Send to Kindle

Value-focused, rapid, open engineering

Voice operating wearable computer

Since last July, I have been working on making a voice operating mini-computer that is meant to be carried with me so I can breathe information in and out while I am on the move. For example, I can search for a short answer simply saying the question: "What is the weather in Mountain View today?" I ask this question to the mini-computer every morning as I get ready to go out. It also reads out the message to my ear when someone sends me an instant message. It actually learns from the inbound texts to enhance the voice recognition capability.

The quality of this computer is far from the level to be called a commercial product. But it is something I can develop little by little based on my own desire and imagination. It is very satisfying to use what I made just as I enjoy eating what I have cooked.

As I use what I create, I also find things that I want to improve. Then I rip the old code, and quickly re-design and re-write the code. I think this computer and other things I create increase their value little by little this way.

Although they are not commercial products, I take a value-driven approach. I want to be benefited by what I code as fast as possible. Or at least I want to come up with a prototype of the idea very quickly so I can learn if it changes my life, and possibly of others.

To build quickly, I am shameless about using other people's work. Probably, the most technologically complicated part of the voice operating computer project is voice recognition. I did not develop this. I first recorded audio, and used Google's speech-to-text web service. But Google recently made a serious limit on the audio file size that can be sent to the server, and it became impossible to parse voice. So I compiled CMU Sphinx 1 on Raspberry Pi 2 and running voice recognition on the mini-computer. Voice recognition requires to build a good language model for a good accuracy. So, I am using the text coming in through the instant messenger and search results as the example sentences to build such model.

Overall, my personal projects typically have these three characteristics:

  1. Focus on delivering value. Something I can try using right away.
  2. Avoid re-inventing the wheel. Use other people's work and focus on designing the applications.
  3. In return, share what I create to the world.

This project has been a lot of fun, and I see bigger potential. I am going to integrate this voice operating computer with my server code to log the information around my life, and feed myself with useful information based on data analysis. It could also be integrated with a robot, I suppose.

Original post: May 9, 2014 | Last updated: May 19, 2014

Previous: Me as a data bee
Next: First person camera for recording creative processes
Read more

comments powered by Disqus