fame_phonetics.py and functions to make quests.hed to tie triphone are added.

acoustic_model/acoustic_model.pyproj

@@ -51,6 +51,9 @@
     <Compile Include="fame_hmm.py" />
     <Compile Include="phoneset\fame_asr.py" />
     <Compile Include="phoneset\fame_ipa.py" />
+    <Compile Include="phoneset\fame_phonetics.py">
+      <SubType>Code</SubType>
+    </Compile>
     <Compile Include="stimmen_functions.py" />
     <Compile Include="stimmen_test.py" />
   </ItemGroup>

acoustic_model/fame_functions.py

@@ -370,7 +370,8 @@ def ipa2asr(ipa):
 def ipa2htk(ipa):
 	curr_dir = os.path.dirname(os.path.abspath(__file__))
 	translation_key_ipa2asr = np.load(os.path.join(curr_dir, 'phoneset', 'fame_ipa2asr.npy')).item(0)
-	
+	#translation_key_ipa2asr = np.load(r'c:\Users\Aki\source\repos\acoustic_model\acoustic_model\phoneset\fame_ipa2asr.npy').item(0)
+
 	ipa_splitted = convert_phoneset.split_word(ipa, fame_ipa.multi_character_phones)
 	ipa_splitted = fame_ipa.phone_reduction(ipa_splitted)
 	asr_splitted = convert_phoneset.convert_phoneset(ipa_splitted, translation_key_ipa2asr)

acoustic_model/fame_hmm.py

@@ -11,7 +11,7 @@ import numpy as np
 import pandas as pd
 
 import fame_functions
-from phoneset import fame_ipa, fame_asr
+from phoneset import fame_ipa, fame_asr, fame_phonetics
 import defaultfiles as default
 sys.path.append(default.toolbox_dir)
 import file_handling as fh
@@ -44,6 +44,9 @@ lexicon_asr = os.path.join(default.fame_dir, 'lexicon', 'lex.asr')
 lexicon_oov = os.path.join(default.fame_dir, 'lexicon', 'lex.oov')
 
 config_dir = os.path.join(default.htk_dir, 'config')
+phonelist_full_txt = os.path.join(config_dir, 'phonelist_full.txt')
+tree_hed   = os.path.join(config_dir, 'tree.hed')
+quest_hed  = os.path.join(config_dir, 'quests.hed')
 
 model_dir     = os.path.join(default.htk_dir, 'model')
 model_mono0_dir    = os.path.join(model_dir, 'mono0')
@@ -57,7 +60,7 @@ lexicon_dir = os.path.join(default.htk_dir, 'lexicon')
 lexicon_htk_asr = os.path.join(lexicon_dir, 'lex.htk_asr')
 lexicon_htk_oov = os.path.join(lexicon_dir, 'lex.htk_oov')
 lexicon_htk     = os.path.join(lexicon_dir, 'lex.htk')
-#lexicon_htk_with_sp = os.path.join(lexicon_dir, 'lex_with_sp.htk')
+lexicon_htk_triphone = os.path.join(lexicon_dir, 'lex_triphone.htk')
 
 feature_dir = os.path.join(default.htk_dir, 'mfc')
 fh.make_new_directory(feature_dir, existing_dir='leave')
@@ -270,7 +273,7 @@ if train_monophone_without_sp:
 		'mfc', 
 		os.path.join(htk_stimmen_dir, 'word_lattice.ltc'), 
 		mlf_file=mlf_file_train, 
-		lexicon_file=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic')
+		lexicon=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic')
 		)
 
 	print("elapsed time: {}".format(time.time() - timer_start))
@@ -290,27 +293,27 @@ if add_sp:
 	modeln_dir_pre = os.path.join(model_mono1_dir, 'iter'+str(niter))
 	modeln_dir	   = os.path.join(model_mono1sp_dir, 'iter0')
 
-	#hmmdefs_pre = os.path.join(modeln_dir_pre, 'hmmdefs')
 	chtk.add_sp(modeln_dir_pre, modeln_dir)
-	print("elapsed time: {}".format(time.time() - timer_start))
-
+	
+	print('>>> re-estimation...')
 	niter = chtk.re_estimation_until_saturated(
 		model_mono1sp_dir, modeln_dir, improvement_threshold, hcompv_scp_train, 
 		os.path.join(htk_stimmen_dir, 'mfc'), 
 		'mfc', 
 		os.path.join(htk_stimmen_dir, 'word_lattice.ltc'), 
 		mlf_file=mlf_file_train_with_sp, 
-		lexicon_file=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
+		lexicon=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
 		model_type='monophone_with_sp'
 		)
-		
+	print("elapsed time: {}".format(time.time() - timer_start))
+	
 
