def main():
args = parser.parse_args()
htk_save_path = mkdir(args.htk_save_path)
path = Path(data_path=args.data_path,
config_path=args.config_path)
# HTK settings
save_config(audio_file_type='nist',
feature_type=args.feature_type,
channels=args.channels,
config_save_path='./config',
sampling_rate=16000,
window=args.window,
slide=args.slide,
energy=bool(args.energy),
delta=bool(args.delta),
deltadelta=bool(args.deltadelta))
# NOTE: 123-dim features are extracted by default
for data_type in ['train', 'dev', 'test']:
wav_paths = path.wav(data_type=data_type)
save_path = mkdir_join(htk_save_path, data_type)
with open('./config/wav2htk_' + data_type + '.scp', 'w') as f:
for wav_path in wav_paths:
speaker = wav_path.split('/')[-2]
utt_index = basename(wav_path).split('.')[0]
save_path_tmp = mkdir_join(
save_path, speaker, utt_index + '.htk')
f.write(wav_path + ' ' + save_path_tmp + '\n')
"""Test for input data (TIMIT corpus)."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import sys
import unittest
sys.path.append('../../')
from timit.path import Path
from timit.input_data import read_audio
from utils.measure_time_func import measure_time
path = Path(data_path='/n/sd8/inaguma/corpus/timit/data',
config_path='../config',
htk_save_path='/n/sd8/inaguma/corpus/timit/htk')
htk_paths = {
'train': path.htk(data_type='train'),
'dev': path.htk(data_type='dev'),
'test': path.htk(data_type='test')
}
wav_paths = {
'train': path.wav(data_type='train'),
'dev': path.wav(data_type='dev'),
'test': path.wav(data_type='test')
}
CONFIG = {
parser.add_argument('--save_format', type=str, choices=['numpy', 'htk', 'wav'])
parser.add_argument('--feature_type', type=str, choices=['fbank', 'mfcc'])
parser.add_argument('--channels', type=int,
help='the number of frequency channels')
parser.add_argument('--window', type=float,
help='window width to extract features')
parser.add_argument('--slide', type=float, help='extract features per slide')
parser.add_argument('--energy', type=int, help='if 1, add the energy feature')
parser.add_argument('--delta', type=int, help='if 1, add the energy feature')
parser.add_argument('--deltadelta', type=int,
help='if 1, double delta features are also extracted')
args = parser.parse_args()
path = Path(data_path=args.data_path,
config_path=args.config_path,
htk_save_path=args.htk_save_path)
CONFIG = {
'feature_type': args.feature_type,
'channels': args.channels,
'sampling_rate': 16000,
'window': args.window,
'slide': args.slide,
'energy': bool(args.energy),
'delta': bool(args.delta),
'deltadelta': bool(args.deltadelta)
}
if args.save_format == 'htk':
assert args.tool == 'htk'
"""Test for phone-level transcript (TIMIT corpus)."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import sys
import unittest
sys.path.append('../../')
from timit.path import Path
from timit.transcript_phone import read_phone
from utils.measure_time_func import measure_time
from utils.util import mkdir_join
path = Path(data_path='/n/sd8/inaguma/corpus/timit/data',
config_path='../config')
label_paths = {
'train': path.phone(data_type='train'),
'dev': path.phone(data_type='dev'),
'test': path.phone(data_type='test')
}
class TestLabelPhone(unittest.TestCase):
def test(self):
self.check()
@measure_time
def check(self):