else:
logging.basicConfig(level=logging.INFO)
logging.getLogger("requests").setLevel(logging.WARNING)
source = options.source
volume = options.volume
aggressiveness = options.aggressiveness
rec = PulseRecorder(source_name=source, volume=volume, rate=sampleRate)
vad = VAD(aggressiveness=aggressiveness,
sample_rate=sampleRate,
max_utt_length=MAX_UTT_LENGTH)
#main
frames = int(sampleRate * BUFFER_DURATION / 1000) #BUFFER_DURATION 30
rec.start_recording(frames_per_buffer=frames)
time_start = time()
print("Please speak.")
total, buff_size, finalize = 0, 0, 0
while True:
samples = rec.get_samples()
audio, finalize = vad.process_audio(samples)
#print(len(samples),total,audio)
if not audio:
continue
#logging.debug ('decoding audio len=%d finalize=%s audio=%s'% (len(audio), repr(finalize), audio[0].__class__))
# decoding stop while finalize is true, max_length 12s
buff_size = len(samples)
#buff = struct.unpack_from('<%dh' % buff_size, audio)
tts = TTS(engine="espeak", voice="en")
#
# main loop
#
print(chr(27) + "[2J")
while True:
#
# record audio, run VAD
#
print "Please speak.",
rec.start_recording()
finalize = False
recording = []
while not finalize:
samples = rec.get_samples()
audio, finalize = vad.process_audio(samples)
if not audio:
continue
recording.extend(audio)
user_utt, confidence = asr.decode(audio, finalize)
vad = VAD(aggressiveness=AGGRESSIVENESS, sample_rate=SAMPLE_RATE)
#
# ASR
#
print("Loading model from %s ..." % MODEL_DIR)
asr = KaldiNNet3OnlineModel(MODEL_DIR, MODEL)
#, acoustic_scale=ACOUSTIC_SCALE, beam=BEAM, frame_subsampling_factor=FRAME_SUBSAMPLING_FACTOR)
print("Loading model from %s, done ..." % MODEL_DIR)
#
# main
#
print("Start recording")
rec.start_recording(FRAMES_PER_BUFFER)
print("Please speak.")
while True:
samples = rec.get_samples()
logging.debug("%d samples, %5.2f s" %
(len(samples), float(len(samples)) / float(SAMPLE_RATE)))
audio, finalize = vad.process_audio(samples)
if not audio:
continue
#
# ASR
#
print "Loading model from %s ..." % model_dir
asr = ASR(engine = ASR_ENGINE_NNET3, model_dir = model_dir,
kaldi_beam = DEFAULT_BEAM, kaldi_acoustic_scale = DEFAULT_ACOUSTIC_SCALE,
kaldi_frame_subsampling_factor = DEFAULT_FRAME_SUBSAMPLING_FACTOR)
#
# main
#
complete_utterance=''
rec.start_recording()
finalize = False
while True:
samples = rec.get_samples()
audio, finalize = vad.process_audio(samples)
if not audio:
continue
logging.debug ('decoding audio len=%d finalize=%s audio=%s' % (len(audio), repr(finalize), audio[0].__class__))
user_utt, confidence = asr.decode(audio, finalize, stream_id=STREAM_ID)
print "\r%s " % user_utt,
if finalize:
