def update(self) -> None:
try:
wf_data = self.stream.read(flags.chunk,
exception_on_overflow=False)
wf_data = np.frombuffer(wf_data, dtype=float)
if np.size(wf_data) == 0:
return
self.input_data = np.hstack(
(self.input_data[wf_data.shape[0]:], wf_data))
input_last = self.input_data[:self.fftWindow]
# Normalization
input_last = input_last / np.sqrt(
(np.sum(np.square(input_last)) + 1e-6))
self.spectral_data = np.fft.fft(input_last)
self.spectral_data = np.abs(self.spectral_data[:self.fftWindow //
2])
if np.sum(self.spectral_data) > 0:
self.spectrum_trace2 = self.spectrum_trace
self.spectrum_trace = self.spectrum_plot.plot(pen='m', width=3)
self.spectrum_trace.setData(self.f, self.spectral_data)
self.spectrum_plot.setLogMode(x=True, y=True)
self.spectrum_plot.setYRange(-13, 2, padding=0.001)
self.spectrum_plot.setXRange(np.log10(Note.C(4).freq),
np.log10(Note.C(8).freq),
padding=0.005)
if self.spectrum_trace2 is not None:
self.spectrum_plot.removeItem(self.spectrum_trace2)
except IOError:
pass # underflow, waiting for data
def next_pitch(self) -> int:
n = Note()
n.pitch = numpy.argmax(
self.net_fn(
[self._encode_network_input(self.past, self.current_chord)]))
self.past.append(n)
self.past = self.past[-self.order:]
return n.pitch
def draw_note_lines(self):
for i in range(-48, 94):
if i % 12 in {0, 2, 4, 5, 7, 9, 11}:
color = 'r'
else:
color = 'b'
if i % 12 == 0:
thickness = 0.5
else:
thickness = 0.35
inf_line = pg.InfiniteLine(movable=True,
angle=90,
label=Note(i).name(),
pen=pg.mkPen(color, width=thickness))
inf_line.setPos([np.log10(Note(i).freq), 0])
self.spectrum_plot.addItem(inf_line)
def midi2piece(file_name):
piece = Piece(file_name)
file_path = MIDI_PATH / Path(file_name + '.mid')
midi = mid.MidiFile(file_path)
has_pedal = check_pedal(midi)
if not has_pedal:
time_ticks = 0
for m, msg in enumerate(midi.tracks[0]):
time_ticks += msg.time
if msg.type == 'note_on':
if msg.velocity != 0:
m_end = m + 1
delta_ticks = 0
while True:
delta_ticks += midi.tracks[0][m_end].time
if midi.tracks[0][
m_end].note == msg.note and midi.tracks[0][
m_end].velocity == 0:
note = Note.from_midi(msg.note, msg.velocity,
time_ticks,
time_ticks + delta_ticks)
piece.append(note)
break
m_end += 1
else:
raise Exception("Pedal not integrated yet")
return piece
def notes_from_file(filename: str) -> List[Note]:
midifile_rel = midi.read_midifile(filename)
midifile_abs = copy.deepcopy(midifile_rel)
midifile_abs.make_ticks_abs()
# Convert MIDI events to our music representation: a list of Note objects
notes = []
active_notes = {}
for ev_rel, ev_abs in zip(midifile_rel[-1], midifile_abs[-1]):
if isinstance(ev_rel, midi.NoteOnEvent) and ev_rel.data[1]:
n = Note()
n.resolution = midifile_rel.resolution
n.tick_abs = ev_abs.tick
n.pitch = ev_rel.data[0]
n.velocity = ev_rel.data[1]
if n.pitch not in active_notes:
active_notes[n.pitch] = {n}
else:
active_notes[n.pitch].add(n)
