Good morning, We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list. I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet. I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play. For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..". All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too: { "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" } We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now. The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it. If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this. Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;) Thanks, William VE4VR
Thanks William for the feedback. I find your implementation on the IRLP controler to be a piece of art.. ;-) Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt? did you implement a way to prevent the AI to respond to request sent by a networked audio feed? I am hyper excited by this and cant wait to test this. Pierre VE2PF Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit :
Good morning,
We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list.
I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet.
I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play.
For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice
Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..".
All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too:
{ "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" }
We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now.
The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it.
If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this.
Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;)
Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
Hi Pierre This was just a fresh IRLP install for testing and nothing more. Did the updates. Added the SDK examples and then a few lines changed for transmitter control. Anything with audio input and output should work fine, so could be an IRLP box with sound card, EchoLink USB adapter, DMK USB Radio Interface, NWDR UDRC/DRAWS, or transcoded input/output from AMBE. The commands back to the system are being done via IFTTT HTTPS webhook request right now and just a script that accepts authenticated commands. I think you'll find it's fairly easy to build this functionality into anything. All of the real work happens back in Google/Amazon's massive data centers and these smart speaker devices are really just voice terminals. They're just a small embedded platform like a Raspberry Pi with a microphone array and speaker(s). On the device they're running wakeword detection "Hey Google" | "Alexa" and then start streaming the audio back to the data center for processing. Depending on the product (speaker, TV, coffee maker) there could be code running there to handle the understood command(s). If you want to see some progress today, I'd start with https://github.com/Picovoice/Porcupine and get Porcupine up and running. It also comes with everything you need to get started including wakeword files for "bumblebee" and other words you won't hear on a repeater in common speech. Later you can generate your own custom wakeword files. When I started with Picovoice all my code did was understand a wakeword and respond with a beep tone over the air. There's quite the debate over Google/Internet services being tied to radio. I think the Picovoice Rhino is probably the best area to focus on for on-device repeater control. It includes the Porcupine library for wakeword detection so "OK Repeater" is possible and the intents (turn on the link etc) are all handled within the controller platform, no internet. That's probably where I'll focus my effort on - it will end up being a hybrid system where the repeater controller has basic commands and I can still say "OK Repeater, ask Google ..." and then we're using Google's web services as a secondary method. One of the first voice apps I've been working on is just a radio check. But rather than asking other people on a repeater "how does this radio sound" I can use a SDR on the repeater receiver and determine if the user requesting a check is on frequency, what kind of deviation, DTMF deviation, signal level in db and all that stuff you can do with a $10 RTL SDR. Getting those voice services up and running will open up a pile of new things you can do over the air with analog/digital radios. Best regards William On Sun, Jan 6, 2019 at 10:08 AM Pierre Martel <petem001@gmail.com> wrote:
Thanks William for the feedback.
I find your implementation on the IRLP controler to be a piece of art.. ;-)
Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt?
did you implement a way to prevent the AI to respond to request sent by a networked audio feed?
I am hyper excited by this and cant wait to test this.
Pierre VE2PF
Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit :
Good morning,
We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list.
I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet.
I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play.
For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice
Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..".
All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too:
{ "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" }
We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now.
The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it.
If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this.
Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;)
Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
Pierre, There is not much stopping you from experimenting on your own right now. It helps if you are familiar with asterisk commands and at least bash scripts. AGI scripting a plus, but you can figure that out, it's all well documented. This has been done with asterisk for at least 6-7 years. I'm sure those folks at nervittles.com have some updated packages now-a-days. A quick look shows http://nerdvittles.com/?p=21703 so there is your start. Adapt from that. As far as not triggering it outside of your intended system, that can be accomplished using app_rpt's event management system, a quick look at 'last keyed node' will give you the data to evaluate if the node that gave the attention keyword is one you want to allow the flow of voice commands and can/will prevent network falsifying. The major hick-ups that I had was from all the dev's and hack's kept overloading the free use of dev tools and google and MS kept changing the criteria for how it worked. The more that were playing, the faster the changes came in. Much like voice.google has finally been shut down as a open interface in the end after frequent changes to eliminate those that were using it for unintended purposes. Those changes were a bit slower since google had contracts for devices that worked from them. The same will probably happen to the voice assistant boxes in the end. I have seen so many hacks for them on the internet now. Enjoy and learn from the freebie while you can. Personally, now, I don't have a interest in any of them that rely on external networks to do the task. But a STT/TTS engine run on a local server is not a uncomplicated task if it is to be accurate. And that's why google, MS, IBM and a few others have this pretty much wrapped up for now, but not forever. They did spend about 5+ years developing this and there was quite a battle over the talent to do it between MS and Google. Anyway, I see you have the bug... really, you should go for it ! It is fun. The more that are playing, the better the chances someone comes up with something really useful, not just novel. ...mike/kb8jnm On 1/6/2019 11:08 AM, Pierre Martel wrote:
Thanks William for the feedback.
I find your implementation on the IRLP controler to be a piece of art.. ;-)
Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt?
did you implement a way to prevent the AI to respond to request sent by a networked audio feed?
I am hyper excited by this and cant wait to test this.
Pierre VE2PF
Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com <mailto:wfranzin@gmail.com>> a écrit :
Good morning,
We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list.
I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet.
I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play.
For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice
Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..".
All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too:
{ "type":"espresso", "size":"small", "numberOfShots":"2", "sugar":"a lot", "milk":"some" }
We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now.
The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it.
If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this.
Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;)
Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org <mailto:App_rpt-users@lists.allstarlink.org> http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
Can I ask how you interfaced the audio from and to porcupine into the irlp node? I am not very aware of the irlp interface. Envoyé depuis mon téléphone intelligent Samsung Galaxy. -------- Message d'origine --------De : William Franzin <wfranzin@gmail.com> Date : 19-01-06 12:31 p.m. (GMT-05:00) À : Users of Asterisk app_rpt <app_rpt-users@lists.allstarlink.org> Objet : Re: [App_rpt-users] Voice control options Hi Pierre This was just a fresh IRLP install for testing and nothing more. Did the updates. Added the SDK examples and then a few lines changed for transmitter control. Anything with audio input and output should work fine, so could be an IRLP box with sound card, EchoLink USB adapter, DMK USB Radio Interface, NWDR UDRC/DRAWS, or transcoded input/output from AMBE. The commands back to the system are being done via IFTTT HTTPS webhook request right now and just a script that accepts authenticated commands. I think you'll find it's fairly easy to build this functionality into anything. All of the real work happens back in Google/Amazon's massive data centers and these smart speaker devices are really just voice terminals. They're just a small embedded platform like a Raspberry Pi with a microphone array and speaker(s). On the device they're running wakeword detection "Hey Google" | "Alexa" and then start streaming the audio back to the data center for processing. Depending on the product (speaker, TV, coffee maker) there could be code running there to handle the understood command(s). If you want to see some progress today, I'd start with https://github.com/Picovoice/Porcupine and get Porcupine up and running. It also comes with everything you need to get started including wakeword files for "bumblebee" and other words you won't hear on a repeater in common speech. Later you can generate your own custom wakeword files. When I started with Picovoice all my code did was understand a wakeword and respond with a beep tone over the air. There's quite the debate over Google/Internet services being tied to radio. I think the Picovoice Rhino is probably the best area to focus on for on-device repeater control. It includes the Porcupine library for wakeword detection so "OK Repeater" is possible and the intents (turn on the link etc) are