/snowboyPi

snowboy setup on raspberry pi

Primary LanguagePython

snowboyPi

snowboy setup on raspberry pi

Start with a fresh install of Raspbian (Lite or Regular, this guide assumes Lite)

  1. Update
sudo apt update && sudo apt -y upgrade && sudo apt-get -y auto-remove && sudo reboot
  1. Install dependencies:
sudo apt -y install python-pyaudio python3-pyaudio sox python3-pip python-pip libatlas-base-dev
  1. Install PortAudio’s Python bindings:

python2

sudo pip install pyaudio

python3

sudo pip3 install pyaudio
  1. Create ~/.asoundrc with correct hw settings. use aplay -l & arecord -l to find out hw cards. "card 0, device 0" is "hw:0,0" In the attached sample, playback is through onboard jack & input is through usb mic
  2. If using RESTful API Calls via python script (per snowboy instructions) need to install "requests" module for python:
pip install requests
  1. Download pre-packaged Snowboy binaries and their Python wrappers for Raspberry Pis: https://s3-us-west-2.amazonaws.com/snowboy/snowboy-releases/rpi-arm-raspbian-8.0-1.1.1.tar.bz2
  2. Extract to Pi & rename directory to "snowboy"
  3. Use speaker-test -c 2 to test audio out
  4. Use arecord -d 3 test.wav to record a 3 second test clip. Use aplay test.wav to verify
  5. If you need to tweak some alsa settings, use alsamixer & then run the following to keep the settings
sudo alsactl store

Prepare Snowboy

  1. copy training_service.py to the snowboy directory
  2. log into https://snowboy.kitt.ai, click on “Profile settings”, and copy your API token
  3. change the appropriate fields (token, hotword, etc.) in training_service.py
  4. use the following to record 3 wav files of your hotword to the same directory
rec -r 16000 -c 1 -b 16 -e signed-integer FILENAME.wav
  1. run the following to generate a pmdl
python training_service.py 1.wav 2.wav 3.wav saved_model.pmdl
  1. move saved_model.pmdl to ~/snowboy/resources/ (rename for easier recall later)
  2. run the demo to make sure everything is working:
python demo.py ~/snowboy/resources/saved_model.pmdl

Customize Snowboy

  1. copy snowboy.py to the snowboy directory
  2. copy the curl requeset from thinger.io to ~/thingeriocurl.sh and make it executable. be sure to replace the "bearer" information with the static device token
  3. run snowboy.py to make sure everything is working:
python snowboy.py

Autostart Snowboy

  1. copy snowboy.service to /lib/systemd/system/
  2. you may need to run this:
sudo systemctl daemon-reload
  1. start the snowboy.service to make sure everything is working:
sudo systemctl start snowboy.service

Adding

Use training_service.py to train more voice models and add pmdl files to ~/snowboy/snowboy.py (comma separated)

To Do

  • incorporate ~/thingeriocurl.sh into ~/snowboy/snowboy.py