_s = win_s / 4 #hop size
samplerate = 0
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
pv = pvoc(win_s, hop_s)
f = filterbank(40, win_s)
f.set_mel_coeffs_slaney(samplerate)
energies = np.zeros((40, ))
o = {}
total_frames = 0
downsample = 2
while True:
samples, read = s()
fftgrain = pv(samples)
new_energies = f(fftgrain)
energies = np.vstack([energies, new_energies])
total_frames += read
if read < hop_s:
break
return energies
def get_waveform_plot(filename, samplerate = 0, block_size = 4096, ax = None, downsample = 2**4):
import matplotlib.pyplot as plt
if not ax:
fig = plt.figure()
ax = fig.add_subplot(111)
hop_s = block_size
allsamples_max = zeros(0,)
downsample = downsample # to plot n samples / hop_s
a = source(filename, samplerate, hop_s) # source file
if samplerate == 0: samplerate = a.samplerate
total_frames = 0
while True:
samples, read = a()
# keep some data to plot it later
new_maxes = (abs(samples.reshape(hop_s//downsample, downsample))).max(axis=0)
allsamples_max = hstack([allsamples_max, new_maxes])
total_frames += read
if read < hop_s: break
allsamples_max = (allsamples_max > 0) * allsamples_max
allsamples_max_times = [ ( float (t) / downsample ) * hop_s for t in range(len(allsamples_max)) ]
ax.plot(allsamples_max_times, allsamples_max, '-b')
ax.plot(allsamples_max_times, -allsamples_max, '-b')
ax.axis(xmin = allsamples_max_times[0], xmax = allsamples_max_times[-1])
set_xlabels_sample2time(ax, allsamples_max_times[-1], samplerate)
return ax
def apply_filter(path, target):
# open input file, get its samplerate
s = aubio.source(path)
samplerate = s.samplerate
# create an A-weighting filter
f = aubio.digital_filter(7)
f.set_a_weighting(samplerate)
# create output file
o = aubio.sink(target, samplerate)
total_frames = 0
while True:
# read from source
samples, read = s()
# filter samples
filtered_samples = f(samples)
# write to sink
o(filtered_samples, read)
# count frames read
total_frames += read
# end of file reached
if read < s.hop_size:
break
# print some info
duration = total_frames / float(samplerate)
input_str = "input: {:s} ({:.2f} s, {:d} Hz)"
output_str = "output: {:s}, A-weighting filtered ({:d} frames total)"
print(input_str.format(s.uri, duration, samplerate))
print(output_str.format(o.uri, total_frames))
def test_wrong_seek_too_large(self):
f = source(self.default_test_sound)
try:
with self.assertRaises(ValueError):
f.seek(f.duration + f.samplerate * 10)
except AssertionError:
self.skipTest('seeking after end of stream failed raising ValueError')
def read_pitch(filename, chunk):
win_s = chunk # fft size
hop_s = win_s # hop size
samplerate = 0
# if len( sys.argv ) > 2: samplerate = int(sys.argv[2])
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
pitch_o = pitch("default", win_s, hop_s, samplerate)
pitch_o.set_unit("freq")
pitches = []
# times = []
# total number of frames read
total_frames = 0
while True:
samples, read = s()
cur_pitch = pitch_o(samples)[0]
# print "%f %f" % (total_frames / float(samplerate), pitch)
pitches.append(Pitch(cur_pitch, total_frames / float(samplerate)))
total_frames += read
if read < hop_s: break
# print pitches
if len(pitches) == 0: return None
return pitches
def get_file_bpm(self, path):
""" Calculate the beats per minute (bpm) of a given file.
