-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathaudio.py
More file actions
1786 lines (1439 loc) Β· 55.7 KB
/
audio.py
File metadata and controls
1786 lines (1439 loc) Β· 55.7 KB
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
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""Core IO, DSP and utility functions."""
from __future__ import annotations
import os
import pathlib
import warnings
import soundfile as sf
import audioread
import numpy as np
import scipy
import scipy.signal
import soxr
import lazy_loader as lazy
from numba import jit, stencil, guvectorize
from .fft import get_fftlib
from .convert import frames_to_samples, time_to_samples
from .._cache import cache
from .. import util
from ..util.exceptions import ParameterError
from ..util.decorators import deprecated
from ..util.deprecation import Deprecated, rename_kw
from .._typing import _FloatLike_co, _IntLike_co, _SequenceLike
from typing import Any, BinaryIO, Callable, Generator, Optional, Tuple, Union
from numpy.typing import DTypeLike
# Lazy-load optional dependencies
samplerate = lazy.load("samplerate")
resampy = lazy.load("resampy")
__all__ = [
"load",
"stream",
"to_mono",
"resample",
"get_duration",
"get_samplerate",
"autocorrelate",
"lpc",
"zero_crossings",
"clicks",
"tone",
"chirp",
"mu_compress",
"mu_expand",
]
# -- CORE ROUTINES --#
# Load should never be cached, since we cannot verify that the contents of
# 'path' are unchanged across calls.
def load(
path: Union[
str, int, os.PathLike[Any], sf.SoundFile, audioread.AudioFile, BinaryIO
],
*,
sr: Optional[float] = 22050,
mono: bool = True,
offset: float = 0.0,
duration: Optional[float] = None,
dtype: DTypeLike = np.float32,
res_type: str = "soxr_hq",
) -> Tuple[np.ndarray, Union[int, float]]:
"""Load an audio file as a floating point time series.
Audio will be automatically resampled to the given rate
(default ``sr=22050``).
To preserve the native sampling rate of the file, use ``sr=None``.
Parameters
----------
path : string, int, pathlib.Path, soundfile.SoundFile, audioread object, or file-like object
path to the input file.
Any codec supported by `soundfile` or `audioread` will work.
Any string file paths, or any object implementing Python's
file interface (e.g. `pathlib.Path`) are supported as `path`.
If the codec is supported by `soundfile`, then `path` can also be
an open file descriptor (int) or an existing `soundfile.SoundFile` object.
Pre-constructed audioread decoders are also supported here, see the example
below. This can be used, for example, to force a specific decoder rather
than relying upon audioread to select one for you.
.. warning:: audioread support is deprecated as of version 0.10.0.
audioread support be removed in version 1.0.
sr : number > 0 [scalar]
target sampling rate
'None' uses the native sampling rate
mono : bool
convert signal to mono
offset : float
start reading after this time (in seconds)
duration : float
only load up to this much audio (in seconds)
dtype : numeric type
data type of ``y``
res_type : str
resample type (see note)
.. note::
By default, this uses `soxr`'s high-quality mode ('HQ').
For alternative resampling modes, see `resample`
.. note::
`audioread` may truncate the precision of the audio data to 16 bits.
See :ref:`ioformats` for alternate loading methods.
Returns
-------
y : np.ndarray [shape=(n,) or (..., n)]
audio time series. Multi-channel is supported.
