def tts(self, text, session_begin_params):
ret = c_int()
sessionID = self.sdk.QTTSSessionBegin(session_begin_params, byref(ret))
self._logger.debug('QTTSSessionBegin => sessionID: %s ret: %s'% (sessionID, ret.value))
#input text
ret = self.sdk.QTTSTextPut(sessionID, text, len(text), None)
if const.MSP_SUCCESS != ret:
self._logger.error("QTTSTextPut failed Error code %d.\n"%ret)
else:
self._logger.debug("QTTSTextPut SUCCESS=> %s"% ret)
#systhesize audio
audio_len = c_uint()
synth_status = c_int()
errorCode = c_int()
lame = LameEncoder(const.RATE, const.CHANNEL, pyaudio.get_sample_size(pyaudio.paInt16))
with tempfile.NamedTemporaryFile(suffix='.%s'%self.audioFormat,mode='w+b', delete=False) as f:
audioFile = None
if self.audioFormat == "mp3":
audioFile = open(f.name, "wb+")
else:
audioFile = wave.open(f, "wb")
# 配置声道数、量化位数、取样频率
audioFile.setnchannels(const.CHANNEL)
audioFile.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
audioFile.setframerate(const.RATE)
self._logger.debug('QTTSAudioGet => ')
while True:
self.sdk.QTTSAudioGet.restype = POINTER(c_ushort * (1024 * 1024))
audio_data = self.sdk.QTTSAudioGet(sessionID, byref(audio_len), byref(synth_status), byref(errorCode))
self._logger.debug('QTTSAudioGet => audio_len: %s synth_status: %s errorCode: %s'% (audio_len, synth_status,errorCode))
if audio_data:
data = string_at(audio_data, audio_len.value)
if self.audioFormat == "mp3":
output = lame.encode(data)
audioFile.write(output)
else:
#将wav data 转换为二进制数据写入wav文件
audioFile.writeframes(data)
if synth_status.value == const.MSP_TTS_FLAG_DATA_END or errorCode.value != const.MSP_SUCCESS:
break
time.sleep(0.1)
if self.audioFormat == "mp3":
output = lame.flush()
audioFile.write(output)
audioFile.close()
ret = self.sdk.QTTSSessionEnd(sessionID, "Normal")
self._logger.debug('QTTSSessionEnd => ret: %s'% ret)
f.seek(0)
return f.name
def __init__(self, recognizer, agent):
super(Listener, self).__init__()
self.recognizer = recognizer
self.agent = agent
self.queue = Queue.Queue()
self.running = True
self.sample_rate = 16000
self.sample_width = pyaudio.get_sample_size(pyaudio.paInt16) * 1
self.channels = 1
config = Decoder.default_config()
config.set_string('-hmm', path.join(PS_MODEL_DIR, 'en-us/en-us'))
config.set_string('-lm', path.join(PS_MODEL_DIR, 'en-us/en-us.lm.dmp'))
config.set_string('-dict', path.join(DATA_DIR, 'pocketsphinx/model/en-us/victoria-en-us.dict'))
config.set_string('-logfn', 'NUL')
config.set_string('-keyphrase', KEY_PHRASE)
config.set_float('-samprate', self.sample_rate)
config.set_float('-kws_threshold', 1e-40)
self.decoder = Decoder(config)
self.decoder.start_utt()
self.frames = collections.deque()
self.listening = -1
self.silence_start = -1
self.last_logged_hyp = None
def __init__(self,
on_audio: FunctionType,
pa_instance: pyaudio.PyAudio = None,
**kwargs):
""" Create a new Listener object. """
# Set defaults for arguments
kwargs.setdefault('format', pyaudio.paInt16)
kwargs.setdefault('channels', 1)
kwargs.setdefault('rate', 44100)
self.chunk = kwargs.pop('chunk', 1024)
self.threshold = kwargs.pop('threshold', 10.0)
self.timeout = kwargs.pop('timeout', 1.0)
self.sample_width = pyaudio.get_sample_size(kwargs['format'])
# Register the function called when audio is captured
self.on_audio = on_audio
# Save the named arguments
self.stream_args = kwargs
# If we were constructed with a connection, use that,
# Otherwise create one.
if pa_instance is not None:
self.connection = pa_instance
else:
self.connection = pyaudio.PyAudio()
self.stream = self.connection.open(input=True, **kwargs)
def __init__(self):
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format) # size of each sample
self.SAMPLE_RATE = 16000 # sampling rate in Hertz
self.CHUNK = 1024 # number of frames stored in each buffer
self.stream = NaoMicStream()
def sigToWav(data, stream, out):
waveFile = wave.open(out, 'wb')
waveFile.setnchannels(stream._channels)
waveFile.setsampwidth(pyaudio.get_sample_size(stream._format))
waveFile.setframerate(stream._rate)
waveFile.writeframes(b''.join(data))
waveFile.close()
def __init__(self, recognizer, agent):
super(Listener, self).__init__()
self.recognizer = recognizer
self.agent = agent
self.queue = Queue.Queue()
self.running = True
self.sample_rate = 16000
self.sample_width = pyaudio.get_sample_size(pyaudio.paInt16) * 1
self.channels = 1
config = Decoder.default_config()
config.set_string('-hmm', path.join(PS_MODEL_DIR, 'en-us/en-us'))
config.set_string('-lm', path.join(PS_MODEL_DIR, 'en-us/en-us.lm.dmp'))
config.set_string(
'-dict',
path.join(DATA_DIR,
'pocketsphinx/model/en-us/victoria-en-us.dict'))
config.set_string('-logfn', 'NUL')
config.set_string('-keyphrase', KEY_PHRASE)
config.set_float('-samprate', self.sample_rate)
config.set_float('-kws_threshold', 1e-40)
self.decoder = Decoder(config)
self.decoder.start_utt()
self.frames = collections.deque()
self.listening = -1
self.silence_start = -1
self.last_logged_hyp = None
def __init__(self, filename, channels, format, rate):
super(WaveFile, self).__init__()
self._frames = []
self._wf = wave.open(filename, 'wb')
self._wf.setnchannels(channels)
self._wf.setsampwidth(pyaudio.get_sample_size(format))
self._wf.setframerate(rate)
def on_command(self, data):
"""
Perform actions after voice input was recorded.
