Having trouble with Lumenvox. Could be a config error. I had this working with an older build. Mod_lumenvox is loaded successfully but the pizza demo keeps hanging up on me with this error: 2008-07-14 10:21:47 [NOTICE] switch_ivr_async.c:1773 switch_ivr_detect_speech() Hangup sofia/internal/fred-sip@200.56.122.8[CS_EXECUTE] [DESTINATION_OUT_OF_ORDER] 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:802 switch_core_session_thread() Session 1 (sofia/internal/fred-sip@200.56.122.8) Ended 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:804 switch_core_session_thread() Close Channel sofia/internal/ fred-sip@200.56.122.8 [CS_HANGUP]
Can you press F8 and let me know what that says? When you try to run the demo. /b On Jul 14, 2008, at 1:27 AM, Frederick Jabre wrote:
Having trouble with Lumenvox. Could be a config error. I had this working with an older build. Mod_lumenvox is loaded successfully but the pizza demo keeps hanging up on me with this error:
2008-07-14 10:21:47 [NOTICE] switch_ivr_async.c:1773 switch_ivr_detect_speech() Hangup sofia/internal/fred- sip@200.56.122.8 [CS_EXECUTE] [DESTINATION_OUT_OF_ORDER] 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:802 switch_core_session_thread() Session 1 (sofia/internal/fred-sip@200.56.122.8 ) Ended 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:804 switch_core_session_thread() Close Channel sofia/internal/fred-sip@200.56.122.8 [CS_HANGUP]
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Brian West sip:brian@freeswitch.org
That did the trick Brian. Grammar files were in the wrong place as reported by debug. Doh..! BTW, freeswitch/lumenvox was below 3%-5% CPU during the entire pizza demo, so pocketsphinx definitely needs to be tweaked.. ;) Thanks On Mon, Jul 14, 2008 at 9:49 AM, Brian West <brian@freeswitch.org> wrote:
Can you press F8 and let me know what that says? When you try to run the demo. /b
On Jul 14, 2008, at 1:27 AM, Frederick Jabre wrote:
Having trouble with Lumenvox. Could be a config error. I had this working with an older build. Mod_lumenvox is loaded successfully but the pizza demo keeps hanging up on me with this error:
2008-07-14 10:21:47 [NOTICE] switch_ivr_async.c:1773 switch_ivr_detect_speech() Hangup sofia/internal/fred-sip@200.56.122.8[CS_EXECUTE] [DESTINATION_OUT_OF_ORDER] 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:802 switch_core_session_thread() Session 1 (sofia/internal/ fred-sip@200.56.122.8) Ended 2008-07-14 10:21:47 [NOTICE] switch_core_session.c:804 switch_core_session_thread() Close Channel sofia/internal/ fred-sip@200.56.122.8 [CS_HANGUP]
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Brian West sip:brian@freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I have only seen it go nutz on the yes and no grammar ... which is weird. It doesn't do it every time. I'll work with david and see if maybe we can nail down what is going on. On Jul 14, 2008, at 10:21 AM, Frederick Jabre wrote:
BTW, freeswitch/lumenvox was below 3%-5% CPU during the entire pizza demo, so pocketsphinx definitely needs to be tweaked.. ;)
Brian West sip:brian@freeswitch.org
What kind of hardware you running it on? This has been going to 100% CPU almost every time it does an interp for me. Tested on my P4 box and my AMD 4200+ Dual both with 1GB ram. Both with the same awful results. This can't be right given this thing was designed to run on an ARM processor. I emailed Dave about it as well. Perhaps we need to incent him more ;) Fred On Mon, Jul 14, 2008 at 11:40 AM, Brian West <brian@freeswitch.org> wrote:
I have only seen it go nutz on the yes and no grammar ... which is weird. It doesn't do it every time. I'll work with david and see if maybe we can nail down what is going on.
On Jul 14, 2008, at 10:21 AM, Frederick Jabre wrote:
BTW, freeswitch/lumenvox was below 3%-5% CPU during the entire pizza demo, so pocketsphinx definitely needs to be tweaked.. ;)
Brian West sip:brian@freeswitch.org <sip%3Abrian@freeswitch.org>
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
It wasn't doing this till he did some fixes for the confidence score. I know when I was working on and testing while we were writing mod_pocketsphinx it sure didn't jump like that.. 30-40% tops with a few spikes to 100%. I personally tossed some money at him for the confidence score stuff which is the root of the evilness 100% cpu I suspect because it didn't do that before those last few changes went in. /b On Jul 14, 2008, at 11:35 AM, Frederick Jabre wrote:
What kind of hardware you running it on? This has been going to 100% CPU almost every time it does an interp for me. Tested on my P4 box and my AMD 4200+ Dual both with 1GB ram. Both with the same awful results. This can't be right given this thing was designed to run on an ARM processor.
I emailed Dave about it as well. Perhaps we need to incent him more ;)
Fred
Brian West sip:brian@freeswitch.org
Very well. Even at 30-40% CPU we'll have problem's scaling. Some more testing reveals that it's probably somewhere between 40-100% for 9 out of 10 of interps BUT interestingly enough the footprint in memory is quite low as compared to lumenvox, hovering around 3-5% mem usage compared to lumenvox's 25%-30% memory for a single interp. Ouch..! Some weird trade-off..? Memory for CPU usage. I hate those. Space/Time tradeoff crap. Brings back memories from comp lab... I'll toss some conditional additional cash at him if he can optimize this further. ;) Fred On Mon, Jul 14, 2008 at 12:44 PM, Brian West <brian@freeswitch.org> wrote:
It wasn't doing this till he did some fixes for the confidence score. I know when I was working on and testing while we were writing mod_pocketsphinx it sure didn't jump like that.. 30-40% tops with a few spikes to 100%. I personally tossed some money at him for the confidence score stuff which is the root of the evilness 100% cpu I suspect because it didn't do that before those last few changes went in.