sr : number > 0 [scalar]
sampling rate of ``y``
Examples
--------
>>> # Load an ogg vorbis file
>>> filename = librosa.ex('trumpet')
>>> y, sr = librosa.load(filename)
>>> y
array([-1.407e-03, -4.461e-04, ..., -3.042e-05, 1.277e-05],
dtype=float32)
>>> sr
22050
>>> # Load a file and resample to 11 KHz
>>> filename = librosa.ex('trumpet')
>>> y, sr = librosa.load(filename, sr=11025)
>>> y
array([-8.746e-04, -3.363e-04, ..., -1.301e-05, 0.000e+00],
dtype=float32)
>>> sr
11025
>>> # Load 5 seconds of a file, starting 15 seconds in
>>> filename = librosa.ex('brahms')
>>> y, sr = librosa.load(filename, offset=15.0, duration=5.0)
>>> y
array([0.146, 0.144, ..., 0.128, 0.015], dtype=float32)
>>> sr
22050
>>> # Load using an already open SoundFile object
>>> import soundfile
>>> sfo = soundfile.SoundFile(librosa.ex('brahms'))
>>> y, sr = librosa.load(sfo)
>>> # Load using an already open audioread object
>>> import audioread.ffdec # Use ffmpeg decoder
>>> aro = audioread.ffdec.FFmpegAudioFile(librosa.ex('brahms'))
>>> y, sr = librosa.load(aro)
"""
if isinstance(path, tuple(audioread.available_backends())):
# Force the audioread loader if we have a reader object already
y, sr_native = __audioread_load(path, offset, duration, dtype)
else:
# Otherwise try soundfile first, and then fall back if necessary
try:
y, sr_native = __soundfile_load(path, offset, duration, dtype)
except sf.SoundFileRuntimeError as exc:
# If soundfile failed, try audioread instead
if isinstance(path, (str, pathlib.PurePath)):
warnings.warn(
"PySoundFile failed. Trying audioread instead.", stacklevel=2
)
y, sr_native = __audioread_load(path, offset, duration, dtype)
else:
raise exc
# Final cleanup for dtype and contiguity
if mono:
y = to_mono(y)
if sr is not None:
y = resample(y, orig_sr=sr_native, target_sr=sr, res_type=res_type)
else:
sr = sr_native
return y, sr
def __soundfile_load(path, offset, duration, dtype):
"""Load an audio buffer using soundfile."""
if isinstance(path, sf.SoundFile):
# If the user passed an existing soundfile object,
# we can use it directly
context = path
else:
# Otherwise, create the soundfile object
context = sf.SoundFile(path)
with context as sf_desc:
sr_native = sf_desc.samplerate
if offset:
# Seek to the start of the target read
sf_desc.seek(int(offset * sr_native))
if duration is not None:
frame_duration = int(duration * sr_native)
else:
frame_duration = -1
# Load the target number of frames, and transpose to match librosa form
y = sf_desc.read(frames=frame_duration, dtype=dtype, always_2d=False).T
return y, sr_native
@deprecated(version="0.10.0", version_removed="1.0")
def __audioread_load(path, offset, duration, dtype: DTypeLike):
"""Load an audio buffer using audioread.
This loads one block at a time, and then concatenates the results.
"""
buf = []
if isinstance(path, tuple(audioread.available_backends())):
# If we have an audioread object already, don't bother opening
reader = path
else:
# If the input was not an audioread object, try to open it
reader = audioread.audio_open(path)
with reader as input_file:
sr_native = input_file.samplerate
n_channels = input_file.channels
s_start = int(sr_native * offset) * n_channels
if duration is None:
s_end = np.inf
else:
s_end = s_start + (int(sr_native * duration) * n_channels)
n = 0
for frame in input_file:
frame = util.buf_to_float(frame, dtype=dtype)
n_prev = n
n = n + len(frame)
if n < s_start:
# offset is after the current frame
# keep reading
continue
if s_end < n_prev:
# we're off the end. stop reading
break
if s_end < n:
# the end is in this frame. crop.
frame = frame[: int(s_end - n_prev)] # pragma: no cover
if n_prev <= s_start <= n:
# beginning is in this frame
frame = frame[(s_start - n_prev) :]
# tack on the current frame
buf.append(frame)
if buf:
y = np.concatenate(buf)
if n_channels > 1:
y = y.reshape((-1, n_channels)).T
else:
y = np.empty(0, dtype=dtype)
return y, sr_native
def stream(
path: Union[str, int, sf.SoundFile, BinaryIO],
*,
block_length: int,
frame_length: int,
hop_length: int,
mono: bool = True,
offset: float = 0.0,
duration: Optional[float] = None,
fill_value: Optional[float] = None,
dtype: DTypeLike = np.float32,
) -> Generator[np.ndarray, None, None]:
"""Stream audio in fixed-length buffers.
