-
Notifications
You must be signed in to change notification settings - Fork 5
/
align.py
247 lines (217 loc) · 8.8 KB
/
align.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
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import codecs
import itertools
import logging
import os
import phonetizer as ph
import subprocess
import sys
def force(phonetizer, utterance, starttime, endtime, wavefile,
soxbinary, hvitebinary, hcopybinary, parameterdir,
basename, hdr=True, code='w'):
"""Wrapper for the _force function that writes the status to a file for
feedback via praat module.
statusses can be:
done - Alignment was successfull.
missox - Sox binary not found.
mishcopy - HCopy binary not found.
mishvite - HVite binary not found.
"""
status = _force(
phonetizer, utterance, starttime, endtime, wavefile,
soxbinary, hvitebinary, hcopybinary, parameterdir,
basename, hdr, code)
with open('{}.status'.format(basename), 'w') as f:
f.write(status)
return not status
def _force(phonetizer, utterance, starttime, duration, wavefile,
soxbinary, hvitebinary, hcopybinary, parameterdir,
basename, hdr, code):
"""
Force aligns the given utterance, all parameters are passed by kwarg
"""
logging.info('Starting to align: {}'.format(code))
logging.info('Removing old files')
for suffix in ['dot', 'htk', 'nis', 'rec', 'slf', 'status']:
try:
os.remove('{}.{}'.format(basename, suffix))
except:
pass
# Open the preconfig and extract the sourcerate
sourcerate = str(1e7/625)
# Load the phonetizer
pron = phonetizer.phonetize(utterance) or [[['<nib>']]]
canonical = [''.join(p[0]) for p in pron]
logging.info('Utterance phonetized spawned')
# Create the graph file
dawg = phonetizer.todawg(pron)
with open('{}.slf'.format(basename), 'w') as f:
f.write(phonetizer.toslf(*dawg))
with open('{}.dot'.format(basename), 'w') as f:
f.write(phonetizer.todot(*dawg))
logging.info('SLF file created')
# Run the sound processing
soxcommand = [
soxbinary, wavefile,
'-t', 'sph',
'-e', 'signed-integer',
'-b', '16', '-c', '1',
'{}.nis'.format(basename),
'trim', starttime, duration,
'rate', '-s', '-a', sourcerate]
logging.info(' '.join(soxcommand))
proc = subprocess.Popen(soxcommand,
stderr=subprocess.PIPE, stdout=subprocess.PIPE)
out, err = proc.communicate()
if proc.returncode == 127:
return 'Sox binary couldn\'t be found...\n'\
'It is searched for in {}'.format(hcopybinary)
logging.info('Sox ran({}):\n\tout: {}\n\terr: {}'.format(
proc.returncode, out, err))
# Run the HCopy process
hcopycommand = [
hcopybinary,
'-T', '0',
'-C', os.path.join(parameterdir, 'PRECONFIGNIST'),
'{}.nis'.format(basename), '{}.htk'.format(basename)]
logging.info(' '.join(hcopycommand))
proc = subprocess.Popen(hcopycommand,
stderr=subprocess.PIPE, stdout=subprocess.PIPE)
out, err = proc.communicate()
if proc.returncode == 127:
return 'HCopy binary couldn\'t be found...\n'\
'It is searched for in {}'.format(hcopybinary)
logging.info('HCopy ran({}):\n\tout: {}\n\terr: {}'.format(
proc.returncode, out, err))
# Run the HVite actual alignment
hvitecommand = [
hvitebinary,
'-C', os.path.join(parameterdir, 'HVITECONF'),
'-w', '-X', 'slf',
'-H', os.path.join(parameterdir, 'MMF'),
'-s', '7.0',
'-p', '6.0',
os.path.join(parameterdir, 'DICT'),
os.path.join(parameterdir, 'MONOPHONES'),
'{}.htk'.format(basename)]
logging.info(' '.join(hvitecommand))
proc = subprocess.Popen(hvitecommand,
stderr=subprocess.PIPE, stdout=subprocess.PIPE)
out, err = proc.communicate()
if proc.returncode == 127:
return 'HVite binary couldn\'t be found...\n'\
'It is searched for in {}'.format(hvitebinary)
logging.info('HVite ran({}):\n\tout: {}\n\terr: {}'.format(