path: path to the file
buf_size length of FFT
hop_size number of frames between two consecutive runs
samplerate sampling rate of the signal to analyze
"""
samplerate, buf_size, hop_size = bpm_slider_settings[
BPMOptionsPage.config.setting["bpm_slider_parameter"]]
mediasource = source(path.encode("utf-8"), samplerate, hop_size)
samplerate = mediasource.samplerate
beattracking = tempo("specdiff", buf_size, hop_size, samplerate)
# List of beats, in samples
beats = []
# Total number of frames read
total_frames = 0
while True:
samples, read = mediasource()
is_beat = beattracking(samples)
if is_beat:
this_beat = beattracking.get_last_s()
beats.append(this_beat)
total_frames += read
if read < hop_size:
break
# Convert to periods and to bpm
bpms = 60. / diff(beats)
return median(bpms)
def decoder(fname):
s = aubio.source(fname, samplerate, hop)
frames = []
while True:
samples, reads = s()
if reads == 0:
break
frames.extend(samples[0:reads])
#print len(frames)
max_amp = max(frames)
#print max_amp
for i in range(0, len(frames)):
frames[i] = frames[i]/max_amp
#print max(frames)
start = 0
while frames[start] < 0.3:
start = start+1
end = len(frames)-1
while frames[end] < 0.3:
end = end-1
samples = frames[start:end]
#print max(samples)
x = 0
#print len(samples)
freqs = []
while x + 2048 < len(samples):
freqs.append(find_freq(samples[x+20:x+20 + 2048]))
# print "-----------"
x = x + 2205
# print freqs
base64 = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/'
bits = ''
for x in freqs:
y = x[0]
z = x[1]
b64 = (y-400)/50
b642 = (z-4500)/50
bits = bits + base64[b64] + base64[b642]
# print len(bits)*6/8
return bits
def test_wrong_samplerate(self):
for p in list_of_sounds:
try:
f = source(p, -1)
except ValueError, e:
pass
else:
self.fail('negative samplerate does not raise ValueError')
def test_wrong_hop_size(self):
for p in list_of_sounds:
try:
f = source(p, 0, -1)
except ValueError, e:
pass
else:
self.fail('negative hop_size does not raise ValueError')
def test_wrong_hop_size(self):
for p in list_of_sounds:
try:
f = source(p, 0, -1)
except Exception, e:
print e
else:
self.fail('does not fail with wrong hop_size %d' % f.hop_size)
def test_wrong_samplerate(self):
for p in list_of_sounds:
try:
f = source(p, -1)
except Exception, e:
print e
else:
self.fail('does not fail with wrong samplerate')
def open_audio_source(input_wav):
"""
Open the audio source file `input_wav` and return a tuple with
- the aubio source object
- its samplerate
"""
s = aubio.source(input_wav, 0, HOP_S)
samplerate = s.samplerate
return s, samplerate
def test_duration(self, p):
total_frames = 0
f = source(p)
duration = f.duration
while True:
_, read = f()
total_frames += read
if read < f.hop_size: break
self.assertEqual(duration, total_frames)
def analyzePitch(grain):
s = source(grain["file"], int(grain["sampleRate"]), int(float(grain["frameCount"])))
samplerate = s.samplerate
tolerance = 0.8
pitch_out = pitch("yin", int(float(grain["frameCount"])), int(float(grain["frameCount"])), samplerate)
samples, read = s()
pitchFreq = pitch_out(samples)[0].item()
del s
return pitchFreq
def get_spectrogram(filename, samplerate=0):
win_s = 512 # fft window size
hop_s = win_s // 2 # hop size
fft_s = win_s // 2 + 1 # spectrum bins
a = source(filename, samplerate, hop_s) # source file
if samplerate == 0: samplerate = a.samplerate
pv = pvoc(win_s, hop_s) # phase vocoder
specgram = zeros([0, fft_s], dtype=float_type) # numpy array to store spectrogram
# analysis
while True:
samples, read = a() # read file
specgram = vstack((specgram, pv(samples).norm)) # store new norm vector
if read < a.hop_size: break
# plotting
fig = plt.imshow(log10(specgram.T + .001), origin='bottom', aspect='auto', cmap=plt.cm.gray_r)
ax = fig.axes
ax.axis([0, len(specgram), 0, len(specgram[0])])
# show axes in Hz and seconds
time_step = hop_s / float(samplerate)
total_time = len(specgram) * time_step
outstr = "total time: %0.2fs" % total_time
print(outstr + ", samplerate: %.2fkHz" % (samplerate / 1000.))
n_xticks = 10
n_yticks = 10
def get_rounded_ticks(top_pos, step, n_ticks):
top_label = top_pos * step
# get the first label
ticks_first_label = top_pos * step / n_ticks
# round to the closest .1
ticks_first_label = round(ticks_first_label * 10.) / 10.