 ## ======================= train model with re-aligned mlf =======================
 if train_monophone_with_re_aligned_mlf:
 	print('==== traina monophone with re-aligned mlf ====')
+	timer_start = time.time()
 
 	print('>>> re-aligning the training data... ')
-	timer_start = time.time()
 	niter = chtk.get_niter_max(model_mono1sp_dir)
 	modeln_dir = os.path.join(model_mono1sp_dir, 'iter'+str(niter))
 	chtk.make_aligned_label(
@@ -326,7 +329,6 @@ if train_monophone_with_re_aligned_mlf:
 		mlf_file_train_with_sp, 
 		hcompv_scp_train, 
 		hcompv_scp_train_updated)
-	print("elapsed time: {}".format(time.time() - timer_start))
 
 	print('>>> re-estimation... ')
 	timer_start = time.time()
@@ -341,7 +343,7 @@ if train_monophone_with_re_aligned_mlf:
 		'mfc', 
 		os.path.join(htk_stimmen_dir, 'word_lattice.ltc'), 
 		mlf_file=mlf_file_train_aligned, 
-		lexicon_file=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
+		lexicon=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
 		model_type='monophone_with_sp'
 		)
 	print("elapsed time: {}".format(time.time() - timer_start))
@@ -350,7 +352,7 @@ if train_monophone_with_re_aligned_mlf:
 ## ======================= train triphone =======================
 if train_triphone:
 	print('==== traina triphone model ====')
-	#model_out_dir = os.path.join(model_dir, 'hmm1_tri', 'iter1')
+	timer_start = time.time()
 
 	triphonelist_txt = os.path.join(config_dir, 'triphonelist.txt')
 	triphone_mlf = os.path.join(default.htk_dir, 'label', 'train_triphone.mlf')
@@ -385,7 +387,7 @@ if train_triphone:
 	#	'mfc', 
 	#	os.path.join(htk_stimmen_dir, 'word_lattice.ltc'), 
 	#	mlf_file=triphone_mlf, 
-	#	lexicon_file=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
+	#	lexicon=os.path.join(htk_stimmen_dir, 'lexicon_recognition.dic'), 
 	#	model_type='triphone'
 	#	)
 	#
@@ -409,8 +411,21 @@ if train_triphone:
 			macros=os.path.join(_modeln_dir_pre, 'macros'),
 			model_type='triphone')
 
+	print("elapsed time: {}".format(time.time() - timer_start))
+
 
 ## ======================= train triphone =======================
 if train_triphone_tied:
 	print('==== traina tied-state triphone ====')
-	
+	timer_start = time.time()
+
+	print('>>> making lexicon for triphone... ')
+	chtk.make_triphone_full(phonelist_full_txt, lexicon_htk_triphone)
+
+	print('>>> making headers... ')
+	chtk.make_tree_header(tree_hed)
+	fame_phonetics.make_quests_hed(quest_hed)
+
+	print("elapsed time: {}".format(time.time() - timer_start))
+
+

acoustic_model/fame_test.py

@@ -109,30 +109,30 @@ np.save(os.path.join('phoneset', 'fame_ipa2asr.npy'), translation_key_ipa2asr)
 
