class TestRawDecoder(TestCase):
def __init__(self, *args, **kwargs):
self.ps = Pocketsphinx()
self.ps.decode()
super(TestRawDecoder, self).__init__(*args, **kwargs)
def test_raw_decoder_lookup_word(self):
self.assertEqual(self.ps.lookup_word('hello'), 'HH AH L OW')
self.assertEqual(self.ps.lookup_word('abcdf'), None)
def test_raw_decoder_hypothesis(self):
self.assertEqual(self.ps.hypothesis(), 'go forward ten meters')
self.assertEqual(self.ps.score(), -7066)
self.assertEqual(self.ps.confidence(), 0.04042641466841839)
def test_raw_decoder_segments(self):
self.assertEqual(
self.ps.segments(),
['<s>', '<sil>', 'go', 'forward', 'ten', 'meters', '</s>'])
def test_raw_decoder_best_hypothesis(self):
self.assertEqual(self.ps.best(), [('go forward ten meters', -28034),
('go for word ten meters', -28570),
('go forward and majors', -28670),
('go forward and meters', -28681),
('go forward and readers', -28685),
('go forward ten readers', -28688),
('go forward ten leaders', -28695),
('go forward can meters', -28695),
('go forward and leaders', -28706),
('go for work ten meters', -28722)])
class TestRawDecoder(TestCase):
def __init__(self, *args, **kwargs):
self.ps = Pocketsphinx()
self.ps.decode()
super(TestRawDecoder, self).__init__(*args, **kwargs)
def test_raw_decoder_lookup_word(self):
self.assertEqual(self.ps.lookup_word('hello'), 'HH AH L OW')
self.assertEqual(self.ps.lookup_word('abcdf'), None)
def test_raw_decoder_hypothesis(self):
self.assertEqual(self.ps.hypothesis(), 'go forward ten meters')
self.assertEqual(self.ps.score(), -7066)
self.assertEqual(self.ps.confidence(), 0.04042641466841839)
def test_raw_decoder_segments(self):
self.assertEqual(self.ps.segments(), [
'<s>', '<sil>', 'go', 'forward', 'ten', 'meters', '</s>'
])
def test_raw_decoder_best_hypothesis(self):
self.assertEqual(self.ps.best(), [
('go forward ten meters', -28034),
('go for word ten meters', -28570),
('go forward and majors', -28670),
('go forward and meters', -28681),
('go forward and readers', -28685),
('go forward ten readers', -28688),
('go forward ten leaders', -28695),
('go forward can meters', -28695),
('go forward and leaders', -28706),
('go for work ten meters', -28722)
])
class SpeechProcessor:
def __init__(self, hmm='data/spanish/CIEMPIESS_Spanish_Models_581h/Models/modelo',
lm='data/spanish/CIEMPIESS_Spanish_Models_581h/Models/leng.lm.bin',
dict='data/spanish/CIEMPIESS_Spanish_Models_581h/Models/dicc.dic',
grammar='data/gramatica-tp2.gram', dataPath='tmp/'):
self.data_path = dataPath
config = {
'hmm': hmm,
'lm': lm,
'dict': dict
}
#model_path = get_model_path()
self.ps = Pocketsphinx(**config)
# Switch to JSGF grammar
jsgf = Jsgf(grammar)
rule = jsgf.get_rule('tp2.grammar')
fsg = jsgf.build_fsg(rule, self.ps.get_logmath(), 7.5)
self.ps.set_fsg('tp2', fsg)
self.ps.set_search('tp2')
# Síntesis
self.tts_authenticator = IAMAuthenticator('cq9_4YcCXxClw2AfgUhbokFktZ-xSRT4kcHS2akcZ05J')
self.tts = TextToSpeechV1(authenticator=self.tts_authenticator)
self.tts.set_service_url('https://stream.watsonplatform.net/text-to-speech/api')
def sintetizar(self, outFileName, msg):
if len(msg) > 0:
with open(outFileName, 'wb') as audio_file:
audio_file.write(
self.tts.synthesize(
msg,
voice='es-LA_SofiaV3Voice',
accept='audio/wav'
).get_result().content)
def reconocer(self, inFileName='audio.wav'):