elif isinstance(ev_rel, midi.NoteOffEvent) or (isinstance(
ev_rel, midi.NoteOnEvent) and ev_rel.data[1] == 0):
n = active_notes[ev_rel.data[0]].pop()
n.duration = ev_abs.tick - n.tick_abs
notes.append(n)
assert not any(active_notes.values()), "Some notes were not released"
return sorted(notes, key=lambda note: note.tick_abs)
def notes_from_file(filename: str) -> List[Note]:
midifile_rel = midi.read_midifile(filename)
midifile_abs = copy.deepcopy(midifile_rel)
midifile_abs.make_ticks_abs()
# Convert MIDI events to our music representation: a list of Note objects
notes = []
active_notes = {}
for ev_rel, ev_abs in zip(midifile_rel[-1], midifile_abs[-1]):
if isinstance(ev_rel, midi.NoteOnEvent) and ev_rel.data[1]:
n = Note()
n.resolution = midifile_rel.resolution
n.tick_abs = ev_abs.tick
n.pitch = ev_rel.data[0]
n.velocity = ev_rel.data[1]
if n.pitch not in active_notes:
active_notes[n.pitch] = {n}
else:
active_notes[n.pitch].add(n)
elif isinstance(ev_rel, midi.NoteOffEvent) or (isinstance(ev_rel, midi.NoteOnEvent) and ev_rel.data[1] == 0):
n = active_notes[ev_rel.data[0]].pop()
n.duration = ev_abs.tick - n.tick_abs
notes.append(n)
assert not any(active_notes.values()), "Some notes were not released"
return sorted(notes, key=lambda note: note.tick_abs)
def __init__(self,
stream,
rate: int,
lowest: Note = Note.C(2),
highest: Note = Note.C(8),
List=[]) -> None:
super().__init__(List)
self.rate = rate
self.stream = stream
self.lowest = lowest
self.highest = highest
self.fftWindow = flags.fftWindow
self.savingRange = 44100 # 1 second
self.timer = Qt.QTimer()
self.timer.timeout.connect(self.update)
self.app = Qt.QApplication(sys.argv)
self.window = pg.GraphicsWindow(title="PyTuner")
self.window.setGeometry(5, 115, 1910, 1070)
sp_labels = [(np.log10(Note(i).freq), Note(i).name())
for i in range(-48, 94)]
sp_x_axis = pg.AxisItem(orientation='bottom')
sp_x_axis.setTicks([sp_labels])
self.spectrum_plot = self.window.addPlot(
title="Sound spectrum",
row=1,
col=1,
axisItems={'bottom': sp_x_axis})
self.spectrum_trace = None
self.spectrum_trace2 = None
self.spectrum_plot.setXRange(np.log10(self.lowest.freq),
np.log10(self.highest.freq),
padding=0.005)
self.f = np.linspace(1, self.rate / 2, self.fftWindow // 2)
self.input_data = np.zeros((self.savingRange, ), dtype=float)
self.spectral_data = np.zeros((self.fftWindow // 2, ), dtype=float)
self.draw_note_lines()
self.start()
def generate(past: List[Note], changes: ChordProgression,
melody_generator: Union[MelodyGenerator, MelodyAndRhythmGenerator,
UniversalGenerator],
rhythm_generator: Optional[RhythmGenerator],
measures: int) -> List[Note]:
""" Improvise a melody using two models for the melody and the rhythm, and one chord progression
:param melody_generator:
:param rhythm_generator:
:param past: the seed
:param changes: the chord progression.
It can be the same as the melody generator, or equivalently None.