:param np.array data: The voice input data
"""
self.recording_state = False
# Simulate action
time.sleep(5)
if self.raspi_mode:
self.light.processing()
# TODO: For now, save the file
filename = 'testapp' + datetime.now().strftime(
'%Y-%m-%d_%H-%M-%S') + '.wav'
wf = wave.open(filename, 'wb')
wf.setnchannels(self.channels)
wf.setsampwidth(
pyaudio.get_sample_size(pyaudio.get_format_from_width(self.width)))
wf.setframerate(self.sample_rate)
wf.writeframes(b''.join(data))
wf.close()
self.recording = []
self.data = np.zeros(self.feed_samples, dtype=self.format)
self.queue.empty()
self.recording_state = False
if self.raspi_mode:
self.light.off()
def __init__(self, wrapped_stream, format, muted=False):
assert wrapped_stream is not None
self.wrapped_stream = wrapped_stream
self.muted = muted
self.SAMPLE_WIDTH = pyaudio.get_sample_size(format)
self.muted_buffer = b''.join([b'\x00' * self.SAMPLE_WIDTH])
def __init__(self):
self.CHUNK = 1024
self.RATE = 16000
self.FORMAT = pyaudio.paInt16
self.DTYPE='Int16'
self.CHANNELS = 1
self.RUN_SECONDS = 1000
self.sampwidth = pyaudio.get_sample_size(self.FORMAT)
self.mic = False
self.wf = None
self.deque_time = deque(maxlen=20)
self.deque_mean = deque(maxlen=3)
self.deque_freq = deque(maxlen=3)
self.posneg_frequencies = np.fft.fftfreq(self.CHUNK, 1.0/self.RATE)
self.freqs_indices = np.where(self.posneg_frequencies >= 0)
self.freqs = self.posneg_frequencies[np.where(self.posneg_frequencies >= 0)]
self.t0 = time.time()
self.ichunk = 0
# plt.ion()
# plt.axis([0,100,0,800])
# fig=plt.figure()
if RPI:
self.led = Lights()
self.led.start()
def record(self,
participant='0',
session='0',
trial=0): # TODO: eliminate this session requirement
# record mark
RECORDON = True
self.RECORDON = RECORDON
# open audio stream
self.STREAM = self.pad.open(format=FORMAT,
channels=CHANNELS,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
self.FRAMES = []
while self.RECORDON: # when being called, the 'finish' method will feed a False value to RECORDON to end loop
self.SOUND = self.STREAM.read(CHUNK)
self.FRAMES.append(self.SOUND)
# close stream and terminate PyAudio object
self.STREAM.stop_stream()
self.STREAM.close()
self.pad.terminate()
# define file name
WAVE_OUTPUT_FILENAME = str(participant) + '_' + str(
session) + '_' + str(trial) + '.wav'
# write sound to disk
waveFile = wave.open(WAVE_OUTPUT_FILENAME, 'wb')
waveFile.setnchannels(CHANNELS)
waveFile.setsampwidth(pyaudio.get_sample_size(FORMAT))
waveFile.setframerate(RATE)
waveFile.writeframes(b''.join(self.FRAMES))
waveFile.close()
return None
def __init__(self, byte_data, sample_rate, bit_width, channels, dtype = None):
"""
byte_data: A byte string containing the raw data.
BIT_WIDTH: bit width in bytes.
"""
assert isinstance(bit_width, (int, long)) and bit_width > 0, \
"`bit_width` must be positive integer."
bit_width = pyaudio.get_sample_size(pyaudio.get_format_from_width(bit_width))
assert isinstance(channels, int) and channels in [1, 2], \
"`channels` can be either 1(mono) or 2(stereo)."
assert channels in (1, 2), \
"`channels` can be either 1(mono) or 2(stereo) only."
assert sample_rate > 0, "`sample_rate` must be positive."
self.__bit_width = bit_width
self.__channels = channels
self.__sample_rate = sample_rate
self.__byte_data = byte_data # a byte string
if dtype is None:
dtype = self._get_dtype_by_bit_width()
if not self._validate_dtype(dtype):
raise ValueError("`dtype` is not compatible with the `bit_width`.")
self.__dtype = dtype
self.format = pyaudio.get_format_from_width(self.BIT_WIDTH)
async def listen(websocket, _):
print('Connected..')
if settings.DEBUG:
print('Debug activated')
frame_data = []
stt = speechtotext.SpeechToText()
while True:
chunk = await websocket.recv()
if chunk == 'transcribe' and frame_data:
audio_data = stt.process_audio(frame_data)
if settings.AUDIO_STORAGE and frame_data:
filename = f'{settings.AUDIO_FOLDER_PATH}/{uuid.uuid4()}.wav'
with wave.open(filename, 'wb') as f:
f.setnchannels(settings.CHANNELS)
f.setsampwidth(pyaudio.get_sample_size(settings.FORMAT))
f.setframerate(settings.RATE)
f.writeframes(audio_data.frame_data)
try:
transcription = stt.recognize(audio_data)
r = response(transcription)
except speechtotext.RecognitionException as e:
print('Error recognizing: {}'.format(str(e)))
r = response(None, error=str(e))
frame_data = []
await websocket.send(r)
else:
frame_data.append(chunk)
class Recorder:
FORMAT = pyaudio.paInt16
CHUNK = 1024
SWIDTH = pyaudio.get_sample_size(pyaudio.paInt16)
def __init__(self, input_name='Microphone', channel_id=1):
self.__create_stream(input_name, channel_id)
def __create_stream(self, input_name, channel_id):
self.__stream = p.open(
format=Recorder.FORMAT,
channels=channel_id,
rate=settings.sampling_rate,
input=True,
input_device_index=get_index_by_name(input_name),
frames_per_buffer=Recorder.CHUNK)
def record(self):
nbits = self.__stream.get_read_available()
try:
raw_data = self.__stream.read(
settings.recording_chunk_size,
exception_on_overflow=False) # TODO catch proper exception
data = np.array(
wave.struct.unpack("%dh" % (len(raw_data) / self.SWIDTH),
raw_data))
except OSError:
print(('skipping audio', nbits))
raw_data = self.__stream.read(settings.recording_chunk_size,
exception_on_overflow=False)
data = np.array(
wave.struct.unpack("%dh" % (len(raw_data) / self.SWIDTH),
raw_data))
return data
def __init__(self,
device_index=None,
sample_rate=16000,
chunk_size=1024):
assert device_index is None or isinstance(
device_index, int), "Device index must be None or an integer"
if device_index is not None: # ensure device index is in range
audio = pyaudio.PyAudio()
count = audio.get_device_count()
audio.terminate() # obtain device count
assert 0 <= device_index < count, "Device index out of range"
assert isinstance(
sample_rate, int
) and sample_rate > 0, "Sample rate must be a positive integer"
assert isinstance(
chunk_size, int
) and chunk_size > 0, "Chunk size must be a positive integer"
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(
self.format) # size of each sample
self.SAMPLE_RATE = sample_rate # sampling rate in Hertz
self.CHANNELS = 1 # mono audio
self.CHUNK = chunk_size # number of frames stored in each buffer
self.audio = None
self.stream = None
def activeListenToAllOptions(self, THRESHOLD=None, LISTEN=True,
MUSIC=False):
"""
Records until a second of silence or times out after 12 seconds
Returns a list of the matching options or None
"""
RATE = 16000
CHUNK = 1024
LISTEN_TIME = 12
# check if no threshold provided
if THRESHOLD is None:
THRESHOLD = self.fetchThreshold()
self.speaker.play(jasperpath.data('audio', 'beep_hi.wav'))
# prepare recording stream
stream = self._audio.open(format=pyaudio.paInt16,
channels=1,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
frames = []
# increasing the range # results in longer pause after command
# generation
lastN = [THRESHOLD * 1.2 for i in range(30)]
for i in range(0, int(RATE / CHUNK * LISTEN_TIME)):
data = stream.read(CHUNK)
frames.append(data)
score = self.getScore(data)
lastN.pop(0)
lastN.append(score)
average = sum(lastN) / float(len(lastN))
