Messages

1

"Open air" and "whole home" microphones / Re: Anyone gonna try this out? Amazon echo

« on: September 23, 2015, 11:41:55 AM »

@shanekuz,
10-4.  I really didn't intend to be negative or anything and I apologize if I rambled on about things that you already know and believe. 

I am familiar with Polycom from work (conference room mics).  I hear what you are saying - as I ran into that exact same issues.  I would say my biggest issue was background noise.  It could be a tv, washing machine, people in another room, dogs, about anything.  Unless I had everything adjusted spot on, had silence in the room, and was positioned properly relative to the microphone (in my case a Kinect or a conference mic I got), my recognition was not good enough.  And rarely did I have the ideal conditions.  I don't really blame anything.  I understand that all sound goes in and I cannot expect it to be smart enough (at this point) to clearly understand my voice amongst all of the other sounds.  That's why I have found VoxWav to be ideal.  I have great success with it, but then my goal is to be able to just say things and not have to get a phone out.

BTW - I have not even voice trained my Echo.  I wonder how much better it would be with training.  It is already blowing my mind with how great it works, even with noise.  I can ask Alexa to tell me the definition of unusual words and I am surprised how often she gets it right.  Pretty much all the time.  Unless the noise in the room is particularly bad.  I've even had success telling her to do things with the TV on in the same room.  I do have to say it louder but it would seem she is able to lock onto my voice and focus on it (my non-technical understanding). 

In your case, or even with the Echo, I think it would be great to be able to train in such a way that the device recognizes voices like fingerprints.  Run through some kind of training routine with them and have it pick up on certain quantifiable values that can be used to know 'this is a voice, this is what I am supposed to hear, and as little of everything else as possible'.  I've seen something similar to this in S.A.R.A.H where you can use a Kinect and it will assemble a profile of you based on imaging, voice pitch, mood, things like that.  But I'm thinking even more specific, and using just the voice.  -- Just something I have thought about.

Finally, even with a wake word, I would get false positives and not really understand how (coming from TV).  A unique wake word should help with that.  On a related note, I have watched YouTube videos on my TV about Alexa development, and when the person in the video issues a command, if it is one available to all Echo users, my Echo would pick it up and respond.  The other fascinating thing is the timing would be such that she would respond to me almost exactly at the same time as she did for the person in the video.

Shanekuz, what are your plans for the future?  Are you planning to try and integrate more with Alexa over time?  Are you planning to add anything new to your environment?  Thanks again for the responses and please share any tips you have about Echo and Vox.  One thing I am interested in knowing is if you are using more than one instance of VoxCommando in your setup and how well routing the packets to the proper UDP listener is working.

Thanks,
tobias.

2

"Open air" and "whole home" microphones / Re: Anyone gonna try this out? Amazon echo

« on: September 20, 2015, 04:22:46 PM »

@shanekuz,
I was re-reading your post and noticed something.  And, please understand I am only saying this to clarify, this is not meant as anything negative.  VoxCommando does not need to improve in speech recognition.  It relies on other components, both software and hardware, for speech recognition.  Vox itself processes the recognized speech, and does so very nicely, but really the quality of the recognition comes down to things like microphones, MS technologies for speech recognition, or any other software for that matter (CMU Sphinx or something like that). 

Dictation speech recognition is extremely difficult to have high accuracy and high confidence, and correctness, with today's technologies.  Without something behind the speech to say 'I think I heard this ... is this ... in the list? Then recognition becomes very difficult.  Google, Amazon, Siri, I would imagine they do so well because 1) the person it typically speaking directly into the mic right in front of their face (good input to begin with) 2) they probably have developed sophisticated matching algorithms to help turn the babble into something recognizable.  Point being, Vox itself does not need to improve in recognition, everything else does.

If you google around you can come up with your own little test to see what I am talking about.  If you use MS speech recognition (not Cortana), and say 'open Word', it will probably open Word.  But if you open Word and then speak a letter to someone (especially if you do not train it first, and are not sitting close to the mic) watch how horrible it is at figuring out what you are saying.  And if you want to dig even deeper, read some of the articles on the net about how speech recognition on computers words.  It is a very complex topic and amazing that it works as well as it does period.

All that being said, yes, I would imagine as time rolls by these technologies will get better and better to the point where some of the issues I mentioned will not be such an issue.

Your response to me suggestions that you are having pretty good success with Echo and Vox.  A proper Alexa Skill will have an utterances file, which as we now know, greatly improves the chance of recognition.  Vox takes care of generating massive amounts of speech recognition into xml (tv shows, movies, music (artist, album, song)).  This is wonderful.  But I wonder how good Echo will work with no utterance file and purely grabbing the speech, turning it to text, parsing it and then sending it to Vox.  Any mistake and the system won't work.  I suppose the best thing I can do is just try.  I'll be surprised if it works anywhere near as good as VoxWav though.