This is primarily useful for processing large files that won't
fit entirely in memory at once.
Instead of loading the entire audio signal into memory (as
in `load`, this function produces *blocks* of audio spanning
a fixed number of frames at a specified frame length and hop
length.
While this function strives for similar behavior to `load`,
there are a few caveats that users should be aware of:
1. This function does not return audio buffers directly.
It returns a generator, which you can iterate over
to produce blocks of audio. A *block*, in this context,
refers to a buffer of audio which spans a given number of
(potentially overlapping) frames.
2. Automatic sample-rate conversion is not supported.
Audio will be streamed in its native sample rate,
so no default values are provided for ``frame_length``
and ``hop_length``. It is recommended that you first
get the sampling rate for the file in question, using
`get_samplerate`, and set these parameters accordingly.
3. Many analyses require access to the entire signal
to behave correctly, such as `resample`, `cqt`, or
`beat_track`, so these methods will not be appropriate
for streamed data.
4. The ``block_length`` parameter specifies how many frames
of audio will be produced per block. Larger values will
consume more memory, but will be more efficient to process
down-stream. The best value will ultimately depend on your
application and other system constraints.
5. By default, most librosa analyses (e.g., short-time Fourier
transform) assume centered frames, which requires padding the
signal at the beginning and end. This will not work correctly
when the signal is carved into blocks, because it would introduce
padding in the middle of the signal. To disable this feature,
use ``center=False`` in all frame-based analyses.
See the examples below for proper usage of this function.
Parameters
----------
path : string, int, sf.SoundFile, or file-like object
path to the input file to stream.
Any codec supported by `soundfile` is permitted here.
An existing `soundfile.SoundFile` object may also be provided.
block_length : int > 0
The number of frames to include in each block.
Note that at the end of the file, there may not be enough
data to fill an entire block, resulting in a shorter block
by default. To pad the signal out so that blocks are always
full length, set ``fill_value`` (see below).
frame_length : int > 0
The number of samples per frame.
hop_length : int > 0
The number of samples to advance between frames.
Note that by when ``hop_length < frame_length``, neighboring frames
will overlap. Similarly, the last frame of one *block* will overlap
with the first frame of the next *block*.
mono : bool
Convert the signal to mono during streaming
offset : float
Start reading after this time (in seconds)
duration : float
Only load up to this much audio (in seconds)
fill_value : float [optional]
If padding the signal to produce constant-length blocks,
this value will be used at the end of the signal.
In most cases, ``fill_value=0`` (silence) is expected, but
you may specify any value here.
dtype : numeric type
data type of audio buffers to be produced
Yields
------
y : np.ndarray
An audio buffer of (at most)
``(block_length-1) * hop_length + frame_length`` samples.
See Also
--------
load
get_samplerate
soundfile.blocks
Examples
--------
Apply a short-term Fourier transform to blocks of 256 frames
at a time. Note that streaming operation requires left-aligned
frames, so we must set ``center=False`` to avoid padding artifacts.
>>> filename = librosa.ex('brahms')
>>> sr = librosa.get_samplerate(filename)
>>> stream = librosa.stream(filename,
... block_length=256,
... frame_length=4096,
... hop_length=1024)
>>> for y_block in stream:
... D_block = librosa.stft(y_block, center=False)
Or compute a mel spectrogram over a stream, using a shorter frame
and non-overlapping windows
>>> filename = librosa.ex('brahms')
>>> sr = librosa.get_samplerate(filename)
>>> stream = librosa.stream(filename,
... block_length=256,
... frame_length=2048,
... hop_length=2048)
>>> for y_block in stream:
... m_block = librosa.feature.melspectrogram(y=y_block, sr=sr,
... n_fft=2048,
... hop_length=2048,
... center=False)
"""
if not util.is_positive_int(block_length):
raise ParameterError(f"block_length={block_length} must be a positive integer")
if not util.is_positive_int(frame_length):
raise ParameterError(f"frame_length={frame_length} must be a positive integer")
if not util.is_positive_int(hop_length):
raise ParameterError(f"hop_length={hop_length} must be a positive integer")
if isinstance(path, sf.SoundFile):
sfo = path
else:
sfo = sf.SoundFile(path)
# Get the sample rate from the file info
sr = sfo.samplerate
# Construct the stream
if offset:
start = int(offset * sr)
else:
start = 0
if duration:
frames = int(duration * sr)
else:
frames = -1
# Seek the soundfile object to the starting frame
sfo.seek(start)
blocks = sfo.blocks(
blocksize=frame_length + (block_length - 1) * hop_length,
overlap=frame_length - hop_length,
frames=frames,
dtype=dtype,
always_2d=False,
fill_value=fill_value,
)
for block in blocks:
if mono:
yield to_mono(block.T)
else:
yield block.T
@cache(level=20)
def to_mono(y: np.ndarray) -> np.ndarray:
"""Convert an audio signal to mono by averaging samples across channels.
