ivanwu/jarvis-models
1
0
Fork 0
mirror of https://github.com/BoardWare-Genius/jarvis-models.git synced 2025-12-14 00:53:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: asr blackblox

Browse Source
This commit is contained in:
superobk
2024-03-19 11:44:51 +08:00
parent 4e2a4ef63c
commit 2dccf5e78d
19 changed files with 1408 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

201
src/asr/rapid_paraformer/kaldifeat/LICENSE Normal file
View File

@ -0,0 +1,201 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

108
src/asr/rapid_paraformer/kaldifeat/README.md Normal file
View File

@ -0,0 +1,108 @@
# KaldiFeat
KaldiFeat is a light-weight Python library for computing Kaldi-style acoustic features based on NumPy. It might be helpful if you want to:
- Test a pre-trained model on new data without writing shell commands and creating a bunch of files.
- Run a pre-trained model in a new environment without installing Kaldi.
## Example
The following codes calculate MFCCs with the same settings in `kaldi/egs/voxceleb/v2`
```
import librosa
from kaldifeat import compute_mfcc_feats, compute_vad, apply_cmvn_sliding
# Assume we have a wav file called example.wav whose sample rate is 16000 Hz
data, _ = librosa.load('example.wav', 16000)
# We adopt 16 bits data, thus we need to transform dtype from float to int16 for librosa
data = (data * 32768).astype(np.int16)
raw_mfcc = compute_mfcc_feats(data, sample_frequency=16000, frame_length=25, frame_shift=10, low_freq=20, high_freq=-400, num_mel_bins=30, num_ceps=30, snip_edges=False)
log_energy = raw_mfcc[:, 0]
vad = compute_vad(log_energy, energy_threshold=5.5, energy_mean_scale=0.5, frames_context=2, proportion_threshold=0.12)
mfcc = apply_cmvn_sliding(raw_mfcc, window=300, center=True)[vad]
```
## Supported Functions
### compute_fbank_feats
Compute (log) Mel filter bank energies (FBanks) in the same way as `kaldi/src/featbin/compute_fbank_feats`
| Parameters | Description |
| :--------- | :---------- |
|blackman_coeff| Constant coefficient for generalized Blackman window. (float, default = 0.42)|
|dither| Dithering constant (0.0 means no dither). If you turn this off, you should set the --energy-floor option, e.g. to 1.0 or 0.1 (float, default = 1)|
|energy_floor| Floor on energy (absolute, not relative) in FBANK computation. Only makes a difference if --use-energy=true; only necessary if --dither=0.0. Suggested values: 0.1 or 1.0 (float, default = 0)|
|frame_length| Frame length in milliseconds (float, default = 25)|
|frame_shift| Frame shift in milliseconds (float, default = 10)|
|high_freq| High cutoff frequency for mel bins (if <= 0, offset from Nyquist) (float, default = 0)|
|low_freq| Low cutoff frequency for mel bins (float, default = 20)|
|num_mel_bins| Number of triangular mel-frequency bins (int, default = 23)|
|preemphasis_coefficient| Coefficient for use in signal preemphasis (float, default = 0.97)|
|raw_energy| If true, compute energy before preemphasis and windowing (bool, default = true)|
|remove_dc_offset| Subtract mean from waveform on each frame (bool, default = true)|
|round_to_power_of_two| If true, round window size to power of two by zero-padding input to FFT. (bool, default = true)|
|sample_frequency| Waveform data sample frequency (must match the waveform file, if specified there) (float, default = 16000)|
|snip_edges| If true, end effects will be handled by outputting only frames that completely fit in the file, and the number of frames depends on the frame-length. If false, the number of frames depends only on the frame-shift, and we reflect the data at the ends. (bool, default = true)|
|use_energy| Add an extra energy output. (bool, default = false)|
