def _non_block_listen(self, hello, lvl, file_path):
max_play_time = 120 # максимальное время воспроизведения приветствия
max_wait_time = 10 # ожидание после приветствия
r = sr.Recognizer()
mic = sr.Microphone(device_index=self.get_mic_index())
with mic as source: # Слушаем шум 1 секунду, потом распознаем, если раздажает задержка можно закомментировать.
energy_threshold = self.energy.correct(r, source)
if self._cfg.gts('alarmtts') and not hello:
self._play.play(self._cfg.path['dong'], lvl)
start_wait = time.time()
if file_path:
self._play.play(file_path, lvl)
# Начинаем фоновое распознавание голосом после того как запустился плей.
listener = NonBlockListener(r=r, source=mic, phrase_time_limit=self._cfg.gts('phrase_time_limit', 15))
if file_path:
while listener.work() and self._play.really_busy() and \
time.time() - start_wait < max_play_time and self._work:
# Ждем пока время не выйдет, голос не распознался и файл играет
time.sleep(0.01)
self._play.quiet()
start_wait2 = time.time()
while listener.work() and time.time() - start_wait2 < max_wait_time and self._work:
# ждем еще секунд 10
time.sleep(0.01)
record_time = time.time() - start_wait
listener.stop()
return listener.audio, record_time, energy_threshold
def __init__(self, cfg, play_, log, tts):
self.log = log
self._cfg = cfg
self.sys_say = Phrases(log, cfg)
self._lock = threading.Lock()
self._work = True
self._play = play_
self._tts = tts
self.energy = utils.EnergyControl(cfg, play_)
self._recognizer = sr.Recognizer()
try:
self.max_mic_index = len(sr.Microphone().list_microphone_names()) - 1
except OSError as e:
self.log('Error get list microphones: {}'.format(e), logger.CRIT)
self.max_mic_index = -2
def start(self):
self._terminate = False
while not self._interrupted():
with sr.Microphone() as source:
r = sr.Recognizer(self._interrupted, self._cfg.gts('sensitivity'), self._hotword_callback)
energy_threshold = self._stt.energy.correct(r, source)
try:
adata = r.listen(source, 5, self._cfg.gts('phrase_time_limit'),
(self._cfg.path['home'], self._cfg.path['models_list']))
except sr.WaitTimeoutError:
self._stt.energy.set(None)
continue
except sr.Interrupted:
self._stt.energy.set(energy_threshold)
continue
model = r.get_model
if model < 1:
continue
self._adata_parse(adata, model, energy_threshold)
def _voice_record(self, hello: str, save_to: str, convert_rate=None, convert_width=None):
lvl = 5 # Включаем монопольный режим
file_path = self._tts(hello)()
self._play.say(file_path, lvl, True, is_file=True, blocking=120)
with sr.Microphone(device_index=self.get_mic_index()) as source:
r = sr.Recognizer()
r.adjust_for_ambient_noise(source)
self._play.play(self._cfg.path['ding'], lvl, blocking=3)
record_time = time.time()
try:
adata = r.listen(source=source, timeout=5, phrase_time_limit=10)
except sr.WaitTimeoutError as e:
return str(e)
if time.time() - record_time < 0.5:
return LNG['err_voice_record']
try:
with open(save_to, "wb") as f:
f.write(adata.get_wav_data(convert_rate, convert_width))
except IOError as err:
return str(err)
else:
return None
def _block_listen(self, hello, lvl, file_path, self_call=False):
with sr.Microphone(device_index=self.get_mic_index()) as source:
r = sr.Recognizer()
if self._cfg.gts('alarmtts') and not hello:
self._play.play(self._cfg.path['dong'], lvl, wait=0.01, blocking=2)
if file_path:
self._play.play(file_path, lvl, wait=0.01, blocking=120)
energy_threshold = self.energy.correct(r, source)
record_time = time.time()
try:
audio = r.listen(source, timeout=10, phrase_time_limit=self._cfg.gts('phrase_time_limit', 15))
except sr.WaitTimeoutError:
audio = None
record_time = time.time() - record_time
if record_time < 0.5 and not self_call:
# Если от инициализации микрофона до записи прошло больше 20-35 сек, то запись ломается
# Игнорируем полученную запись и запускаем новую, без приветствий
self.log('Long ask fix!', logger.DEBUG)
return self._block_listen(hello=True, lvl=lvl, file_path=None, self_call=True)
else:
return audio, record_time, energy_threshold