This repository has been archived by the owner on Apr 18, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 127
/
mtts.py
271 lines (230 loc) · 9.81 KB
/
mtts.py
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
import os
import re
import logging
from pypinyin import pinyin, Style, load_phrases_dict
import textgrid as tg
from mandarin_frontend import txt2label
consonant = [
'b', 'p', 'm', 'f', 'd', 't', 'n', 'l', 'g', 'k', 'h', 'j', 'q', 'x', 'zh',
'ch', 'sh', 'r', 'z', 'c', 's', 'y', 'w'
]
def _pre_pinyin_setting():
''' fix pinyin error'''
load_phrases_dict({'嗯': [['ēn']]})
load_phrases_dict({'风云变幻': [['fēng'], ['yún'], ['bià'], ['huàn']]})
load_phrases_dict({'不破不立': [['bù'], ['pò'], ['bù'], ['lì']]})
def _add_lab(txtlines, wav_dir_path):
logger = logging.getLogger('mtts')
for line in txtlines:
numstr, txt = line.split(' ')
txt = re.sub('#\d', '', txt)
pinyin_list = pinyin(txt, style=Style.TONE3)
new_pinyin_list = []
for item in pinyin_list:
if not item:
logger.warning(
'{file_num} do not generate right pinyin'.format(numstr))
if not item[0][-1].isdigit():
phone = item[0] + '5'
else:
phone = item[0]
new_pinyin_list.append(phone)
lab_file = os.path.join(wav_dir_path, numstr + '.lab')
with open(lab_file, 'w') as oid:
oid.write(' '.join(new_pinyin_list))
def _add_pinyin(txtlines, output_path):
''' txt2pinyin in one file '''
logger = logging.getLogger('mtts')
all_pinyin = []
for line in txtlines:
numstr, txt = line.split(' ')
txt = re.sub('#\d', '', txt)
pinyin_list = pinyin(txt, style=Style.TONE3)
new_pinyin_list = []
for item in pinyin_list:
if not item:
logger.warning(
'{file_num} do not generate right pinyin'.format(numstr))
if not item[0][-1].isdigit():
phone = item[0] + '5'
else:
#phone = item[0]
phone = item[0].replace('v', 'u')
new_pinyin_list.append(phone)
all_pinyin.append(numstr + ' ' + ' '.join(new_pinyin_list))
all_pinyin_file = os.path.join(output_path, 'all_pinyin.lab')
with open(all_pinyin_file, 'w') as oid:
for item in all_pinyin:
oid.write(item + '\n')
def _txt_preprocess(txtfile, output_path):
# 去除所有标点符号(除非是韵律标注#1符号),报错,如果txt中含有数字和字母(报错并跳过)
logger = logging.getLogger('mtts')
with open(txtfile) as fid:
txtlines = [x.strip() for x in fid.readlines()]
valid_txtlines = []
error_list = [] # line which contain number or alphabet
pattern = re.compile('(?!#(?=\d))[\W]')
for line in txtlines:
num, txt = line.split(' ', 1)
if bool(re.search('[A-Za-z]', txt)) or bool(
re.search('(?<!#)\d', txt)):
error_list.append(num)
else:
txt = re.sub('[,.,。]', '#4', txt)
txt = pattern.sub('', txt)
# 去除除了韵律标注'#'之外的所有非中文文本, 数字, 英文字符符号
if txt:
valid_txtlines.append(num + ' ' + txt)
else:
logger.warning('txt error, check your txt %ts' % num)
if error_list:
for item in error_list:
logger.warning(
'Line %s contain number and alphabet! Skip this txt' % item)
return valid_txtlines
def _standard_sfs(csv_list):
'''Change csv_list like "0 0.21 sil phones" to standard format like "2100000 s" '''
def change2absd(phone, csv_list):
if phone in consonant:
return 'a'
elif phone == 'sil' or phone == 'sp':
if float(csv_list[1]) - float(csv_list[0]) > 0.1:
return 's'
else:
#return 'd'
return 's'
else: #phone is vowel
return 'b'
standard_sfs_list = list((str(int(float(csv_list[1]) * 10e6)),
change2absd(csv_list[2], csv_list)))
return standard_sfs_list
def _mfa_align(txtlines, wav_dir_path, output_path, acoustic_model_path):
logger = logging.getLogger('mtts')
logger.info('Start montreal forced align')
base_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
os.makedirs('%s/wav' % output_path, exist_ok=True)
wav_dir_real_path = os.path.realpath(wav_dir_path)
symbolic_path = '%s/wav/mandarin_voice' % output_path
if not os.path.exists(symbolic_path):
os.system('ln -s %s %s' % (wav_dir_real_path, symbolic_path))
mfa_align_path = os.path.join(
base_dir, 'tools/montreal-forced-aligner/bin/mfa_align')
lexicon_path = os.path.join(base_dir, 'misc/mandarin_mtts.lexicon')
acoustic_model_path = os.path.join(base_dir, acoustic_model_path)
exec_result = os.system('%s %s/wav %s %s %s/textgrid' %
(mfa_align_path, output_path, lexicon_path, acoustic_model_path,