|use_log_fbank| If true, produce log-filterbank, else produce linear. (bool, default = true)|
|use_power| If true, use power, else use magnitude. (bool, default = true)|
|window_type| Type of window ("hamming"\|"hanning"\|"povey"\|"rectangular"\|"sine"\|"blackmann") (string, default = "povey")|
|dtype| Type of array (np.float32\|np.float64) (dtype or string, default=np.float32)|
### compute_mfcc_feats
Compute Mel-frequency cepstral coefficients (MFCCs) in the same way as `kaldi/src/featbin/compute_mfcc_feats`
| Parameters | Description |
| :--------- | :---------- |
|blackman_coeff| Constant coefficient for generalized Blackman window. (float, default = 0.42)|
|cepstral_lifter| Constant that controls scaling of MFCCs (float, default = 22)|
|dither| Dithering constant (0.0 means no dither). If you turn this off, you should set the --energy-floor option, e.g. to 1.0 or 0.1 (float, default = 1)|
|energy_floor| Floor on energy (absolute, not relative) in MFCC computation. Only makes a difference if --use-energy=true; only necessary if --dither=0.0. Suggested values: 0.1 or 1.0 (float, default = 0)|
|frame_length| Frame length in milliseconds (float, default = 25)|
|frame_shift| Frame shift in milliseconds (float, default = 10)|
|high_freq| High cutoff frequency for mel bins (if <= 0, offset from Nyquist) (float, default = 0)|
|low_freq| Low cutoff frequency for mel bins (float, default = 20)|
|num_ceps| Number of cepstra in MFCC computation (including C0) (int, default = 13)|
|num_mel_bins| Number of triangular mel-frequency bins (int, default = 23)|
|preemphasis_coefficient| Coefficient for use in signal preemphasis (float, default = 0.97)|
|raw_energy| If true, compute energy before preemphasis and windowing (bool, default = true)|
|remove_dc_offset| Subtract mean from waveform on each frame (bool, default = true)|
|round_to_power_of_two| If true, round window size to power of two by zero-padding input to FFT. (bool, default = true)|
|sample_frequency| Waveform data sample frequency (must match the waveform file, if specified there) (float, default = 16000)|
|snip_edges| If true, end effects will be handled by outputting only frames that completely fit in the file, and the number of frames depends on the frame-length. If false, the number of frames depends only on the frame-shift, and we reflect the data at the ends. (bool, default = true)|
|use_energy| Use energy (not C0) in MFCC computation (bool, default = true)|
|window_type| Type of window ("hamming"\|"hanning"\|"povey"\|"rectangular"\|"sine"\|"blackmann") (string, default = "povey")|
|dtype| Type of array (np.float32\|np.float64) (dtype or string, default=np.float32)|
### apply_cmvn_sliding
Apply sliding-window cepstral mean (and optionally variance) normalization in the same way as `kaldi/src/featbin/apply_cmvn_sliding`
| Parameters | Description |
| :--------- | :---------- |
|center| If true, use a window centered on the current frame (to the extent possible, modulo end effects). If false, window is to the left. (bool, default = false)|
|window| Window in frames for running average CMN computation (int, default = 600)|
|min_window| Minimum CMN window used at start of decoding (adds latency only at start). Only applicable if center == false, ignored if center==true (int, default = 100)|
|norm_vars| If true, normalize variance to one. (bool, default = false)|
### compute_vad
Apply energy-based voice activity detection in the same way as `kaldi/src/ivectorbin/compute_vad`
| Parameters | Description |
| :--------- | :---------- |
|energy_mean_scale| If this is set to s, to get the actual threshold we let m be the mean log-energy of the file, and use s\*m + vad-energy-threshold (float, default = 0.5)|