# TODO: 0.8 should not be a MAGIC NUMBER!
if average < THRESHOLD * 0.8:
break
self.speaker.play(jasperpath.data('audio', 'beep_lo.wav'))
# save the audio data
stream.stop_stream()
stream.close()
with tempfile.SpooledTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(RATE)
wav_fp.writeframes(b''.join(frames))
wav_fp.close()
f.seek(0)
return self.active_stt_engine.transcribe(f)
def __init__(self, file_handler, to_text=True, stop_event=None):
self.file_handler = file_handler
self.sample_width = pyaudio.get_sample_size(
settings.STREAM_AUDIO_FORMAT)
self.to_text = to_text
self.recognizer = sr.Recognizer()
self.stop_event = stop_event
def __init__(self, wrapped_stream, format, muted=False):
assert wrapped_stream is not None
self.wrapped_stream = wrapped_stream
self.muted = muted
self.SAMPLE_WIDTH = pyaudio.get_sample_size(format)
self.muted_buffer = b''.join([b'\x00' * self.SAMPLE_WIDTH])
def __init__(self, device_index, sample_rate, bit_width, chunk_size = 8092, channels = 1):
audio = pyaudio.PyAudio()
## Checking the device_index is valid or not.
assert isinstance(device_index, (int, long)), "Device index must be an integer."
device_count = audio.get_device_count()
assert 0 <= device_index < device_count, "`device_index` out of range: {} out of {}".format(device_index, count)
audio.terminate()
self.__device_index = device_index
if not self.device_info["maxInputChannels"] > 0:
raise DeviceTypeError("Can not source from a non-input device.")
self.__format = pyaudio.get_format_from_width(bit_width)
self.__bit_width = pyaudio.get_sample_size(self.FORMAT)
assert isinstance(sample_rate, (int, long)), "`sample_rate` must be integer."
max_sample_rate = self.device_info["defaultSampleRate"]
assert 0 < sample_rate <= max_sample_rate, "`sample_rate` out of range: {} out of {}".format(sample_rate, max_sample_rate)
self.__sample_rate = sample_rate
assert isinstance(chunk_size, (int, long)), "`chunk_size` must be integer."
self.__chunk_size = chunk_size
assert channels in [1, 2], '`channels` can be either 1 or 2. 1 for mono audio, 2 for stereo.'
self.__channels = channels
# audio resource and streams.
self.__audio = None
self.__input_stream = None
def __init__(self, path_prefix):
assert len(self.COLOR_GRADIENT_WHEEL) == self.FRAME_HEIGHT, \
"Need exactly {} colors in 'COLOR_GRADIENT_WHEEL'".format(self.FRAME_HEIGHT)
# Convert hex string (for easy programmer modification) to bytearrays in 'COLOR_GRADIENT_WHEEL'
for i, color_str in enumerate(self.COLOR_GRADIENT_WHEEL):
self.COLOR_GRADIENT_WHEEL[i] = np.frombuffer(
bytes.fromhex(color_str), dtype=np.uint8)
self.template = np.zeros(
(self.FRAME_HEIGHT, self.FRAME_WIDTH, self.NUM_COLOR_CHANNELS),
dtype=np.uint8)
self.pyaudio = PyAudio()
audio_device_index = settings(
path_prefix).get_selected_audio_device_index()
self.audio_device_info = self.pyaudio.get_device_info_by_index(
audio_device_index)
if self.audio_device_info[
'maxOutputChannels'] < self.NUM_AUDIO_CHANNELS:
raise Exception("Audio output device should be at least stereo.")
self.format = pyaudio.paInt16
self.sample_size = pyaudio.get_sample_size(self.format)
self.stream = None
self.raw_audio_frames = b'\x00' * (self.NUM_AUDIO_CHANNELS *
self.NUM_AUDIO_FRAMES_PER_BUFFER *
self.sample_size)
def activeListenToAllOptions(self, THRESHOLD=None, LISTEN=True, MUSIC=False):
"""
Records until a second of silence or times out after 12 seconds
Returns a list of the matching options or None
"""
RATE = 16000
CHUNK = 1024
LISTEN_TIME = 12
# check if no threshold provided
if THRESHOLD == None:
THRESHOLD = self.fetchThreshold()
self.speaker.play(jasperpath.data('audio', 'beep_hi.wav'))
# prepare recording stream
stream = self._audio.open(format=pyaudio.paInt16,
channels=1,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
frames = []
# increasing the range # results in longer pause after command generation
lastN = [THRESHOLD * 1.2 for i in range(30)]
for i in range(0, RATE / CHUNK * LISTEN_TIME):
data = stream.read(CHUNK)
frames.append(data)
score = self.getScore(data)
lastN.pop(0)
lastN.append(score)
average = sum(lastN) / float(len(lastN))
