Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Throwing another (sort of hybrid) voice control option out there
#11
The argument for something like CQC Voice is that it's designed specifically for CQC integration needs and doesn't require much technical effort to get it set up. The Echo, and I would assume Google, are much more general purpose, which is good and bad. In order to get something nice and flexible, the Echo setup is currently pretty complex. I don't know what's involved for the equivalent Google Home, but I would imagine it's something similar, either that or you stick with some very limited set of commands it comes with.

In order to test the new Amazon scheme it wouldn't be necessary to create a new CQC Voice immediately. Just something hacked together sufficient enough to speak back the commands spoken to it would be a good enough quick test I guess. I could then give that to anyone who wants to see how it works with their microphone setup.
Dean Roddey
Software Geek Extraordinaire
Reply
#12
Yeah, Agreed.  There is a certain beauty about CQC voice and it is very easy to set up.  If we could get around the Microphone complexities and add some range to what it can hear accurately it would be amazing.  From what a friend tells me his Google home has really good range and even hears what he says with the music at a high volume. Not sure if that is the array itself or voice processing tech.  I'm going to borrow it to compare at some point here and see for myself. 

I only wish I could buy one and just use it as a standalone microphone for CQC voice.  A simple box that only needs a power connection and sends the voice commands via network to CQC would be perfect.  I guess that's what I was trying to say is my holy grail at present for CQC voice.  Use the Google\Amazon hardware without actually signing up on a Google\Alexa account and use CQC voice with it.

When I get some time I'll have to look around at what array mics are out there, my concern is dropping hundreds on a mic and mini PC and it not being much better than the Kinect.  This thing sounds interesting although I'm not savvy enough to tell if it could be useful or not.  http://linuxgizmos.com/rpi-friendly-home...n-support/  Likely would need some driver/firmware work perhaps.



Happy to be a tester for anything you hack together.  I'd like to see CQC voice pushed a bit further and if we could get the hardware a bit less constraining it would be great.
Reply
#13
https://aws.amazon.com/blogs/ai/export-y...d=41934110
|Z-Wave|Sonos|Tivo|Hue|Plex|Roku|MyMovies|Echo|
Nest|Harmony|Neeo|LG TV|Smarthings|
Reply
#14
I just went through the setup of a simple bot configuration on AWS. For our purposes there's not even any need for any javascript, since all we want is to get the final values. So all that's really required is the bot with the intents and slots. This was done by hand using the console, not one built on the fly to include room config data.

I'm going to now see if I can access this guy via HTTP in a little test program and make sure that works as indicated.
Dean Roddey
Software Geek Extraordinaire
Reply
#15
It requires some authentication stuff I'll need to implement, so that'll probably take a day or two.
Dean Roddey
Software Geek Extraordinaire
Reply
#16
I got the SHA256 hashing algorithm implemented, which is the basis of the authentication, well authorization really, algorithm. So I have just implement an HMAC thingie on top of that and the rest will be easy enough. It's good to get SHA256 in place, since we are likely to run into other stuff that uses in the future. SHA1 is still fine for non-security stuff, but it's considered broken for security these days. I use it to generate unique ids for cover art and the metadata database in the repos/client service, but that's just for unique identification, not for security.
Dean Roddey
Software Geek Extraordinaire
Reply
#17
Whew... I've implemented a generic HMAC algorithm. Since I have polymorphic hashers and hashes, it can work in terms of any hashing algorithm I have implemented, which is nice. In this case it's a HMAC-SHA256.

And I've gone through what I think is the correct tedious set of steps to create the correct data to do a signed AWS request. That's all done via a helper that I'll eventually move to a probably an 'AWS helper' type library for this sort of stuff once I prove it all out via my little hacked up test program, since we may access other AWS services in the future that require it.

So now I should in theory be able to try a request. I just HMAC the built up signing data and and then set up the HTTP operation that includes all that stuff.
Dean Roddey
Software Geek Extraordinaire
Reply
#18
And still going. It's pretty psycho. You have to write an entire program just to send a message. But I'm just about there now where I can try to send one.
Dean Roddey
Software Geek Extraordinaire
Reply
#19
I've gone through all of the steps up to, but not actually sending the request, checking the data along the way and it looks basically correct. So I'll give it a try tomorrow. It is quite crazy how much is involved.

BTW, I also realized another advantage this would have. I don't have to have it return text for me to speak. I can have have it return audio, and therefore you can configure your bot to use one of their (quite nice) voices to speak the output. And that also means that much easier to run all of this as a background service, because it's not trying to do either speech recognition or text to speech, neither of which work in the background. It's all just moving around raw PCM audio. That's a bit more network overhead, but in the grand scheme of things these days, for the typically fairly short replies involved, it's going to be pretty trivial.
Dean Roddey
Software Geek Extraordinaire
Reply
#20
Well, despite seemingly insurmountable odds, I finally just got a valid request done. Any time there's layer after layer of content being built up and hashed and each one feeds the downstream processes, it's so difficult to figure out what's going on. And I wasted way too much time by having accidentally called the wrong POST method, the one that takes a set of key/value pairs that it URL encodes in the standard way, so it was changing my content on the way out. Oy!

But, anyhoo, that's the big step. The rest is grunt work basically. The big thing is the building up of the intents and plugging in the user's room config data, which will be a lot of steps. I'll have to come up with some XML file that can be read and used to drive the process or something. It won't be like in the other Echo support, where you set up the intents and utterances yourself. I'll have to create them for you on the fly.

So, I guess before doing all that, I should update this little test program to actually do audio input/output and test it's recognition quality on my test bot. It's not worth taking more steps unless we know for sure it's really good with a non-Echo array mic. That's probably a good couple of days to work out, even just to do the least possible hacked version just for testing. But , assuming it works well, that will pay off later in having worked out the audio processing stuff up front in this simpler environment.
Dean Roddey
Software Geek Extraordinaire
Reply


Possibly Related Threads...
Thread Author Replies Views Last Post
  Control of Epson Projector Using IR Jnetto 6 467 08-08-2017, 05:34 PM
Last Post: Jnetto
  Announcing CQCVoice - CQC's All Local Voice Control System Dean Roddey 66 7,753 06-28-2017, 10:03 AM
Last Post: Dean Roddey
  Any way to control Wolf and Subzero brand appliances ghurty 8 1,133 06-29-2016, 06:53 PM
Last Post: znelbok
  A/V Component Control rbejr 20 2,127 03-25-2016, 09:03 AM
Last Post: Bugman
  A 'Find Widgets' option Dean Roddey 14 1,028 01-15-2016, 02:20 PM
Last Post: Dean Roddey
  IV remote control dlmorgan999 5 1,111 07-04-2015, 08:23 AM
Last Post: dlmorgan999
  Minor sort issue with states dlmorgan999 2 759 06-03-2015, 07:52 AM
Last Post: dlmorgan999
  Generic app control - 4.5.0 Mark Stega 33 3,153 10-26-2014, 05:07 PM
Last Post: Dean Roddey
  CQC/USB-UIRT to control Comcast STB pbohannon 4 1,289 09-08-2014, 07:56 AM
Last Post: zaccari
  How to tell a popup light switch which one to control ghurty 1 854 05-06-2014, 07:43 PM
Last Post: Dean Roddey

Forum Jump:


Users browsing this thread: 1 Guest(s)