-
Notifications
You must be signed in to change notification settings - Fork 0
/
transcribe.py
361 lines (287 loc) · 12.9 KB
/
transcribe.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
# Copyright (c) 2011 Michael Scott Cuthbert and the music21 Project
# Copyright (c) 2015 Joel Robichaud
# All rights reserved.
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
# * Redistributions of source code must retain the above copyright notice, this
# list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
# * Neither the name of [project] nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
import copy
import math
import wave
import numpy
import scipy.signal
from music21 import stream, note, pitch, scale
def interpolation(correlation, peak):
"""Interpolation for estimating the true position of an inter-sample
maximum when nearby samples are known."""
curr = correlation[peak]
prev = correlation[peak - 1] if peak - 1 >= 0 else curr
next = correlation[peak + 1] if peak + 1 < len(correlation) else curr
vertex = (prev - next) / (prev - 2.0 * curr + next)
vertex = vertex * 0.5 + peak
return vertex
def getFrequenciesFromAudioFile(filename, blocksize=512):
"""Retrieve a list of frequencies from an audio file."""
wv = wave.open(filename, 'r')
srate = wv.getframerate()
blocks = []
for i in range(int(wv.getnframes() / blocksize)):
data = wv.readframes(blocksize)
blocks.append(data)
wv.close()
freqs = []
for data in blocks:
samples = numpy.fromstring(data, dtype=numpy.int16)
freqs.append(autocorrelationFunction(samples, srate))
return freqs
def autocorrelationFunction(signal, srate):
"""Convert a signal from the time domain into the frequency domain."""
signal = numpy.array(signal)
correlation = scipy.signal.fftconvolve(signal, signal[::-1], mode='full')
lengthCorrelation = len(correlation) / 2
correlation = correlation[lengthCorrelation:]
difference = numpy.diff(correlation) # Calculate the difference between slots
positiveDifferences, = numpy.nonzero(numpy.ravel(difference > 0))
if len(positiveDifferences) == 0:
finalResult = 10 # Rest
else:
beginning = positiveDifferences[0]
peak = numpy.argmax(correlation[beginning:]) + beginning
vertex = interpolation(correlation, peak)
finalResult = srate / vertex
return finalResult
def detectPitchFrequencies(freqFromAQList, useScale=None):
"""Detect the pitches of the notes from a list of frequencies."""
if useScale is None:
useScale = scale.ChromaticScale()
(thresholds, pitches) = prepareThresholds(useScale)
detectedPitchesFreq = []
for i in range(len(freqFromAQList)): # Find thresholds and frequencies
inputPitchFrequency = freqFromAQList[i]
unused_freq, pitch_name = normalizeInputFrequency(inputPitchFrequency, thresholds, pitches)
detectedPitchesFreq.append(pitch_name.frequency)
return detectedPitchesFreq
def normalizeInputFrequency(inputPitchFrequency, thresholds=None, pitches=None):
"""Return a tuple of the normalized frequency and the pitch detected."""
if ((thresholds is None and pitches is not None)
or (thresholds is not None and pitches is None)):
raise AudioSearchException("Cannot normalize input frequency if both thresholds and pitches are not given.")
elif thresholds == None:
(thresholds, pitches) = prepareThresholds()
inputPitchLog2 = math.log(inputPitchFrequency, 2)
(remainder, octave) = math.modf(inputPitchLog2)
octave = int(octave)
for i in range(len(thresholds)):
threshold = thresholds[i]
if remainder < threshold:
returnPitch = copy.deepcopy(pitches[i])
returnPitch.octave = octave - 4
name_note = pitch.Pitch(str(pitches[i]))
return name_note.frequency, returnPitch
returnPitch = copy.deepcopy(pitches[-1])
returnPitch.octave = octave - 3
returnPitch.inputFrequency = inputPitchFrequency
name_note = pitch.Pitch(str(pitches[-1]))
return name_note.frequency, returnPitch
def prepareThresholds(useScale=None):
"""Return a tuple of two lists consisting of the threshold values and the
pitches of a scale."""
if useScale is None:
useScale = scale.ChromaticScale('C4')
scPitches = useScale.pitches
scPitchesRemainder = []
for p in scPitches:
pLog2 = math.log(p.frequency, 2)
scPitchesRemainder.append(math.modf(pLog2)[0])
scPitchesRemainder[-1] += 1
scPitchesThreshold = []
for i in range(len(scPitchesRemainder) - 1):
scPitchesThreshold.append((scPitchesRemainder[i] + scPitchesRemainder[i + 1]) / 2)
return scPitchesThreshold, scPitches
def smoothFrequencies(detectedPitchesFreq, smoothLevels=7, inPlace=True):
"""Smooth the shape of the signal in order to avoid false detections of
the fundamental frequency."""
dpf = detectedPitchesFreq
if inPlace == True:
detectedPitchesFreq = dpf
else:
detectedPitchesFreq = copy.copy(dpf)
beginning = 0.0
ends = 0.0
for i in range(smoothLevels):
beginning = beginning + float(detectedPitchesFreq[i])
ends = ends + detectedPitchesFreq[len(detectedPitchesFreq) - 1 - i]
beginning = beginning / smoothLevels
ends = ends / smoothLevels
for i in range(len(detectedPitchesFreq)):
if i < int(math.floor(smoothLevels / 2.0)):
detectedPitchesFreq[i] = beginning
elif i > len(detectedPitchesFreq) - int(math.ceil(smoothLevels / 2.0)) - 1:
detectedPitchesFreq[i] = ends
else:
t = 0
for j in range(smoothLevels):
t = t + detectedPitchesFreq[i + j - int(math.floor(smoothLevels / 2.0))]
detectedPitchesFreq[i] = t / smoothLevels
return [int(round(fq)) for fq in detectedPitchesFreq]
def pitchFrequenciesToObjects(detectedPitchesFreq, useScale=None):
"""Return a list of the pitches that best match the input frequencies."""