# TODO: 0.8 should not be a MAGIC NUMBER!
if average < THRESHOLD * 0.8:
break
self.speaker.play(jasperpath.data('audio', 'beep_lo.wav'))
# save the audio data
stream.stop_stream()
stream.close()
with tempfile.SpooledTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(RATE)
wav_fp.writeframes(''.join(frames))
wav_fp.close()
f.seek(0)
mode = TranscriptionMode.MUSIC if MUSIC else TranscriptionMode.NORMAL
transcribed = self.active_stt_engine.transcribe(f, mode=mode)
return transcribed
def __init__(self, rt):
super().__init__(rt)
self.chunk_size = self.config['chunk_size']
self.format = pyaudio.paInt16
self.sample_width = pyaudio.get_sample_size(self.format)
self.sample_rate = self.config['sample_rate']
self.channels = self.config['channels']
self.p = pyaudio.PyAudio()
self.stream = self.p.open(format=self.format,
channels=self.channels,
rate=self.sample_rate,
input=True,
frames_per_buffer=self.chunk_size)
self.talking_volume_ratio = self.config['talking_volume_ratio']
self.required_integral = self.config['required_noise_integral']
self.max_di_dt = self.config['max_di_dt']
self.noise_max_out_sec = self.config['noise_max_out_sec']
self.recording_timeout = self.config['recording_timeout']
self.energy_weight = 1.0 - pow(
1.0 - self.config['ambient_adjust_speed'],
self.chunk_size / self.sample_rate)
# For convenience
self.chunk_sec = self.chunk_size / self.sample_rate
self.av_energy = None
self.integral = 0
self.noise_level = 0
self._intercept = None
self._has_activated = False
self.engine = WakeWordService(
rt, self.on_activation) # type: WakeWordEnginePlugin
self.engine.startup()
def predict_file(dec,
pyaudio,
path,
frames,
args,
rate=16000,
format=pyaudio.paInt16,
save=False):
wf = wave.open(path, 'wb')
wf.setnchannels(1)
wf.setsampwidth(pyaudio.get_sample_size(format))
wf.setframerate(rate)
#this code works for only for pulseaudio
#wf.writeframes(b''.join(frames))
wf.writeframes(frames)
wf.close()
results = dec.predict_file(path,
feat_mode=args.feat_mode,
feat_dim=args.feat_dim,
three_d=args.three_d)
if save == False:
os.remove(path)
if args.predict_mode == 0:
task_outputs = dec.returnDiff(results)
elif args.predict_mode == 1:
task_outputs = dec.returnLabel(results)
else:
task_outputs = dec.returnClassDist(results)
return task_outputs
def save(self):
wf = wave.open(self._name, 'wb')
wf.setnchannels(self._channels)
wf.setsampwidth(pyaudio.get_sample_size(self._format))
wf.setframerate(self._rate)
wf.writeframes(b''.join(self._frames))
wf.close()
def sample_size(self) -> int:
"""
Get size of a single audio sample.
:return: Size in bytes
"""
return pyaudio.get_sample_size(self._sample_fmt)
def __init__(self, device_index = None): self.format = pyaudio.paInt16 self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format) self.CHUNK = 1024 self.audio = None self.stream = None
def __init__(self, args, CHUNK = 1024, FORMAT = pyaudio.paInt16, CHANNELS = 1, RATE = 16000, THRESHOLD = 2500, SILENCE_LIMIT = 2, PREV_AUDIO = 1):
"""
Initialization method for class AudioUtils.
Defines the constants needed throughout the program.
Keyword Arguments:
CHUNK {number} -- CHUNKS of bytes to read each time from mic (default: {1024})
FORMAT {[type]} -- [description] (default: {pyaudio.paInt16})
CHANNELS {number} -- [description] (default: {1})
RATE {number} -- [description] (default: {16000})
THRESHOLD {number} -- The threshold intensity that defines silence and noise signal (an int. lower than THRESHOLD is silence) (default: {2500})
SILENCE_LIMIT {number} -- Silence limit in seconds. The max ammount of seconds where
only silence is recorded. When this time passes the
recording finishes and the file is delivered. (default: {2})
PREV_AUDIO {number} -- Previous audio (in seconds) to prepend. When noise
is detected, how much of previously recorded audio is
prepended. This helps to prevent chopping the beggining
of the phrase. (default: {1})
"""
vc_logging.init_logger(level = args.log_level, verbose = args.verbose)
self.log = logging.getLogger("vc_logger")
self.WIDTH = pyaudio.get_sample_size(FORMAT)
self.CHUNK = CHUNK
self.FORMAT = FORMAT
self.CHANNELS = CHANNELS
self.RATE = RATE
self.THRESHOLD = THRESHOLD
self.SILENCE_LIMIT = SILENCE_LIMIT
self.PREV_AUDIO = PREV_AUDIO
self.audioQueue = Queue()
def readMic(utteranceToneQ, utteranceSpeechQ, audioInputDevice):
# setup
DEVICE_IP_HW = audioInputDevice # this usually is hw:2,0
# DEVICE_IP_HW = audioInput
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 16000
CHUNK = 4096
BASELINE_SECONDS = 3
CHECK_SILENCE_SECONDS = 1
UTTERANCE_SECONDS = 5
OUTPUT_DIR = os.path.join(os.path.dirname(os.path.realpath(__file__)),
"test")
try:
with noalsaerr():
p = pyaudio.PyAudio() # start the PyAudio class
# open stream with this device
stream = p.open(format=FORMAT,
channels=CHANNELS,
rate=RATE,
input_device_index=get_ip_device_index(
p, DEVICE_IP_HW),
input=True,
frames_per_buffer=CHUNK)