 
 ## check which letters are not coded in ascii. 
-print('asr phones which cannot be coded in ascii:\n')
-for i in fame_asr.phoneset_short:
-	try:
-		i_encoded = i.encode("ascii")
-		#print("{0} --> {1}".format(i, i.encode("ascii")))
-	except UnicodeEncodeError:
-		print(">>> {}".format(i))
-
-print("letters in the scripts which is not coded in ascii:\n")
-for dataset in ['train', 'devel', 'test']:
-	timer_start = time.time()
-
-	script_list = os.path.join(default.fame_dir, 'data', dataset, 'text')
-	with open(script_list, "rt", encoding="utf-8") as fin:
-		scripts = fin.read().split('\n')
-
-	for line in scripts:
-		sentence  = ' '.join(line.split(' ')[1:])
-		sentence_htk = fame_functions.word2htk(sentence)
-
-		#if len(re.findall(r'[âêôûč\'àéèúćäëïöü]', sentence))==0:
-		try:
-			sentence_htk = bytes(sentence_htk, 'ascii')
-		except UnicodeEncodeError:
-			print(sentence)
-			print(sentence_htk)
+#print('asr phones which cannot be coded in ascii:\n')
+#for i in fame_asr.phoneset_short:
+#	try:
+#		i_encoded = i.encode("ascii")
+#		#print("{0} --> {1}".format(i, i.encode("ascii")))
+#	except UnicodeEncodeError:
+#		print(">>> {}".format(i))
+
+#print("letters in the scripts which is not coded in ascii:\n")
+#for dataset in ['train', 'devel', 'test']:
+#	timer_start = time.time()
+
+#	script_list = os.path.join(default.fame_dir, 'data', dataset, 'text')
+#	with open(script_list, "rt", encoding="utf-8") as fin:
+#		scripts = fin.read().split('\n')
+
+#	for line in scripts:
+#		sentence  = ' '.join(line.split(' ')[1:])
+#		sentence_htk = fame_functions.word2htk(sentence)
+
+#		#if len(re.findall(r'[âêôûč\'àéèúćäëïöü]', sentence))==0:
+#		try:
+#			sentence_htk = bytes(sentence_htk, 'ascii')
+#		except UnicodeEncodeError:
+#			print(sentence)
+#			print(sentence_htk)
 

acoustic_model/phoneset/fame_asr.py

@@ -80,8 +80,11 @@ def phone_reduction(phones):
 	Args:
 		phones (list): list of phones.
 	"""
+	if sum([phone in phones for phone in phones_to_be_removed]) != 0:
+		print('input includes phone(s) which is not defined in fame_asr.')
+		print('those phone(s) are removed.')		
 	return [reduction_key.get(i, i) for i in phones
-				  if not i in phones_to_be_removed]
+		 if i not in phones_to_be_removed]
 
 phoneset_short = list(set(phone_reduction(phoneset)))
 phoneset_short.sort()
@@ -96,7 +99,7 @@ translation_key_asr2htk = {
 	'ṷ': 'u_',
 
 	# on the analogy of German umlaut, 'e' is used.
-	'ö': 'oe', 'ö:': 'oe:',
+	'ö': 'oe', 'ö:': 'oe:', ''
 	'ü': 'ue', 'ü:': 'ue:',
 
 	# on the analogy of Chinese...

acoustic_model/phoneset/fame_ipa.py

@@ -61,7 +61,7 @@ phoneset = [
 	'ɔⁿ',
 	'ɔ:',
 	'ɔ:ⁿ',
-	#'ɔ̈', # not included in lex.ipa 
+	'ɔ̈', # not included in lex.ipa 
 	'ɔ̈.',
 	'ɔ̈:',
 