# compute all labels from the first rounded one
ticks_labels = [ticks_first_label * n for n in range(n_ticks)] + [top_label]
# get the corresponding positions
ticks_positions = [ticks_labels[n] / step for n in range(n_ticks)] + [top_pos]
# convert to string
ticks_labels = ["%.1f" % x for x in ticks_labels]
# return position, label tuple to use with x/yticks
return ticks_positions, ticks_labels
# apply to the axis
x_ticks, x_labels = get_rounded_ticks(len(specgram), time_step, n_xticks)
y_ticks, y_labels = get_rounded_ticks(len(specgram[0]), (samplerate / 1000. / 2.) / len(specgram[0]), n_yticks)
ax.set_xticks(x_ticks)
ax.set_yticks(y_ticks)
ax.set_xticklabels(x_labels)
ax.set_yticklabels(y_labels)
ax.set_ylabel('Frequency (kHz)')
ax.set_xlabel('Time (s)')
ax.set_title(os.path.basename(filename))
for item in ([ax.title, ax.xaxis.label, ax.yaxis.label] +
ax.get_xticklabels() + ax.get_yticklabels()):
item.set_fontsize('x-small')
return fig
def test_duration(self, p):
total_frames = 0
f = source(p)
duration = f.duration
while True:
_, read = f()
total_frames += read
if read < f.hop_size: break
self.assertEqual(duration, total_frames)
def calculate_song_bpm(path: str, params: typing.Dict = None):
path = f'{DATA_FOLDER}/{path}'
if params is None:
params = {}
# default:
samplerate, win_s, hop_s = 44100, 1024, 512
if 'mode' in params:
if params.mode in ['super-fast']:
# super fast
samplerate, win_s, hop_s = 4000, 128, 64
elif params.mode in ['fast']:
# fast
samplerate, win_s, hop_s = 8000, 512, 128
elif params.mode in ['default']:
pass
else:
print("unknown mode {:s}".format(params.mode))
# manual settings
if 'samplerate' in params:
samplerate = params.samplerate
if 'win_s' in params:
win_s = params.win_s
if 'hop_s' in params:
hop_s = params.hop_s
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
o = tempo("specdiff", win_s, hop_s, samplerate)
# List of beats, in samples
beats = []
# Total number of frames read
total_frames = 0
while True:
samples, read = s()
is_beat = o(samples)
if is_beat:
this_beat = o.get_last_s()
beats.append(this_beat)
#if o.get_confidence() > .2 and len(beats) > 2.: # break
total_frames += read
if read < hop_s: break
def beats_to_bpm(beats, path):
if len(beats) > 1:
if len(beats) < 4:
print("few beats found in {:s}".format(path))
bpms = 60. / diff(beats) #bpm is an array
medinbpm = median(bpms) #medinbpm is a float number
return medinbpm #needs to be understood
else:
print("not enough beats found in {:s}".format(path))
return 0
# print "beats-in-bpm: ", beats
return beats_to_bpm(beats, path)
def read_file_aubio(filename):
import aubio
f = aubio.source(filename, hop_size = 1024)
total_frames = 0
while True:
_, read = f()
total_frames += read
if read < f.hop_size: break
return total_frames, f.samplerate
def get_spectrogram(filename, samplerate = 0):
win_s = 512 # fft window size
hop_s = win_s // 2 # hop size
fft_s = win_s // 2 + 1 # spectrum bins
a = source(filename, samplerate, hop_s) # source file
if samplerate == 0: samplerate = a.samplerate
pv = pvoc(win_s, hop_s) # phase vocoder
specgram = zeros([0, fft_s], dtype=float_type) # numpy array to store spectrogram
# analysis
while True:
samples, read = a() # read file
specgram = vstack((specgram,pv(samples).norm)) # store new norm vector
if read < a.hop_size: break
# plotting
fig = plt.imshow(log10(specgram.T + .001), origin = 'bottom', aspect = 'auto', cmap=plt.cm.gray_r)
ax = fig.axes
ax.axis([0, len(specgram), 0, len(specgram[0])])
# show axes in Hz and seconds
time_step = hop_s / float(samplerate)
total_time = len(specgram) * time_step
outstr = "total time: %0.2fs" % total_time
print(outstr + ", samplerate: %.2fkHz" % (samplerate / 1000.))
n_xticks = 10
n_yticks = 10
def get_rounded_ticks( top_pos, step, n_ticks ):
top_label = top_pos * step
# get the first label
ticks_first_label = top_pos * step / n_ticks
# round to the closest .1
ticks_first_label = round ( ticks_first_label * 10. ) / 10.
# compute all labels from the first rounded one
ticks_labels = [ ticks_first_label * n for n in range(n_ticks) ] + [ top_label ]
# get the corresponding positions
ticks_positions = [ ticks_labels[n] / step for n in range(n_ticks) ] + [ top_pos ]
# convert to string
ticks_labels = [ "%.1f" % x for x in ticks_labels ]
# return position, label tuple to use with x/yticks
return ticks_positions, ticks_labels
# apply to the axis
x_ticks, x_labels = get_rounded_ticks ( len(specgram), time_step, n_xticks )
y_ticks, y_labels = get_rounded_ticks ( len(specgram[0]), (samplerate / 1000. / 2.) / len(specgram[0]), n_yticks )
ax.set_xticks( x_ticks )
ax.set_yticks ( y_ticks )
ax.set_xticklabels( x_labels )
ax.set_yticklabels ( y_labels )
ax.set_ylabel('Frequency (kHz)')
ax.set_xlabel('Time (s)')
ax.set_title(os.path.basename(filename))
for item in ([ax.title, ax.xaxis.label, ax.yaxis.label] +
ax.get_xticklabels() + ax.get_yticklabels()):
item.set_fontsize('x-small')
return fig
def get_pitch(self):
if path.exists(self.filename) == False:
raise Exception(f"File Path to {self.filename} does not exist")
else:
downsample = 1
samplerate = 44100 // downsample
if len(sys.argv) > 2: samplerate = int(sys.argv[2])
win_s = 4096 // downsample # fft size
hop_s = 512 // downsample # hop size
s = source(self.filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("midi")
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# Total number of frames read
total_frames = 0
while True:
samples, read = s()
pitch_midi = pitch_o(samples)[0]
pitch_midi = int(round(pitch_midi))
confidence = pitch_o.get_confidence()
if confidence < 0.9: pitch_midi = 0.