# Reconocimiento
print(self.data_path)
self.ps.decode(
audio_file=os.path.join(self.data_path,inFileName),
buffer_size=2048,
no_search=False,
full_utt=False
)
return self.ps.segments(), self.ps.best(count=3)
class TestPhoneme(TestCase):
def __init__(self, *args, **kwargs):
self.ps = Pocketsphinx(
lm=False,
dic=False,
allphone='deps/pocketsphinx/model/en-us/en-us-phone.lm.bin',
lw=2.0,
pip=0.3,
beam=1e-200,
pbeam=1e-20,
mmap=False
)
self.ps.decode()
super(TestPhoneme, self).__init__(*args, **kwargs)
def test_phoneme_hypothesis(self):
self.assertEqual(
self.ps.hypothesis(),
'SIL G OW F AO R W ER D T AE N M IY IH ZH ER Z S V SIL'
)
def test_phoneme_best_phonemes(self):
self.assertEqual(self.ps.segments(), [
'SIL',
'G',
'OW',
'F',
'AO',
'R',
'W',
'ER',
'D',
'T',
'AE',
'N',
'M',
'IY',
'IH',
'ZH',
'ER',
'Z',
'S',
'V',
'SIL'
])
class TestPhoneme(TestCase):
def __init__(self, *args, **kwargs):
self.ps = Pocketsphinx(
lm=False,
dic=False,
allphone='deps/pocketsphinx/model/en-us/en-us-phone.lm.bin',
lw=2.0,
pip=0.3,
beam=1e-200,
pbeam=1e-20,
mmap=False
)
self.ps.decode()
super(TestPhoneme, self).__init__(*args, **kwargs)
def test_phoneme_hypothesis(self):
self.assertEqual(
self.ps.hypothesis(),
'SIL G OW F AO R D T AE N NG IY ZH ER S SIL'
)
def test_phoneme_best_phonemes(self):
self.assertEqual(self.ps.segments(), [
'SIL',
'G',
'OW',
'F',
'AO',
'R',
'D',
'T',
'AE',
'N',
'NG',
'IY',
'ZH',
'ER',
'S',
'SIL'
])
data_path = get_data_path()
config = {
'hmm': os.path.join(model_path, 'en-us'),
'lm': os.path.join(model_path, 'en-us.lm.bin'),
'dict': os.path.join(model_path, 'cmudict-en-us.dict')
}
ps = Pocketsphinx(**config)
ps.decode(audio_file=os.path.join(data_path, 'output.raw'),
buffer_size=2048,
no_search=False,
full_utt=False)
# => ['<s>', '<sil>', 'go', 'forward', 'ten', 'meters', '</s>']
print(ps.segments())
print('Detailed segments:', *ps.segments(detailed=True), sep='\n') # => [
# word, prob, start_frame, end_frame
# ('<s>', 0, 0, 24)
# ('<sil>', -3778, 25, 45)
# ('go', -27, 46, 63)
# ('forward', -38, 64, 116)
# ('ten', -14105, 117, 152)
# ('meters', -2152, 153, 211)
# ('</s>', 0, 212, 260)
# ]
print(ps.hypothesis()) # => go forward ten meters
print(ps.probability()) # => -32079
print(ps.score()) # => -7066
print(ps.confidence()) # => 0.04042641466841839
model_path = get_model_path()
data_path = get_data_path()
config = {
'hmm': os.path.join(model_path, 'en-us'),
'lm': r'C:\Users\BZT\Desktop\speech_segment\5446.lm',
'dict': r'C:\Users\BZT\Desktop\speech_segment\5446.dic'
}
ps = Pocketsphinx(**config)
ps.decode(
audio_file=
r'C:\Users\BZT\Desktop\speech_segment\speech_segment\Ses01F_impro01_M013.wav',
buffer_size=2048,
no_search=False,
full_utt=False)
# print(ps.segments()) # => ['<s>', '<sil>', 'go', 'forward', 'ten', 'meters', '</s>']
# print('Detailed segments:', *ps.segments(detailed=True), sep='\n')
for sample in ps.segments(detailed=True):
for subsample in sample:
print(subsample, end='\t')
print()
print(ps.hypothesis()) # => go forward ten meters
# print(ps.probability()) # => -32079
# print(ps.score()) # => -7066
# print(ps.confidence()) # => 0.04042641466841839
# print(*ps.best(count=10), sep='\n')
class Sphinx(Thread):
def __init__(self):
Thread.__init__(self)
self.ready = False
def run(self):
print_important("Info! Thread sphinx started.")
