Skip to content
This repository has been archived by the owner on Jul 19, 2020. It is now read-only.

Latest commit

 

History

History
239 lines (177 loc) · 7.29 KB

README.md

File metadata and controls

239 lines (177 loc) · 7.29 KB
Raw

AIY Voice Only

Google AIY Voice Kit is a cool project. But it locks you into its custom hardware. I have separated its software to work on Raspberry Pi (3B and 3B+) independently, just using a normal speaker and microphone.

The following instructions aim at:

Raspberry Pi (3B, 3B+)
Raspbian Stretch
Python 3

Additionally, you need:

  • a Speaker to plug into Raspberry Pi's headphone jack
  • a USB Microphone

Plug them in. Let's go.

Find your speaker and mic

Locate your speaker in the list of playback hardware devices. Normally, it is at card 0, device 0, as indicated by the sample output below.

$ aplay -l

**** List of PLAYBACK Hardware Devices ****
card 0: ALSA [bcm2835 ALSA], device 0: bcm2835 ALSA [bcm2835 ALSA]
  Subdevices: 8/8
  Subdevice #0: subdevice #0
  Subdevice #1: subdevice #1
  Subdevice #2: subdevice #2
  Subdevice #3: subdevice #3
  Subdevice #4: subdevice #4
  Subdevice #5: subdevice #5
  Subdevice #6: subdevice #6
  Subdevice #7: subdevice #7
card 0: ALSA [bcm2835 ALSA], device 1: bcm2835 ALSA [bcm2835 IEC958/HDMI]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

Locate your USB microphone in the list of capture hardware devices. Normally, it is at card 1, device 0, as indicated by the sample output below.

$ arecord -l

**** List of CAPTURE Hardware Devices ****
card 1: Device [USB PnP Audio Device], device 0: USB Audio [USB Audio]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

Your hardware's number might be different from mine. Adapt accordingly.

Make them the defaults

Create a new file named .asoundrc in the home directory (/home/pi). Put in the following contents. Adjust the card,device number if needed.

pcm.!default {
  type asym
  capture.pcm "mic"
  playback.pcm "speaker"
}
pcm.mic {
  type plug
  slave {
    pcm "hw:1,0"   # card number, device number
  }
}
pcm.speaker {
  type plug
  slave {
    pcm "hw:0,0"   # card number, device number
  }
}

Make sure sound output to headphone jack

Sound may be output via HDMI or headphone jack. We want to use the headphone jack.

Enter sudo raspi-config. Select Advanced Options, then Audio. Choose Force 3.5mm (headphone) jack.

Turn up the volume

A lot of times when sound applications seem to fail, it is because we forget to turn up the volume.

Volume adjustment can be done with alsamixer. This program makes use of some function keys (F1, F2, etc). For function keys to function properly on PuTTY, we need to change some settings (click on the top-left corner of the PuTTY window, then select Change Settings ...):

  1. Go to Terminal / Keyboard
  2. Look for section The Function keys and keypad
  3. Select Xterm R6
  4. Press button Apply

Now, we are ready to turn up the volume, for both the speaker and the mic:

$ alsamixer

F6 to select between sound cards
F3 to select playback volume (for speaker)
F4 to select capture volume (for mic)
⬆ ⬇ arrow keys to adjust
Esc to exit

If you unplug the USB microphone at any moment, all volume settings (including that of the speaker) may be reset. Make sure to check the volume again.

Hardware all set, let's test them.

Test the speaker

$ speaker-test -t wav

Press Ctrl-C when done.

Record a WAV file

$ arecord --format=S16_LE --duration=5 --rate=16000 --file-type=wav out.wav

Play a WAV file

$ aplay out.wav

Register for Google Assistant or Google Cloud Speech

Although we are not using Google's hardware, there is no escaping from its software. We still rely on Google Assistant or Google Cloud Speech API to perform voice recognition. To use these cloud services, you have to go through a series of registration steps:

Which one to use depends on what you need. Google Assistant can recognize speech and talk back intelligently, but supports fewer languages. Google Cloud Speech only recognizes speech (no talk-back), but supports far more languages.

Here is a summary of the steps for using Google Assistant, as of 2019-11-27. Always pay attention to the Project ID on top of the page to make sure you are in the right project.

  1. Create a Project

  2. Enable Google Assistant API

  3. Configure OAuth consent screen (must fill in Support email)

  4. Enable activity controls

  5. Register device model, Download credentials file (check project_id)

  6. Install system dependencies:

    $ sudo apt-get install portaudio19-dev libffi-dev libssl-dev

  7. Install Python packages:

    $ sudo pip3 install --upgrade pip setuptools wheel
$ sudo pip3 install google-assistant-library==1.0.1 \
                    google-assistant-grpc==0.2.0 \
                    google-assistant-sdk[samples]==0.5.1 \
                    google-auth-oauthlib[tool] \
                    google-cloud-speech

    On Raspbian Buster, you likely have to downgrade to google-assistant-library==1.0.0 for it to work.

  8. Use google-oauthlib-tool to authenticate:

    $ google-oauthlib-tool --scope https://www.googleapis.com/auth/assistant-sdk-prototype \
   --save --headless --client-secrets /path/to/client_secret_client-id.json

  9. Use googlesamples-assistant-devicetool to register your Raspberry Pi. A few useful commands may be:

    $ googlesamples-assistant-devicetool --project-id <Project ID> register-device \
--model <Model ID> \
--device <Make up a new Device ID> \
--client-type LIBRARY

$ googlesamples-assistant-devicetool --project-id <Project ID> list --model

$ googlesamples-assistant-devicetool --project-id <Project ID> list --device

How to use this library

I used to have it uploaded to PYPI for easy installation. But Google Assistant is changing too rapidly. I find it more informing to download and try to integrate it manually:

  1. Download the aiy directory

  2. Set environment variable PYTHONPATH so Python can find the aiy package

  3. You may have to install the Pico text-to-speech engine, libttspico-utils, to allow it to generate speech dynamically (If apt-get install libttspico-utils doesn't work, manually download and install deb packages libttspico-data, libttspico0, and libttspico-utils separately)

The best way to experience the software is to try it. Let's go to the examples.

Changes to original library

Here is an outline of the changes I have made to the original AIY Voice Kit source code:

  1. No Vision stuff: The AIY project actually includes the Vision Kit and associated software, which are of no concern to this project. I have removed those.

  2. No Voice Hat stuff: This project does not rely on the Voice Hat. The aiy.board module has been removed.

  3. The class Led and Button have been moved to the aiy.util module.