47 lines
1.1 KiB
Python
47 lines
1.1 KiB
Python
import struct
|
|
|
|
import pvporcupine
|
|
import pyaudio
|
|
import speech_recognition as sr
|
|
|
|
from jarvis.utils import server_utils
|
|
|
|
wake_word_handler = pvporcupine.create(keywords=['jarvis'])
|
|
|
|
|
|
def wake_word_listening():
|
|
pa = pyaudio.PyAudio()
|
|
|
|
audio_stream = pa.open(
|
|
rate=wake_word_handler.sample_rate,
|
|
channels=1,
|
|
format=pyaudio.paInt16,
|
|
input=True,
|
|
frames_per_buffer=wake_word_handler.frame_length
|
|
)
|
|
|
|
while True:
|
|
pcm = audio_stream.read(wake_word_handler.frame_length)
|
|
pcm = struct.unpack_from("h" * wake_word_handler.frame_length, pcm)
|
|
|
|
keyword_index = wake_word_handler.process(pcm)
|
|
|
|
if keyword_index >= 0:
|
|
print("Recognized")
|
|
record()
|
|
|
|
|
|
def record():
|
|
# obtain audio from the microphone
|
|
r = sr.Recognizer()
|
|
with sr.Microphone() as source:
|
|
print("Say something!")
|
|
r.adjust_for_ambient_noise(source=source, duration=0.5)
|
|
audio = r.listen(source, timeout=2, phrase_time_limit=5)
|
|
|
|
server_utils.send_record_to_server(audio.frame_data)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
wake_word_listening()
|