Voice LLM Application - Experiment 01: Voice Recognition

Experiment Preparation:

  1. Install ALSA tools (for recording and playback)

sudo apt-get install alsa-utils

  1. Install required dependencies
  • pip install -r requirements.txt
  • python -m pip install websocket-client
  • sudo apt-get update && sudo apt-get install -y ffmpeg

  1. Register iFlytek account

  2. Login to iFlytek Open Platform: https://www.xfyun.com.cn

  3. Click to enter console

  4. Register and login

  5. Create application

  6. Enable speech recognition - speech dictation service

  7. Get API ID, API Secret, API Key, and speech dictation interface URL

  8. Save the four pieces of information (to be filled in config.py later)

Experiment Steps:

  1. Check voice module (ensure voice module is connected to RDK mainboard and speaker)

Terminal: arecord -l # Identify microphone card number and device number

Terminal: aplay -l # Check speaker/output devices

Terminal: sudo arecord -f S16_LE -r 16000 -c 1 -d 5 /tmp/test_mic.wav # Use default device to record for 5 seconds

Terminal: aplay /tmp/test_mic.wav # Play audio

  1. Fill in API ID, API Secret, API Key, and speech dictation interface URL in config.py
TOOL
  1. cd AI_online_voice # Enter package
  2. python examples/01_voice_chat.py # Run example program, enter r to start testing

Terminal operation effect:

TOOL

Experiment Effect: Start recording (default 5 seconds), after recording play audio, upload audio to iFlytek speech dictation large model, return recognition result to Linux terminal