acoustic_model/phoneset/fame_phonetics.py

@@ -0,0 +1,197 @@
+import sys
+import os
+os.chdir(r'C:\Users\Aki\source\repos\acoustic_model\acoustic_model')
+
+import fame_functions
+from phoneset import fame_ipa, fame_asr
+import convert_phoneset
+
+
+## general
+stop	  = 'p, b, t, d, k, g'  
+nasal	  = 'm, n, ŋ'
+fricative = 's, z, f, v, h, x, j'
+liquid	  = 'l, r'
+vowel	  = 'a, a:, e:, i, i:, i̯, o, o:, u, u:, ṷ, ö, ö:, ü, ü:, ɔ, ɔ:, ɔ̈, ə, ɛ, ɛ:, ɪ, ɪ:'
+
+## consonant
+c_front   = 'p, b, m, f, v'
+c_central = 't, d, n, s, z, l, r' 
+c_back	  = 'k, g, ŋ, h, x, j'
+
+fortis = 'p, t, k, f, s'
+lenis  = 'b, d, g, v, z, j'
+neither_fortis_nor_lenis = 'm, n, ŋ, h, l, r, x'
+
+coronal		= 't, d, n, s, z, l, r, j'
+non_coronal = 'p, b, m, k, g, ŋ, f, v, h, x'
+
+anterior	 = 'p, b, m, t, d, n, f, v, s, z, l'
+non_anterior = 'k, g, ŋ, h, x, j, r'
+
+continuent	   = 'm, n, ŋ, f, v, s, z, h, l, r'
+non_continuent = 'p, b, t, d, k, g, x, j'
+
+strident	 = 's, z, j'
+non_strident = 'f, v, h'
+unstrident	 = 'p, b, t, d, m, n, ŋ, k, g, r, x'
+
+glide	 = 'h, l, r'
+syllabic = 'm, l, ŋ'
+
+unvoiced = 'p, t, k, s, f, x, h'
+voiced	 = 'b, d, g, z, v, m, n, ŋ, l, r, j'
+
+#affricate: ???
+non_affricate = 's, z, f, v'
+
+voiced_stop   = 'b, d, g'
+unvoiced_stop = 'p, t, k'
+front_stop	  = 'p, b'
+central_stop  = 't, d'
+back_stop	  = 'k, g'
+
+voiced_fricative   = 'z, v'
+unvoiced_fricative = 's, f'
+front_fricative	   = 'f, v'
+central_fricative  = 's, z'
+back_fricative	   = 'j'
+
+
+## vowel
+v_front   = 'i, i:, i̯,  ɪ, ɪ:, e:, ə, ɛ, ɛ:, a, a:'
+v_central = 'ə, ɛ, ɛ:, a, a:'
+v_back	  = 'u, u:, ü, ü:, ṷ, ɔ, ɔ:, ɔ̈,  ö, ö:, o, o:'
+
+long  = 'a:, e:, i:, o:, u:, ö:, ü:, ɔ:, ɛ:, ɪ:' 
+short = 'a, i, i̯, o, u, ṷ, ö, ü, ɔ, ɔ̈, ə, ɛ, ɪ'
+
+#Dipthong: ???
+#Front-Start: ???
+#Fronting: ???
+
+high   = 'i, i:, i̯,  ɪ, ɪ: u, u:, ṷ, ə, e:, o, o:, ö, ö:, ü, ü:'
+medium = 'e:, ə, ɛ, ɛ:, ɔ, ɔ:, ɔ̈, o, o:, ö, ö:'
+low	   = 'a, a:, ɛ, ɛ:, ɔ, ɔ:, ɔ̈'
+
+rounded   = 'a, a:, o, o:, u, u:, ṷ, ö, ö:, ü, ü:, ɔ, ɔ:, ɔ̈' 
+unrounded = 'i, i:, i̯, e:, ə, ɛ, ɛ:, ɪ, ɪ:' 
+
+i_vowel = 'i, i:, i̯, ɪ, ɪ:'
+e_vowel = 'e:,ə, ɛ, ɛ:'
+a_vowel = 'a, a:'
+o_vowel = 'o, o:, ö, ö:, ɔ, ɔ:, ɔ̈'
+u_vowel = 'u, u:, ṷ, ü, ü:'
+
+## htk phoneset
+phoneset = fame_asr.phoneset_htk
+
+## convert ipa group to htk format for quests.hed.
+def _ipa2quest(R_or_L, ipa_text):
+	assert R_or_L in ['R', 'L'], print('the first argument should be either R or L.')
+	ipa_list = ipa_text.replace(' ', '').split(',')
+	if R_or_L == 'R':
+		quests_list = ['*+' + fame_functions.ipa2htk(ipa) for ipa in ipa_list]
+	else:
+		quests_list = [fame_functions.ipa2htk(ipa) + '-*' for ipa in ipa_list]
+	return ','.join(quests_list)
+
+
+def make_quests_hed(quest_hed):
+	def _add_quests_item(R_or_L, item_name_, ipa_text):
+		assert R_or_L in ['R', 'L'], print('the first argument should be either R or L.')
+		item_name = R_or_L + '_' + item_name_
+		with open(quest_hed, 'ab') as f:
+			f.write(bytes('QS  "' + item_name + '"\t{ ' + _ipa2quest(R_or_L, ipa_text) + ' }\n', 'ascii'))
+
+	if os.path.exists(quest_hed):
+		os.remove(quest_hed)
+
+	for R_or_L in ['R', 'L']:
+		_add_quests_item(R_or_L, 'NonBoundary', '*')
