This is the forum archive of Homey. For more information about Homey, visit the Official Homey website.

The Homey Community has been moved to https://community.athom.com.

This forum is now read-only for archive purposes.
Closed

Speech recognition services quality comparison

This is a question for Athom so I hope someone picks it up. I was just wondering something. I know Homey uses various services for speech recognition and that the voice command is send to all these services but the one that responds the fastest is used for further processing.

My question is if Athom is keeping statistics about this in any way and if the responses are compared between the different services. The OK Homey trigger had been working for me quite well from the start but the interpretation of the voice command varies from quite good to something completely different then what I said. And it just occurred to me that this could be related to the quality of which voice recognition service responds the fastest.

I'd rather have a slow services that takes it time but which actually understands my voice command then one that reacts fast but is incorrect and I need to repeat myself a couple of times.

Comments

  • Fire69Fire69 Member
    edited April 2016
    I've been wondering about this too. 
    Sometimes speech is understood quite well, sometimes it returns something that doesn't even resemble the words I used. 
    When this happens, I wonder if the cause is the recognition service or the microphones... 
    (could be my English, but when something as basic as 'lights off' returns 'I also', I don't think it's my fault) 
  • This is a very interesting question. I'm looking forward to the answer on this.
  • A strange Idea maybe, but would it help to create your own basic voice file. A file where you can safe your own speech. 
    So home gives you a lot ow words sentence hat you can pronounce yourself, that Homey later can use to recognize your commands?

  • A strange Idea maybe, but would it help to create your own basic voice file. A file where you can safe your own speech. 
    So home gives you a lot ow words sentence hat you can pronounce yourself, that Homey later can use to recognize your commands?

    Not such a strange idea at all. This is what Jeroen from Athom said on Slack:

    In fact the default trigger and the latest trained personal trigger of each homey user are combined into the trigger that homey is using upon boot.

    0.8.24 introduced a new trigger backend, and 0.8.27 introduced the new default “OK Homey” trigger, but if you did not retrain since 0.8.24, the “old” user-defined trigger will get combined with the new "OK Homey" trigger.

    I think the removal of triggers and the possibility to disable the default trigger will be added to the user interface with a future update
    This topic is not about the trigger however but about the voice commands to actually control stuff. These voice commands need to be interpreted by a voice recognition service so Homey knows what to do.
This discussion has been closed.