:param measures: The number of measures to generate
"""
if rhythm_generator is None:
rhythm_generator = melody_generator # type: RhythmGenerator
universal = isinstance(melody_generator, UniversalGenerator)
melody = past
beat = melody[-1].beat
chord = changes[beat]
melody_generator.start(beat)
while beat < measures * Note.meter:
n = Note()
n.resolution = past[0].resolution
rest = None
if universal:
n.pitch, tsbq, dq, *rest = melody_generator.next(
) # in LSTM case, rest[0] is the beat diff
else:
tsbq, dq = rhythm_generator.next_rhythm()
tsbq *= Note.ticks_quantisation_rate
n.duration = dq * Note.duration_quantisation_rate
if melody and (rest or melody[-1].ticks_since_beat > tsbq):
beat_diff = rest[
0] if rest else 1 + melody[-1].duration // n.resolution
for _ in range(beat_diff):
beat += 1
# If the chord changed, inform the melody generator
newchord = changes[beat]
if newchord != chord:
melody_generator.start(beat)
chord = newchord
# Prevent overlapping notes
n.tick_abs = tsbq + n.resolution * \
(beat if rest else max(beat, math.floor((melody[-1].tick_abs + melody[-1].duration) / n.resolution)))
if not universal:
n.pitch = melody_generator.next_pitch()
melody.append(n)
if melody_generator == rhythm_generator:
melody_generator.add_past(n)
return melody
def convert_song(song: SongMetadata):
""" Combine the quantized and the unquantized MIDI files
into one that aligns to measures but attempts to retain the original phrasing """
from file_handlers import notes_to_file
from file_handlers import notes_from_file
Note.default_resolution = 960
quantized = notes_from_file('weimardb/midi_from_db_quant/{}.mid'.format(song.name))
original = notes_from_file('weimardb/midi_from_db/{}.mid'.format(song.name))
lilypond = notes_from_file('weimardb/midi_from_ly/{}.mid'.format(song.name))
d = {}
for n, m in zip(quantized, lilypond):
if m.measure - n.measure not in d:
d[m.measure - n.measure] = 1
else:
d[m.measure - n.measure] += 1
if len(d) == 2:
break
meas_start = min(d.keys()) # type: int
meas_size = Note.meter * Note.default_resolution
meas_no = 0
a, b, c = [], [], [] # quantized measure, original measure, combined output
lp = [] # lilypond measure: only needed for durations
for q, o, l in zip(quantized, original, lilypond):
if q.measure != meas_no:
if len(a) > 1:
r = (a[-1].tick_abs - a[0].tick_abs) / (b[-1].tick_abs - b[0].tick_abs) # stretch ratio
a_m = (meas_no + 0.5) * meas_size # middle of quantized measure
b_m = b[0].tick_abs + (a_m - a[0].tick_abs) / r # estimated middle of unquantized measure
for a_j, b_j, l_j in zip(a, b, lp):
n = Note()
n.pitch = b_j.pitch
n.resolution = b_j.resolution
n.velocity = b_j.velocity
n.tick_abs = int(a_m + r * (b_j.tick_abs - b_m)) + (meas_start * meas_size)
n.duration = int(r * b_j.duration) or a_j.duration or l_j.duration
c.append(n)
else:
c += a
meas_no = q.measure
a, b, lp = [], [], []
a.append(q)
b.append(o)
lp.append(l)
notes_to_file(sorted(c, key=lambda p: p.tick_abs), 'weimardb/midi_combined/{}.mid'.format(song.name))
def generate(past: List[Note], changes: ChordProgression,
melody_generator: Union[MelodyGenerator, MelodyAndRhythmGenerator, UniversalGenerator],
rhythm_generator: Optional[RhythmGenerator], measures: int) -> List[Note]:
""" Improvise a melody using two models for the melody and the rhythm, and one chord progression
:param melody_generator:
:param rhythm_generator:
:param past: the seed
:param changes: the chord progression.
It can be the same as the melody generator, or equivalently None.