|energy_threshold| Constant term in energy threshold for VAD (also see energy_mean_scale) (float, default = 5)|
|frames_context| Number of frames of context on each side of central frame, in window for which energy is monitored (int, default = 0)|
|proportion_threshold| Parameter controlling the proportion of frames within the window that need to have more energy than the threshold (float, default = 0.6)|
### Related Projects
- [python_speech_features](https://github.com/jameslyons/python_speech_features)
- [python_kaldi_features](https://github.com/ZitengWang/python_kaldi_features)

3
src/asr/rapid_paraformer/kaldifeat/__init__.py Normal file
View File

@ -0,0 +1,3 @@
# -*- encoding: utf-8 -*-
from .feature import compute_fbank_feats, compute_mfcc_feats, apply_cmvn_sliding
from .ivector import compute_vad

459
src/asr/rapid_paraformer/kaldifeat/feature.py Normal file
View File

@ -0,0 +1,459 @@
import numpy as np
from scipy.fftpack import dct
# ---------- feature-window ----------
def sliding_window(x, window_size, window_shift):
shape = x.shape[:-1] + (x.shape[-1] - window_size + 1, window_size)
strides = x.strides + (x.strides[-1],)
return np.lib.stride_tricks.as_strided(x, shape=shape, strides=strides)[::window_shift]
def func_num_frames(num_samples, window_size, window_shift, snip_edges):
if snip_edges:
if num_samples < window_size:
return 0
else:
return 1 + ((num_samples - window_size) // window_shift)
else:
return (num_samples + (window_shift // 2)) // window_shift
def func_dither(waveform, dither_value):
if dither_value == 0.0:
return waveform
waveform += np.random.normal(size=waveform.shape).astype(waveform.dtype) * dither_value
return waveform
def func_remove_dc_offset(waveform):
return waveform - np.mean(waveform)
def func_log_energy(waveform):
return np.log(np.dot(waveform, waveform).clip(min=np.finfo(waveform.dtype).eps))
def func_preemphasis(waveform, preemph_coeff):
if preemph_coeff == 0.0:
return waveform
assert 0 < preemph_coeff <= 1
waveform[1:] -= preemph_coeff * waveform[:-1]
waveform[0] -= preemph_coeff * waveform[0]
return waveform
def sine(M):
if M < 1:
return np.array([])
if M == 1:
return np.ones(1, float)
n = np.arange(0, M)
return np.sin(np.pi*n/(M-1))
def povey(M):
if M < 1:
return np.array([])
if M == 1:
return np.ones(1, float)
n = np.arange(0, M)
return (0.5 - 0.5*np.cos(2.0*np.pi*n/(M-1)))**0.85
def feature_window_function(window_type, window_size, blackman_coeff):
assert window_size > 0
if window_type == 'hanning':
return np.hanning(window_size)
elif window_type == 'sine':
return sine(window_size)
elif window_type == 'hamming':
return np.hamming(window_size)
elif window_type == 'povey':
return povey(window_size)
elif window_type == 'rectangular':
return np.ones(window_size)
elif window_type == 'blackman':
window_func = np.blackman(window_size)
if blackman_coeff == 0.42:
return window_func
else:
return window_func - 0.42 + blackman_coeff
else:
raise ValueError('Invalid window type {}'.format(window_type))
def process_window(window, dither, remove_dc_offset, preemphasis_coefficient, window_function, raw_energy):
if dither != 0.0:
window = func_dither(window, dither)
if remove_dc_offset:
window = func_remove_dc_offset(window)
if raw_energy:
log_energy = func_log_energy(window)
if preemphasis_coefficient != 0.0:
window = func_preemphasis(window, preemphasis_coefficient)
window *= window_function
if not raw_energy:
log_energy = func_log_energy(window)
return window, log_energy
def extract_window(waveform, blackman_coeff, dither, window_size, window_shift,
preemphasis_coefficient, raw_energy, remove_dc_offset,
snip_edges, window_type, dtype):
num_samples = len(waveform)
