def main():
from nara_wpe import project_root
import soundfile as sf
from nara_wpe.utils import stft
from nara_wpe.utils import istft as istft
from nara_wpe.utils import get_stft_center_frequencies
from tqdm import tqdm
from librosa.core.audio import resample
channels = 8
parameter_set = 'Katka'
if parameter_set == 'Katka':
sampling_rate = 16000
stft_size, stft_shift = 512, 128
delay = 3
iterations = 5
def get_K(f):
return 10
elif parameter_set == 'Yoshioka2012GeneralWPE':
sampling_rate = 8000
stft_size, stft_shift = 128, 64
delay = 2
iterations = 2
def get_K(f):
if center_frequencies[f] < 800:
K = 18
elif center_frequencies[f] < 1500:
K = 15
else:
K = 12
return K
else:
raise ValueError
file_template = 'AMI_WSJ20-Array1-{}_T10c0201.wav'
signal_list = [
sf.read(str(project_root / 'data' / file_template.format(d + 1)))[0]
for d in range(channels)
]
signal_list = [resample(x_, 16000, sampling_rate) for x_ in signal_list]
y = np.stack(signal_list, axis=0)
center_frequencies = get_stft_center_frequencies(stft_size, sampling_rate)
Y = stft(y, size=stft_size, shift=stft_shift)
X = np.copy(Y)
D, T, F = Y.shape
for f in tqdm(range(F), total=F):
K = get_K(f)
X[:, :, f] = wpe_v5(Y[:, :, f],
K=K,
delay=delay,
iterations=iterations)
x = istft(X, size=stft_size, shift=stft_shift)
sf.write(str(project_root / 'data' / 'wpe_out.wav'),
x[0],
samplerate=sampling_rate)
def main(channels, sampling_rate, file_template, taps_frequency_dependent,
delay, iterations):
"""
User interface for WPE. The defaults of the command line interface are
suited for example audio files of nara_wpe.
'Yoshioka2012GeneralWPE'
sampling_rate = 8000
delay = 2
iterations = 2
"""
from nara_wpe import project_root
import soundfile as sf
from nara_wpe.utils import stft
from nara_wpe.utils import istft
from nara_wpe.utils import get_stft_center_frequencies
from tqdm import tqdm
from librosa.core.audio import resample
stft_options = dict(size=512,
shift=128,
window_length=None,
fading=True,
pad=True,
symmetric_window=False)
def get_taps(f, mode=taps_frequency_dependent):
if mode:
if center_frequencies[f] < 800:
taps = 18
elif center_frequencies[f] < 1500:
taps = 15
else:
taps = 12
else:
taps = 10
return taps
if file_template == 'AMI_WSJ20-Array1-{}_T10c0201.wav':
signal_list = [
sf.read(str(project_root / 'data' /
file_template.format(d + 1)))[0]
for d in range(channels)
]
else:
signal = sf.read(file_template)[0].transpose(1, 0)
signal_list = list(signal)
signal_list = [resample(x_, 16000, sampling_rate) for x_ in signal_list]
y = np.stack(signal_list, axis=0)
center_frequencies = get_stft_center_frequencies(stft_options['size'],
sampling_rate)
Y = stft(y, **stft_options)
X = np.copy(Y)
D, T, F = Y.shape
for f in tqdm(range(F), total=F):
taps = get_taps(f)
X[:, :, f] = wpe_v7(Y[:, :, f],
taps=taps,
delay=delay,
iterations=iterations)
x = istft(X, size=stft_options['size'], shift=stft_options['shift'])
sf.write(str(project_root / 'data' / 'wpe_out.wav'),
x[0],
samplerate=sampling_rate)
print('Output in {}'.format(str(project_root / 'data' / 'wpe_out.wav')))