:param measures: The number of measures to generate
"""
if rhythm_generator is None:
rhythm_generator = melody_generator # type: RhythmGenerator
universal = isinstance(melody_generator, UniversalGenerator)
melody = past
beat = melody[-1].beat
chord = changes[beat]
melody_generator.start(beat)
while beat < measures * Note.meter:
n = Note()
n.resolution = past[0].resolution
rest = None
if universal:
n.pitch, tsbq, dq, *rest = melody_generator.next() # in LSTM case, rest[0] is the beat diff
else:
tsbq, dq = rhythm_generator.next_rhythm()
tsbq *= Note.ticks_quantisation_rate
n.duration = dq * Note.duration_quantisation_rate
if melody and (rest or melody[-1].ticks_since_beat > tsbq):
beat_diff = rest[0] if rest else 1 + melody[-1].duration // n.resolution
for _ in range(beat_diff):
beat += 1
# If the chord changed, inform the melody generator
newchord = changes[beat]
if newchord != chord:
melody_generator.start(beat)
chord = newchord
# Prevent overlapping notes
n.tick_abs = tsbq + n.resolution * \
(beat if rest else max(beat, math.floor((melody[-1].tick_abs + melody[-1].duration) / n.resolution)))
if not universal:
n.pitch = melody_generator.next_pitch()
melody.append(n)
if melody_generator == rhythm_generator:
melody_generator.add_past(n)
return melody
def _build_net(self) -> keras.models.Model:
dummy_input = self._encode_network_input(
[Note()] * self.order, [Chord('C7')] * self.chord_order,
self.changes)
in_notes = keras.layers.Input(batch_shape=(1,) + dummy_input[0].shape) if self.stateful\
else keras.layers.Input(shape=dummy_input[0].shape)
in_chords = keras.layers.Input(batch_shape=(1,) + dummy_input[1].shape) if self.stateful\
else keras.layers.Input(shape=dummy_input[1].shape)
lstm_out = keras.layers.LSTM(
512, stateful=self.stateful,
implementation=self._implementation)(in_notes)
x = keras.layers.concatenate([lstm_out, in_chords])
x = keras.layers.Dense(512)(x)
pitch_tensor = keras.layers.Dense(12, activation=softmax)(x)
tsbq_tensor = keras.layers.Dense(self.maxtsbq + 1,
activation=softmax)(x)
dq_tensor = keras.layers.Dense(self.maxdq + 1, activation=softmax)(x)
octave_tensor = keras.layers.Dense(NUM_OCTAVES, activation=softmax)(x)
beatdiff_tensor = keras.layers.Dense(self.maxbeatdiff + 1,
activation=softmax)(x)
model = keras.models.Model(inputs=[in_notes, in_chords],
outputs=[
pitch_tensor, tsbq_tensor, dq_tensor,
octave_tensor, beatdiff_tensor
])
model.compile(optimizer=rmsprop(), loss=categorical_crossentropy)
self.octave_model = keras.models.Model(inputs=model.inputs,
outputs=model.outputs[3])
self.epochs = 30
self.outfuns = (
sampler(.1), ) + (weighted_nlargest(2), ) * 2 + (np.argmax, ) * 2
return model
from music import Note, Pitch
from music import Buzzer, BuzzerBand
import oak
def main():
b1 = Buzzer(oak.PINS[5])
b2 = Buzzer(oak.PINS[6])
band = BuzzerBand(b1, b2)
band.play_tunes(IN_THE_HALL_OF_THE_MOUNTAIN_KING_1,
IN_THE_HALL_OF_THE_MOUNTAIN_KING_2, speed=2)
IN_THE_HALL_OF_THE_MOUNTAIN_KING_1 = (
Note(Pitch.E[3], 10, 4, 0),
Note(Pitch.A[2], 10, 0.5, 1), Note(Pitch.B[2], 10, 0.5, 1),
Note(Pitch.C[3], 10, 0.5, 1), Note(Pitch.D[3], 10, 0.5, 1),