all handled within the controller platform, no internet. That's probably where I'll focus my effort on - it will end up being a hybrid system where the repeater controller has basic commands and I can still say "OK Repeater, ask Google ..." and then we're using Google's web services as a secondary method. One of the first voice apps I've been working on is just a radio check. But rather than asking other people on a repeater "how does this radio sound" I can use a SDR on the repeater receiver and determine if the user requesting a check is on frequency, what kind of deviation, DTMF deviation, signal level in db and all that stuff you can do with a $10 RTL SDR. Getting those voice services up and running will open up a pile of new things you can do over the air with analog/digital radios. Best regards William On Sun, Jan 6, 2019 at 10:08 AM Pierre Martel <petem001@gmail.com> wrote: Thanks William for the feedback. I find your implementation on the IRLP controler to be a piece of art.. ;-) Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt? did you implement a way to prevent the AI to respond to request sent by a networked audio feed? I am hyper excited by this and cant wait to test this. Pierre VE2PF Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit : Good morning, We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list. I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet. I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play. For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..". All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too: { "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" } We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now. The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it. If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this. Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;) Thanks,William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem. _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
IRLP boards have been around for about 20 years and it's just a DTMF decoder wired to the parallel port and interfacing for PTT, COS, PL etc. You don't need anything fancy here. Anything that works with alsa audio in Linux will work fine. Control the PTT using a script and your choice of GPIO. I would run this outside of Asterisk (on the side) and then use it to tell Asterisk which commands I want to execute. Asterisk audio required transcoding. So there was zero hardware work required to add this to the IRLP node. Same would apply anywhere you've got a way to control the PTT output. If you want to try something today use a Raspberry Pi or Linux desktop (Ubuntu etc) and connect a USB headset up. You'll be able to speak and detect. Because this doesn't require COS or PL detect its just listening for the wakeword then listening to speech & executing commands. Just need PTT output. An EchoLink USB adapter is probably the easiest way to do this as the audio hardware is just a USB sound card and they also emulate a serial port for transmitter/receiver interfacing. That would be just a few lines of code needed to turn the serial lines on/off for PTT control. When you're using a wakeword to activate any of these voice services it's continuously listening so doesn't depend on COS or CTCSS. Dan VE4DRK and I were talking about this and he suggested listening for signals without the PL might be a nice way to keep the voice commands away from the audio path when connected to links. Program the same repeater frequency twice into your radio and one has a PL for normal repeater use and the other no PL. Then wakeword still works and repeater keys up with PL encode and nobody hears the voice commands at all, completely silent to links, On Sun, Jan 6, 2019 at 4:42 PM petem001 <petem001@gmail.com> wrote:
Can I ask how you interfaced the audio from and to porcupine into the irlp node?
I am not very aware of the irlp interface.
Envoyé depuis mon téléphone intelligent Samsung Galaxy.
-------- Message d'origine -------- De : William Franzin <wfranzin@gmail.com> Date : 19-01-06 12:31 p.m. (GMT-05:00) À : Users of Asterisk app_rpt <app_rpt-users@lists.allstarlink.org> Objet : Re: [App_rpt-users] Voice control options
Hi Pierre
This was just a fresh IRLP install for testing and nothing more. Did the updates. Added the SDK examples and then a few lines changed for transmitter control. Anything with audio input and output should work fine, so could be an IRLP box with sound card, EchoLink USB adapter, DMK USB Radio Interface, NWDR UDRC/DRAWS, or transcoded input/output from AMBE. The commands back to the system are being done via IFTTT HTTPS webhook request right now and just a script that accepts authenticated commands.
I think you'll find it's fairly easy to build this functionality into anything. All of the real work happens back in Google/Amazon's massive data centers and these smart speaker devices are really just voice terminals. They're just a small embedded platform like a Raspberry Pi with a microphone array and speaker(s). On the device they're running wakeword detection "Hey Google" | "Alexa" and then start streaming the audio back to the data center for processing. Depending on the product (speaker, TV, coffee maker) there could be code running there to handle the understood command(s).
If you want to see some progress today, I'd start with https://github.com/Picovoice/Porcupine and get Porcupine up and running. It also comes with everything you need to get started including wakeword files for "bumblebee" and other words you won't hear on a repeater in common speech. Later you can generate your own custom wakeword files. When I started with Picovoice all my code did was understand a wakeword and respond with a beep tone over the air.