# THRESHOLD = getThreshold(stream, RATE, CHUNK, BASELINE_SECONDS) +3000 # just to be safe
THRESHOLD = 20000 # set for testing
print("________________________________________")
print("RECORDER -> Threshold : " + str(THRESHOLD))
print("________________________________________")
utteranceCount = 0
while (True):
utteranceData = getUtterance(stream, RATE, CHUNK, THRESHOLD,
CHECK_SILENCE_SECONDS,
UTTERANCE_SECONDS)
# print("-----------------------------------------------")
# set up the wav container to store the recorded 5 second utterances
wavFile = wave.open(
os.path.join(OUTPUT_DIR,
"mic_" + str(utteranceCount) + ".wav"), "w")
wavFile.setnchannels(CHANNELS)
wavFile.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wavFile.setframerate(RATE)
wavFile.writeframes(utteranceData)
wavFile.close()
# print("saved " + os.path.join(OUTPUT_DIR, "mic_" + str(utteranceCOunt) + ".wav"))
utteranceToneQ.put(utteranceData)
utteranceSpeechQ.put(utteranceData)
# print("AUDIO RECORDER -> Utterance " + str(utteranceCount) + " recorded")
# print("-----------------------------------------------")
utteranceCount += 1
except:
pass
def pyaudio_scb(self, rate, fmt=pyaudio.paInt16):
samp_size = pyaudio.get_sample_size(fmt)
maxint = (1 << (8*samp_size)) - 1
dtype = ['!', 'h', 'i', '!', 'l', '!', '!', '!', 'q'][samp_size]
def __callback(data, frames, time, status, self=self, rate=rate, maxint=maxint, dtype=dtype):
return struct.pack(dtype*frames, *[maxint*int(i) for i in self.data(frames, self.freq, rate)])
return __callback
def getAudio():
#set properties of audio
CHUNK = 2**11
RATE = 44100
CHANNELS = 2
#open stream
p = pyaudio.PyAudio()
stream = p.open(format=pyaudio.paInt16,
channels=1,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
##Listen to Microphone
listening = False
while (True):
frames = []
i = 0
peak = 0
while (True):
data = np.fromstring(stream.read(CHUNK), dtype=np.int16)
peak = np.average(np.abs(data)) * 2
#check if silent
if (peak < 600): ##do some tuning here boi
##Cut words and save words from microphone
if (listening):
##Save the audio to disk
filename = 'downloads/test' + str(random.randint(
0, 10000)) + '.wav'
wavefile = wave.open(filename, 'wb')
wavefile.setnchannels(1)
wavefile.setsampwidth(
pyaudio.get_sample_size(pyaudio.paInt16))
wavefile.setframerate(RATE)
wavefile.writeframes(b''.join(frames))
wavefile.close
frames = []
listening = False
#print("SAVED")
##Start a new thread
t = Thread(target=recognizeAudio, args=(filename, ))
t.start()
#print("Not listening")
else:
listening = True
#print("Listening")
#data_audio=stream.read(CHUNK)
frames.append(data)
##display words
stream.stop_stream()
stream.close()
p.terminate()
def save_audio(self, fp, frames):
sample_width = pyaudio.get_sample_size(self.format)
f = open_audio(fp, 'wb')
f.setsampwidth(sample_width)
f.setframerate(self.rate)
f.setnchannels(1)
f.writeframes(''.join(frames))
f.close()
def write_wav(frames, file_name):
waveFile1 = wave.open(file_name, 'wb')
waveFile1.setnchannels(1)
waveFile1.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
waveFile1.setframerate(44100)
waveFile1.writeframes(b''.join(frames))
waveFile1.close()
def save_file(self):
c=Configure()
wf = wave.open(self.output_file_name, 'wb')
wf.setnchannels(c.CHANNELS)
wf.setsampwidth(pyaudio.get_sample_size(c.FORMAT))
wf.setframerate(c.RATE)
wf.writeframes(b''.join(self.wave_data_after))
wf.close()
def save_audio(self, fp, frames):
sample_width = pyaudio.get_sample_size(self.format)
f = open_audio(fp, 'wb')
f.setsampwidth(sample_width)
f.setframerate(self.rate)
f.setnchannels(1)
f.writeframes(''.join(frames))
f.close()
def __callback(self, in_data, # recorded data if input=True; else None
frame_count, # number of frames
time_info, # dictionary
status_flags): # PaCallbackFlags
n = self.buffer.write_chunk(in_data, len(in_data))
self.dropped_samples += frame_count - (n / (self.channels * pyaudio.get_sample_size(self.format)))
self.rec_bytes += n
return None, pyaudio.paContinue
def __init__(self):
self.pa = pa = pyaudio.PyAudio()
self.stream = pa.open(format=self.FORMAT,
channels=self.CHANNELS,
rate=self.RATE,
input=True,
frames_per_buffer=self.CHUNK)
self.packer = Pcm2Wave(self.RATE, pyaudio.get_sample_size(self.FORMAT),
self.CHANNELS)
def __init__(self):
self._sound = pyaudio.PyAudio()
self.format = pyaudio.paInt16
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.stream = self._sound.open(rate=self.SAMPLE_RATE,
channels=1,
format=self.format,
input=True,
frames_per_buffer=self.CHUNK)
def write_chunks(path, frames, rate=44100, channels=2, format=pyaudio.paInt16):
""" write audio to hard disk """
wf = wave.open(path, "wb")
wf.setnchannels(channels)
wf.setsampwidth(pyaudio.get_sample_size(format))
wf.setframerate(rate)
wf.writeframes(b"".join(frames))
wf.close()
def ActiveListening(this):
threshold = None
textout.SystemPrint("Started to listen actively")
#RECORD A WAV FILE, CUTOFF AT 12s OR FALLS BELOW THRESHOLD
#SEND TO WITAI
#RECIEVE INPUT
#SEND TO PROCESSOR AND RETURN INTENT
#FIND ACTION.