if useScale is None:
useScale = scale.ChromaticScale()
detectedPitchObjects = []
(thresholds, pitches) = prepareThresholds(useScale)
for i in range(len(detectedPitchesFreq)):
inputPitchFrequency = detectedPitchesFreq[i]
unused_freq, pitch_name = normalizeInputFrequency(inputPitchFrequency, thresholds, pitches)
detectedPitchObjects.append(pitch_name)
i = 0
while i < len(detectedPitchObjects) - 1:
name = detectedPitchObjects[i].name
hold = i
tot_octave = 0
while i < len(detectedPitchObjects) - 1 and detectedPitchObjects[i].name == name:
tot_octave = tot_octave + detectedPitchObjects[i].octave
i = i + 1
tot_octave = round(tot_octave / (i - hold))
for j in range(i - hold):
detectedPitchObjects[hold + j - 1].octave = tot_octave
return detectedPitchObjects
def joinConsecutiveIdenticalPitches(detectedPitchObjects):
"""Return a tuple of two lists consisting of a list of note and rest
objects (each of quarterLength 1.0) and a list of how many pitches were
joined together to make that object."""
REST_FREQUENCY = 10
detectedPitchObjects[0].frequency = REST_FREQUENCY
j = 0
good = 0
bad = 0
valid_note = False
total_notes = 0
total_rests = 0
notesList = []
durationList = []
while j < len(detectedPitchObjects):
fr = detectedPitchObjects[j].frequency
# Detect consecutive instances of the same frequency
while j < len(detectedPitchObjects) and fr == detectedPitchObjects[j].frequency:
good = good + 1
# If more than 6 consecutive identical samples, it might be a note
if good >= 6:
valid_note = True
# If we've gone 15 or more samples without getting something constant, assume it's a rest
if bad >= 15:
durationList.append(bad)
total_rests = total_rests + 1
notesList.append(note.Rest())
bad = 0
j = j + 1
if valid_note == True:
durationList.append(good)
total_notes = total_notes + 1
n = note.Note()
n.pitch = detectedPitchObjects[j - 1]
notesList.append(n)
else:
bad = bad + good
good = 0
valid_note = False
j = j + 1
return notesList, durationList
def notesAndDurationsToStream(notesList, durationList, removeRestsAtBeginning=True):
"""Take a list of objects or rests and an equally long list of how long
each ones lasts in terms of samples and return a Stream using the information
from quarterLengthEstimation and quantizeDurations."""
qle = quarterLengthEstimation(durationList)
part = stream.Part()
for i in range(len(durationList)):
actualDuration = quantizeDuration(durationList[i] / qle)
notesList[i].quarterLength = actualDuration
if removeRestsAtBeginning and notesList[i].name == "rest":
pass
else:
part.append(notesList[i])
removeRestsAtBeginning = False
return part
def quarterLengthEstimation(durationList, mostRepeatedQuarterLength=1.0):
"""Take a list of lengths of notes (measured in audio samples) and try to
estimate what the length of a quarter note should be in this list."""
dl = copy.copy(durationList)
dl.append(0)
pdf, bins = histogram(dl,8.0)
i = len(pdf) - 1
while pdf[i] != max(pdf):
i = i - 1
qle = (bins[i] + bins[i + 1]) / 2.0
if mostRepeatedQuarterLength == 0:
mostRepeatedQuarterLength = 1.0
binPosition = 0 - math.log(mostRepeatedQuarterLength, 2)
qle = qle * math.pow(2, binPosition) # Normalize the length to a quarter note
return qle
def histogram(data, bins):
"""Partition a list into a number of bins and return the number of elements
in each bins and a set of elements where the first element is the start of
the first bin, the last element is the end of the last bin, and every remaining
element is the dividing point between one bin and another."""
maxValue = max(data)
minValue = min(data)
lengthEachBin = (maxValue-minValue) / bins
container = []
for i in range(int(bins)):
container.append(0)
for i in data:
count = 1
while i > minValue + count*lengthEachBin:
count += 1
container[count - 1] += 1
binsLimits = []
binsLimits.append(minValue)
count = 1
for i in range(int(bins)):
binsLimits.append(minValue+count*lengthEachBin)
count +=1
return container, binsLimits
def quantizeDuration(length):
"""Round an approximated quarterLength duration to better one."""
length = length * 100
typicalLengths = [25.00, 50.00, 100.00, 150.00, 200.00, 400.00]
thresholds = []
for i in range(len(typicalLengths) - 1):
thresholds.append((typicalLengths[i] + typicalLengths[i + 1]) / 2)
finalLength = typicalLengths[0]
for i in range(len(thresholds)):
threshold = thresholds[i]
if length > threshold:
finalLength = typicalLengths[i + 1]
return finalLength / 100
def polyphonicStreamFromFiles(filenames):
"""Generate a multi-part score using each file as a part."""
parts = [monophonicStreamFromFile(filename) for filename in filenames]
score = stream.Score()
for part in parts:
score.append(part)
return score
def monophonicStreamFromFile(filename):
"""Generate a score part from a wav file."""
freqFromAQList = getFrequenciesFromAudioFile(filename, 256)
detectedPitchesFreq = detectPitchFrequencies(freqFromAQList)
detectedPitchesFreq = smoothFrequencies(detectedPitchesFreq)
detectedPitchObjects = pitchFrequenciesToObjects(detectedPitchesFreq)
(notesList, durationList) = joinConsecutiveIdenticalPitches(detectedPitchObjects)
part = notesAndDurationsToStream(notesList, durationList, removeRestsAtBeginning=True)
return part