There's quite the debate over Google/Internet services being tied to radio. I think the Picovoice Rhino is probably the best area to focus on for on-device repeater control. It includes the Porcupine library for wakeword detection so "OK Repeater" is possible and the intents (turn on the link etc) are all handled within the controller platform, no internet. That's probably where I'll focus my effort on - it will end up being a hybrid system where the repeater controller has basic commands and I can still say "OK Repeater, ask Google ..." and then we're using Google's web services as a secondary method.
One of the first voice apps I've been working on is just a radio check. But rather than asking other people on a repeater "how does this radio sound" I can use a SDR on the repeater receiver and determine if the user requesting a check is on frequency, what kind of deviation, DTMF deviation, signal level in db and all that stuff you can do with a $10 RTL SDR. Getting those voice services up and running will open up a pile of new things you can do over the air with analog/digital radios.
Best regards William
On Sun, Jan 6, 2019 at 10:08 AM Pierre Martel <petem001@gmail.com> wrote:
Thanks William for the feedback.
I find your implementation on the IRLP controler to be a piece of art.. ;-)
Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt?
did you implement a way to prevent the AI to respond to request sent by a networked audio feed?
I am hyper excited by this and cant wait to test this.
Pierre VE2PF
Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit :
Good morning,
We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list.
I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet.
I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play.
For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice
Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..".
All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too:
{ "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" }
We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now.
The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it.
If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this.
Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;)
Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
Nice info from you Mike. I was aware of the event system in the asterisk software. And the worst is that I spoke of this a few days ago about the multiple ctcss code question. Damn it is true that you dont see the forest that is behind a tree some times ;-) Envoyé depuis mon téléphone intelligent Samsung Galaxy. -------- Message d'origine --------De : Mike <mm@midnighteng.com> Date : 19-01-06 1:58 p.m. (GMT-05:00) À : app_rpt-users@lists.allstarlink.org Objet : Re: [App_rpt-users] Voice control options Pierre, There is not much stopping you from experimenting on your own right now. It helps if you are familiar with asterisk commands and at least bash scripts. AGI scripting a plus, but you can figure that out, it's all well documented. This has been done with asterisk for at least 6-7 years. I'm sure those folks at nervittles.com have some updated packages now-a-days. A quick look shows http://nerdvittles.com/?p=21703 so there is your start. Adapt from that. As far as not triggering it outside of your intended system, that can be accomplished using app_rpt's event management system, a quick look at 'last keyed node' will give you the data to evaluate if the node that gave the attention keyword is one you want to allow the flow of voice commands and can/will prevent network falsifying. The major hick-ups that I had was from all the dev's and hack's kept overloading the free use of dev tools and google and MS kept changing the criteria for how it worked. The more that were playing, the faster the changes came in. Much like voice.google has finally been shut down as a open interface in the end after frequent changes to eliminate those that were using it for unintended purposes. Those changes were a bit slower since google had contracts for devices that worked from them. The same will probably happen to the voice assistant boxes in the end. I have seen so many hacks for them on the internet now. Enjoy and learn from the freebie while you can. Personally, now, I don't have a interest in any of them that rely on external networks to do the task. But a STT/TTS engine run on a local server is not a uncomplicated task if it is to be accurate. And that's why google, MS, IBM and a few others have this pretty much wrapped up for now, but not forever. They did spend about 5+ years developing this and there was quite a battle over the talent to do it between MS and Google. Anyway, I see you have the bug... really, you should go for it ! It is fun. The more that are playing, the better the chances someone comes up with something really useful, not just novel. ...mike/kb8jnm On 1/6/2019 11:08 AM, Pierre Martel wrote: Thanks William for the feedback. I find your implementation on the IRLP controler to be a piece of art.. ;-) Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt? did you implement a way to prevent the AI to respond to request sent by a networked audio feed? I am hyper excited by this and cant wait to test this. Pierre VE2PF Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit : Good morning, We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list. I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet. I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play. For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..". All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too: { "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" } We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now. The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it. If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this. Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;) Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem. _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