#print("%f %f %f" % (total_frames / float(samplerate), pitch, confidence))
if len(pitches) == 0 and pitch_midi != 0:
pitches.append(pitch_midi)
elif len(pitches) > 0:
if pitch_midi != pitches[-1] and pitch_midi != 0:
pitches.append(pitch_midi)
else:
pass
#print(pitches)
confidences += [confidence]
total_frames += read
if read < hop_s: break
if 0: sys.exit(0)
notes = []
for midi in pitches:
note = midi2note(midi)
notes.append(note.strip("0123456789"))
print(notes)
self.pitch_text = str(notes)
self.q.put(notes)
raise Exception("Thread Terminated")
def disp(filename, samplerate = 44100, tuning = None):
from aubio import source, pitch
from numpy import mean, std
# READ SETUP
unit = "cent"
downsample = 1
samplerate //= downsample
win_s = 4096 // downsample # fft size
hop_s = 512 // downsample # hop size
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit(unit)
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# READING
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
confidence = pitch_o.get_confidence()
pitches += [pitch]
confidences += [confidence]
if read < hop_s: break
# OUTPUT
if type(tuning) is str:
tuning = nameNote(tuning)
mistakes = {}
for i, pitch in enumerate(pitches):
if confidences[i] > 0.1:
correct = int(pitch + 0.5) #### CENT/MIDI ONLY ####
if correct > 0:
if not correct in mistakes:
mistakes[correct] = []
mistakes[correct].append(pitch - correct)
offset = 0
if tuning != None:
offset = mean(mistakes[tuning])
means = {i:100 * (mean(mistakes[i]) - offset) for i in mistakes}
stds = {i:100 * std(mistakes[i]) for i in mistakes}
print({i:len(mistakes[i]) for i in sorted(mistakes)})
return "\n".join([noteName(i) + ": " + "%.2f" % abs(means[i]) + " cents " + ["flat", "sharp"][means[i] > 0] + ", " + "%.2f" % stds[i] + " cents variation" for i in sorted(mistakes)])
def play(file_name, handle_beat, sample_rate=0):
win_s = 1024 # fft size
hop_s = win_s // 2 # hop size
a_source = aubio.source(file_name, sample_rate, hop_s) # create aubio source
sample_rate = a_source.samplerate
# create aubio tempo detection
a_tempo = aubio.tempo("default", win_s, hop_s, sample_rate)
global last_beat_time
last_beat_time = time.time()
# pyaudio callback
def callback(_in_data, _frame_count, _time_info, _status):
samples, read = a_source()
is_beat = a_tempo(samples)
global last_beat_time
now = time.time()
if is_beat:
beat_length = now - last_beat_time
last_beat_time = now
print("tick")
t = threading.Thread(target=handle_beat, args=[beat_length])
t.start()
audiobuf = samples.tobytes()
if read < hop_s:
beat_length = now - last_beat_time
handle_beat(beat_length)
#t = threading.Thread(target=handle_beat, args=[beat_length])
#t.start()
return audiobuf, pyaudio.paComplete
return audiobuf, pyaudio.paContinue
# create pyaudio stream with frames_per_buffer=hop_s and format=paFloat32
p = pyaudio.PyAudio()
pyaudio_format = pyaudio.paFloat32
frames_per_buffer = hop_s
n_channels = 1
stream = p.open(format=pyaudio_format, channels=n_channels, rate=sample_rate,
output=True, frames_per_buffer=frames_per_buffer,
stream_callback=callback)
# start pyaudio stream
stream.start_stream()
# wait for stream to finish
while stream.is_active():
time.sleep(0.1)
# stop pyaudio stream
stream.stop_stream()
stream.close()
# close pyaudio
p.terminate()
def audio_to_midi(filename, midioutput, samplerate=44100, downsample=1):
samplerate = 44100 // downsample
