This project is just a proof of concept for a voice based computer interface. It uses a combination of Google Speech Recognition, pocketsphinx, and julius to interpret speech and Google Text To Speech and pyttsx to speak back. There is a shell fallback also, just in case.

The interface can do the following:

• Sing 'Daisy, Daisy, give me an answer, do...'
• Look for information in Wikipedia
• Control the Plex Media Server
• Call the Police
• Chitchat

The framework is flexible enough for other actions to be incorporated with relative ease.

To launch HAL, run the script /src/root/hal.py. It works well under eclipse, but you may need to modify some imports to get it to run from other environments.

There is a sample output in /01 Output/20150921195817/20150921195817.txt. The video of that output is available at https://youtu.be/YKlop6htKGM. Enjoy!

HAL has two operating modes:

• Inactive: in this mode, HAL will listen to specific commands that will wake the interface up. Those commands are:

  • Okay HAL
  • HAL, wake up
  • HAL, are you there?
  • HAL, can you help me?
  • HAL, do you have a minute?

  From inactive mode, the command 'HAL, shut down' can be issued in order to completely shut down the HAL interface.

• Active: in this mode, HAL pushes the input through a layer of interpretation based on regex matching. If a command can be identified, it gets executed; otherwise, the input gets sent to the chatbot, based on A.L.I.C.E.

  Samples of Wikipedia commands would be:

  • What can you gather about little green men from wikipedia?
  • Would you be kind enough to get me a summary on Chernobyl?
  • Do you know anything about banjos?

  Samples of Plex commands would be:

  • Can you play 'The Big Bang Theory'?
  • Play 'Dead Man'
  • HAL, pause Plex
  • Stop the movie

  To put HAL back into inactive mode, the following commands can be used:

  • Goodbye!
  • Go to sleep
  • See you later!
  • That is all for now

In order to work, the following needs to be available in the target system:

• Pocketsphinx, Julius

  Some paths related to these will need to be updated in the configuration file /parameters/alarm_conf.py.

• Python packages:

  • subprocess, re, random, requests, tempfile, os
  • pyaudio, pocketsphinx, speech_recognition, gtts, pyttsx, aiml, plexapi, wikipedia

I cannot promise anything, but do not hesitate getting in touch if you have questions or comments!