proc.returncode, out, err))
if proc.returncode != 0:
return 'HVite failed with the following error: {}'.format(err)
# Open the output file
out = 'praat_temp_out'
ortwords = utterance.replace(",", "").split(' ')
with open(out, code) as fileio:
logging.info('Output file selected')
with open('{}.rec'.format(basename), 'r') as f:
# Write if necessary the header
if hdr != 'False':
fileio.write('start,end,label,type\n')
logging.info('Header written')
word = None
for d in itertools.imap(lambda x: x.split(), f):
# Parse the time parameters and convert them to seconds
start = float(starttime) + int(d[0]) / 1e7
end = float(starttime) + int(d[1]) / 1e7
# Detect word boundaries
if d[2] == '<':
word = (end, '')
# If the end of a word is reached write the word to file
elif d[2] == '#':
fileio.write('{:f},{:f},{},w\n'.format(
word[0], start, word[1]))
fileio.write('{:f},{:f},{},c\n'.format(
word[0], start, canonical.pop(0)))
fileio.write('{:f},{:f},{},o\n'.format(
word[0], start, ortwords.pop(0).encode('utf-8')))
word = (end, '')
# Else add the current phone to the current word
else:
word = (word[0], word[1] + d[2])
# If the length is non zero write the phone to the file
if end - start > 0:
fileio.write('{:f},{:f},{},p\n'.format(
start, end, d[2]))
fileio.write('{:f},{:f},{},l\n'.format(
start, end, float(d[3])))
logging.info('Datafile written')
logging.info('Finished')
return ''
def parsesettings(filepath):
with open(filepath, 'r') as f:
settings = {k: v.strip() for k, v in (x.split(': ') for x in f)}
return settings
if __name__ == '__main__':
# Load the sett
sett = parsesettings('isettings')
sett.update(parsesettings('settings'))
# Initialize the logger
logging.basicConfig(filename=sett['LOG'], level=20,
format='%(created)f: %(message)s')
# Load the phonetizer
try:
phone = ph.getphonetizer(
sett['LAN'], sett['PHO'], sett['DCT'], sett['RUL'])
except UnicodeError:
with open('temp.status', 'w') as f:
f.write('Unicode error. Check if the files are utf-8')
exit()
except IOError as e:
if e.filename == sett['DCT']:
error = 'Dictionary file couldn\'t be found\n'\
'It is searched for in {}'.format(e.filename)
elif e.filename == sett['RUL']:
error = 'Ruleset file couldn\'t be found\n'\
'It is searched for in {}'.format(e.filename)
elif e.filename == sett['PHO']:
error = 'Universal phonetizer file couldn\'t be found\n'\
'It is searched for in {}'.format(e.filename)
else:
error = 'Some io error: ' + str(e)
with open('temp.status', 'w') as f:
f.write(error)
exit()
p = 'par.{}'.format(sett['MOD'])
if sys.argv[1] == 'tier':
# Read the data
with codecs.open(sett['OUT'], 'r', 'utf-8') as f:
data = f.readlines()
# Setup the code for writing and header sett
first = 0
code = 'w'
hdr = 'True'
# Parse the data
data = map(lambda x: x.strip().split('\t'), data[1:])
# Setup and align all data
for i, (start, _, utt, end) in enumerate(data):
if first == 0:
first += 1
# If neccesary unset the header and change write mode to append
elif first == 1:
hdr = 'False'
code = 'a'
# Check for empty pre and post annotations for extended bounds option
if i > 0:
start = max(float(start)-float(sett['THR']),
float(data[i-1][3]))
if i < len(data) - 1:
end = min(float(end) + float(sett['THR']),
float(data[i+1][0]))
dur = str(float(end)-float(start))
# Do the actual alignment
if not force(
phone, utt, str(start), dur, sett['WAV'],
sett['SOX'], sett['HVB'], sett['HCB'], p, 'temp',
hdr=hdr, code=code):
break
elif sys.argv[1] == 'annotation':
force(phone, sett['UTT'], sett['STA'], sett['DUR'], sett['WAV'],
sett['SOX'], sett['HVB'], sett['HCB'], p, 'temp')