Amazon Echo to HA Controllers

Someone managed to emulate the Hue bridge in software to allow the Amazon Echo to interface with existing HA platforms.
 
http://forum.micasaverde.com/index.php/topic,31920.html
 
http://community.smartthings.com/t/playing-around-with-amazon-echo/14887/50
 
It looks like this could be an "easy" way to get simple voice control into our existing platforms.

My amazon echo invitation just came through. So now the question is, what's my confidence level that using a custom phrase (ie living room lights on) can be translated into a custom url http://192.168.1.100:11001/CMLBin/User/NFC?Cmd=LIVLGHTON) using the echo.

For Google Glass, I decided to decline the invite as it didn't look like that would ever happen. Not sure about this device, and why would Amazon allow this in the future. The only interesting bit is the promise of far field mics,everything else I can do via a plugged in Nexus 7.

The Amazon 2 is still in the monolith box here... I did move it to it's own space in my office now ...mountain terrain is an illustrative add and is not present in my home office.
 
 

 

If you're going to go through that much trouble to get the Echo to work, you should just get CastleOS...it does everything you want, better than the Echo, and it's less expensive per voice node too  

Isn't castle just software? I already have that, I just need a vr triggering mechanism.

I'm not worried about $150/room, as long as it works. I'd pay $250/room for 100% reliability, $50 for 80%= not remotely WAF. I have 95% via Nexus 7 which are cheap and re-usable but you need to stand close.

IVB said:

Isn't castle just software? I already have that, I just need a vr triggering mechanism.

I'm not worried about $150/room, as long as it works. I'd pay $250/room for 100% reliability, $50 for 80%= not remotely WAF. I have 95% via Nexus 7 which are cheap and re-usable but you need to stand close.

Click to expand...

 
CastleOS is software, but we leverage the Microsoft Kinect's microphone array to provide 99% reliability with a virtually zero false positive rate. 

Only CastleOS and the Echo allow for device-less voice recognition. In other words, you can just speak out loud.
 
However, only CastleOS does this without requiring a prompt. For example, on the Echo, it would be "Alexa" pause while it activates "turn on the lights".
 
With CastleOS, it's "House turn on the lights" without any pause needed. 
 
Echo can turn Hue on and off, but only CastleOS can change the color by voice "House make the living room blue"
 
https://www.youtube.com/watch?v=CmNGt4tND64
 
Edit: the hot word "house" can also be changed. By default, you can say "house" or "computer", or add a third option of your choosing, like "Alexa" or "Jarvis"

got it.  Dumb question: Wouldn't that require a $100 kinect in every room? I have an xBone / Kinect, it doesn't really work that well from more than 15 feet away, at least for my teenage daughters. I can generally make it work from 15-20 feet but I have to project my voice. That doesn't really sound like a cost savings over an Android device. 
 
Given that it sounds like the Kinect is an externally controllable device, it sounds like all I need is to get a CQC driver / integration for it.  I already have a stable and pervasive setup for CQC, an alternative VR execution mechanism (ie Kinect) isn't valuable enough to switch.
 
Thanks for the thoughts though.

IVB said:

got it.  Dumb question: Wouldn't that require a $100 kinect in every room? I have an xBone / Kinect, it doesn't really work that well from more than 15 feet away, at least for my teenage daughters. I can generally make it work from 15-20 feet but I have to project my voice. That doesn't really sound like a cost savings over an Android device. 
 
Given that it sounds like the Kinect is an externally controllable device, it sounds like all I need is to get a CQC driver / integration for it.  I already have a stable and pervasive setup for CQC, an alternative VR execution mechanism (ie Kinect) isn't valuable enough to switch.
 
Thanks for the thoughts though.

Click to expand...

 
You're comparing apples and oranges. If you want to be tied to your device, sure that's a cheaper way. We're comparing device-less voice recognition, and there are only two choices. The Echo costs $199, it's not inexpensive, and yes, you need one per room. Regarding distance, we've tested the Kinect past 50 feet. It all depends on the acoustics of your room, just like the Echo as well.

But getting down to your comments about building a driver for Kinect, that's not how it works We built the entire natural language processor, the false positive filters, nevermind everything that goes with the automation interface. That's all our IP. So I wouldn't expect it to be so easily recreated. Just look at the Echo - they have a huge team and budget and have only scratched the surface of what we can do with voice controlled home automation... it's not as easy as it looks
 
Edit: Oh and did I mention, unlike the Echo we don't require the cloud and we don't listen in on what you're doing/saying in your home?  

Honestly that's the reason to use a trigger phrase. I had auto voice set up to not require one so I could use quasi natural language (no fixed order or syntax). But last night we were watching law and order SVU and the TV triggered the lighting command. Goddammit detective Benson.

I got a psychotic quote on the subpanel. We're at 80% capacity on the main so I let it be. Anyone want to buy my brultech, I was going to install that at the same time

IVB said:

Honestly that's the reason to use a trigger phrase. I had auto voice set up to not require one so I could use quasi natural language (no fixed order or syntax). But last night we were watching law and order SVU and the TV triggered the lighting command. Goddammit detective Benson.

Click to expand...

 
We solved those problems. It is indeed possible to have natural language, reliability, and lack of false positives. Give CastleOS a try, I think you'll be impressed! Just look at the video I've posted - no other system can do anything like that...

