Angelos Italianknows Blog

Monday, 13 April 2015

Building a voice actuated IOT controller

Ive been meaning to use one of my RaspberryPIs for something  "cool" recently.. And Ive decided to create myself a raspberry PI version of Siri.... Recently Amazon announced the Amazon Echo. A very nice looking package... However nothing beats homebrew right??

Initial research

Ive discovered three principle voice-control frameworks

  1. Jasper
  2. CMU Sphinx aka PocketSphinx
    and finally
  3. VoiceControl by Steven Hickson


I like Jasper , it appears to be well thought out and appears to have a lot of extensions, compatibility with different voice recognition/ text to speech systems... 
- Documentation is a bit hit-n-miss
- Rather large and not sure why....

CMU Sphinx/PocketSphinx

I recently followed the tutorial by Wolf Paulus on his blog and got it all working. When compiling everything I had to ignore the "make check" step, it works regardless, and after I created some dictionaries then it all worked fine.. Except... I could only get it to detect words Id created using the dictionary approach, I couldnt get it to detect/parse a sentence which wasnt in the vocab list.. 
I kept getting "input overrun, read calls are too rare (non-fatal)".. From my googling at 2am I think this is because the PI has run out of steam.. Wolf appears to be using a PI v2 so I think Im gonna need to upgrade..... 


Havent tried this one yet.. will do and let you know


  1. Same (input overrun, read calls are too rare (non-fatal)) on RPI2.

  2. Same (input overrun, read calls are too rare (non-fatal)) on RPI2.