win_s = 512 // downsample # fft size
hop_s = 128 // downsample # hop size
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
notes_o = notes("default", win_s, hop_s, samplerate)
print("%8s" % "time","[ start","vel","last ]")
# create a midi file
mid = MidiFile()
track = MidiTrack()
mid.tracks.append(track)
ticks_per_beat = mid.ticks_per_beat # default: 480
bpm = 120 # default midi tempo
tempo = bpm2tempo(bpm)
track.append(MetaMessage('set_tempo', tempo=tempo))
track.append(MetaMessage('time_signature', numerator=4, denominator=4))
def frames2tick(frames, samplerate=samplerate):
sec = frames / float(samplerate)
return int(second2tick(sec, ticks_per_beat, tempo))
last_time = 0
# total number of frames read
total_frames = 0
while True:
samples, read = s()
new_note = notes_o(samples)
if (new_note[0] != 0):
note_str = ' '.join(["%.2f" % i for i in new_note])
print("%.6f" % (total_frames/float(samplerate)), new_note)
delta = frames2tick(total_frames) - last_time
if new_note[2] > 0:
track.append(Message('note_off', note=int(new_note[2]),
velocity=127, time=delta)
)
track.append(Message('note_on',
note=int(new_note[0]),
velocity=int(new_note[1]),
time=delta)
)
last_time = frames2tick(total_frames)
total_frames += read
if read < hop_s: break
mid.save(midioutput)
def count_samples_in_file(file_path):
from aubio import source
hopsize = 256
s = source(file_path, 0, hopsize)
total_frames = 0
while True:
_, read = s()
total_frames += read
if read < hopsize: break
return total_frames
def get_pitches(filename):
pitches = []
s = source(filename, SIMPLE_RATE, HOP_SIZE)
p = pitch(PITCH_METHOD, FRAME_SIZE, HOP_SIZE, SIMPLE_RATE)
while True:
samples, read = s()
pit = p(samples)[0]
pitches += [pit]
if read < HOP_SIZE: break
return pitches
def test_read_after_close(self):
samplerate = 0 # use native samplerate
hop_size = 256
f = source(default_test_sound, samplerate, hop_size)
read, frames = f()
f.close()
with assert_raises(RuntimeError):
read, frames = f()
with assert_raises(RuntimeError):
read, frames = f.do_multi()
def test_seek_to_half(self, p):
from random import randint
f = source(p, 0, 0)
assert f.samplerate != 0
assert f.hop_size != 0
a = self.read_from_source(f)
c = randint(0, a)
f.seek(c)
b = self.read_from_source(f)
assert a == b + c
def get_spectrogram(filename, samplerate=0):
win_s = 512 # fft window size
hop_s = win_s / 2 # hop size
fft_s = win_s / 2 + 1 # spectrum bins
a = source(filename, samplerate, hop_s) # source file
if samplerate == 0: samplerate = a.samplerate
pv = pvoc(win_s, hop_s) # phase vocoder
specgram = zeros([0, fft_s],
dtype='float32') # numpy array to store spectrogram
# analysis
while True:
samples, read = a() # read file
specgram = vstack(
(specgram, pv(samples).norm)) # store new norm vector
if read < a.hop_size: break
# plotting
imshow(log10(specgram.T + .001),
origin='bottom',
aspect='auto',
cmap=cm.gray_r)
axis([0, len(specgram), 0, len(specgram[0])])
# show axes in Hz and seconds
time_step = hop_s / float(samplerate)
total_time = len(specgram) * time_step
print "total time: %0.2fs" % total_time,
print ", samplerate: %.2fkHz" % (samplerate / 1000.)
n_xticks = 10
n_yticks = 10
def get_rounded_ticks(top_pos, step, n_ticks):
top_label = top_pos * step
# get the first label
ticks_first_label = top_pos * step / n_ticks
# round to the closest .1
ticks_first_label = round(ticks_first_label * 10.) / 10.