RATE = 16000
CHUNK = 1024
p = pyaudio.PyAudio()
stream = p.open(format=pyaudio.paInt16,
channels=1,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
frames = []
lastN = [
144 * 1.2 for i in range(15)
] #changing array length will determine if average will change faster or not
for i in range(0, int(RATE / CHUNK *
5)): #RATE(16000) / CHUNK(1024) * TIME (12s)
data = stream.read(CHUNK)
frames.append(data)
score = getScore(data)
lastN.pop(0)
lastN.append(score)
average = sum(lastN) / float(len(lastN))
print(str(average))
if average < 144 - 20:
textout.SystemPrint("Listening stopped, below threshold.")
break
textout.SystemPrint("Listening Timeout!")
stream.stop_stream()
stream.close()
p.terminate()
with tempfile.NamedTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(16000)
wav_fp.writeframes(b''.join(frames))
wav_fp.close()
f.seek(0)
transcriber.TranscribeAudiofile(this, f)
textout.SystemPrint("Stopped listening actively")
def __init__(self, device_index = None):
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.RATE = 16000 # sampling rate in Hertz
self.CHANNELS = 1 # mono audio
self.CHUNK = 1024 # number of frames stored in each buffer
self.audio = None
self.stream = None
def write_wave(arrs, filepath,
sample_width=pyaudio.get_sample_size(PYAUDIO_FORMAT),
channels=CHANNELS, sample_rate=SAMPLE_RATE):
f = wave.open(filepath, 'w')
f.setnchannels(channels)
f.setsampwidth(sample_width)
f.setframerate(sample_rate)
for arr in _structpack(arrs):
f.writeframes(arr)
f.close()
def __init__(self, device_index=None):
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.RATE = 16000 # sampling rate in Hertz
self.CHANNELS = 1 # mono audio
self.CHUNK = 1024 # number of frames stored in each buffer
self.audio = None
self.stream = None
def __init__(self, device_index = None):
self.device_index = device_index
self.format = pyaudio.paInt16
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.RATE = 44100
self.CHANNELS = 1
self.CHUNK = 2205
self.audio = None
self.stream = None
def passiveListen(self, persona):
rate = self.PASSIVE_RATE
chunk = self.PASSIVE_CHUNK
LISTEN_TIME = 10
THRESHOLD = self.THRESHOLD_MULTIPLIER * self.fetchThreshold()
stream = self._audio.open(format=pyaudio.paInt16,
channels=1,
rate=rate,
input=True,
frames_per_buffer=chunk)
frames = []
didDetect = False
for i in range(0, rate / chunk * LISTEN_TIME):
data = stream.read(chunk)
frames.append(data)
score = self.getScore(data)
if score > THRESHOLD:
didDetect = True
break
# no use continuing if no flag raised
if not didDetect:
print "No disturbance detected"
stream.stop_stream()
stream.close()
return (None, None)
# cutoff any recording before this disturbance was detected
frames = frames[-20:]
# otherwise, let's keep recording for few seconds and save the file
DELAY_MULTIPLIER = 1.5
for i in range(0, rate / chunk * DELAY_MULTIPLIER):
data = stream.read(chunk)
frames.append(data)
with tempfile.NamedTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(rate)
wav_fp.writeframes(''.join(frames))
wav_fp.close()
f.seek(0)
# check if PERSONA was said
transcribed = self.passive_transcribe(f)
if any(persona in phrase for phrase in transcribed):
return (THRESHOLD, persona)
return (False, transcribed)
def activeListenToAllOptions(self, THRESHOLD=None, LISTEN=True,
MUSIC=False):
"""
Records until a second of silence or times out after 12 seconds
Returns a list of the matching options or None
"""
# check if no threshold provided
if THRESHOLD is None:
THRESHOLD = self.fetchThreshold()
playing = 'playing' in check_output(['mpc', '-h', '[email protected]']).split('\n')[1]
if playing:
call(['mpc', '-h', '[email protected]', 'pause'])
# self.speaker.play(jasperpath.data('audio', 'beep_hi.wav'))
self.say(random.choice(BEFORE))
frames = []
# increasing the range # results in longer pause after command
# generation
lastN = [THRESHOLD * 1.2 for i in range(30)]
for i in range(0, RATE / CHUNK * LISTEN_TIME):
data = self.queue.get()
frames.append(data)
score = self.getScore(data)
lastN.pop(0)
lastN.append(score)
average = sum(lastN) / float(len(lastN))
# TODO: 0.8 should not be a MAGIC NUMBER!
if average < THRESHOLD * 0.8:
break
# self.speaker.play(jasperpath.data('audio', 'beep_lo.wav'))
self.say(random.choice(AFTER))
if playing:
call(['mpc', '-h', '[email protected]', 'play'])
# save the audio data
with tempfile.SpooledTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(RATE)
wav_fp.writeframes(''.join(frames))
wav_fp.close()
f.seek(0)
return self.active_stt_engine.transcribe(f)
def __init__(self, device_index = None):
assert device_index is None or isinstance(device_index, int), "Device index must be None or an integer"
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.RATE = 16000 # sampling rate in Hertz
self.CHANNELS = 1 # mono audio
self.CHUNK = 1024 # number of frames stored in each buffer
self.audio = None
self.stream = None
def __init__(self, hmm_type=1, vad_threshold=3.5, pl_window=10, wip=1e-4,
silprob=0.3, bestpath=True, remove_dc=True, do_keyphrase=False, keyphrase="NAVSA", kws_threshold=1e-4):
self.CHUNK = 1024
self.RATE = 16000
self.FORMAT = pyaudio.paInt16
self.CHANNELS = 1
self.RUN_SECONDS = 1000
self.sampwidth = pyaudio.get_sample_size(self.FORMAT)
self.do_trigger = False
# Create a decoder with certain model
self.config = Decoder.default_config()
if hmm_type == 0:
self.config.set_string('-hmm', '/usr/local/share/pocketsphinx/model/en-us/en-us')
elif hmm_type == 1:
self.config.set_string('-hmm', 'model/cmusphinx-en-us-5.2')
self.config.set_string('-dict', 'model/7705.dic')
if do_keyphrase:
self.config.set_string('-keyphrase', keyphrase)
self.config.set_float('-kws_threshold', kws_threshold)
else:
self.config.set_string('-lm', 'model/7705.lm')
self.config.set_string('-logfn', '/dev/null')
self.config.set_string('-debug', '1')
# http://cmusphinx.sourceforge.net/wiki/pocketsphinxhandhelds
self.config.set_boolean('-bestpath', bestpath) # default is true
self.config.set_float('-vad_threshold', vad_threshold) # default is 2
self.config.set_float("-pl_window", pl_window) # default is 5, range is 0 to 10
self.config.set_float('-wip', wip) # 0.005 Silence word transition probability
self.config.set_float('-silprob', silprob) # 0.65 Word insertion penalty
self.config.set_string('-remove_dc', 'yes' if remove_dc else 'no')
