Kinect and Speaker client..? Fundamental questions

[hjk]

After working briefly with these, I have a couple of fundamental questions.

First, I noted that HS's built-in VR includes some default syntax. For example, "Run Event ___" will run an event. Status, timers etc - all have default syntax.

Does Kinect borrow this syntax, or do you program ALL syntax for Kinect?

Second question -- does Kinect use the standard Microsoft SAPI, or does it have its own speech recognition engine coming from the Kinect SDK?

Finally, do people run both the speaker client AND Kinect on the same remote systems? If I've got a remote PC/tablet/headless unit, with a mic (provided by Kinect) and possibly a speaker attached, are there advantages/disadvantages to using both on the same system? (Is there any reason to - for example, to gain the default syntax provided by the Speaker client..?)

Thanks for your help -- just a little confused!!

hjk
---

[spud]

First, I noted that HS's built-in VR includes some default syntax. For example, "Run Event ___" will run an event. Status, timers etc - all have default syntax.

Does Kinect borrow this syntax, or do you program ALL syntax for Kinect?

No, the Kinect plugin does not inherit from these default commands. You have to either use the kinect trigger "A phrase was recognized" from an event or use "device voice commands" to associate a command with a device control:

have a look at the user guide for more information:
http://board.homeseer.com/showthread.php?t=162315

Second question -- does Kinect use the standard Microsoft SAPI, or does it have its own speech recognition engine coming from the Kinect SDK?

the Kinect plugin use the server version of SAPI whereas HS's built-in VR use the desktop version. They are similar but use different recognizers. One difference is that the server version does not need to be trained.

Also, the kinect SDK offers some options to configure the Audio source coming from the kinect microphone array: noise suppression, echo cancellation, etc...
The kinect plugin use these options internally for enhancing speech recognition performance.

Finally, do people run both the speaker client AND Kinect on the same remote systems? If I've got a remote PC/tablet/headless unit, with a mic (provided by Kinect) and possibly a speaker attached, are there advantages/disadvantages to using both on the same system? (Is there any reason to - for example, to gain the default syntax provided by the Speaker client..?)

IMO, using both the kinect plugin VR and HS built-in VR on the same system would be confusing as they don't use the same way to define voice commands. But you can use the Kinect VR and speaker client (with voice recognition disabled) on the same remote machine to provide VR and TTS for this machine.

[hjk]

Spud, thank you for the explanations.

I completely missed the Device Voice Commands section in the user guide post! I have also noticed that the VR that Kinect uses seems much more accurate than the desktop/standard VR. I can speak quickly and elide words and it still picks it up with a high degree of confidence.

At this point, I'm still a bit torn. I purchased an xAP800 from eBay, intending to use it as a central solution for whole-house audio. It can take an input (say, from my receiver) and filter it out. So if I'm watching Star Trek and Picard says "Computer! Do x/y/z," Homeseer won't resopnd.

However... I think I read that you're working to modify the plugin so that it can accept input from a non-Kinect source. If that's the case, then I suppose one could still use the xAP800 simply as an input source.

Still trying to conceptualize all this.

Thanks again! I find the plugin to be fast and effective!

hjk
---

[LeoS]

I would like to follow this topic as well, as I'm on similar boat...

I have an AP800 too (bought years ago, plan to wire... this weekend!). But using kinect would be more interesting as it has superior array mic, has awesome depth camera and humanoid recognizer... and also seems to have access to a superior voice recognition engine.

The downside to that would be more limited command vocabulary and not integrated with HS3's VR engine.

So I wonder if it's possible to make use of Kinect's voice recognition engine, then output the resulting text to HS3 to process with HS3's engine as if the command was spoken directly to HS3?

[UltimateInternet]

I'm hoping to see the internet sourced api become available (the one that cortana uses for the phone/windows 10) and hook into THAT.

With xbox one and windows 10 becoming much more voice centric, I think we either need to hook into it as well or write an app that hooks into the ecosystem...