def audio2phoneme(audio_file):
wave_read = wave.open(audio_file, 'rb')
length = wave_read.getnframes() / wave_read.getframerate()
wave_read.close()
# Decode streaming data.
decoder = Decoder(config)
buf = bytearray(1024)
with open(audio_file, 'rb') as f:
decoder.start_utt()
while f.readinto(buf):
decoder.process_raw(buf, False, False)
decoder.end_utt()
nframes = decoder.n_frames()
phonemes = []
offset = None
for seg in decoder.seg():
if offset is None:
offset = seg.start_frame
start_frame = seg.start_frame - offset
end_frame = seg.end_frame - offset
phonemes.append((seg.word, start_frame / nframes * length,
end_frame / nframes * length))
return phonemes
def process_file(self, audiofile):
"""
processes audio file and returns the text
"""
with open(audiofile, 'rb') as audiofile:
decoder = Decoder(self.config)
decoder.start_utt()
while True:
buf = audiofile.read(1024)
if buf:
decoder.process_raw(buf, False, False)
else:
break
decoder.end_utt()
hyp = decoder.hyp()
print "Hyp:", hyp
if hyp != None:
print "Hyp Score", (hyp.prob, hyp.best_score)
average_score = 0
seg_count = 0
for seg in decoder.seg():
if seg.word != "<sil>":
seg_count += 1
average_score += seg.ascore
print(seg.word, seg.ascore, seg.lscore)
print "hyp:", hyp.hypstr
print average_score / seg_count
return hyp.hypstr
return None
def audio2phoneme(audio_file):
wave_read = wave.open(audio_file, 'rb')
length = wave_read.getnframes()/wave_read.getframerate()
wave_read.close()
# Decode streaming data.
decoder = Decoder(config)
buf = bytearray(1024)
with open(audio_file, 'rb') as f:
decoder.start_utt()
while f.readinto(buf):
decoder.process_raw(buf, False, False)
decoder.end_utt()
nframes = decoder.n_frames()
phonemes = []
offset = None
for seg in decoder.seg():
if offset is None:
offset = seg.start_frame
start_frame = seg.start_frame - offset
end_frame = seg.end_frame - offset
phonemes.append((
seg.word, start_frame/nframes*length, end_frame/nframes*length))
return phonemes
class PocketSphinxASREngine(ASREngine):
"""https://pypi.org/project/pocketsphinx/"""
def __init__(self):
# https://github.com/cmusphinx/pocketsphinx-python/blob/master/example.py
config = Decoder.default_config()
config.set_string('-logfn', '/dev/null')
config.set_string('-hmm', os.path.join(get_model_path(), 'en-us'))
config.set_string('-lm', os.path.join(get_model_path(),
'en-us.lm.bin'))
config.set_string('-dict',
os.path.join(get_model_path(), 'cmudict-en-us.dict'))
self._decoder = Decoder(config)
def transcribe(self, path):
pcm, sample_rate = soundfile.read(path)
assert sample_rate == 16000
pcm = (np.iinfo(np.int16).max * pcm).astype(np.int16).tobytes()
self._decoder.start_utt()
self._decoder.process_raw(pcm, no_search=False, full_utt=True)
self._decoder.end_utt()
words = []
for seg in self._decoder.seg():
word = seg.word
# Remove special tokens.
if word == '<sil>' or word == '<s>' or word == '</s>':
continue
word = ''.join([x for x in word if x.isalpha()])
words.append(word)
return ' '.join(words)
def __str__(self):
return 'PocketSphinx'
rate=16000,
input=True,
frames_per_buffer=1024)
stream.start_stream()
print('start...')
while True:
buf = stream.read(1024)
if buf:
decoder.process_raw(buf, False, False)
else:
break
hypothesis = decoder.hyp()
if hypothesis:
print('\nhypothesis: %s, score: %d' %
(hypothesis.hypstr, hypothesis.best_score))
print([(seg.word, seg.prob, seg.start_frame, seg.end_frame)
for seg in decoder.seg()])
print("Detected keyword, restarting search")
os.system('mpg123 ' + os.path.join(script_dir, 'hi.mp3'))
print('restart...')
decoder.end_utt()
decoder.start_utt()
print('ok')
# break
stream.close()
stream = open(sys.argv[1], "rb")
else:
p = pyaudio.PyAudio()
stream = p.open(format=pyaudio.paInt16, channels=1, rate=16000, input=True, frames_per_buffer=1024)
stream.start_stream()
print('start...')
while True:
buf = stream.read(1024)
if buf:
decoder.process_raw(buf, False, False)
else:
break
hypothesis = decoder.hyp()
if hypothesis:
print('\nhypothesis: %s, score: %d' % (hypothesis.hypstr, hypothesis.best_score))
print ([(seg.word, seg.prob, seg.start_frame, seg.end_frame) for seg in decoder.seg()])
print ("Detected keyword, restarting search")
os.system('mpg123 ' + os.path.join(script_dir, 'hi.mp3'))
print('restart...')
decoder.end_utt()
decoder.start_utt()
print('ok')
# break
stream.close()