num_frames = func_num_frames(num_samples, window_size, window_shift, snip_edges)
num_samples_ = (num_frames - 1) * window_shift + window_size
if snip_edges:
waveform = waveform[:num_samples_]
else:
offset = window_shift // 2 - window_size // 2
waveform = np.concatenate([
waveform[-offset - 1::-1],
waveform,
waveform[:-(offset + num_samples_ - num_samples + 1):-1]
])
frames = sliding_window(waveform, window_size=window_size, window_shift=window_shift)
frames = frames.astype(dtype)
log_enery = np.empty(frames.shape[0], dtype=dtype)
for i in range(frames.shape[0]):
frames[i], log_enery[i] = process_window(
window=frames[i],
dither=dither,
remove_dc_offset=remove_dc_offset,
preemphasis_coefficient=preemphasis_coefficient,
window_function=feature_window_function(
window_type=window_type,
window_size=window_size,
blackman_coeff=blackman_coeff
).astype(dtype),
raw_energy=raw_energy
)
return frames, log_enery
# ---------- feature-window ----------
# ---------- feature-functions ----------
def compute_spectrum(frames, n):
complex_spec = np.fft.rfft(frames, n)
return np.absolute(complex_spec)
def compute_power_spectrum(frames, n):
return np.square(compute_spectrum(frames, n))
def apply_cmvn_sliding_internal(feat, center=False, window=600, min_window=100, norm_vars=False):
num_frames, feat_dim = feat.shape
std = 1
if center:
if num_frames <= window:
mean = feat.mean(axis=0, keepdims=True).repeat(num_frames, axis=0)
if norm_vars:
std = feat.std(axis=0, keepdims=True).repeat(num_frames, axis=0)
else:
feat1 = feat[:window]
feat2 = sliding_window(feat.T, window, 1)
feat3 = feat[-window:]
mean1 = feat1.mean(axis=0, keepdims=True).repeat(window // 2, axis=0)
mean2 = feat2.mean(axis=2).T
mean3 = feat3.mean(axis=0, keepdims=True).repeat((window - 1) // 2, axis=0)
mean = np.concatenate([mean1, mean2, mean3])
if norm_vars:
std1 = feat1.std(axis=0, keepdims=True).repeat(window // 2, axis=0)
std2 = feat2.std(axis=2).T
std3 = feat3.mean(axis=0, keepdims=True).repeat((window - 1) // 2, axis=0)
std = np.concatenate([std1, std2, std3])
else:
if num_frames <= min_window:
mean = feat.mean(axis=0, keepdims=True).repeat(num_frames, axis=0)
if norm_vars:
std = feat.std(axis=0, keepdims=True).repeat(num_frames, axis=0)
else:
feat1 = feat[:min_window]
mean1 = feat1.mean(axis=0, keepdims=True).repeat(min_window, axis=0)
feat2_cumsum = np.cumsum(feat[:window], axis=0)[min_window:]
cumcnt = np.arange(min_window + 1, min(window, num_frames) + 1, dtype=feat.dtype)[:, np.newaxis]
mean2 = feat2_cumsum / cumcnt
mean = np.concatenate([mean1, mean2])
if norm_vars:
std1 = feat1.std(axis=0, keepdims=True).repeat(min_window, axis=0)
feat2_power_cumsum = np.cumsum(np.square(feat[:window]), axis=0)[min_window:]
std2 = np.sqrt(feat2_power_cumsum / cumcnt - np.square(mean2))
std = np.concatenate([std1, std2])
if num_frames > window:
feat3 = sliding_window(feat.T, window, 1)
mean3 = feat3.mean(axis=2).T
mean = np.concatenate([mean, mean3[1:]])
if norm_vars:
std3 = feat3.std(axis=2).T
std = np.concatenate([std, std3[1:]])
feat = (feat - mean) / std
return feat
# ---------- feature-functions ----------
# ---------- mel-computations ----------
def inverse_mel_scale(mel_freq):
return 700.0 * (np.exp(mel_freq / 1127.0) - 1.0)
def mel_scale(freq):
return 1127.0 * np.log(1.0 + freq / 700.0)
def compute_mel_banks(num_bins, sample_frequency, low_freq, high_freq, n):
""" Compute Mel banks.
:param num_bins: Number of triangular mel-frequency bins
:param sample_frequency: Waveform data sample frequency
:param low_freq: Low cutoff frequency for mel bins
:param high_freq: High cutoff frequency for mel bins (if <= 0, offset from Nyquist)
:param n: Window size
:return: Mel banks.