Parameters
----------
y : np.ndarray [shape=(..., n)]
audio time series. Multi-channel is supported.
Returns
-------
y_mono : np.ndarray [shape=(n,)]
``y`` as a monophonic time-series
Notes
-----
This function caches at level 20.
Examples
--------
>>> y, sr = librosa.load(librosa.ex('trumpet', hq=True), mono=False)
>>> y.shape
(2, 117601)
>>> y_mono = librosa.to_mono(y)
>>> y_mono.shape
(117601,)
"""
# Validate the buffer
util.valid_audio(y)
if y.ndim > 1:
y = np.mean(y, axis=tuple(range(y.ndim - 1)))
return y
@cache(level=20)
def resample(
y: np.ndarray,
*,
orig_sr: float,
target_sr: float,
res_type: str = "soxr_hq",
fix: bool = True,
scale: bool = False,
axis: int = -1,
**kwargs: Any,
) -> np.ndarray:
"""Resample a time series from orig_sr to target_sr
By default, this uses a high-quality method (`soxr_hq`) for band-limited sinc
interpolation. The alternate ``res_type`` values listed below offer different
trade-offs of speed and quality.
Parameters
----------
y : np.ndarray [shape=(..., n, ...)]
audio time series, with `n` samples along the specified axis.
orig_sr : number > 0 [scalar]
original sampling rate of ``y``
target_sr : number > 0 [scalar]
target sampling rate
res_type : str (default: `soxr_hq`)
resample type
'soxr_vhq', 'soxr_hq', 'soxr_mq' or 'soxr_lq'
`soxr` Very high-, High-, Medium-, Low-quality FFT-based bandlimited interpolation.
``'soxr_hq'`` is the default setting of `soxr`.
'soxr_qq'
`soxr` Quick cubic interpolation (very fast, but not bandlimited)
'kaiser_best'
`resampy` high-quality mode
'kaiser_fast'
`resampy` faster method
'fft' or 'scipy'
`scipy.signal.resample` Fourier method.
'polyphase'
`scipy.signal.resample_poly` polyphase filtering. (fast)
'linear'
`samplerate` linear interpolation. (very fast, but not bandlimited)
'zero_order_hold'
`samplerate` repeat the last value between samples. (very fast, but not bandlimited)
'sinc_best', 'sinc_medium' or 'sinc_fastest'
`samplerate` high-, medium-, and low-quality bandlimited sinc interpolation.
.. note::
Not all options yield a bandlimited interpolator. If you use `soxr_qq`, `polyphase`,
`linear`, or `zero_order_hold`, you need to be aware of possible aliasing effects.
.. note::
`samplerate` and `resampy` are not installed with `librosa`.
To use `samplerate` or `resampy`, they should be installed manually::
$ pip install samplerate
$ pip install resampy
.. note::
When using ``res_type='polyphase'``, only integer sampling rates are
supported.
fix : bool
adjust the length of the resampled signal to be of size exactly
``ceil(target_sr * len(y) / orig_sr)``
scale : bool
Scale the resampled signal so that ``y`` and ``y_hat`` have approximately
equal total energy.
axis : int
The target axis along which to resample. Defaults to the trailing axis.