# compute all labels from the first rounded one
ticks_labels = [ticks_first_label * n
for n in range(n_ticks)] + [top_label]
# get the corresponding positions
ticks_positions = [ticks_labels[n] / step
for n in range(n_ticks)] + [top_pos]
# convert to string
ticks_labels = ["%.1f" % x for x in ticks_labels]
# return position, label tuple to use with x/yticks
return ticks_positions, ticks_labels
# apply to the axis
xticks(*get_rounded_ticks(len(specgram), time_step, n_xticks))
yticks(*get_rounded_ticks(len(specgram[0]), (samplerate / 2. / 1000.) /
len(specgram[0]), n_yticks))
ylabel('Frequency (kHz)')
xlabel('Time (s)')
def main():
parser = aubio_parser()
args = parser.parse_args()
if 'show_version' in args and args.show_version:
sys.stdout.write('aubio version ' + aubio.version + '\n')
sys.exit(0)
elif 'verbose' in args and args.verbose > 3:
sys.stderr.write('aubio version ' + aubio.version + '\n')
if 'command' not in args or args.command is None:
# no command given, print help and return 1
parser.print_help()
sys.exit(1)
elif not args.source_uri and not args.source_uri2:
sys.stderr.write("Error: a source is required\n")
parser.print_help()
sys.exit(1)
elif args.source_uri2 is not None:
args.source_uri = args.source_uri2
try:
# open source_uri
with aubio.source(args.source_uri,
hop_size=args.hop_size,
samplerate=args.samplerate) as a_source:
# always update args.samplerate to native samplerate, in case
# source was opened with args.samplerate=0
args.samplerate = a_source.samplerate
# create the processor for this subcommand
processor = args.process(args)
frames_read = 0
while True:
# read new block from source
block, read = a_source()
# execute processor on this block
res = processor(block)
# print results for this block
if args.verbose > 0:
processor.repr_res(res, frames_read, a_source.samplerate)
# increment total number of frames read
frames_read += read
# exit loop at end of file
if read < a_source.hop_size: break
# flush the processor if needed
processor.flush(frames_read, a_source.samplerate)
if args.verbose > 1:
fmt_string = "read {:.2f}s"
fmt_string += " ({:d} samples in {:d} blocks of {:d})"
fmt_string += " from {:s} at {:d}Hz\n"
sys.stderr.write(
fmt_string.format(frames_read / float(a_source.samplerate),
frames_read,
frames_read // a_source.hop_size + 1,
a_source.hop_size, a_source.uri,
a_source.samplerate))
except KeyboardInterrupt:
sys.exit(1)
def track(fn):
sr = 44100
n_fft = 4096
hop_size = 512
tolerance = .8
pitch_o = aubio.pitch('yinfft', n_fft, hop_size, sr)
pitch_o.set_unit('midi')
pitch_o.set_tolerance(tolerance)
strings = []
pitches = []
confidences = []
time_plt = []
pitch_plt = []
s = aubio.source(fn, sr, hop_size)
total_frames = 0
while True:
samples, read = s()
pitch = int(pitch_o(samples)[0])
confidence = pitch_o.get_confidence()
time = total_frames / sr
# print("%f %f %f" % (time, pitch, confidence))
pitches.append(pitch)
confidences.append(confidence)
if confidence > .95 and pitch >= 1:
time_plt.append(time)
pitch_plt.append(pitch)
total_frames += read
if read < hop_size:
break
# Processing
pitch_plt = np.array(pitch_plt)
pitch_median = np.median(pitch_plt[:50])
pitch_plt -= pitch_median
time_plt = np.array(time_plt)
time_plt -= time_plt[0]
strings.append(np.array_str(pitch_plt)[1:pitch_plt.size-1])
# intervals = np.zeros(pitch_plt.size-1)
# for j in range(intervals.size):
# intervals[j] = pitch_plt[j+1]-pitch_plt[j]
#plt.plot(time_plt, pitch_plt)
#plt.show()
return pitch_plt
def analyzeMFCC(grain):
windowSize = int(float(grain["frameCount"]))
s = source(grain["file"], int(grain["sampleRate"]), windowSize)
sampleRate = s.samplerate
p = pvoc(windowSize, windowSize)
m = mfcc(windowSize, 40, 13, s.samplerate)
samples, read = s()
spec = p(samples)
mfcc_out = m(spec)
mfccs = mfcc_out.tolist()
return mfccs
def get_file_bpm(path, params = None):
""" Calculate the beats per minute (bpm) of a given file.
path: path to the file
param: dictionary of parameters
"""
if params is None:
params = {}
try:
win_s = params['win_s']
samplerate = params['samplerate']
hop_s = params['hop_s']
except KeyError:
"""
# super fast
samplerate, win_s, hop_s = 4000, 128, 64
# fast
samplerate, win_s, hop_s = 8000, 512, 128
"""
# default:
samplerate, win_s, hop_s = 44100, 1024, 512
s = source(path, samplerate, hop_s)
samplerate = s.samplerate
o = tempo("specdiff", win_s, hop_s, samplerate)
# List of beats, in samples
beats = []
# Total number of frames read
total_frames = 0
while True:
samples, read = s()
is_beat = o(samples)
if is_beat:
this_beat = o.get_last_s()
beats.append(this_beat)
#if o.get_confidence() > .2 and len(beats) > 2.:
# break
total_frames += read
