#!/usr/bin/env python2.6
from optparse import OptionParser
import sys
from htk2.units import HTK_transcription
usage = "usage: %prog [options] wordsmlf referencetrn"
parser = OptionParser(usage=usage)
parser.add_option('--num-speaker-chars', dest='numspeakerchars', type='int', default=3)
options, args = parser.parse_args()
if len(args) < 2:
sys.exit("Need at least to arguments")
mlf,trn = args[:2]
tr = HTK_transcription()
tr.read_mlf(mlf,HTK_transcription.WORD)
tr.write_trn(trn,options.numspeakerchars)
#!/usr/bin/env python2.6
from optparse import OptionParser
import sys
from htk2.units import HTK_transcription
usage = "usage: %prog [options] wordsmlf referencetrn"
parser = OptionParser(usage=usage)
parser.add_option('--num-speaker-chars',
dest='numspeakerchars',
type='int',
default=3)
options, args = parser.parse_args()
if len(args) < 2:
sys.exit("Need at least to arguments")
mlf, trn = args[:2]
tr = HTK_transcription()
tr.read_mlf(mlf, HTK_transcription.WORD)
tr.write_trn(trn, options.numspeakerchars)
def initialize_new(self, scp_list, word_mlf, dict, remove_previous=False):
System.set_log_dir(self.name)
if remove_previous:
for f in glob.iglob(System.get_log_dir() + '/*'):
os.remove(f)
if not remove_previous and (
os.path.exists(self.train_files_dir) or
len(glob.glob(self.model_dir + '/' + self.name + '.*')) > 0):
raise ExistingFilesException
if os.path.exists(self.train_files_dir):
shutil.rmtree(self.train_files_dir)
for f in glob.iglob(self.model_dir + '/' + self.name + '.*'):
os.remove(f)
os.mkdir(self.train_files_dir)
# handle dictionary
dic = HTK_dictionary()
if isinstance(dict, basestring):
dic.read_dict(dict)
elif all(isinstance(d, basestring) for d in dict):
for d in dict:
dic.read_dict(d)
else:
raise TypeError
dic.write_dict(self.training_dict)
self.phones = dic.get_phones()
# handle transcription
trans = HTK_transcription()
# if isinstance(word_mlf,basestring):
# trans.read_mlf(word_mlf, HTK_transcription.WORD)
# elif all(isinstance(w,basestring) for w in word_mlf):
# for w in word_mlf:
# trans.read_mlf(w, HTK_transcription.WORD)
# else:
# raise TypeError
word_mlf = word_mlf.strip().split(',')
for w in word_mlf:
trans.read_mlf(w, HTK_transcription.WORD)
self.id = 1
phones_list = self._get_model_name_id() + '.hmmlist'
with open(phones_list, 'w') as phones_desc:
for p in self.phones:
print(p, file=phones_desc)
# handle scp files
scp_list = scp_list.strip().split(',')
# if isinstance(scp_list,basestring):
# scp_list = [scp_list]
real_trans = HTK_transcription()
real_trans.transcriptions[real_trans.WORD] = {}
with open(self.training_scp, 'w') as scp_desc:
for scp in scp_list:
for file in open(scp):
id = os.path.splitext(os.path.basename(file.strip()))[0]
if not file.startswith('/'):
file = os.path.join(os.path.dirname(scp), file.strip())
ok = True
for word in trans.transcriptions[
HTK_transcription.WORD][id]:
if not dic.word_in_dict(word):
print("%s skipped, because has missing word %s" %
(file.strip(), word))
ok = False
break
if ok:
print(file.strip(), file=scp_desc)
real_trans.transcriptions[real_trans.WORD][
id] = trans.transcriptions[real_trans.WORD][id]
real_trans.write_mlf(self.training_word_mlf,
target=HTK_transcription.WORD)
self.expand_word_transcription()
sp = parts[0]
neighbors = [n for n in parts[2:]]
if sp in neighbors:
neighbors.remove(sp)
neighbour_dict[sp] = neighbors[:options.num_neighbours]
transform_files = {}
for line in open(t_scp):
sp = basename(line.strip())[:3]
if sp not in transform_files:
transform_files[sp] = []
transform_files[sp].append(line.strip())
mlf = HTK_transcription()
mlf.read_mlf(t_mlf, target=HTK_transcription.WORD)
#trans_mlf = HTK_transcription()
with open(transform_scp, 'w') as transform_desc:
for sp in neighbour_dict.keys():
neighbors = neighbour_dict[sp]
t_files = []
for n in neighbors:
t_files.extend(transform_files[n])
shuffle(t_files)
# if options.num_adaptation_files > 0:
# t_files = t_files[:options.num_adaptation_files]
sp = parts[0]
neighbors = [n for n in parts[2:]]
if sp in neighbors:
neighbors.remove(sp)
neighbour_dict[sp] = neighbors[:options.num_neighbours]
transform_files = {}
for line in open(t_scp):
sp = basename(line.strip())[:3]
if sp not in transform_files:
transform_files[sp] = []
transform_files[sp].append(line.strip())
mlf = HTK_transcription()
mlf.read_mlf(t_mlf,target=HTK_transcription.WORD)
#trans_mlf = HTK_transcription()
with open(transform_scp, 'w') as transform_desc:
for sp in neighbour_dict.keys():
neighbors = neighbour_dict[sp]
t_files = []
for n in neighbors:
t_files.extend(transform_files[n])
shuffle(t_files)
# if options.num_adaptation_files > 0:
# t_files = t_files[:options.num_adaptation_files]