"""
assert num_bins >= 3, 'Must have at least 3 mel bins'
num_fft_bins = n // 2
nyquist = 0.5 * sample_frequency
if high_freq <= 0:
high_freq = nyquist + high_freq
assert 0 <= low_freq < high_freq <= nyquist
fft_bin_width = sample_frequency / n
mel_low_freq = mel_scale(low_freq)
mel_high_freq = mel_scale(high_freq)
mel_freq_delta = (mel_high_freq - mel_low_freq) / (num_bins + 1)
mel_banks = np.zeros([num_bins, num_fft_bins + 1])
for i in range(num_bins):
left_mel = mel_low_freq + mel_freq_delta * i
center_mel = left_mel + mel_freq_delta
right_mel = center_mel + mel_freq_delta
for j in range(num_fft_bins):
mel = mel_scale(fft_bin_width * j)
if left_mel < mel < right_mel:
if mel <= center_mel:
mel_banks[i, j] = (mel - left_mel) / (center_mel - left_mel)
else:
mel_banks[i, j] = (right_mel - mel) / (right_mel - center_mel)
return mel_banks
def compute_lifter_coeffs(q, M):
""" Compute liftering coefficients (scaling on cepstral coeffs)
the zeroth index is C0, which is not affected.
:param q: Number of lifters
:param M: Number of coefficients
:return: Lifters.
"""
if M < 1:
return np.array([])
if M == 1:
return np.ones(1, float)
n = np.arange(0, M)
return 1 + 0.5*np.sin(np.pi*n/q)*q
# ---------- mel-computations ----------
# ---------- compute-fbank-feats ----------
def compute_fbank_feats(
waveform,
blackman_coeff=0.42,
dither=1.0,
energy_floor=1.0,
frame_length=25,
frame_shift=10,
high_freq=0,
low_freq=20,
num_mel_bins=23,
preemphasis_coefficient=0.97,
raw_energy=True,
remove_dc_offset=True,
round_to_power_of_two=True,
sample_frequency=16000,
snip_edges=True,
use_energy=False,
use_log_fbank=True,
use_power=True,
window_type='povey',
dtype=np.float32):
""" Compute (log) Mel filter bank energies
:param waveform: Input waveform.
:param blackman_coeff: Constant coefficient for generalized Blackman window. (float, default = 0.42)
:param dither: Dithering constant (0.0 means no dither). If you turn this off, you should set the --energy-floor option, e.g. to 1.0 or 0.1 (float, default = 1)
:param energy_floor: Floor on energy (absolute, not relative) in FBANK computation. Only makes a difference if --use-energy=true; only necessary if --dither=0.0. Suggested values: 0.1 or 1.0 (float, default = 0)
:param frame_length: Frame length in milliseconds (float, default = 25)
:param frame_shift: Frame shift in milliseconds (float, default = 10)
:param high_freq: High cutoff frequency for mel bins (if <= 0, offset from Nyquist) (float, default = 0)
:param low_freq: Low cutoff frequency for mel bins (float, default = 20)
:param num_mel_bins: Number of triangular mel-frequency bins (int, default = 23)
:param preemphasis_coefficient: Coefficient for use in signal preemphasis (float, default = 0.97)
:param raw_energy: If true, compute energy before preemphasis and windowing (bool, default = true)
:param remove_dc_offset: Subtract mean from waveform on each frame (bool, default = true)
:param round_to_power_of_two: If true, round window size to power of two by zero-padding input to FFT. (bool, default = true)
:param sample_frequency: Waveform data sample frequency (must match the waveform file, if specified there) (float, default = 16000)
:param snip_edges: If true, end effects will be handled by outputting only frames that completely fit in the file, and the number of frames depends on the frame-length. If false, the number of frames depends only on the frame-shift, and we reflect the data at the ends. (bool, default = true)
:param use_energy: Add an extra energy output. (bool, default = false)
:param use_log_fbank: If true, produce log-filterbank, else produce linear. (bool, default = true)
:param use_power: If true, use power, else use magnitude. (bool, default = true)
:param window_type: Type of window ("hamming"|"hanning"|"povey"|"rectangular"|"sine"|"blackmann") (string, default = "povey")
:param dtype: Type of array (np.float32|np.float64) (dtype or string, default=np.float32)
:return: (Log) Mel filter bank energies.