**kwargs : additional keyword arguments
If ``fix==True``, additional keyword arguments to pass to
`librosa.util.fix_length`.
Returns
-------
y_hat : np.ndarray [shape=(..., n * target_sr / orig_sr, ...)]
``y`` resampled from ``orig_sr`` to ``target_sr`` along the target axis
Raises
------
ParameterError
If ``res_type='polyphase'`` and ``orig_sr`` or ``target_sr`` are not both
integer-valued.
See Also
--------
librosa.util.fix_length
scipy.signal.resample
resampy
samplerate.converters.resample
soxr.resample
Notes
-----
This function caches at level 20.
Examples
--------
Downsample from 22 KHz to 8 KHz
>>> y, sr = librosa.load(librosa.ex('trumpet'), sr=22050)
>>> y_8k = librosa.resample(y, orig_sr=sr, target_sr=8000)
>>> y.shape, y_8k.shape
((117601,), (42668,))
"""
# First, validate the audio buffer
util.valid_audio(y)
if orig_sr == target_sr:
return y
ratio = float(target_sr) / orig_sr
n_samples = int(np.ceil(y.shape[axis] * ratio))
if res_type in ("scipy", "fft"):
y_hat = scipy.signal.resample(y, n_samples, axis=axis)
elif res_type == "polyphase":
if int(orig_sr) != orig_sr or int(target_sr) != target_sr:
raise ParameterError(
"polyphase resampling is only supported for integer-valued sampling rates."
)
# For polyphase resampling, we need up- and down-sampling ratios
# We can get those from the greatest common divisor of the rates
# as long as the rates are integrable
orig_sr = int(orig_sr)
target_sr = int(target_sr)
gcd = np.gcd(orig_sr, target_sr)
y_hat = scipy.signal.resample_poly(
y, target_sr // gcd, orig_sr // gcd, axis=axis
)
elif res_type in (
"linear",
"zero_order_hold",
"sinc_best",
"sinc_fastest",
"sinc_medium",
):
# Use numpy to vectorize the resampler along the target axis
# This is because samplerate does not support ndim>2 generally.
y_hat = np.apply_along_axis(
samplerate.resample, axis=axis, arr=y, ratio=ratio, converter_type=res_type
)
elif res_type.startswith("soxr"):
# Use numpy to vectorize the resampler along the target axis
# This is because soxr does not support ndim>2 generally.
y_hat = np.apply_along_axis(
soxr.resample,
axis=axis,
arr=y,
in_rate=orig_sr,
out_rate=target_sr,
quality=res_type,
)
else:
y_hat = resampy.resample(y, orig_sr, target_sr, filter=res_type, axis=axis)
if fix:
y_hat = util.fix_length(y_hat, size=n_samples, axis=axis, **kwargs)
if scale:
y_hat /= np.sqrt(ratio)
# Match dtypes
return np.asarray(y_hat, dtype=y.dtype)
def get_duration(
*,
y: Optional[np.ndarray] = None,
sr: float = 22050,
S: Optional[np.ndarray] = None,
n_fft: int = 2048,
hop_length: int = 512,
center: bool = True,
path: Optional[Union[str, os.PathLike[Any]]] = None,
filename: Optional[Union[str, os.PathLike[Any], Deprecated]] = Deprecated(),
) -> float:
"""Compute the duration (in seconds) of an audio time series,
feature matrix, or filename.
Examples
--------
>>> # Load an example audio file
>>> y, sr = librosa.load(librosa.ex('trumpet'))
>>> librosa.get_duration(y=y, sr=sr)
5.333378684807256
>>> # Or directly from an audio file
>>> librosa.get_duration(filename=librosa.ex('trumpet'))
5.333378684807256
>>> # Or compute duration from an STFT matrix
>>> y, sr = librosa.load(librosa.ex('trumpet'))
>>> S = librosa.stft(y)
>>> librosa.get_duration(S=S, sr=sr)
5.317369614512471
>>> # Or a non-centered STFT matrix
>>> S_left = librosa.stft(y, center=False)
>>> librosa.get_duration(S=S_left, sr=sr)
5.224489795918367
Parameters
----------
y : np.ndarray [shape=(..., n)] or None
audio time series. Multi-channel is supported.
sr : number > 0 [scalar]
audio sampling rate of ``y``
S : np.ndarray [shape=(..., d, t)] or None
STFT matrix, or any STFT-derived matrix (e.g., chromagram
or mel spectrogram).