self.config = {
'verbose': True,
'hmm': os.path.join('s2m', 'core', 'sphinx', 'fr'),
'lm': os.path.join('s2m', 'core', 'sphinx', 'fr.lm.dmp'),
'dict': os.path.join('s2m', 'core', 'sphinx', 's2m.dict'),
'jsgf': os.path.join('s2m', 'core', 'sphinx', 's2m.jsgf'),
}
self.pocketsphinx = Pocketsphinx(**self.config)
self.ready = True
def get_silence(self, duration):
if duration < 0.25:
return '[veryshortsil]'
elif duration < 0.5:
return '[shortsil]'
elif duration < 1.5:
return '[sil]'
elif duration < 3.:
return '[longsil]'
else:
return '[verylongsil]'
def get_segment_string(self, segments):
segment_list = []
last_silence = 0
spoken_duration = 0
word_count = 0
for segment in segments:
if segment.word in ['<s>', '</s>']:
continue
elif segment.word == '<sil>':
last_silence += segment.end_frame - segment.start_frame
else:
if last_silence > 0:
segment_list.append(last_silence)
last_silence = 0
spoken_duration += segment.end_frame - segment.start_frame
segment_list.append(segment.word)
word_count += 1
if word_count == 0:
return ''
avg_word_duration = spoken_duration / word_count
return ' '.join((self.get_silence(s / avg_word_duration)
if type(s) is int else nobrackets(s))
for s in segment_list)
def to_text(self, filename, erase=False):
if not self.ready:
raise EnvironmentError('Initialization of sphinx not finished.')
FILLER_WORDS = ['<s>', '<sil>', '</s>']
try:
self.pocketsphinx.decode(filename)
except Exception as e:
print("An error was raised by sphinx while decoding file '%r', parsing aborted" % filename)
text = " ".join(
[s for s in self.pocketsphinx.segments() if s not in FILLER_WORDS])
text = nobrackets(text)
segment_string = self.get_segment_string(self.pocketsphinx.seg())
nbest = [nobrackets(w[0])
for w in self.pocketsphinx.best(count=10)[1:]]
if erase:
os.remove(loc)
return segment_string, nbest
data_path = get_data_path()
print "config set"
config = {
'hmm': os.path.join(model_path, 'en-us'),
'lm': os.path.join(model_path, 'en-us.lm.bin'),
'dict': os.path.join(model_path, 'cmudict-en-us.dict')
}
ps = Pocketsphinx(**config)
print "decoging"
ps.decode(audio_file=sys.argv[1],
buffer_size=2048,
no_search=False,
full_utt=False)
print "decoded"
print(ps.segments()
) # => ['<s>', '<sil>', 'go', 'forward', 'ten', 'meters', '</s>']
print "DETAILED SHIT"
print(ps.segments(detailed=True)) # => [
# word, prob, start_frame, end_frame
# ('<s>', 0, 0, 24)
# ('<sil>', -3778, 25, 45)
# ('go', -27, 46, 63)
# ('forward', -38, 64, 116)
# ('ten', -14105, 117, 152)
# ('meters', -2152, 153, 211)
# ('</s>', 0, 212, 260)
# ]
#print(ps.hypothesis()) # => go forward ten meters
#print(ps.probability()) # => -32079
buffer_size=2048,
no_search=False,
full_utt=False
)
#print(ps.segments())
#save the detailed segments of the words,
#which will contain details word, probablity, start_time and end_time
#print('Detailed segments:', *ps.segments(detailed=True), sep='\n')
# with open('output_segments_obama_farewell_speech.txt', 'a') as f:
# print(*ps.segments(detailed=True), sep='\n', file=f)
with open(filename_output_segments, 'a') as f:
print(*ps.segments(detailed=True), sep='\n', file=f)
#convert from audio to text and save
text = ps.hypothesis()
file1 = open(filename_sphinx,"w")#write mode
file1.write(text)
file1.close()
#load into dataframe
# For the above saved file, modify manually by removing '(',')',' and then save as modified fie
#df = pd.read_csv('output_segments_donaldTrump_modified.txt', sep=",", header=None)
df = pd.read_csv(filename_output_segments_mod, sep=",", header=None)
df.columns = ["word", "prob","startTime", "endTime"]
df.head()
#instantiate PyAudio
p = pyaudio.PyAudio()
#open stream
stream = p.open(format = 8,
channels = 1,
rate = 16000,
output = True)
#read data
data = f.read(chunk*2)
frames = []
phones = {}
phoneme_pred = ps.segments(detailed=True)
fps = 100
for idx,s in enumerate(phoneme_pred):
if (idx < len(phoneme_pred)-1):
phones[phoneme_pred[idx+1][2]/fps] = phoneme_pred[idx][0]
else:
phones[phoneme_pred[idx][3]/fps] = phoneme_pred[idx][0]
mouth_list = []
pre_mouth_shape = 'M'
#lip sync style intensity variation
if (opt.level == 0):
intensity = 0.2
elif (opt.level == 1):