Thanks again for the explaination William. I do understand the principle behind the porcupine software.. And as you just told, to test its fonctionality someone just need a mic and speaker that run on the linux sound system. The part that I need to know is how did you pass the ptt info to the irlp board? I will try to run porcupine in a small VM. Next I will need to find how to pipe the audio to and from the Analog_bridge software to Picovoice suite. That will be on the dvswitch mailling list as it has not much to do with allstarlink. I think I am getting somewhere. The planning in such type of project is what makes it fun or a pain and frustration magnet ;-) Le dim. 6 janv. 2019 à 18:11, William Franzin <wfranzin@gmail.com> a écrit :
IRLP boards have been around for about 20 years and it's just a DTMF decoder wired to the parallel port and interfacing for PTT, COS, PL etc. You don't need anything fancy here. Anything that works with alsa audio in Linux will work fine. Control the PTT using a script and your choice of GPIO. I would run this outside of Asterisk (on the side) and then use it to tell Asterisk which commands I want to execute. Asterisk audio required transcoding. So there was zero hardware work required to add this to the IRLP node. Same would apply anywhere you've got a way to control the PTT output.
If you want to try something today use a Raspberry Pi or Linux desktop (Ubuntu etc) and connect a USB headset up. You'll be able to speak and detect. Because this doesn't require COS or PL detect its just listening for the wakeword then listening to speech & executing commands. Just need PTT output. An EchoLink USB adapter is probably the easiest way to do this as the audio hardware is just a USB sound card and they also emulate a serial port for transmitter/receiver interfacing. That would be just a few lines of code needed to turn the serial lines on/off for PTT control.
When you're using a wakeword to activate any of these voice services it's continuously listening so doesn't depend on COS or CTCSS. Dan VE4DRK and I were talking about this and he suggested listening for signals without the PL might be a nice way to keep the voice commands away from the audio path when connected to links. Program the same repeater frequency twice into your radio and one has a PL for normal repeater use and the other no PL. Then wakeword still works and repeater keys up with PL encode and nobody hears the voice commands at all, completely silent to links,
On Sun, Jan 6, 2019 at 4:42 PM petem001 <petem001@gmail.com> wrote:
Can I ask how you interfaced the audio from and to porcupine into the irlp node?
I am not very aware of the irlp interface.
Envoyé depuis mon téléphone intelligent Samsung Galaxy.
-------- Message d'origine -------- De : William Franzin <wfranzin@gmail.com> Date : 19-01-06 12:31 p.m. (GMT-05:00) À : Users of Asterisk app_rpt <app_rpt-users@lists.allstarlink.org> Objet : Re: [App_rpt-users] Voice control options
Hi Pierre
This was just a fresh IRLP install for testing and nothing more. Did the updates. Added the SDK examples and then a few lines changed for transmitter control. Anything with audio input and output should work fine, so could be an IRLP box with sound card, EchoLink USB adapter, DMK USB Radio Interface, NWDR UDRC/DRAWS, or transcoded input/output from AMBE. The commands back to the system are being done via IFTTT HTTPS webhook request right now and just a script that accepts authenticated commands.
I think you'll find it's fairly easy to build this functionality into anything. All of the real work happens back in Google/Amazon's massive data centers and these smart speaker devices are really just voice terminals. They're just a small embedded platform like a Raspberry Pi with a microphone array and speaker(s). On the device they're running wakeword detection "Hey Google" | "Alexa" and then start streaming the audio back to the data center for processing. Depending on the product (speaker, TV, coffee maker) there could be code running there to handle the understood command(s).
If you want to see some progress today, I'd start with https://github.com/Picovoice/Porcupine and get Porcupine up and running. It also comes with everything you need to get started including wakeword files for "bumblebee" and other words you won't hear on a repeater in common speech. Later you can generate your own custom wakeword files. When I started with Picovoice all my code did was understand a wakeword and respond with a beep tone over the air.