Thank you again for your post and sharing.

tobias.

3

"Open air" and "whole home" microphones / Re: Anyone gonna try this out? Amazon echo

« on: September 19, 2015, 02:23:50 PM »

Hi shanekuz,
Just wondering if you have any code you could share with me?  Maybe not... I was kind of wondering if you have setup a node server type setup to route the voice commands to the appropriate VC destination.  Also, is there an ISY device in your setup?  Possibly running the beta v5 firmware?  And if so, just wondering if you have setup a node server with that.  I've been waiting for them to release their demo code but since you are functional I just wondered if you had anything I could look at to maybe help give me a jump start.

All in all glad to hear it is working?  Are you finding the speech recognition to be pretty accurate.  I've not been able to do anything since around the time of my above post.  I had to go out of town for a training course and I just go back.

If I come up with anything that would be of value to you I'll be happy to share.

Thanks,
tobias.

4

"Open air" and "whole home" microphones / Re: Anyone gonna try this out? Amazon echo

« on: September 04, 2015, 03:37:34 PM »

Hi all,
I looked through the posts and I don't see much interest in using an Echo as a mic for VC.  Despite that I would still like to run this by you.  If I could successfully use VC with Echo, it has the potential to be very cool.  The Kinect drives me crazy and VoxWav is by far the best way to communicate but that requires using the phone which ideally I'd like to just be able to speak and not get the phone out.

To keep it short as possible I'm not going to go into details on the Alexa Skills Kit (what Amazon setup to allow developers to extend what the Echo can do). 

Consider this an experiment.  Will it work kind of thing.  Skipping over how Echo works for custom development I will say this.  There is code on GitHub where a person has bypassed the whole Amazon way and is just having the Echo send back what was heard as text to his HTTP endpoint on his home network.  From there he parses the text to figure out the intent looking for keywords.  From there a module is loaded to handle the speech designed for the particular intent.  My thoughts right now are on using this with Emby.  So here is what I am thinking.

I say Alexa, ask Emby to open (that is sent to Amazon AWS, a Lambda function I create does what I want, and the text is sent back to my listening web server). Alexa can do conversation so she responds, 'ok, which room?'.  I say 'game room' or 'bedroom' (I have multiple instances of Emby and VC).  This will then give my code on my endpoint the info it needs to know which address and port to send commands to the VC UDP listener.  Next, I say Alexa, ask Emby to launch Media Browser Theater. The text comes back to my endpoint, which now knows what is going on and where to send, and evaluates the text to clean it up if needed and send to the VC listener.  Hopefully, VC then opens Emby theater.  Without grinding through the details, that is the basic idea.  Triggering a particular intent, coding on my end (the http listener box) to know how to deal with the text coming back, and then trying to send the proper text to the VC listener which will then do what it does.

Technically this should be possible, as far as I know.  Speaking from the Amazon side of things at least.  All that I said can be done.  I do not know yet how well the recognition of dictation speech will work out.  I also realize that what code I need to develop may be a pain but based on the examples I've reviewed, I've seen where others are doing this with (according to them) good success.  I know that the text going into VC has to be exactly correct to work. 

You may say, instead of doing that with an Echo, why don't you do this or this.  I am open to ideas.  I'm also open to 'did you think about this?' or 'it probably won't work because...'.  But in addition my Emby goal, I am helping a person in another state who has all kinds of home automation devices and wants me to do many things for him utilizing these devices.  I'm doing it for fun because it is a great learning opportunity for me and with all that he has, there is plenty of room for experimentation.

He has asked me to allow him to control any of 6 Sonos systems with Amazon Echo.  I've already seen where people have done this.  But I'm hoping to create a web service for the home that doesn't just do one thing based on something someone else did, but look at the big picture and come up with an extensible foundation for the home web service where the Echo can be used in home automation (beyond the built in support).  Actually, I'd be extremely happy just to accomplish being able to send viable speech commands as text to VoxCommando.

I'd like to leverage VC again for his Sonos request.  So this post is sort of being driven by trying to solve his request, but at the same time I'd love to work with Emby for myself.  And I know that in both cases there is the issue of speaking to a device while tv or music is playing, and how hard that can/will be.  If you tell the Echo to play Pandora station and then you want to say something else to it, it will hear you.  But I imagine that will be tougher once the sound is coming from somewhere other than the echo device.  Not to ramble, but she has impressed me with her ability to hear with the tv on, or the dogs all barking, or other noise.  Things that I could never manage to do with the kinect.  So the Echo seems to be good at picking out my voice in all of the noise.

Thank you all and again, thank you for VoxCommando.  It is a truly awesome program.

tobias.

