Amazon Echo to HA Controllers

[pete_c]

Curious; Chris is your software just using MS SAPI VR / TTS with the Kinect?
 
Is your software using the rest of the Kinect features?  (gestures, movement et al).
 
Reason I ask here is that I have put together a little test environment using a 17" wide screen (for in wall mounting) which is a dual touch capacitance screen, Kinect (which would be mounted above the screen) and an Atom Baytrail PC.
 
Would I be able to use hand gestures (no VR) to navigate on the touch screen rather than touch?
 
I purchased a few Kinects (early 2015) but at this time do not want to test it with my LCD TVs (well don't really integrate my automation with my LCD watching or music or whatever - totally autonmous today). 
 
The touch screens today though can stream live TV / recorded media / CCTV in a little window plus serve as touch consoles to the automation showing buttons and status of variables.  

 

[ChrisCicc]

Curious; Chris is your software just using MS SAPI VR / TTS with the Kinect?
 
Is your software using the rest of the Kinect features?  (gestures, movement et al).
 
Reason I ask here is that I have put together a little test environment using a 17" wide screen (for in wall mounting) which is a dual touch capacitance screen, Kinect (which would be mounted above the screen) and an Atom Baytrail PC.
 
Would I be able to use hand gestures (no VR) to navigate on the touch screen rather than touch?
 
I purchased a few Kinects (early 2015) but at this time do not want to test it with my LCD TVs (well don't really integrate my automation with my LCD watching or music or whatever - totally autonmous today). 
 
The touch screens today though can stream live TV / recorded media / CCTV in a little window plus serve as touch consoles to the automation showing buttons and status of variables.  

 
We use both the old and new versions of TTS. Any SAPI compatible voice can be used with CastleOS.
 
Hand gestures are a different story. We are intentionally not shipping them, because they require far too much manual programming by the end user. That said, we don't lock down the sensor, so you can create a hand gesture app that plugs into our API with ease. 

 

[IVB]

he issue is whether you want natural language processing or not. If you just want to issue a fixed command and get a custom response, that's easy. Natural language processing within the context of home automation is hard, and for that CastleOS needs a full protocol driver.

 
Can you please explain that, i'm not following.  I currently have a fixed command and custom response on near-field using any Android device.  With Kinect, I can expand from 18 inches to 15 feet max (assuming 100% reliability is the goal, I can get maybe 6 feet from android or further from Kinect, but reliability drops below 95% which is unacceptable in our house. 
 
Speaking just for me personally, I'd rather have fixed language + 100% reliable instead of natural language and 94% reliable. I just need something to trap what i'm saying and tell CQC, as it has RS232 or IP control over all my devices. (I sold off all IR devices, too flaky).  Does that mean I need a full protocol driver? If so, what exactly does that mean?
 
BTW, I just realized: I can't use my current Kinect as "double-duty" for xBone & COS, i'd need another one for the H/T room. Plus, more for other rooms. Are these things able to be stripped down to just whats needed for VR?  
 
Wife doesn't mind small android devices, I can get a brand new 4" moto e for $120 (or cheaper for older phones via eBay) that could blend in, not sure about hiding a Kinect. (Although maybe hiding below a windowsill?)

 

[jkmonroe]

What you could do is pick up the TW700 tablet that I am such a huge fan of.  They are like $60 at Microcenter brand new and run full Windows 8.  You could run a CQC IV on that *and* plug in a Kinect.  Right at your $120 mark for full 7" CQC Interface Tablet + Kinect.

 

[IVB]

What you could do is pick up the TW700 tablet that I am such a huge fan of.  They are like $60 at Microcenter brand new and run full Windows 8.  You could run a CQC IV on that *and* plug in a Kinect.  Right at your $120 mark for full 7" CQC Interface Tablet + Kinect.

 
dude that is one fugly tablet though, bezel bigger than my watch.  plus it would be a fixed location device, in my house we removed both fixed location tablets as they were getting minimal use and took up wall or table space.  Every person has a smartphone (2 iPhone 2 Android), we have 2 iPads, 2 Nexus 7, and google accidentally shipped an extra Nexus 6.  We don't need any more input devices.

 

[jkmonroe]

maybe not. the point being that it's a $60 full PC which you could use to run the kinects. although i use them as both fixed and portable tablets, amazon has right angle USB that makes them work.

 

[ChrisCicc]

Can you please explain that, i'm not following.  I currently have a fixed command and custom response on near-field using any Android device.  With Kinect, I can expand from 18 inches to 15 feet max (assuming 100% reliability is the goal, I can get maybe 6 feet from android or further from Kinect, but reliability drops below 95% which is unacceptable in our house. 
 
Speaking just for me personally, I'd rather have fixed language + 100% reliable instead of natural language and 94% reliable. I just need something to trap what i'm saying and tell CQC, as it has RS232 or IP control over all my devices. (I sold off all IR devices, too flaky).  Does that mean I need a full protocol driver? If so, what exactly does that mean?

 
A couple things to note... with non-dictation VR systems fixed commands are actually prone to less reliability than grammar-based natural language commands. The longer the individual phrase, or in the case of fixed commands, sentence, the more processing time is required, and the less effective the algorithms are. 
 
The difference between the fixed command and NL command can be illustrated this way. Take the command "turn on the lights". The fixed command, would be simply, "house turn on the lights". But that's not the only way to say turn on the lights. For instance, you can say:

1. turn on lights
2. turn the lights on
3. turn lights on
4. lights on

It gets more complex when you add device and group names into the mix. A command like "house turn on the bedroom lights" can have many grammatical variations. 
 