+		_add_quests_item(R_or_L, 'Silence', 'sil')
+
+		_add_quests_item(R_or_L, 'Stop', stop)
+		_add_quests_item(R_or_L, 'Nasal', nasal)
+		_add_quests_item(R_or_L, 'Fricative', fricative)
+		_add_quests_item(R_or_L, 'Liquid', liquid)
+		_add_quests_item(R_or_L, 'Vowel', vowel)
+
+		_add_quests_item(R_or_L, 'C-Front', c_front)
+		_add_quests_item(R_or_L, 'C-Central', c_central)
+		_add_quests_item(R_or_L, 'C-Back', c_back)
+
+		_add_quests_item(R_or_L, 'V-Front', v_front)
+		_add_quests_item(R_or_L, 'V-Central', v_central)
+		_add_quests_item(R_or_L, 'V-Back', v_back)
+
+		_add_quests_item(R_or_L, 'Front', c_front + v_front)
+		_add_quests_item(R_or_L, 'Central', c_central + v_central)
+		_add_quests_item(R_or_L, 'Back', c_front + v_back)
+
+		_add_quests_item(R_or_L, 'Fortis', fortis)
+		_add_quests_item(R_or_L, 'Lenis', lenis)
+		_add_quests_item(R_or_L, 'UnFortLenis', neither_fortis_nor_lenis)
+
+		_add_quests_item(R_or_L, 'Coronal', coronal)
+		_add_quests_item(R_or_L, 'NonCoronal', non_coronal)
+
+		_add_quests_item(R_or_L, 'Anterior', anterior)
+		_add_quests_item(R_or_L, 'NonAnterior', non_anterior)
+
+		_add_quests_item(R_or_L, 'Continuent', continuent)
+		_add_quests_item(R_or_L, 'NonContinuent', non_continuent)
+
+		_add_quests_item(R_or_L, 'Strident', strident)
+		_add_quests_item(R_or_L, 'NonStrident', non_strident)
+		_add_quests_item(R_or_L, 'UnStrident', unstrident)
+
+		_add_quests_item(R_or_L, 'Glide', glide)
+		_add_quests_item(R_or_L, 'Syllabic', syllabic)
+
+		_add_quests_item(R_or_L, 'Unvoiced-Cons', unvoiced)
+		_add_quests_item(R_or_L, 'Voiced-Cons', voiced)
+		_add_quests_item(R_or_L, 'Unvoiced-All', unvoiced + ', sil')
+
+		_add_quests_item(R_or_L, 'Long', long)
+		_add_quests_item(R_or_L, 'Short', short)
+
+		#_add_quests_item(R_or_L, 'Dipthong', xxx)
+		#_add_quests_item(R_or_L, 'Front-Start', xxx)
+		#_add_quests_item(R_or_L, 'Fronting', xxx)
+
+		_add_quests_item(R_or_L, 'High', high)
+		_add_quests_item(R_or_L, 'Medium', medium)
+		_add_quests_item(R_or_L, 'Low', low)
+
+		_add_quests_item(R_or_L, 'Rounded', rounded)
+		_add_quests_item(R_or_L, 'UnRounded', unrounded)
+
+		#_add_quests_item(R_or_L, 'Affricative', rounded)
+		_add_quests_item(R_or_L, 'NonAffricative', non_affricate)
+
+		_add_quests_item(R_or_L, 'IVowel', i_vowel)
+		_add_quests_item(R_or_L, 'EVowel', e_vowel)
+		_add_quests_item(R_or_L, 'AVowel', a_vowel)
+		_add_quests_item(R_or_L, 'OVowel', o_vowel)
+		_add_quests_item(R_or_L, 'UVowel', u_vowel)
+
+		_add_quests_item(R_or_L, 'Voiced-Stop', voiced_stop)
+		_add_quests_item(R_or_L, 'UnVoiced-Stop', unvoiced_stop)
+
+		_add_quests_item(R_or_L, 'Front-Stop', front_stop)
+		_add_quests_item(R_or_L, 'Central-Stop', central_stop)
+		_add_quests_item(R_or_L, 'Back-Stop', back_stop)
+
+		_add_quests_item(R_or_L, 'Voiced-Fric', voiced_fricative)
+		_add_quests_item(R_or_L, 'UnVoiced-Fric', unvoiced_fricative)
+
+		_add_quests_item(R_or_L, 'Front-Fric', front_fricative)
+		_add_quests_item(R_or_L, 'Central-Fric', central_fricative)
+		_add_quests_item(R_or_L, 'Back-Fric', back_fricative)
+
+		for p in phoneset:
+			_add_quests_item(R_or_L, p, p)
+
+	return
+