if read < hop_s:
break
# Convert to periods and to bpm
if len(beats) > 1:
if len(beats) < 4:
print("few beats found in {:s}".format(path))
bpms = 60./diff(beats)
for element in bpms:
print element
bpms2.append(bpms)
b = median(bpms)
else:
b = 0
print("not enough beats found in {:s}".format(path))
return b
def test_read_from_mono(self, filename):
total_frames = 0
hop_size = 2048
with source(filename, 0, hop_size) as input_source:
assert_equal(input_source.hop_size, hop_size)
#assert_equal(input_source.samplerate, samplerate)
total_frames = 0
for frames in input_source:
total_frames += frames.shape[-1]
# check we read as many samples as we expected
assert_equal(total_frames, input_source.duration)
def load_aubio(fp):
f = aubio.source(fp, hop_size=1024)
sig = np.zeros(f.duration, dtype=aubio.float_type)
total_frames = 0
while True:
samples, read = f()
sig[total_frames:total_frames + read] = samples[:read]
total_frames += read
if read < f.hop_size:
break
return sig
def data_load_aubio(**kwargs):
if 'filename' in kwargs:
filename = kwargs['filename']
else:
filename = '/home/src/QK/data/sound-arglaaa-2018-10-25/24.wav'
samplerate = 0
src = aubio.source(filename, samplerate, channels=1)
src.seek(0)
return src
def analyzePitch(grain):
s = source(grain["file"], int(grain["sampleRate"]),
int(float(grain["frameCount"])))
samplerate = s.samplerate
tolerance = 0.8
pitch_out = pitch("yin", int(float(grain["frameCount"])),
int(float(grain["frameCount"])), samplerate)
samples, read = s()
pitchFreq = pitch_out(samples)[0].item()
return pitchFreq
def get_length_frames(filename, samplerate = 512, hop_size = 256):
f = source(filename, samplerate, hop_size)
total_frames, read = 0, f.hop_size
while read:
vec, read = f()
total_frames += read
if read < f.hop_size: break
return total_frames
def play_music(filename, handle_beat, samplerate=0):
win_s = 1024 # fft size
hop_s = win_s // 2 # hop size
#samplerate = 0
#if len(sys.argv) > 2: samplerate = int(sys.argv[2])
# create aubio source
a_source = aubio.source(filename, samplerate, hop_s)
samplerate = a_source.samplerate
# create aubio tempo detection
a_tempo = aubio.tempo("default", win_s, hop_s, samplerate)
# create a simple click sound
click = 0.7 * np.sin(
2. * np.pi * np.arange(hop_s) / hop_s * samplerate / 3000.)
# pyaudio callback
def pyaudio_callback(_in_data, _frame_count, _time_info, _status):
samples, read = a_source()
is_beat = a_tempo(samples)
if is_beat:
#samples += click
handle_beat()
audiobuf = samples.tobytes()
if read < hop_s:
return (audiobuf, pyaudio.paComplete)
return (audiobuf, pyaudio.paContinue)
# create pyaudio stream with frames_per_buffer=hop_s and format=paFloat32
p = pyaudio.PyAudio()
pyaudio_format = pyaudio.paFloat32
frames_per_buffer = hop_s
n_channels = 1
stream = p.open(format=pyaudio_format,
channels=n_channels,
rate=samplerate,
output=True,
frames_per_buffer=frames_per_buffer,
stream_callback=pyaudio_callback)
# start pyaudio stream
stream.start_stream()
# wait for stream to finish
while stream.is_active():
time.sleep(0.1)
# stop pyaudio stream
stream.stop_stream()
stream.close()
stream.get
# close pyaudio
p.terminate()
def _extract_frequency_bands(self):
"""
Helper function to extract frequency bands from audio
ARGS:
None
RETURNS:
None
"""
s = source(filename, samplerate, hop_s)
samplerate = s.samplerate
pass
def test_samplerate_hopsize(self, hop_size, samplerate, soundfile):
orig_samplerate = parse_file_samplerate(soundfile)
try:
if orig_samplerate is not None and orig_samplerate < samplerate:
# upsampling should emit a warning
with assert_warns(UserWarning):
f = source(soundfile, samplerate, hop_size)
else:
f = source(soundfile, samplerate, hop_size)
except RuntimeError as e:
err_msg = 'failed opening with hop_s={:d}, samplerate={:d} ({:s})'
skipTest(err_msg.format(hop_size, samplerate, str(e)))
assert f.samplerate != 0
read_frames = self.read_from_source(f)
if 'f_' in soundfile and samplerate == 0:
import re
f = re.compile(r'.*_\([0:9]*f\)_.*')
match_f = re.findall('([0-9]*)f_', soundfile)
if len(match_f) == 1:
expected_frames = int(match_f[0])
assert_equal(expected_frames, read_frames)
def __init__(self, song, samplerate=44100, hop_size=256):
self.src = aubio.source(song)
self.total_frames = 0
self.hop_size = hop_size
while True:
samples, read = self.src()
self.total_frames += read
if read < self.hop_size:
break
self.lengthInSecs = self.total_frames / self.src.samplerate
def get_max_time(filename,
samplerate=0,
block_size=4096,
ax=None,
downsample=2**4,
stop_at=None):
import matplotlib.pyplot as plt
if not ax:
fig = plt.figure()
ax = fig.add_subplot(111)
hop_s = block_size