5

XML Exchange / Re: Basic maths with Python -- Times tables quiz

« on: May 09, 2015, 11:10:22 PM »

Cool   Thanks nime5ter!

6

General Discussion / Re: alternate to Python (or Iron Python) for custom coding

« on: May 01, 2015, 08:57:23 AM »

Hi nime5ter,
All helpful responses, and good starting point for my wife.  And I just looked at that wiki entry and yeah, it answers the question.  I don't know what to say.  I looked at forgot?  eeyeeyyyee.  Yeah, no issues there.

I've provided plenty of help and guidance to my question.  I appreciate the time you spent on this and I apologize for apparently forgetting or getting confused, or just being idiot.

I think I may benefit from not trying to half-way do 5 different things and stick to one thing and give it full attention.

If you don't hear from me for a while on here that's probably a good thing. 

Again, much help and I appreciate your time.  I agree with you that having my wife learn on her own we may be able to help each other.

Thanks,
tobias.

7

General Discussion / Re: alternate to Python (or Iron Python) for custom coding

« on: April 30, 2015, 05:29:58 PM »

Hi nime5ter,
Thank you for your response.  And what you said is fine.  I am ignorant, I admit.  While I do understand somethings in VC, I know there is a ton I don't.  It is not that I have not made an effort.  I think I just have difficultly sometimes understanding what is simplistic for you or others.

I'll give an example or two and probably make myself look like a fool. 

1) Adding a conditional - Unfortunately I'm at work and I don't have it up in front of me.  But something like (A)==(B).  I'm not sure that I understand what determines the values of A and B.  LastActionSuccess - what is that based on?  If I make a command to turn on a light, the command is sent, is VC always checking that the light or whatever else it might be, returned success? 

You are absolutely correct that I'm reaching for what I know rather than learning what I don't.  I think this is not the first time the message has been sent to me.  I think part of the issue is I've spent weeks trying to learn so many new things I'm getting lazy or tired. 

Thanks for the response.  I'll try to quit whining and just try harder and learning everything VC has to offer.

8

Off Topic (not related to VC) / Experience with OpenRemote?

« on: April 30, 2015, 03:42:34 PM »

Hello,
I think many of us are on a similar journey.  Home Automation is becoming more popular, cheaper, more available, more powerful, and what we have now is probably going to seem primitive in a few years.  On top of that many of us want speech recognition in our homes.  It seems to me we are in the early stages where new standards and protocols are popping up all the time, change is constant, and it makes this difficult. 

Fortunately we have the diamonds in the rough.  Like VoxCommando. 

To my point... I feel like my implementation is moving kinda slow, requiring a lot of trying this and that, lots of reading and testing, patience.  While I try to get things integrated and working better and better, I need some options to fall back on so the family doesn't bail on me.  I want them to get into this and not 'just use the tv remote'.  So I want to make a custom, simple to use app that anyone at the house can fall back on to control a light, or open Vox Command on a media center where it is not running, these kinds of things.

I started working with OpenRemote.  I'm to the point where I am getting going but I would say still early stages.  I just saw on here iRule so when I am done with this I am going to check it out to see how it compares.

I have some old tablets at the house I can use, my can put it on my wife's phone.  So if there is a hiccup in the system they won't get frustrated with my projects, they can pull out the tablet or phone and be ok.

I was wondering if anyone had recommendations, feedback, advice?  OpenRemote seems like it should do what I need.  And it is free. 

Any thoughts?

Thanks,
tobias.

9

General Discussion / alternate to Python (or Iron Python) for custom coding

« on: April 30, 2015, 03:29:21 PM »

Hello,
As I have stated before, I am not a professional programmer, and I can always stand to learn more.  I first started scripting about 15 years ago.  First languages were Perl, PHP, VBScript, and Javascript.  I got a book on Python and read online.  I am surprised at how many times I have read that Python is a great beginner language.  IMHO, Python is different.  Aside from Python ettiquite, the way things are just really seem to make Python unique to me in comparison to what I just listed.  IMHO the four items I listed all share similarities.

I'm trying to get my wife on board with VoxCommando, home automation, having some fun with technology.  When I was testing the SR program SARAH I had made a learning multiplication tables speech program for the kids.  She would ask the what x times y is and then they would answer and get it right or wrong and be told the correct answer.  My wife wants to get involved more and I really want that to happen.  She wants to make speech math learning stuff for the kids.  I told her she would have to use Python (this was all yesterday).  She looked at Python and then on the way home we talked about it and she told me she didn't care for it and I shared my feelings.

I am wondering, without creating more work for the VC team, how additional scripting/programming support might be accomplished with VC.  When I was testing Castle OS it allowed for C# scripts.  That was great.  I believe that was accomplished by using Mono Develop.  Working with Sarah I was using Javascript.  That is fine.  I would prefer to use something script for most things.  Plus I have to teach my wife all of this (she will self learn too and she is smart so no problem).