Note(Pitch.E[3], 20, 0.5, 1), Note(Pitch.C[3], 10, 0.5, 1),
Note(Pitch.E[3], 10, 1, 1),
Note(Pitch.Eb[3], 20, 0.5, 1), Note(Pitch.B[2], 10, 0.5, 1),
Note(Pitch.Eb[3], 10, 1, 1),
Note(Pitch.D[3], 20, 0.5, 1), Note(Pitch.Bb[3], 10, 0.5, 1),
Note(Pitch.D[3], 10, 1, 1),
Note(Pitch.A[2], 10, 0.5, 1), Note(Pitch.B[2], 10, 0.5, 1),
Note(Pitch.C[3], 10, 0.5, 1), Note(Pitch.D[3], 10, 0.5, 1),
Note(Pitch.E[3], 20, 0.5, 1), Note(Pitch.C[3], 10, 0.5, 1),
Note(Pitch.E[3], 10, 0.5, 1), Note(Pitch.A[3], 10, 0.5, 1),
'Db': ['Db', 'Eb', 'F', 'Gb', 'Ab', 'Bb', 'C', 'Db'],
'Ab': ['Ab', 'Bb', 'C', 'Db', 'Eb', 'F', 'G', 'Ab'],
'Eb': ['Eb', 'F', 'G', 'Ab', 'Bb', 'C', 'D', 'Eb'],
'Bb': ['Bb', 'C', 'D', 'Eb', 'F', 'G', 'A', 'Bb'],
'F': ['F', 'G', 'A', 'Bb', 'C', 'D', 'E', 'F'],
'C': ['C', 'D', 'E', 'F', 'G', 'A', 'B', 'C'],
'G': ['G', 'A', 'B', 'C', 'D', 'E', 'F#', 'G'],
'D': ['D', 'E', 'F#', 'G', 'A', 'B', 'C#', 'D'],
'A': ['A', 'B', 'C#', 'D', 'E', 'F#', 'G#', 'A'],
'E': ['E', 'F#', 'G#', 'A', 'B', 'C#', 'D#', 'E'],
'B': ['B', 'C#', 'D#', 'E', 'F#', 'G#', 'A#', 'B'],
'F#': ['F#', 'G#', 'A#', 'B', 'C#', 'D#', 'E#', 'F#'],
'C#': ['C#', 'D#', 'E#', 'F#', 'G#', 'A#', 'B#', 'C#'],
}
major_scales = {
key: [Note(n) for n in notes]
for key, notes in major_scales.items()
}
@pytest.mark.parametrize(('key', 'notes'), major_scales.items())
def test_major_scales_mode_I(key, notes):
assert Scale(key, 'major', mode=1).notes == notes
@pytest.mark.parametrize(('key', 'notes'), major_scales.items())
@pytest.mark.parametrize('mode', range(1, 8))
def test_major_scales_modes(key, notes, mode):
assert set(Scale(key, 'major', mode=1).notes) == set(
Scale(notes[mode - 1], 'major', mode=mode).notes)
def inputshape(self) -> Tuple[int]:
""" Generates a dummy input matrix for the network and returns its shape. """
return self._encode_network_input([Note()] * self.order,
[Chord('C7')] * self.chord_order,
self.changes)[0].shape
from music import Note
import sys
#Create note using music library constants 0-127
note1 = Note(int(sys.argv[1]), 1)
print('\n--Result--')
print(note1.getPitch())
print('--EndResult--')
sys.exit()
def next_pitch(self) -> int:
n = Note()
n.pitch = numpy.argmax(self.net_fn([self._encode_network_input(self.past, self.current_chord)]))
self.past.append(n)
self.past = self.past[-self.order:]
return n.pitch
from music import Phrase
from music import Note
import sys
import time
phr = Phrase()
note1 = Note(10, 1)
note2 = Note(20, 1)
note3 = Note(30, 1)
note4 = Note(40, 1)
phr.addNote(note1)
phr.addNote(note2)
phr.addNote(note3)
phr.addNote(note4)
phr.empty()
time.sleep(1)
print('\n--Result--')
print(phr.getSize())
print('--EndResult--')
sys.exit()
from music import Phrase
from music import Note
import sys
phr = Phrase()
note1 = Note(10, 5)
phr.addNote(note1)
print('--Result--')
print(phr.getEndTime() == 5.0)
print('--EndResult--')
sys.exit()
def get_note(self, transform=Fft, **transform_args):
freq = self.get_freq(transform, **transform_args)
return Note.from_frequency(freq)
def detectNotes(self):
"""Detect the notes in the audio recording self.data. Plots a
spectrogram with detected notes labelled."""
