From: Jim D. <ji...@du...> - 2008-03-11 03:07:59
|
I pushed some updates to the Pocketsphinx voice recognition code to svn. Basically, I moved the bulk of the control software down into a PocketSphinx.pm library routine. This allows me to create different listener threads in different code modules. I needed to do this such that I could control a separate listener "on demand" in an asterisk interface code module. When a phone extension is dialed, a new listener is started to handle the call. When the call ends, the listener is stopped. This was necessary such that the listener is only started when the phone is active, since pocketsphinx must first analyze the background sound level upon starting. If you are using pocketsphinx now, I don't expect you to have to change anything after updating from svn. It should just work the same. (I've heard these claims from you before Jim :-) Let me know if any problems arise. The next thing I plan to do is make the library smarter and prevent the compilation of the language models in the event no new voice commands have been created since the last restart. This saves some computer cycles at each startup which aren't necessary if there aren't any new voice commands. Let me know if anyone is interested in how I interfaced asterisk to pocketsphinx and I'll do a brain dump in another message. Can anyone add stuff to the MH wiki? Jim |
From: Jim D. <ji...@du...> - 2012-03-12 00:55:41
|
I updated the pocketsphinx implementation today in both the trunk and the insteon-branch. The updates support the latest 0.7 version of Sphinxbase, Pocketsphinx, and the CMU Language Tookit. If you haven't previously installed the CMU Language Toolkit, you will need to do so with these updates. The new version uses the Language Toolkit to build a binary language model. The new implementation is so much better, so much faster, and appears more reliable IMHO. Please let me know if the changes I pushed to either the trunk or insteon-branch have any issues and I will fix them immediately. I don't believe I broke anything outside of Pocketsphinx. Regards, Jim |
From: Tom M. <tma...@sa...> - 2008-03-11 13:39:16
|
Hi Jim, Yes, I am most interested in how you interfaced pocketsphinx to asterisk! Please dump! I much appreciate your work in this area. I'm only getting started with MH, but I have been running Trixbox for almost a year. Cheers, tom -----Original Message----- From: mis...@li... [mailto:mis...@li...] On Behalf Of Jim Duda Sent: Monday, March 10, 2008 11:08 PM To: mis...@li... Subject: [mh] Pocketsphinx Update I pushed some updates to the Pocketsphinx voice recognition code to svn. Basically, I moved the bulk of the control software down into a PocketSphinx.pm library routine. This allows me to create different listener threads in different code modules. I needed to do this such that I could control a separate listener "on demand" in an asterisk interface code module. When a phone extension is dialed, a new listener is started to handle the call. When the call ends, the listener is stopped. This was necessary such that the listener is only started when the phone is active, since pocketsphinx must first analyze the background sound level upon starting. If you are using pocketsphinx now, I don't expect you to have to change anything after updating from svn. It should just work the same. (I've heard these claims from you before Jim :-) Let me know if any problems arise. The next thing I plan to do is make the library smarter and prevent the compilation of the language models in the event no new voice commands have been created since the last restart. This saves some computer cycles at each startup which aren't necessary if there aren't any new voice commands. Let me know if anyone is interested in how I interfaced asterisk to pocketsphinx and I'll do a brain dump in another message. Can anyone add stuff to the MH wiki? Jim ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ ________________________________________________________ To unsubscribe from this list, go to: http://sourceforge.net/mail/?group_id=1365 |
From: Bulek <bu...@em...> - 2008-03-12 12:12:10
|
----- Original Message ----- From: "Jim Duda" <ji...@du...> To: <mis...@li...> Sent: Tuesday, March 11, 2008 4:07 AM Subject: [mh] Pocketsphinx Update >I pushed some updates to the Pocketsphinx voice recognition code to svn. > > Basically, I moved the bulk of the control software down into a > PocketSphinx.pm library routine. This allows me to create different > listener threads in different code modules. > > I needed to do this such that I could control a separate listener "on > demand" in an asterisk interface code module. When a phone extension is > dialed, a new listener is started to handle the call. When the call > ends, the listener is stopped. This was necessary such that the > listener is only started when the phone is active, since pocketsphinx > must first analyze the background sound level upon starting. > > If you are using pocketsphinx now, I don't expect you to have to change > anything after updating from svn. It should just work the same. (I've > heard these claims from you before Jim :-) > > Let me know if any problems arise. > > The next thing I plan to do is make the library smarter and prevent the > compilation of the language models in the event no new voice commands > have been created since the last restart. This saves some computer > cycles at each startup which aren't necessary if there aren't any new > voice commands. > > Let me know if anyone is interested in how I interfaced asterisk to > pocketsphinx and I'll do a brain dump in another message. > I'm surely interested... Currently I'm not aware of any convenient method to connect speech recognition to Asterisk (beside commercial Lumenvox solution)... So more info you write, more we will learn.... :-) Thanks for great effort... I'll take a look and see if I can help somewhere down the line... BTW, heard for project Simon Listens ? ( http://www.simon-listens.org/ ) Regards, Rob. |
From: Jim D. <ji...@du...> - 2008-03-12 15:23:07
|
Yes, I have seen the simon-listens project. It's no my todo list to investigate in more detail :-) I will try to get some asterisk pocketsphinx information posted this weekend. Jim "Bulek" <bu...@em...> wrote in message news:002901c8843a$41055fe0$c800a8c0@Lapitopi2... > > ----- Original Message ----- > From: "Jim Duda" <ji...@du...> > To: <mis...@li...> > Sent: Tuesday, March 11, 2008 4:07 AM > Subject: [mh] Pocketsphinx Update > > >>I pushed some updates to the Pocketsphinx voice recognition code to svn. >> >> Basically, I moved the bulk of the control software down into a >> PocketSphinx.pm library routine. This allows me to create different >> listener threads in different code modules. >> >> I needed to do this such that I could control a separate listener "on >> demand" in an asterisk interface code module. When a phone extension is >> dialed, a new listener is started to handle the call. When the call >> ends, the listener is stopped. This was necessary such that the >> listener is only started when the phone is active, since pocketsphinx >> must first analyze the background sound level upon starting. >> >> If you are using pocketsphinx now, I don't expect you to have to change >> anything after updating from svn. It should just work the same. (I've >> heard these claims from you before Jim :-) >> >> Let me know if any problems arise. >> >> The next thing I plan to do is make the library smarter and prevent the >> compilation of the language models in the event no new voice commands >> have been created since the last restart. This saves some computer >> cycles at each startup which aren't necessary if there aren't any new >> voice commands. >> >> Let me know if anyone is interested in how I interfaced asterisk to >> pocketsphinx and I'll do a brain dump in another message. >> > I'm surely interested... Currently I'm not aware of any convenient method to > connect speech recognition to Asterisk (beside commercial Lumenvox > solution)... > So more info you write, more we will learn.... :-) > > Thanks for great effort... I'll take a look and see if I can help somewhere > down the line... > > BTW, heard for project Simon Listens ? ( http://www.simon-listens.org/ ) > > Regards, > > Rob. > > > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > ________________________________________________________ > To unsubscribe from this list, go to: http://sourceforge.net/mail/?group_id=1365 > > |