allsamples = np.zeros(0, )
downsample = downsample # to plot n samples / hop_s
should_stop_early = True
a = source(filename, samplerate, hop_s) # source file
if samplerate == 0: samplerate = a.samplerate
if stop_at == None: should_stop_early = False
total_frames = 0
while True:
samples, read = a()
# keep some data to plot it later
new_samples = abs(samples)
allsamples = np.hstack([allsamples, new_samples])
total_frames += read
if total_frames > stop_at * samplerate and should_stop_early: break
if read < hop_s: break
offset_samples = np.hstack([allsamples[1:], [0]])
derivative_samples = offset_samples - allsamples
t_max_idx = np.argmax(allsamples, axis=0)
derivative_max_idx = np.argmax(derivative_samples, axis=0)
# normalize
derivative_samples = derivative_samples / derivative_samples[
derivative_max_idx]
allsamples_times = [(float(t)) for t in range(len(allsamples))]
t_max = allsamples_times[t_max_idx]
derivative_t_max = allsamples_times[derivative_max_idx]
print("max is reached at %02d:%02d:%03d" %
(t_max / float(samplerate) / 60, (t_max / float(samplerate)) % 60,
(t_max * 1000.0 / float(samplerate)) % 1000))
ax.plot(allsamples_times, allsamples, '-b', derivative_samples, '-g')
ax.axis(xmin=allsamples_times[0], xmax=allsamples_times[-1])
set_xlabels_sample2time(ax, allsamples_times[-1], samplerate)
return ax, t_max / float(samplerate), derivative_t_max / float(samplerate)
def analyse_notes(self):
"""
Taken from https://github.com/aubio/aubio/tree/master/python/demos and slightly modified
This uses the pitch method in aubio. It reads the audio file and gives a list of frequencies and the proability
that it is that note.
:returns A list of notes and the frequecies of those notes
"""
downsample = 1
samplerate = 44100 // downsample
win_s = 4096 // downsample # fft size
hop_s = 512 // downsample # hop size
# TAKES A STR ARG FOR THE FILE PATH, NOT THE FILE ITSELF
try:
s = aubio.source(self.filepath, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
# uses the yin algorithm to determine pitch
pitch_o = aubio.pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("Hz")
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# total number of frames read
total_frames = 0
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
pitch = int(round(pitch))
confidence = pitch_o.get_confidence()
# set the threshold high to ignore frequencies such as random string noise before audio start playing
if confidence < 0.99: pitch = 0.
pitches += [pitch]
confidences += [confidence]
total_frames += read
if read < hop_s: break
pitch_list_minus_duplicates, freq_list_minus_duplicates = self.remove_consecutive_duplicates(
pitches)
return pitch_list_minus_duplicates, freq_list_minus_duplicates
except Exception as e:
error_str = "Could not open file:{}. Error: {}".format(
self.filepath, e)
print(error_str)
def detect(filename):
downsample = 8
samplerate = 44100 // downsample
win_s = 4096 // downsample # fft size
hop_s = 512 // downsample # hop size
s = aubio.source(filename, samplerate, hop_s)
samplerate = s.samplerate
tolerance = 0.8
pitch_o = aubio.pitch("yin", win_s, hop_s, samplerate)
pitch_o.set_unit("freq")
pitch_o.set_tolerance(tolerance)
pitches = []
confidences = []
# total number of frames read
total_frames = 0
counter = 0
while True:
samples, read = s()
pitch = pitch_o(samples)[0]
confidence = pitch_o.get_confidence()
#print "%f %f %f" % (total_frames / float(samplerate), pitch, confidence)
pitches += [pitch]
confidences += [confidence]
total_frames += read
if read < hop_s: break
letterList = []
newPitches = []
amplitudes = []
for index in range(0, len(pitches)):
totalFreq = 0
if(index+5 <= len(pitches)):
for freq in range(index, index+5):
totalFreq += pitches[freq]
averageFreq = totalFreq/5
newPitches.append(averageFreq)
else:
counter = 0
for index in range(index, len(pitches)):
totalFreq += pitches[freq]
counter += 1
averageFreq = totalFreq/counter
newPitches.append(averageFreq)
# loudness
amp = getSection(filename,index, index + 5)
amplitudes.append(amp.rms)
for pi in newPitches:
letterName = findPitchLetterName(pi)
letterList.append(pi) # letterName -> pi for frequency numbers
newList = modifyList(letterList)
note = findModeInList(newList)
#print(letterList, amplitudes)
return letterList, amplitudes
def load_file_aubio(filename):
import aubio
f = aubio.source(filename, hop_size=1024)
y = np.zeros(f.duration, dtype=aubio.float_type)
total_frames = 0
while True:
samples, read = f()
y[total_frames:total_frames + read] = samples[:read]
total_frames += read
if read < f.hop_size: break
assert len(y) == total_frames
#print y.mean(), y.shape
return total_frames, f.samplerate