self.decoder = Decoder(self.config)
self.deque_time = deque(maxlen=20)
self.deque_mean = deque(maxlen=50)
self.mic = False
self.wf = None
self.vad = False
self.rec_trigger = False
self.rec_frames = []
self.sec_since_kw = 999.9
self.sec_since_vad = 999.9
def read(self, buf, source_channels):
source_sample_width = pyaudio.get_sample_size(pyaudio.paInt16) * source_channels
audio = buf[3:]
try:
# sometimes the data received is incomplete so reusing state
# data from ratecv() sometimes results in errors
(audio, _) = audioop.ratecv(audio, source_sample_width, source_channels, 48000, self.listener.sample_rate, None)
audio = audioop.tomono(audio, self.listener.sample_width, 0.5, 0.5)
self.listener.read(audio)
except audioop.error, e:
logger.warn("Error preparing sample", exc_info=True)
def __init__(self):
self.CHUNK = 1024
self.FORMAT = pyaudio.paInt16
self.CHANNELS = 2
self.RATE = 44100
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.FORMAT)
self.energy_threshold = 1500
self.pause_threshold = 0.8
self.quiet_duration = 0.5
self.seconds_per_buffer = self.CHUNK / self.RATE
self.pause_buffer_count = math.ceil(self.pause_threshold / self.seconds_per_buffer)
self.quiet_buffer_count = math.ceil(self.quiet_duration / self.seconds_per_buffer)
def __init__(self, device_index = None):
assert device_index is None or isinstance(device_index, int), "Device index must be None or an integer"
if device_index is not None: # ensure device index is in range
audio = pyaudio.PyAudio(); count = audio.get_device_count(); audio.terminate() # obtain device count
assert 0 <= device_index < count, "Device index out of range"
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.RATE = 16000 # sampling rate in Hertz
self.CHANNELS = 1 # mono audio
self.CHUNK = 1024 # number of frames stored in each buffer
self.audio = None
self.stream = None
def __init__(self, device_index = None, sample_rate = 16000, chunk_size = 1024):
assert device_index is None or isinstance(device_index, int), "Device index must be None or an integer"
if device_index is not None: # ensure device index is in range
audio = pyaudio.PyAudio(); count = audio.get_device_count(); audio.terminate() # obtain device count
assert 0 <= device_index < count, "Device index out of range"
assert isinstance(sample_rate, int) and sample_rate > 0, "Sample rate must be a positive integer"
assert isinstance(chunk_size, int) and chunk_size > 0, "Chunk size must be a positive integer"
self.device_index = device_index
self.format = pyaudio.paInt16 # 16-bit int sampling
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format) # size of each sample
self.SAMPLE_RATE = sample_rate # sampling rate in Hertz
self.CHUNK = chunk_size # number of frames stored in each buffer
self.audio = None
self.stream = None
def _start_wave_file(self):
''' Open a new wave file for writing in the specified output directory.
Files are named using the UTC timestamp and a c_ prefix
:return:
'''
# figure out the new filename
now = datetime.utcnow()
file_name = "c_{0}.wav".format(now.strftime("%Y%m%dT%H%M%SZ"))
file_path = os.path.join(self.output_directory, file_name)
self._current_wave_file = wave.open(file_path, mode='wb')
self._current_wave_file.setnchannels(self._num_channels)
self._current_wave_file.setsampwidth(pyaudio.get_sample_size(self._sample_format))
self._current_wave_file.setframerate(self._sample_rate)
self.log.info("Starting continuous recording: " + file_name)
def __init__(self, length, n_channels=1):
"""
:param length: length of the buffer in samples
:param n_channels: Number of channels present in the audio samples.
"""
self._n_channels = n_channels
self._length = length * self._n_channels # Length in samples
self._sample_size = pyaudio.get_sample_size(pyaudio.paFloat32)
self._sample_format = self._format[pyaudio.paFloat32]
# Intialize state variables
self._size = 0
self._write_start = 0
self._read_start = 0
# Instantiate buffer
self._buffer = np.zeros(self._length, dtype=np.float32)
print self._buffer.shape
# Setup blocking interface events
self._setup_events()
def store_data(self, indata):
print('store_data(): len(frames): ' + str(len(self.frames)))
self.frames.append(indata)
if len(self.frames) >= self.chunks_per_file:
# TODO: Move file I/O to an async thread
wave_file = wave.open(self.directory +
str(self.file_num),
'wb')
wave_file.setnchannels(self.channels)
wave_file.setsampwidth(pyaudio.get_sample_size(self.audio_format))
wave_file.setframerate(self.rate)
wave_file.writeframes(b''.join(self.frames[:self.chunks_per_file]))
wave_file.close()
self.file_num += 1
# move frame
# TODO: Use a circular buffer or other appropriate data structure,
# to avoid doing this move in the callback thread.
temp_frame_buffer = self.frames[self.chunks_per_file:]
self.frames = temp_frame_buffer
def __init__(self, fmt=pa.paInt16, rate=44100):
self.fmt = fmt
self.rate = rate
self.amp = 1 if fmt == pa.paFloat32 else 128**pa.get_sample_size(
fmt) / 2 - 1
if fmt == pa.paFloat32:
self.dtype = np.float32 # pylint: disable=E1101
elif fmt == pa.paInt32:
self.dtype = np.Int32 # pylint: disable=E1101
elif fmt == pa.paInt24:
raise SampleFormatNotSuportedException('paInt24')
elif fmt == pa.paInt16:
self.dtype = np.int16
elif fmt == pa.paInt8:
self.dtype = np.int8
elif fmt == pa.paUInt8:
self.dtype = np.uint8
else:
raise SampleFormatNotSuportedException('paCustomFormat')
self.reset()
self.data = b''
def trigger_recording(self, filename=None):
'''
:param filename: If specified, this filename will be used to save the triggered recording. If not, a
filename will be generated by appending the UTC timestamp to t_.
:return:
'''
# figure out the new filename, if not specified
if filename is None:
now = datetime.utcnow()
filename = "t_{0}.wav".format(now.strftime("%Y%m%dT%H%M%SZ"))
file_path = os.path.join(self.output_directory, filename)
triggered_wave_file = wave.open(file_path, mode='wb')
triggered_wave_file.setnchannels(self._num_channels)
triggered_wave_file.setsampwidth(pyaudio.get_sample_size(self._sample_format))
triggered_wave_file.setframerate(self._sample_rate)
triggered_wave_file.writeframes(''.join(self._ring_buffer))
triggered_wave_file.close()
self.log.info("Wrote triggered file: " + filename)
def __init__(self):
"""
Initiates the pocketsphinx instance.
