Files
pepperplus-ri/src/robot_interface/endpoints/audio_sender.py
2026-01-14 14:26:38 +00:00

99 lines
3.2 KiB
Python

from __future__ import unicode_literals # So that `logging` can use Unicode characters in names
import threading
import logging
import pyaudio
import zmq
from robot_interface.endpoints.socket_base import SocketBase
from robot_interface.state import state
from robot_interface.utils.microphone import choose_mic
from robot_interface.core.config import settings
logger = logging.getLogger(__name__)
class AudioSender(SocketBase):
"""
Audio sender endpoint, responsible for sending microphone audio data.
:param zmq_context: The ZeroMQ context to use.
:type zmq_context: zmq.Context
:param port: The port to use.
:type port: int
:ivar thread: Thread used for sending audio.
:vartype thread: threading.Thread | None
:ivar audio: PyAudio instance.
:vartype audio: pyaudio.PyAudio | None
:ivar microphone: Selected microphone information.
:vartype microphone: dict | None
"""
def __init__(self, zmq_context, port=settings.agent_settings.audio_sender_port):
super(AudioSender, self).__init__(str("audio")) # Convert future's unicode_literal to str
self.create_socket(zmq_context, zmq.PUB, port)
self.thread = None
try:
self.audio = pyaudio.PyAudio()
self.microphone = choose_mic(self.audio)
except IOError as e:
logger.warning("PyAudio is not available.", exc_info=e)
self.audio = None
self.microphone = None
def start(self):
"""
Start sending audio in a different thread.
Will not start if no microphone is available.
"""
if not self.microphone:
logger.info("Not listening: no microphone available.")
return
logger.info("Listening with microphone \"{}\".".format(self.microphone["name"]))
self.thread = threading.Thread(target=self._stream)
self.thread.start()
def wait_until_done(self):
"""
Wait until the audio thread is done.
Will block until `state.exit_event` is set. If the thread is not running, does nothing.
"""
if not self.thread: return
self.thread.join()
self.thread = None
def _stream(self):
"""
Internal method to continuously read audio from the microphone and send it over the socket.
"""
audio_settings = settings.audio_config
chunk = audio_settings.chunk_size # 320 at 16000 Hz is 20ms, 512 is required for Silero-VAD
# Docs say this only raises an error if neither `input` nor `output` is True
stream = self.audio.open(
format=pyaudio.paFloat32,
channels=audio_settings.channels,
rate=audio_settings.sample_rate,
input=True,
input_device_index=self.microphone["index"],
frames_per_buffer=chunk,
)
try:
while not state.exit_event.is_set():
data = stream.read(chunk)
if (state.is_speaking): continue # Do not send audio while the robot is speaking
self.socket.send(data)
except IOError as e:
logger.error("Stopped listening: failed to get audio from microphone.", exc_info=e)
finally:
stream.stop_stream()
stream.close()