Skip to content

troubleshooting.md

Latest commit

 

History

History
204 lines (159 loc) · 5.44 KB

troubleshooting.md

File metadata and controls

204 lines (159 loc) · 5.44 KB
Raw

Troubleshooting

Audio Issues

ReSpeaker HAT Not Detected

If the ReSpeaker HAT is not detected as an audio device:

  1. Check hardware connections:

    • Ensure the HAT is properly seated on the GPIO pins
    • Verify the HAT has its own power supply connected
    • Check that the case isn't putting pressure on the HAT

  2. Reboot required:

    • After adding device tree overlays, a reboot is mandatory
    • Run: sudo reboot

Audio Playback Issues

Problem: No sound or very quiet audio output

Solution - Adjust Playback Volume:

# Open alsamixer for volume adjustment
alsamixer

Recommended Playback Settings:

  1. Press F6 to select output device
  2. Select 'seeed2micvoicec' from the device list
  3. Set PCM to 100% (use arrow keys or mouse wheel)
  4. Set Line DAC to 100% (use arrow keys or mouse wheel)
  5. Press ESC to exit

Audio Recording Issues

Problem: Microphone not working or recording too quiet/loud

Solution - Adjust Recording Volume:

# Open alsamixer for recording volume adjustment
alsamixer

Recommended Recording Settings:

  1. Press F6 to select input device
  2. Select 'seeed2micvoicec' from the device list
  3. Press F4 to filter capture settings
  4. Set Capture (PGA) to 30% (recommended - avoid going too high to prevent clipping)
  5. Press ESC to exit

Audio Device Detection and Voice Command Recognition

Check available audio devices:

# List playback devices
aplay -l

# List recording devices
arecord -l

Expected output for ReSpeaker HAT:

card 2: seeed2micvoicec [seeed2micvoicec], device 0: ...

If device not found:

  1. Ensure you've rebooted after adding device tree overlays
  2. Check /boot/firmware/config.txt contains: 
    [all]
# ReSpeaker HAT
dtoverlay=respeaker-2mic-v2_0-overlay
dtoverlay=i2s-mmap
dtparam=spi=on

Poor Voice Command Recognition:

Problem: Voice commands are not being recognized accurately or consistently

Symptoms:

  • Commands are misinterpreted or not understood
  • System responds to unintended phrases
  • Recognition works inconsistently
  • Background noise interferes with command detection

Troubleshooting Steps:

  1. Adjust Audio Preprocessing Settings:

    • Review and modify preprocessing parameters in your configuration file
    • Experiment with different noise reduction and gain settings
    • Fine-tune voice activity detection thresholds

    See config.md for more details.

  2. Enable Audio Logging for Analysis:

    • Enable audio_logs_path in your configuration. This will record all voice commands for quality review
    • Review recorded audio files to assess command clarity
    • Check for background noise, speech volume, and pronunciation issues
    • Identify patterns in failed recognition attempts

    See config.md for more details.

  3. Use Development Tools for Recognition Analysis:

    # Run development tools to analyze recognition patterns
make dev-tools
    • Examine what phrases the system is actually recognizing
    • Compare expected vs. actual transcriptions
    • Identify common recognition errors and patterns

  4. Disable Speech2Text Streaming for Better Accuracy:

    • If recognition accuracy is poor despite other adjustments, try disabling streaming mode
    • This will add approximately 500ms of latency but may significantly improve accuracy
    • In your configuration file, set speech2text.streaming to false:
    "speech2text": {
  "streaming": false,
  ...
}
    • Restart the Yova service after making this change:
    sudo systemctl restart yova.service

Additional Recommendations:

  • Speak clearly and at a consistent volume
  • Minimize background noise when giving commands
  • Use consistent phrasing for voice commands
  • Consider adjusting microphone sensitivity if commands are too quiet or loud
  • Test recognition in different acoustic environments

ALSA Configuration Issues

To ensure Yova runs correctly as a system service, you must configure ALSA with a global /etc/asound.conf file. This allows audio access for background services (not just your user session).

Check current ALSA config:

cat /etc/asound.conf

Expected configuration:

pcm.!default {
    type plug
    slave.pcm "hw:2,0"
    slave.rate 16000
}
ctl.!default {
    type hw
    card 2
}

Note: The card and device numbers (2,0) may vary. Use the numbers from aplay -l output.

Audio Testing Commands

Test playback:

# Download test audio file
curl -L -o test_sound.wav https://github.com/jamro/yova/raw/refs/heads/main/yova_shared/assets/test_sound.wav

# Play test sound (replace X,Y with your card,device numbers)
aplay -D "plughw:2,0" test_sound.wav

Test recording:

# Record 5 seconds of audio
arecord -D "plughw:2,0" -f S16_LE -c1 -r 16000 -d 5 rec_test.wav

# Play back recorded audio
aplay -D "plughw:2,0" rec_test.wav

Service Issues

Check YOVA Service Status

# Check if service is running
sudo systemctl status yova.service

# Check if service is enabled (starts on boot)
sudo systemctl is-enabled yova.service

# View service logs
sudo journalctl -u yova.service -f

Restart YOVA Service

# Restart service (after configuration changes)
sudo systemctl restart yova.service

# Start service if stopped
sudo systemctl start yova.service

# Stop service
sudo systemctl stop yova.service