def main():
args = parser.parse_args()
htk_save_path = mkdir(args.htk_save_path)
path = Path(data_path=args.data_path, config_path='./config')
# HTK settings
save_config(audio_file_type='wav',
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))
data_types = ['eval1', 'eval2', 'eval3']
if bool(args.subset):
data_types += ['train_subset']
if bool(args.fullset):
data_types += ['train_fullset']
for data_type in data_types:
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 = basename(wav_path).split('.')[0]
save_path_tmp = join(save_path, speaker + '.htk')
f.write(wav_path + ' ' + save_path_tmp + '\n')
from utils.measure_time_func import measure_time
path = Path(data_path='/n/sd8/inaguma/corpus/csj/data',
config_path='../config',
htk_save_path='/n/sd8/inaguma/corpus/csj/htk')
htk_paths = {
'train': path.htk(data_type='train_subset'),
'dev': path.htk(data_type='dev'),
'eval1': path.htk(data_type='eval1'),
'eval2': path.htk(data_type='eval2'),
'eval3': path.htk(data_type='eval3'),
}
wav_paths = {
'train': path.wav(data_type='train_subset'),
'dev': path.wav(data_type='dev'),
'eval1': path.wav(data_type='eval1'),
'eval2': path.wav(data_type='eval2'),
'eval3': path.wav(data_type='eval3'),
}
label_paths = {
'train': path.trans(data_type='train_subset'),
'dev': path.trans(data_type='dev'),
'eval1': path.trans(data_type='eval1'),
'eval2': path.trans(data_type='eval2'),
'eval3': path.trans(data_type='eval3'),
}
CONFIG = {