fig = pylab.figure()
fig.suptitle('Track Detection')
# SPECTROGRAM
# documentation at
# http://matplotlib.org/api/pyplot_api.html?highlight=specgram#matplotlib.pyplot.specgram
#
# this call to specgram is precise with regard to frequencies, but
# blurry in time domain
smoothAudio = Transcriber.smooth( self.data )
(Pxx, freqs, bins, im) = pylab.specgram( smoothAudio, Fs=self.rate,
NFFT=2**12, noverlap=2**8, sides='onesided', scale_by_freq=True)
print "SHAPE OF freqs", freqs.shape
# ------------------------
# [BEGIN] identify runs of notes
# ------------------------
# how many instantaneous spectra did we calculate
(numBins, numSpectra) = Pxx.shape
print "SHAPE OF Pxx:", Pxx.shape
# how many seconds in entire audio recording
numSeconds = float(self.data.size) / self.rate
scaledPxx = 10 * np.log10(Pxx)
def findPeaks( sample, minPeakVal=None):
peakPos = []
peakVal = []
lastSlopeNeg = None
for i in range(0, len(sample)-1):
# we will be comparing two points on the frequency spectrum at
# a time
# If the two points are equal, the slope is 0, so we will move
# on to the next point. For example, if we have the list [0 1 2
# 2 1 0], there is clearly a peak, but it is between the two 2
# values
if (sample[i] != sample[i+1]):
thisSlopeNeg = sample[i] > sample[i+1]
if lastSlopeNeg != None and (not lastSlopeNeg) and thisSlopeNeg:
if minPeakVal != None and sample[i] >= minPeakVal:
peakPos.append(i)
peakVal.append(sample[i+1])
lastSlopeNeg = thisSlopeNeg
# TODO some sort of statistical analysis of the significance of the
# peaks found. For now, I'll just have a magic number representing
# the minimum peak value.
return (peakPos, peakVal)
# TODO find the optimal value of minPeakVal based on the spectrogram
# determine thresholding value
thresh = np.mean( Pxx )
FREQ_THRESH = 10
lastNotes = []
prevF0NoteNamesSciPitchQueue = deque(maxlen=5)
def recentlySaw( noteNameSciPitch ):
#print prevF0NoteNamesSciPitchQueue
for prevF0NoteNames in prevF0NoteNamesSciPitchQueue:
if noteNameSciPitch in prevF0NoteNames:
return True
return False
for t in range(0, numSpectra):
print "-----------------------"
# extract a block from the spectrogram
sample = Pxx[:, t]
sample = Transcriber.smooth( sample )
#print "SHAPE OF sample:", sample.shape
# find the peaks in this profile (peaks represent notes)
(peakPos, peakVal) = findPeaks(sample, minPeakVal=thresh)
noteNames = []
def freqsAreSameNote( f1, oct1, f2, oct2):
# scale freqs
f1 = f1 / 2**oct1
f2 = f2 / 2**oct2
return abs(f1-f2) < FREQ_THRESH
prevF0NoteNames = []
# Go through notes backwards, from high to low.
# Variable i represents peak number.
for i in reversed(range(len(peakPos))):
# Variable pos represents at which y-value in spectrogram this
# peak was found. Variable intensity contains the intensity at
# that peak; how much energy at that frequency.
pos = peakPos[i]
intensity = peakVal[i]
f = freqs[pos]
if 20 <= f <= 20000: # if it is audible to a human...