There's quite the debate over Google/Internet services being tied to radio. I think the Picovoice Rhino is probably the best area to focus on for on-device repeater control. It includes the Porcupine library for wakeword detection so "OK Repeater" is possible and the intents (turn on the link etc) are all handled within the controller platform, no internet. That's probably where I'll focus my effort on - it will end up being a hybrid system where the repeater controller has basic commands and I can still say "OK Repeater, ask Google ..." and then we're using Google's web services as a secondary method.
One of the first voice apps I've been working on is just a radio check. But rather than asking other people on a repeater "how does this radio sound" I can use a SDR on the repeater receiver and determine if the user requesting a check is on frequency, what kind of deviation, DTMF deviation, signal level in db and all that stuff you can do with a $10 RTL SDR. Getting those voice services up and running will open up a pile of new things you can do over the air with analog/digital radios.
Best regards William
On Sun, Jan 6, 2019 at 10:08 AM Pierre Martel <petem001@gmail.com> wrote:
Thanks William for the feedback.
I find your implementation on the IRLP controler to be a piece of art.. ;-)
Can I ask you what method did you used for the irlp controler? was it the original irlp software or did you used the irlp driver in app_rpt?
did you implement a way to prevent the AI to respond to request sent by a networked audio feed?
I am hyper excited by this and cant wait to test this.
Pierre VE2PF
Le dim. 6 janv. 2019 à 08:53, William Franzin <wfranzin@gmail.com> a écrit :
Good morning,
We're also keen AllStarLink/app_rpt users/experimenters here in Winnipeg and Dan mentioned a discussion about the work I did on here, so thought I'd fill in the blanks. Also joined this list.
I started with Google Assistant and Amazon Alexa and really just wanted to control devices using voice, the "what's the weather" was just an added bonus, but all that requires internet.
I also liked the speed of running the assistant code in real-time, as you speak that's actually a stream to Google/Amazon so the results you get are immediate vs rec/upload/download/play.
For the majority of ham radio actual "needs" in voice control the Picovoice Rhino project speech-to-intent would do a great job. If you haven't check it out: https://github.com/Picovoice
Also watch their video: https://www.youtube.com/watch?v=WadKhfLyqTQ which I think is exactly what we'd want from a ham radio setup. Wakework "Hey Repeater" "Turn on ..".
All of that great capability runs entirely on the device/server/reflector, no internet or web service call to Google/Amazon is needed. And it produces this simple output too:
{ "type": "espresso", "size": "small", "numberOfShots": "2", "sugar": "a lot", "milk": "some" }
We should do this, my 2010 Ford Escape has had Microsoft Sync voice control and it's nothing new. Everything has voice control these days, and it actually works really well now.
The only challenge in ham is we need to come up with some standard commands so if you're visiting here you know the repeater has this capability and how to work it.
If anything I'm just glad the videos I posted a while back got this topic going again. Certainly not the first person to poke around with voice - it's just really easy now to implement this.
Thanks to all the folks working on AllStarLink/app_rpt too. I just built a firmware image with AllStarLink for a Sierra Radio Systems CommServer board and it was apt-get-easy --let-me-drink-my-coffee-while-the-hard-work-gets-done ;)
Thanks, William VE4VR _______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
_______________________________________________ App_rpt-users mailing list App_rpt-users@lists.allstarlink.org http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users
To unsubscribe from this list please visit http://lists.allstarlink.org/cgi-bin/mailman/listinfo/app_rpt-users and scroll down to the bottom of the page. Enter your email address and press the "Unsubscribe or edit options button" You do not need a password to unsubscribe, you can do it via email confirmation. If you have trouble unsubscribing, please send a message to the list detailing the problem.
participants (4)
-
Mike -
petem001 -
Pierre Martel
-
William Franzin