Does anyone have any suggestions for me?  Know any paths I could take to use something other than Python?  If I have to use Python I will but I doubt my wife will want to mess with it and I won't really be enjoying it.

If I was asked to recommend... based on my limited knowledge I would lean towards C# script.  That will allow for the use of the .Net library with all the namespaces, importable items through nuget, and should be a very versatile and functional approach.  If not that, then something that can be extended with additional libraries fairly easily.  If someone makes a Wake-On-Lan library and puts it on GitHub, it would be nice to be using a language that would allow us to grab that and use it.

Those of you who are much more knowledgeable in this area may have better ideas or suggestions. 

Thank you,
tobias.

10

General Discussion / Re: Wink Hub - anyone using this device?

« on: April 29, 2015, 04:48:43 PM »

Hi DHHJ,
All of that sounds great to me!  I'm looking forward to its release and seeing what kind of experience people have with it.  Thank you for sharing your input.  tobias.

11

Media Browser (Emby) / Re: Track current status of different MB clients

« on: April 29, 2015, 04:27:19 PM »

It was good news.  I have not had much success with MBT.  A few things work but not enough to really enjoy it.  I hope they settle on something and VoxCommando can then just work with it.  I think Classic still works but I've been trying to get away from Classic.

12

General Discussion / Re: Wink Hub - anyone using this device?

« on: April 27, 2015, 01:06:26 PM »

Great.  The GE bulbs are cheaper and work with the Hue bridge like you said.  Only limitations I have noticed are they will not dim as low as Lux bulbs do.  Otherwise, pretty similar to the Lux bulb. 

Thank you.  I am glad they Hue bridge supports them so I could get away from the Wink hub

13

General Discussion / Re: Wink Hub - anyone using this device?

« on: April 24, 2015, 04:42:20 PM »

Just curious why you ask where I am from.  Seems funny... Or out of the blue...
Greenwood, IN, USA

It is seemingly overpriced because it is basically a custom PC.  In addition, even their software is expensive.  $200 and it is far from being complete...

But if the hub is really a compact win8 machine with custom software, I can see why they are charging a lot. Especially considering they are all about voice integration.

I only mentioned it because for some, it might be an interesting option.

I'd rather have usb sticks that support protocols and let the computer and some software handle everything rather than buying a 'hub'.  I understand why the clould is being used, but I don't like it and I don't want it.  I want control over the data, and I want control over the 'hub'.  I've been told the SmartThings 2 is supposed to have a LAN mode. 

I looked at Insteon.  They have a good reputation for stability and reliability.  But Insteon only wants to play with certain devices and I'd much rather have more options.  That's one reason I went ahead and got a Hue bridge.  It will work with my lights, and now I can have Hue lights if I want.  And it seems to be better designed.

I will probably do the instructions this weekend on changing the DNS servers on my router and getting new firmware pushed to my Wink hub (I actually have two).  But I don't really know what I am going to do with it after that.  The second one I have still has the original firmware with the SQL injection vulnerability.  I got it in case the community started blazing a trail with rooted Wink hubs.  I knew eventually the store shelves would have patched firmware units.

It is too bad that the protocols used in HA are sprouting and splitting and causing limitations for us, the consumer.  Although I want the technology to advance and get better, which may happen through competing protocols, at the same time it would be nice for a interoperable standard so that consumers don't wind up with Betamax players, or HD DVDs.  The mesh network topology seems effective.  It has allowed me to control lights outdoors where my wireless signal will not reach.  I thought it interesting that people exploring the Wink hub (which offers both Zigbee and Z-wave support) has a whitelist for products.  So devices that would work are intentionally disabled. 

14

General Discussion / Re: Noob question (sorry) but I think it will answer a lot for me

« on: April 23, 2015, 10:31:28 PM »

Thank you jitterjames.  Everything is in order and working.  I don't know if it was lack of attention or what but in the groups GUI I didn't realize I could create the groups there.  Instead I did it by opening the groups.xml in the payload editor and did it that way.  Oops.

I appreciate your time and effort.  On my plugin, all lights say 'error reading light status' and that is ok.  I assume it is because all but two are GE Link lights.  Additionally, two of my named lights do not show up (spots 5 and 6 just do not show the name of the light).  Yet those lights do show up in the group editor.  And even though it says error reading light status (in edit lights), if I turn the light on, the checkbox will reflect that.  That message is not concerning to me.

Thank you again,
tobias.

15

Media Browser (Emby) / Re: Track current status of different MB clients

« on: April 23, 2015, 09:41:39 PM »

Hi,
I just wanted to let you know that I updated Emby Theater after getting home and yes it is working.  I did not spend a tremendous amount of time with it confirming what all is working, but from what I can tell much better than before