›
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.

There is *so much* potential!

AbidDinAbidDin Member
in Questions & Help
So after a long and anxious wait, I finally received my Homey and was very excited to get it up and running.

Initial thoughts: It has a lot of potential. Setup was reasonably seamless, the flow editor is beautiful, installing apps works great and the UX from using the trigger word to speech being interpreted (i.e. the change in colors and pulse during speech input) is lovely.

Right now, the only thing holding Homey back is speech recognition. Given that in one of the KS updates, it was mentioned that Athom are currently partnering with "the best Speech Processing companies there are available", I'm surprised at how inaccurate it is right now. By comparison, I've done tests using the HTML5 Speech API and that appears to be much more accurate at interpreting my speech.

I'm sure there are technical reasons for this that I'm unaware of, but my message to the people of Athom is simply:

Great work on the product so far. It is a true technical achievement for such a small team. It looks beautiful and when it works, does feel magical. Please (if not already) make speech recognition your top priority. That will be the deciding factor between Homey being a useless toy, or a truly great (and helpful) device.

Thank you!

Comments

  • DieterKoblenzDieterKoblenz Member
    Actually I am quite satisfied with voice recognition in the room. When I am close to Homey it is >90% accurate (comparable to google's voice recognition engine behind my computer) and when I am at the dinner table at +- 4mtr it is a lot less reliable +- 50% accurate. Especially English words I probably don't pronounce correctly are not recognized correctly. 

    Voice recognition is important but Athom has a lot more on their "top priority" list like z-wave rewrite, generic 433mhz support, the bluetooth stack... etc etc. Voice recognition is only used for fun when home automation itself cant be used.
  • AbidDinAbidDin Member
    Maybe there is a problem with the English interpretation, then. As I'd say 99% of what I say isn't understood correctly.

    Even if the same words are used, Homey hears them differently nearly every time. So for me, at this point voice control isn't even usable.

    That's a problem. I think it's an impressive product; but with this issue fixed, would be even better.
  • 2Be2Be Member
    You should try the voice recognition from the iOS or Android app. That's much better, so I guess a lot is due to the settings of the microphone. Things like ambient noise, TV etc.

    For example, when I am watching TV I have a flow that mutes my tuner when Homey start listening. This makes recognition much better.
  • tim1990tim1990 Member
    Suggestion form me. Sentences are more aquarate than soms words.
  • MatjaLipuMatjaLipu Member
    edited June 2016
    I'll stick to the title and not just to one subject. Sure voice control is one of the hottest features Homey advertise, but actually it has far more potential with it's openness and extensibility.

    With the broad wireless support and open API, which enables Homey extensibility, possibilities are really great.
    We all know Homey currently have technical issues and I am sure they all have a priority on the list. Stability should definitely be top on the list, but evolving apps support should also be high on the list. For example when Homey enables support for low level BLE to apps, community can step in and connect BT devices. Think, what we can do when Homey will provide raw speech data, RF access etc.

    Community already proved there are good use cases Homey is able to manage with creative flows and clever apps.
    And that's what Homey makes different.

    There is *so much* potential!

  • KoenMartensKoenMartens Member
    AbidDin said:
    Maybe there is a problem with the English interpretation, then. As I'd say 99% of what I say isn't understood correctly.

    Even if the same words are used, Homey hears them differently nearly every time. So for me, at this point voice control isn't even usable.

    That's a problem. I think it's an impressive product; but with this issue fixed, would be even better.
    No, there's a problem with interpreting voice when there is a lot of background. This is also the case with products of f.e. Google, but you never notice it, because you never use a microphone that's connected to Google's services that is more than 30 cm away from your mouth.

    If you would have the Homey Android app, and use the speech feature of that, you will also see that the voice recognition is nearly perfect, while the same speech to text methods are used as for the normal Homey microphone. The major difference is just that you are speaking a lot closer to the microphone; it's appearantly very hard to remove background noise and echoes, alongside the fact that the volume of your voice will go down when you are further away.
  • AbidDinAbidDin Member
    KoenMartens said:
    AbidDin said:
    Maybe there is a problem with the English interpretation, then. As I'd say 99% of what I say isn't understood correctly.

    Even if the same words are used, Homey hears them differently nearly every time. So for me, at this point voice control isn't even usable.

    That's a problem. I think it's an impressive product; but with this issue fixed, would be even better.
    No, there's a problem with interpreting voice when there is a lot of background. This is also the case with products of f.e. Google, but you never notice it, because you never use a microphone that's connected to Google's services that is more than 30 cm away from your mouth.

    If you would have the Homey Android app, and use the speech feature of that, you will also see that the voice recognition is nearly perfect, while the same speech to text methods are used as for the normal Homey microphone. The major difference is just that you are speaking a lot closer to the microphone; it's appearantly very hard to remove background noise and echoes, alongside the fact that the volume of your voice will go down when you are further away.
    To clarify, I have been testing in conditions where there is no background noise. "Ok, Homey" triggers just fine, even when I am across the room. If I monitor the interpreted speech via /manager/speech-input, I can see that most of the time, Homey doesn't correctly interpret what I've said. Although, this may be an issue with using English over Dutch.

    I have the iPhone app and have tried using that for voice instead, but get similar results.

    As I mentioned above, under the same conditions, the HTML5 speech API seems to be much more accurate at speech interpretation than what I'm finding with Homey right now. Everything else works perfectly, it's just voice that I'm struggling with.

    Athom did state in an update that speech recognition is working great for some and not others. So maybe I fall into the category of it not working out so well :smile: 

    I look forward to future firmware updates that will hopefully remedy this!
Sign In or Register to comment.