Durations calculated from spectrogram inputs are only accurate
up to the frame resolution. If high precision is required,
it is better to use the audio time series directly.
n_fft : int > 0 [scalar]
FFT window size for ``S``
hop_length : int > 0 [ scalar]
number of audio samples between columns of ``S``
center : boolean
- If ``True``, ``S[:, t]`` is centered at ``y[t * hop_length]``
- If ``False``, then ``S[:, t]`` begins at ``y[t * hop_length]``
path : str, path, or file-like
If provided, all other parameters are ignored, and the
duration is calculated directly from the audio file.
Note that this avoids loading the contents into memory,
and is therefore useful for querying the duration of
long files.
As in ``load``, this can also be an integer or open file-handle
that can be processed by ``soundfile``.
filename : Deprecated
Equivalent to ``path``
.. warning:: This parameter has been renamed to ``path`` in 0.10.
Support for ``filename=`` will be removed in 1.0.
Returns
-------
d : float >= 0
Duration (in seconds) of the input time series or spectrogram.
Raises
------
ParameterError
if none of ``y``, ``S``, or ``path`` are provided.
Notes
-----
`get_duration` can be applied to a file (``path``), a spectrogram (``S``),
or audio buffer (``y, sr``). Only one of these three options should be
provided. If you do provide multiple options (e.g., ``path`` and ``S``),
then ``path`` takes precedence over ``S``, and ``S`` takes precedence over
``(y, sr)``.
"""
path = rename_kw(
old_name="filename",
old_value=filename,
new_name="path",
new_value=path,
version_deprecated="0.10.0",
version_removed="1.0",
)
if path is not None:
try:
return sf.info(path).duration # type: ignore
except sf.SoundFileRuntimeError:
warnings.warn(
"PySoundFile failed. Trying audioread instead."
"\n\tAudioread support is deprecated in librosa 0.10.0"
" and will be removed in version 1.0.",
stacklevel=2,
category=FutureWarning,
)
with audioread.audio_open(path) as fdesc:
return fdesc.duration # type: ignore
if y is None:
if S is None:
raise ParameterError("At least one of (y, sr), S, or path must be provided")
n_frames = S.shape[-1]
n_samples = n_fft + hop_length * (n_frames - 1)
# If centered, we lose half a window from each end of S
if center:
n_samples = n_samples - 2 * int(n_fft // 2)
else:
n_samples = y.shape[-1]
return float(n_samples) / sr
def get_samplerate(path: Union[str, int, sf.SoundFile, BinaryIO]) -> float:
"""Get the sampling rate for a given file.
Parameters
----------
path : string, int, soundfile.SoundFile, or file-like
The path to the file to be loaded
As in ``load``, this can also be an integer or open file-handle
that can be processed by `soundfile`.
An existing `soundfile.SoundFile` object can also be supplied.
Returns
-------
sr : number > 0
The sampling rate of the given audio file
Examples
--------
Get the sampling rate for the included audio file
>>> path = librosa.ex('trumpet')
>>> librosa.get_samplerate(path)
22050
"""
try:
if isinstance(path, sf.SoundFile):
return path.samplerate # type: ignore
return sf.info(path).samplerate # type: ignore
except sf.SoundFileRuntimeError:
warnings.warn(
"PySoundFile failed. Trying audioread instead."
"\n\tAudioread support is deprecated in librosa 0.10.0"
" and will be removed in version 1.0.",
stacklevel=2,
category=FutureWarning,
)
with audioread.audio_open(path) as fdesc:
return fdesc.samplerate # type: ignore
@cache(level=20)
def autocorrelate(
y: np.ndarray, *, max_size: Optional[int] = None, axis: int = -1
) -> np.ndarray:
"""Bounded-lag auto-correlation
Parameters
----------
y : np.ndarray
array to autocorrelate
max_size : int > 0 or None
maximum correlation lag.