"""
window_size = int(frame_length * sample_frequency * 0.001)
window_shift = int(frame_shift * sample_frequency * 0.001)
frames, log_energy = extract_window(
waveform=waveform,
blackman_coeff=blackman_coeff,
dither=dither,
window_size=window_size,
window_shift=window_shift,
preemphasis_coefficient=preemphasis_coefficient,
raw_energy=raw_energy,
remove_dc_offset=remove_dc_offset,
snip_edges=snip_edges,
window_type=window_type,
dtype=dtype
)
if round_to_power_of_two:
n = 1
while n < window_size:
n *= 2
else:
n = window_size
if use_power:
spectrum = compute_power_spectrum(frames, n)
else:
spectrum = compute_spectrum(frames, n)
mel_banks = compute_mel_banks(
num_bins=num_mel_bins,
sample_frequency=sample_frequency,
low_freq=low_freq,
high_freq=high_freq,
n=n
).astype(dtype)
feat = np.dot(spectrum, mel_banks.T)
if use_log_fbank:
feat = np.log(feat.clip(min=np.finfo(dtype).eps))
if use_energy:
if energy_floor > 0.0:
log_energy.clip(min=np.math.log(energy_floor))
return feat, log_energy
return feat
# ---------- compute-fbank-feats ----------
# ---------- compute-mfcc-feats ----------
def compute_mfcc_feats(
waveform,
blackman_coeff=0.42,
cepstral_lifter=22,
dither=1.0,
energy_floor=0.0,
frame_length=25,
frame_shift=10,
high_freq=0,
low_freq=20,
num_ceps=13,
num_mel_bins=23,
preemphasis_coefficient=0.97,
raw_energy=True,
remove_dc_offset=True,
round_to_power_of_two=True,
sample_frequency=16000,
snip_edges=True,
use_energy=True,
window_type='povey',
dtype=np.float32):
""" Compute mel-frequency cepstral coefficients
:param waveform: Input waveform.
:param blackman_coeff: Constant coefficient for generalized Blackman window. (float, default = 0.42)
:param cepstral_lifter: Constant that controls scaling of MFCCs (float, default = 22)
:param dither: Dithering constant (0.0 means no dither). If you turn this off, you should set the --energy-floor option, e.g. to 1.0 or 0.1 (float, default = 1)
:param energy_floor: Floor on energy (absolute, not relative) in MFCC computation. Only makes a difference if --use-energy=true; only necessary if --dither=0.0. Suggested values: 0.1 or 1.0 (float, default = 0)
:param frame_length: Frame length in milliseconds (float, default = 25)
:param frame_shift: Frame shift in milliseconds (float, default = 10)
:param high_freq: High cutoff frequency for mel bins (if <= 0, offset from Nyquist) (float, default = 0)
:param low_freq: Low cutoff frequency for mel bins (float, default = 20)
:param num_ceps: Number of cepstra in MFCC computation (including C0) (int, default = 13)
:param num_mel_bins: Number of triangular mel-frequency bins (int, default = 23)
:param preemphasis_coefficient: Coefficient for use in signal preemphasis (float, default = 0.97)
:param raw_energy: If true, compute energy before preemphasis and windowing (bool, default = true)
:param remove_dc_offset: Subtract mean from waveform on each frame (bool, default = true)
:param round_to_power_of_two: If true, round window size to power of two by zero-padding input to FFT. (bool, default = true)
:param sample_frequency: Waveform data sample frequency (must match the waveform file, if specified there) (float, default = 16000)
:param snip_edges: If true, end effects will be handled by outputting only frames that completely fit in the file, and the number of frames depends on the frame-length. If false, the number of frames depends only on the frame-shift, and we reflect the data at the ends. (bool, default = true)
:param use_energy: Use energy (not C0) in MFCC computation (bool, default = true)
:param window_type: Type of window ("hamming"|"hanning"|"povey"|"rectangular"|"sine"|"blackmann") (string, default = "povey")
:param dtype: Type of array (np.float32|np.float64) (dtype or string, default=np.float32)
:return: Mel-frequency cespstral coefficients.