(noteName, octave, sciPitchNoteName) = Note.getNoteName( f )
noteNames.append(sciPitchNoteName)
# -------------------------------------------------------
# attempt to find fundamental frequency of this note
f0 = None
f0Pos = None
for j in range( i ):
otherFreq = freqs[ peakPos[j] ]
otherIntensity = peakVal[j]
(otherNoteName, otherOctave, otherSciPitch) = Note.getNoteName( otherFreq )
if freqsAreSameNote( f, octave, otherFreq, otherOctave) \
and intensity <= 0.5*otherIntensity \
and 20 <= otherFreq <= 20000:
# compute frequency of f moved down some number of
# octaves
f0 = f / 2.0**(octave-otherOctave)
f0Pos = j
break
# -------------------------------------------------------
if f0 != None :
(f0NoteName, f0Octave, f0SciPitch) = Note.getNoteName(f0)
print "f0 = %f" % f0
print "recently saw %s? %s" % (f0SciPitch, recentlySaw(f0SciPitch))
# If we haven't recently handled this f0
if (f0SciPitch not in prevF0NoteNames) \
and (not recentlySaw( f0SciPitch )) \
and (f0SciPitch not in lastNotes):
prevF0NoteNames.append(f0SciPitch)
#print "%s @ %.2fHz" % (f0SciPitch, f0)
# Annotate the spectrogram. Note that circles plotted actually
# appear as horizontal lines thanks to our logarithmic y scale.
time = (1.0*t / numSpectra) * numSeconds;
circle = plt.Circle( (time, f), 0.01, color='w')
fig.gca().add_artist(circle)
xPos = 1.0*t / numSpectra
yPos = f / freqs[-1] # y position = f0 / max freq
plt.text( xPos, yPos, f0SciPitch, transform=fig.gca().transAxes)
lastNotes = noteNames
# save the F0s encountered in this time slice
prevF0NoteNamesSciPitchQueue.append( prevF0NoteNames )
# ------------------------
# END [identify runs of notes]
# ------------------------
pylab.show()
def get_random_note(lower=None, higher=None):
lower = lower or Note('A', 0)
higher = higher or Note('C', 8)
n = random.randint(0, higher - lower)
Player().perform([lower + Semitones(n)])
from music import Phrase
from music import Note
import sys
inary = sys.argv[1].split(' ')
phr = Phrase()
note1 = Note(int(inary[0]), 1)
note2 = Note(int(inary[1]), 1)
note3 = Note(int(inary[2]), 1)
note4 = Note(int(inary[3]), 1)
phr.addNote(note1)
phr.addNote(note2)
phr.addNote(note3)
phr.addNote(note4)
print('\n--Result--')
print(phr.getSize())
print('--EndResult--')
sys.exit()
velocity = None
start_seconds = None
# THE loop
piece_reconstructed = Piece(piece.name + ' reconstructed')
for f in tqdm(range(close_onsets.shape[0])):
for t in range(close_onsets.shape[1] - 1):
if close_onsets[f, t] < threshold_detection <= close_onsets[f, t + 1]:
# Note on
frequency = int(round(hz_to_midi(FREQUENCIES[f])))
velocity = db_to_velocity(close_onsets[f, t + 1])
start_seconds = float(time_vector[t + 1])
elif close_onsets[f, t] >= threshold_detection > close_onsets[f,
t + 1]:
# Note off
# TODO: decide between t and t+1 index to note off
end_seconds = float(time_vector[t + 1])
note = Note(frequency,
velocity=velocity,
start_seconds=start_seconds,
end_seconds=end_seconds)
piece_reconstructed.append(note)
# Re-synthesize piece
signal_reconstructed = samples_set.synthesize(piece_reconstructed)
sd.play(signal_reconstructed, FS)
if __name__ == '__main__':
pass