If unspecified, defaults to ``y.shape[axis]`` (unbounded)
axis : int
The axis along which to autocorrelate.
By default, the last axis (-1) is taken.
Returns
-------
z : np.ndarray
truncated autocorrelation ``y*y`` along the specified axis.
If ``max_size`` is specified, then ``z.shape[axis]`` is bounded
to ``max_size``.
Notes
-----
This function caches at level 20.
Examples
--------
Compute full autocorrelation of ``y``
>>> y, sr = librosa.load(librosa.ex('trumpet'))
>>> librosa.autocorrelate(y)
array([ 6.899e+02, 6.236e+02, ..., 3.710e-08, -1.796e-08])
Compute onset strength auto-correlation up to 4 seconds
>>> import matplotlib.pyplot as plt
>>> odf = librosa.onset.onset_strength(y=y, sr=sr, hop_length=512)
>>> ac = librosa.autocorrelate(odf, max_size=4 * sr // 512)
>>> fig, ax = plt.subplots()
>>> ax.plot(ac)
>>> ax.set(title='Auto-correlation', xlabel='Lag (frames)')
"""
if max_size is None:
max_size = y.shape[axis]
max_size = int(min(max_size, y.shape[axis]))
fft = get_fftlib()
real = not np.iscomplexobj(y)
# Pad out the signal to support full-length auto-correlation
n_pad = scipy.fft.next_fast_len(2 * y.shape[axis] - 1, real=real)
if real:
# Compute the power spectrum along the chosen axis
powspec = util.abs2(fft.rfft(y, n=n_pad, axis=axis))
# Convert back to time domain
autocorr = fft.irfft(powspec, n=n_pad, axis=axis)
else:
# Compute the power spectrum along the chosen axis
powspec = util.abs2(fft.fft(y, n=n_pad, axis=axis))
# Convert back to time domain
autocorr = fft.ifft(powspec, n=n_pad, axis=axis)
# Slice down to max_size
subslice = [slice(None)] * autocorr.ndim
subslice[axis] = slice(max_size)
autocorr_slice: np.ndarray = autocorr[tuple(subslice)]
return autocorr_slice
def lpc(y: np.ndarray, *, order: int, axis: int = -1) -> np.ndarray:
"""Linear Prediction Coefficients via Burg's method
This function applies Burg's method to estimate coefficients of a linear
filter on ``y`` of order ``order``. Burg's method is an extension to the
Yule-Walker approach, which are both sometimes referred to as LPC parameter
estimation by autocorrelation.
It follows the description and implementation approach described in the
introduction by Marple. [#]_ N.B. This paper describes a different method, which
is not implemented here, but has been chosen for its clear explanation of
Burg's technique in its introduction.
.. [#] Larry Marple.
A New Autoregressive Spectrum Analysis Algorithm.
IEEE Transactions on Acoustics, Speech, and Signal Processing
vol 28, no. 4, 1980.
Parameters
----------
y : np.ndarray [shape=(..., n)]
Time series to fit. Multi-channel is supported..
order : int > 0
Order of the linear filter
axis : int
Axis along which to compute the coefficients
Returns
-------
a : np.ndarray [shape=(..., order + 1)]
LP prediction error coefficients, i.e. filter denominator polynomial.
Note that the length along the specified ``axis`` will be ``order+1``.
Raises
------
ParameterError
- If ``y`` is not valid audio as per `librosa.util.valid_audio`
- If ``order < 1`` or not integer
FloatingPointError
- If ``y`` is ill-conditioned
See Also
--------
scipy.signal.lfilter
Examples
--------
Compute LP coefficients of y at order 16 on entire series
>>> y, sr = librosa.load(librosa.ex('libri1'))
>>> librosa.lpc(y, order=16)
Compute LP coefficients, and plot LP estimate of original series
>>> import matplotlib.pyplot as plt
>>> import scipy