ChrisCicc said:

 
You're comparing apples and oranges. If you want to be tied to your device, sure that's a cheaper way. We're comparing device-less voice recognition, and there are only two choices. The Echo costs $199, it's not inexpensive, and yes, you need one per room. Regarding distance, we've tested it past 50 feet. It all depends on the acoustics of your room, just like the Echo as well.

Click to expand...

Sorry if I'm being dense, but I'm still tied to a device, it's just the kinect and not the echo, right? All I'm doing is triggering events. I feel like I'm not understanding what you're saying despite your using simple words, sorry about that.

Understood on acoustics. Although honestly that's a strong business case to go with the echo. I can put that in the center of a room, whereas the kinect would need to be near a TV which is best suited against the wall. My main room doesn't work with kinect past 15 feet (28 foot room), so either 2-3 androids or one echo.

IVB said:

Sorry if I'm being dense, but I'm still tied to a device, it's just the kinect and not the echo, right? All I'm doing is triggering events. I feel like I'm not understanding what you're saying despite your using simple words, sorry about that.

Understood on acoustics. Although honestly that's a strong business case to go with the echo. I can put that in the center of a room, whereas the kinect would need to be near a TV which is best suited against the wall. My main room doesn't work with kinect past 15 feet (28 foot room), so either 2-3 androids or one echo.

Click to expand...

 
I'll go step by step...
 
Your way:

1. go find phone or tablet
2. turn on phone or tablet
3. unlock phone or tablet
4. open app (if on android), press Siri button (if on iOS)
5. give voice command (if on android, press voice button first)

Your way with Echo:

1. Speak out loud "Alexa"
2. Wait for Alexa to enter "command listening mode"
3. Give voice command

The CastleOS Way:

1. Speak voice command out loud, i.e. "house dim the lights to 45%"

 
 
Regarding the distance, I can only speak to CastleOS and Kinect. Some apps are optimized to exclude speech outside certain ranges. Some apps are optimized to exclude speech if a person isn't visible. CastleOS is optimized for home automation, which means it accepts input from all angles and ranges. It works far greater than just 15 feet...
 
Edit: Here's the Discovery Channel video showing our voice control in action at much greater than 15 feet in many cases. And this was in 2013...everyone is still playing catch up
 
https://www.youtube.com/watch?v=_NzVyJHk5qw

ChrisCicc said:

I'll go step by step...
 
Your way:

1. go find phone or tablet
2. turn on phone or tablet
3. unlock phone or tablet
4. open app (if on android), press Siri button (if on iOS)
5. give voice command (if on android, press voice button first)

Your way with Echo:

1. Speak out loud "Alexa"
2. Wait for Alexa to enter "command listening mode"
3. Give voice command

The CastleOS Way:

1. Speak voice command out loud, i.e. "house dim the lights to 45%"

Click to expand...

 
Actually with Android 5.1 its far far far simpler. No need to unlock. My way (which happened last night as my phone was locked and on my nightstand):
1. Speak voice command out loud. Magic happens.
 
Although after last night, i'm re-enabling the trigger phrase. I had it set to the new way of doing continuous listening (Blanking on what its called but its the low-power version of what AutoVoice continuous listening was). As soon as the TV said something about the lights being off being scary, my lights turned off .  Now i'll have to just say "Okay google" wherever I am. It instantly snaps into command listening mode, no delay that I've never noticed. I never pause, I just say "okay google lights off". Or if i'm driving, "okay google send a text to the wife". 
 
But, your description of Echo makes me want to *not* buy it, so thank you for that. Sounds like sticking with just Android is the trick. (And, once the motorola phones with the dedicated language chip get newer versions, I can pick up the 2014 models for pennies).
 
I've got 2 soccer games tonight, but I can try taking a video tomorrow night. 

IVB said:

Anyone want to buy my brultech, I was going to install that at the same time

Click to expand...

 
The GEM?  What package/CTs and interface?  It is next on my list, too.

IVB said:

Actually with Android 5.1 its far far far simpler. No need to unlock. My way (which happened last night as my phone was locked and on my nightstand):
1. Speak voice command out loud. Magic happens.
 
Although after last night, i'm re-enabling the trigger phrase. I had it set to the new way of doing continuous listening (Blanking on what its called but its the low-power version of what AutoVoice continuous listening was). As soon as the TV said something about the lights being off being scary, my lights turned off .  Now i'll have to just say "Okay google" wherever I am. It instantly snaps into command listening mode, no delay that I've never noticed. I never pause, I just say "okay google lights off". Or if i'm driving, "okay google send a text to the wife". 
 
But, your description of Echo makes me want to *not* buy it, so thank you for that. Sounds like sticking with just Android is the trick. (And, once the motorola phones with the dedicated language chip get newer versions, I can pick up the 2014 models for pennies).
 
I've got 2 soccer games tonight, but I can try taking a video tomorrow night. 

Click to expand...

To be fair, I described the normal app voice path. You've hacked something together that cannot be made a commercial production product. It's another apples to oranges comparison. All the voice commands you have to manually enter as well. CastleOS does it all automatically just be seeing the groups and devices you have in the system.

Most people have lock screens. Most people don't use tasker. They use apps. Wink app. Vera app. HomeKit app. As you've found, you need a trigger word. So now you've built an Amazon Echo equivalent, not a CastleOS equivalent. Not to mention the lag inherent in the cloud...