output_path))
if exec_result:
raise OSError('Failed to run forced align tools, check if you install'
'montreal-forced-aligner correctly')
def _textgrid2sfs(txtlines, output_path):
logger = logging.getLogger('mtts')
textgrid_path = os.path.join(output_path, 'textgrid/mandarin_voice')
sfs_path = os.path.join(output_path, 'sfs')
csv_path = os.path.join(output_path, 'csv')
os.system('mkdir -p %s' % sfs_path)
os.system('mkdir -p %s' % csv_path)
for line in txtlines:
numstr, txt = line.split(' ', 1)
textgrid_file = os.path.join(textgrid_path, numstr + '.TextGrid')
csv_file = os.path.join(csv_path, numstr + '.csv')
sfs_file = os.path.join(sfs_path, numstr + '.sfs')
if os.path.exists(textgrid_file):
# textgrid to csv
tgrid = tg.read_textgrid(textgrid_file)
tg.write_csv(tgrid, csv_file, sep=' ', header=False, meta=False)
# csv to sfs
total_list = []
with open(csv_file) as fid:
for line in fid.readlines():
#start, end, name, label = line.strip().split(' ')
csv_list = line.strip().split(' ')
if csv_list[3] == 'phones':
total_list.append(_standard_sfs(csv_list))
with open(sfs_file, 'w') as fid:
for item in total_list:
fid.write(' '.join(item) + '\n')
else:
logger.warning('--Miss: %s' % textgrid_file)
with open(os.path.join(output_path, 'error.log'), 'a+') as fid:
fid.write('--Miss: %s \n' % textgrid_file)
def _sfs2label(txtlines, output_path):
logger = logging.getLogger('mtts')
sfs_path = os.path.join(output_path, 'sfs')
label_path = os.path.join(output_path, 'labels')
os.system('mkdir -p %s/labels' % output_path)
sfs_list = [x.replace('.sfs', '') for x in os.listdir(sfs_path)]
process_num = 0
for line in txtlines:
numstr, txt = line.split()
if numstr in sfs_list:
process_num += 1
logger.info('processing %s, file %s' % (process_num, numstr))
sfs_file = os.path.join(sfs_path, numstr + '.sfs')
label_file = os.path.join(label_path, numstr + '.lab')
try:
label_line = txt2label(txt, sfsfile=sfs_file)
except Exception:
logger.error(
'Error at %s, please check your txt %s' % (numstr, txt))
else:
with open(label_file, 'w') as oid:
for item in label_line:
oid.write(item + '\n')
def _set_logger(output_path):
formater = logging.Formatter('%(levelname)-8s: %(message)s')
logger = logging.getLogger('mtts')
console_handler = logging.StreamHandler()
console_handler.setFormatter(
logging.Formatter('%(levelname) -8s: %(message)s'))
console_handler.setLevel(logging.INFO)
log_path = os.path.join(output_path, 'mtts.log')
file_handler = logging.FileHandler(log_path, mode='w')
file_handler.setFormatter(
logging.Formatter('%(asctime)s %(levelname)-8s: %(message)s'))
file_handler.setLevel(logging.WARNING)
logger.addHandler(file_handler)
logger.addHandler(console_handler)
logger.setLevel(logging.DEBUG)
def generate_label(txtfile, wav_dir_path, output_path, acoustic_model_path):
_set_logger(output_path)
txtlines = _txt_preprocess(txtfile, output_path)
_pre_pinyin_setting()
_add_lab(txtlines, wav_dir_path)
#_add_pinyin(txtlines, output_path)
_mfa_align(txtlines, wav_dir_path, output_path, acoustic_model_path)
_textgrid2sfs(txtlines, output_path)
_sfs2label(txtlines, output_path)
logger = logging.getLogger('mtts')
logger.info('the label files are in {}/labels'.format(output_path))
logger.info('the error log is in {}/mtts.log'.format(output_path))
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser(
description="convert mandarin_txt and wav to label for merlin.")
parser.add_argument(
"txtfile",
help=
"Full path to txtfile which each line contain num and txt (seperated by a white space) "
)
parser.add_argument(
"wav_dir_path",
help=
"Full path to a wav directory contain wav (sampling frequency should larger than 16khz)"
)
parser.add_argument(
"output_path",
help=
"Full path to output directory, will be created if it doesn't exist")
parser.add_argument(
'-a',
'--acoustic_model_path',
type=str,
default='misc/thchs30.zip',
help=
'Full path to acoustic model for forced aligner, default is misc/thchs30.zip'
)
args = parser.parse_args()
os.system('mkdir -p %s' % args.output_path)
generate_label(args.txtfile, args.wav_dir_path, args.output_path,
args.acoustic_model_path)