Arguments:
acive_stt_engine -- performs STT while Jasper is in active listen mode
"""
self._logger = logging.getLogger(__name__)
self._logger.info("Initializing PyAudio. ALSA/Jack error messages " +
"that pop up during this process are normal and " +
"can usually be safely ignored.")
self._audio = pyaudio.PyAudio()
self._logger.info("Initialization of PyAudio completed.")
self.format = pyaudio.paInt16
self.SAMPLE_WIDTH = pyaudio.get_sample_size(self.format)
self.energy_threshold = 300 # minimum audio energy to consider for recording
self.dynamic_energy_threshold = True
self.dynamic_energy_adjustment_damping = 0.15
self.dynamic_energy_ratio = 1.5
self.pause_threshold = 0.8 # seconds of non-speaking audio before a phrase is considered complete
self.phrase_threshold = 0.3 # minimum seconds of speaking audio before we consider the speaking audio a phrase - values below this are ignored (for filtering out clicks and pops)
self.non_speaking_duration = 0.5 # seconds of non-speaking audio to keep on both sides of the recording
from __future__ import print_function, division
import pyaudio, os, sys, time, inspect
FILE_PATH = os.path.dirname(inspect.getfile(inspect.currentframe()))
##########################################################################
# SAMPLING PARAMETERS
FRAME_RATE = 44100 # The sound card sampling rate in Hz
DOWNSAMPLED_RATE = 16000 # Hz (MIT REDD uses 15kHz but 16kHz is a standard
# rate and so increases compatibility)
RECORD_SECONDS = 1 # Seconds to record per queue item
N_CHANNELS = 2 # one for voltage, one for current
FRAMES_PER_BUFFER = 1024
SAMPLE_FORMAT = pyaudio.paInt16
SAMPLE_WIDTH = pyaudio.get_sample_size(SAMPLE_FORMAT)
N_READS_PER_QUEUE_ITEM = int(round(FRAME_RATE / FRAMES_PER_BUFFER
* RECORD_SECONDS))
##########################################################################
# MAINS PARAMETERS
MAINS_HZ = 50
SAMPLES_PER_MAINS_CYCLE = FRAME_RATE / MAINS_HZ
PHASE_DIFF_TOLERANCE = SAMPLES_PER_MAINS_CYCLE / 4
SAMPLES_PER_DEGREE = SAMPLES_PER_MAINS_CYCLE / 360
##########################################################################
# PORT TO BROADCAST MEASURES ON
BROADCAST_PORT = 5556
##########################################################################
import threading import math import matplotlib.pyplot as plt from mpl_toolkits.mplot3d import Axes3D from pa_tools.audiohelper import AudioHelper from pa_tools.audiobuffer import AudioBuffer from pa_tools.stftmanager import StftManager from pa_tools.audiolocalizer import AudioLocalizer from pa_tools.distributionlocalizer import DistributionLocalizer # Setup constants SAMPLE_TYPE = pyaudio.paFloat32 DATA_TYPE = np.float32 SAMPLE_SIZE = pyaudio.get_sample_size(SAMPLE_TYPE) SAMPLE_RATE = 44100 FRAMES_PER_BUF = 4096 # For 44100 Fs, be careful going over 4096, loud sounds may occur... FFT_LENGTH = FRAMES_PER_BUF WINDOW_LENGTH = FFT_LENGTH HOP_LENGTH = WINDOW_LENGTH / 2 NUM_CHANNELS_IN = 7 NUM_CHANNELS_OUT = 2 N_THETA = 20 N_PHI = N_THETA / 2 PLOT_CARTES = False PLOT_POLAR = True EXTERNAL_PLOT = False PLAY_AUDIO = True TIMEOUT = 1 # Setup mics
def passiveListen(self, PERSONA):
"""
Listens for PERSONA in everyday sound. Times out after LISTEN_TIME, so
needs to be restarted.
"""
THRESHOLD_MULTIPLIER = 1.8
RATE = 16000
CHUNK = 1024
# number of seconds to allow to establish threshold
THRESHOLD_TIME = 1
# number of seconds to listen before forcing restart
LISTEN_TIME = 10
# prepare recording stream
stream = self._audio.open(format=pyaudio.paInt16,
channels=1,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
# stores the audio data
frames = []
# stores the lastN score values
lastN = [i for i in range(30)]
# calculate the long run average, and thereby the proper threshold
for i in range(0, RATE / CHUNK * THRESHOLD_TIME):
data = stream.read(CHUNK)
frames.append(data)
# save this data point as a score
lastN.pop(0)
lastN.append(self.getScore(data))
average = sum(lastN) / len(lastN)
# this will be the benchmark to cause a disturbance over!
THRESHOLD = average * THRESHOLD_MULTIPLIER
# save some memory for sound data
frames = []
# flag raised when sound disturbance detected
didDetect = False
# start passively listening for disturbance above threshold
for i in range(0, RATE / CHUNK * LISTEN_TIME):
data = stream.read(CHUNK)
frames.append(data)
score = self.getScore(data)
if score > THRESHOLD:
didDetect = True
break
# no use continuing if no flag raised
if not didDetect:
print "No disturbance detected"
stream.stop_stream()
stream.close()
return (None, None)
# cutoff any recording before this disturbance was detected
frames = frames[-20:]
# otherwise, let's keep recording for few seconds and save the file
DELAY_MULTIPLIER = 1
for i in range(0, RATE / CHUNK * DELAY_MULTIPLIER):
data = stream.read(CHUNK)
frames.append(data)
# save the audio data
stream.stop_stream()
stream.close()
with tempfile.NamedTemporaryFile(mode='w+b') as f:
wav_fp = wave.open(f, 'wb')
wav_fp.setnchannels(1)
wav_fp.setsampwidth(pyaudio.get_sample_size(pyaudio.paInt16))
wav_fp.setframerate(RATE)
wav_fp.writeframes(''.join(frames))
wav_fp.close()
f.seek(0)
# check if PERSONA was said
json = self.passive_stt_engine.transcribe(f)
if json['_text'] and json['outcomes'][0]:
transcribed = []
transcribed.append(json['outcomes'][0]['intent'].upper())
if any(PERSONA in phrase for phrase in transcribed):
return (THRESHOLD, PERSONA)
else:
return (False, transcribed)
return (False, '')