"""
feat, log_energy = compute_fbank_feats(
waveform=waveform,
blackman_coeff=blackman_coeff,
dither=dither,
energy_floor=energy_floor,
frame_length=frame_length,
frame_shift=frame_shift,
high_freq=high_freq,
low_freq=low_freq,
num_mel_bins=num_mel_bins,
preemphasis_coefficient=preemphasis_coefficient,
raw_energy=raw_energy,
remove_dc_offset=remove_dc_offset,
round_to_power_of_two=round_to_power_of_two,
sample_frequency=sample_frequency,
snip_edges=snip_edges,
use_energy=use_energy,
use_log_fbank=True,
use_power=True,
window_type=window_type,
dtype=dtype
)
feat = dct(feat, type=2, axis=1, norm='ortho')[:, :num_ceps]
lifter_coeffs = compute_lifter_coeffs(cepstral_lifter, num_ceps).astype(dtype)
feat = feat * lifter_coeffs
if use_energy:
feat[:, 0] = log_energy
return feat
# ---------- compute-mfcc-feats ----------
# ---------- apply-cmvn-sliding ----------
def apply_cmvn_sliding(feat, center=False, window=600, min_window=100, norm_vars=False):
""" Apply sliding-window cepstral mean (and optionally variance) normalization
:param feat: Cepstrum.
:param center: If true, use a window centered on the current frame (to the extent possible, modulo end effects). If false, window is to the left. (bool, default = false)
:param window: Window in frames for running average CMN computation (int, default = 600)
:param min_window: Minimum CMN window used at start of decoding (adds latency only at start). Only applicable if center == false, ignored if center==true (int, default = 100)
:param norm_vars: If true, normalize variance to one. (bool, default = false)
:return: Normalized cepstrum.
"""
# double-precision
feat = apply_cmvn_sliding_internal(
feat=feat.astype(np.float64),
center=center,
window=window,
min_window=min_window,
norm_vars=norm_vars
).astype(feat.dtype)
return feat
# ---------- apply-cmvn-sliding ----------

43
src/asr/rapid_paraformer/kaldifeat/ivector.py Normal file
View File

@ -0,0 +1,43 @@
import numpy as np
from .feature import sliding_window
# ---------- compute-vad ----------
def compute_vad(log_energy, energy_mean_scale=0.5, energy_threshold=0.5, frames_context=0, proportion_threshold=0.6):
""" Apply voice activity detection
:param log_energy: Log mel energy.
:param energy_mean_scale: If this is set to s, to get the actual threshold we let m be the mean log-energy of the file, and use s*m + vad-energy-threshold (float, default = 0.5)
:param energy_threshold: Constant term in energy threshold for VAD (also see energy_mean_scale) (float, default = 5)
:param frames_context: Number of frames of context on each side of central frame, in window for which energy is monitored (int, default = 0)
:param proportion_threshold: Parameter controlling the proportion of frames within the window that need to have more energy than the threshold (float, default = 0.6)
:return: A vector of boolean that are True if we judge the frame voiced and False otherwise.
"""
assert len(log_energy.shape) == 1
assert energy_mean_scale >= 0
assert frames_context >= 0
assert 0 < proportion_threshold < 1
dtype = log_energy.dtype
energy_threshold += energy_mean_scale * log_energy.mean()
if frames_context > 0:
num_frames = len(log_energy)
window_size = frames_context * 2 + 1
log_energy_pad = np.concatenate([
np.zeros(frames_context, dtype=dtype),
log_energy,
np.zeros(frames_context, dtype=dtype)
])
log_energy_window = sliding_window(log_energy_pad, window_size, 1)
num_count = np.count_nonzero(log_energy_window > energy_threshold, axis=1)
den_count = np.ones(num_frames, dtype=dtype) * window_size
max_den_count = np.arange(frames_context + 1, min(window_size, num_frames) + 1, dtype=dtype)
den_count[:-(frames_context + 2):-1] = max_den_count
den_count[:frames_context + 1] = np.min([den_count[:frames_context + 1], max_den_count], axis=0)
vad = num_count / den_count >= proportion_threshold
else:
vad = log_energy > energy_threshold
return vad
# ---------- compute-vad ----------