This also becomes important when you are doing this at a whole-home scale. Quite simply, the system can only load a few hundred fixed commands before it starts to become unusable. With our natural language design, a typical home can accept several hundred thousand different command variations. And while no VR system is 100% reliable, I promise you a couple nines
 

BTW, I just realized: I can't use my current Kinect as "double-duty" for xBone & COS, i'd need another one for the H/T room. Plus, more for other rooms. Are these things able to be stripped down to just whats needed for VR?  
 
Wife doesn't mind small android devices, I can get a brand new 4" moto e for $120 (or cheaper for older phones via eBay) that could blend in, not sure about hiding a Kinect. (Although maybe hiding below a windowsill?)

 
The Kinect blends in far better than you might think. This is not trying to hide it, and IMHO still looks good. If I put it on top of the black floor speaker in the corner, no one notices it. Window sills work too. Media centers are great. Bookshelves make it easy. For TVs that are using their integrated stands, it can stand on the stand but below the screen in most cases.  
 

 

[ChrisCicc]

maybe not. the point being that it's a $60 full PC which you could use to run the kinects. although i use them as both fixed and portable tablets, amazon has right angle USB that makes them work.

 
I'm concerned the 1GB of RAM isn't enough. I'm going to pick one up to test it out... Thanks for the tip! Even if the Kinect doesn't work well with it, the Core Service should run great. It's a $69 solution for getting started with the free version of CastleOS

 

[IVB]

uh, not sure how what you just said relates to "full protocol driver".  Does that mean "free" versus $199/full-thing?  Meaning, to get the five 9s requires the full thing? For me, cost is less of an issue, as long as I can predictably tell CQC what to do from far-field. 
 
As previously said, in my house the Kinect just doesn't work further than 15 feet with remotely sufficient reliability (technically 13-14 feet). We've tried, and have to repeat ourselves. For kinect related commands thats fine since we're there, but for HA its inadequate. As shown above, I have 5 android tablet/phones, plus a Moto 360, so i'm never really that far from a device.  My placement wouldn't be a location where other electronics exist, like the TV stand. I'd need it in the dining room where nothing electronic is, or other odd locations.
 
One more question: Lets say I have 5 of these in various locations due to acoustics. Would I need to do that programming 5 times like I have to do right now with Tasker? Its not a huge deal with tasker since I can export/import, but its definitely a hassle.
 
(BTW JKMonroe, I have pissed away infinite sums on experimental HA things that don't live up to promises. I'm really asking all this so you can try to get VR working with CQC. Once you have it working at 99.999% i'll just by whatever you have. I'm still trying to get my damn VRUSB working. 2 more Intermatic wall switches lost zWave capability in the last 2 weeks, I need that working soon so I don't mind dumping $80/each on the Vizia RF+ replacements.)

 

[ChrisCicc]

uh, not sure how what you just said relates to "full protocol driver".  Does that mean "free" versus $199/full-thing?  Meaning, to get the five 9s requires the full thing? For me, cost is less of an issue, as long as I can predictably tell CQC what to do from far-field. 

 
The custom commands are totally free. For the natural language (and custom protocol drivers), that requires the license fee if using with greater than 10 smart devices. That said, the new API that would allow that driver to be written is still in testing. Should be out in a couple weeks if all goes well. Maybe Dean will be willing to whip up an interface real quick
 

As previously said, in my house the Kinect just doesn't work further than 15 feet with remotely sufficient reliability (technically 13-14 feet). We've tried, and have to repeat ourselves. For kinect related commands thats fine since we're there, but for HA its inadequate. As shown above, I have 5 android tablet/phones, plus a Moto 360, so i'm never really that far from a device.  My placement wouldn't be a location where other electronics exist, like the TV stand. I'd need it in the dining room where nothing electronic is, or other odd locations.

 
Try it with CastleOS...like I said previously apps have a HUGE influence on that. I'm not kidding I've tested it to over 50 feet in low ceiling carpeted rooms without having to yell.
 
 

One more question: Lets say I have 5 of these in various locations due to acoustics. Would I need to do that programming 5 times like I have to do right now with Tasker? Its not a huge deal with tasker since I can export/import, but its definitely a hassle.

 
Nah just copy the config files over, and tweak as necessary. The beauty of our system is it's designed to be as simple as possible so anyone can configure it, but also be pretty hackable for advanced users. All the config is in XML and openly accessible to system admins on Windows. 

 

[jkmonroe]

(BTW JKMonroe, I have pissed away infinite sums on experimental HA things that don't live up to promises. I'm really asking all this so you can try to get VR working with CQC. Once you have it working at 99.999% i'll just by whatever you have. I'm still trying to get my damn VRUSB working. 2 more Intermatic wall switches lost zWave capability in the last 2 weeks, I need that working soon so I don't mind dumping $80/each on the Vizia RF+ replacements.)

 
*sigh*, fine.  i guess i'll go get grab a Kinect and start playing.
 
:horse:

 

[IVB]

*sigh*, fine.  i guess i'll go get grab a Kinect and start playing.
 
:horse:

good man.  I wonder if I should pick up an echo since I got that invite and we can see which works better.

 

[ChrisCicc]

*sigh*, fine.  i guess i'll go get grab a Kinect and start playing.
 
:horse:

 
If you do want to do this I can get you an internal build as soon as the API is usable with plug-in protocols... 

 

[ChrisCicc]

I wonder if I should pick up an echo since I got that invite and we can see which works better.

 
They have 3+ month backorder right now as of a week ago... 

 

[IVB]

I effing hate you for starting this thread, jkmonroe:
 

Order Number: 102-3265044-WHOSYOURDADDY

• Amazon Echo will be shipped to Vivek Bhatia by Amazon.com. 
  Estimated delivery: July 8, 2015