/b
On Jul 14, 2008, at 11:35 AM, Frederick Jabre wrote:
What kind of hardware you running it on? This has been going to 100% CPU almost every time it does an interp for me. Tested on my P4 box and my AMD 4200+ Dual both with 1GB ram. Both with the same awful results. This can't be right given this thing was designed to run on an ARM processor.
I emailed Dave about it as well. Perhaps we need to incent him more ;)
Fred
Brian West sip:brian@freeswitch.org <sip%3Abrian@freeswitch.org>
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I think the issue we have is we keep feeding ps_process_raw when you stop talking. We need to not do that I think :P /b On Jul 14, 2008, at 12:30 PM, Frederick Jabre wrote:
Very well. Even at 30-40% CPU we'll have problem's scaling. Some more testing reveals that it's probably somewhere between 40-100% for 9 out of 10 of interps BUT interestingly enough the footprint in memory is quite low as compared to lumenvox, hovering around 3-5% mem usage compared to lumenvox's 25%-30% memory for a single interp. Ouch..!
Some weird trade-off..? Memory for CPU usage. I hate those. Space/ Time tradeoff crap. Brings back memories from comp lab...
I'll toss some conditional additional cash at him if he can optimize this further. ;)
Fred
Brian West sip:brian@freeswitch.org
Please SVN update you'll see 1-6% cpu on the detector also cd libs; rm -rf pocketsphinx* sphinxbase* I'm pretty sure that was the issue .. I made it stop feeding the recognizer when we stopped talking. /b On Jul 14, 2008, at 12:30 PM, Frederick Jabre wrote:
Very well. Even at 30-40% CPU we'll have problem's scaling. Some more testing reveals that it's probably somewhere between 40-100% for 9 out of 10 of interps BUT interestingly enough the footprint in memory is quite low as compared to lumenvox, hovering around 3-5% mem usage compared to lumenvox's 25%-30% memory for a single interp. Ouch..!
Some weird trade-off..? Memory for CPU usage. I hate those. Space/ Time tradeoff crap. Brings back memories from comp lab...
I'll toss some conditional additional cash at him if he can optimize this further. ;)
Fred
On Mon, Jul 14, 2008 at 12:44 PM, Brian West <brian@freeswitch.org> wrote: It wasn't doing this till he did some fixes for the confidence score. I know when I was working on and testing while we were writing mod_pocketsphinx it sure didn't jump like that.. 30-40% tops with a few spikes to 100%. I personally tossed some money at him for the confidence score stuff which is the root of the evilness 100% cpu I suspect because it didn't do that before those last few changes went in.
/b
On Jul 14, 2008, at 11:35 AM, Frederick Jabre wrote:
What kind of hardware you running it on? This has been going to 100% CPU almost every time it does an interp for me. Tested on my P4 box and my AMD 4200+ Dual both with 1GB ram. Both with the same awful results. This can't be right given this thing was designed to run on an ARM processor.
I emailed Dave about it as well. Perhaps we need to incent him more ;)
Fred
Brian West sip:brian@freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Brian West sip:brian@freeswitch.org
That did help a bit. It will still spike to 60-100% occasionally but its a noticeable difference. Now its mostly between 3-20% CPU. In lumenvox there is a way to tell the speech engine to only detect when the soundfile is finished playing. Is there a way to do that here? or is it set to interrupt-speaker mode all the time..? Fred On Mon, Jul 14, 2008 at 2:06 PM, Brian West <brian@freeswitch.org> wrote:
Please SVN update you'll see 1-6% cpu on the detector also cd libs; rm -rf pocketsphinx* sphinxbase* I'm pretty sure that was the issue .. I made it stop feeding the recognizer when we stopped talking.
/b
On Jul 14, 2008, at 12:30 PM, Frederick Jabre wrote:
Very well. Even at 30-40% CPU we'll have problem's scaling. Some more testing reveals that it's probably somewhere between 40-100% for 9 out of 10 of interps BUT interestingly enough the footprint in memory is quite low as compared to lumenvox, hovering around 3-5% mem usage compared to lumenvox's 25%-30% memory for a single interp. Ouch..!
Some weird trade-off..? Memory for CPU usage. I hate those. Space/Time tradeoff crap. Brings back memories from comp lab...
I'll toss some conditional additional cash at him if he can optimize this further. ;)
Fred
On Mon, Jul 14, 2008 at 12:44 PM, Brian West <brian@freeswitch.org> wrote:
It wasn't doing this till he did some fixes for the confidence score. I know when I was working on and testing while we were writing mod_pocketsphinx it sure didn't jump like that.. 30-40% tops with a few spikes to 100%. I personally tossed some money at him for the confidence score stuff which is the root of the evilness 100% cpu I suspect because it didn't do that before those last few changes went in.
/b
On Jul 14, 2008, at 11:35 AM, Frederick Jabre wrote:
What kind of hardware you running it on? This has been going to 100% CPU almost every time it does an interp for me. Tested on my P4 box and my AMD 4200+ Dual both with 1GB ram. Both with the same awful results. This can't be right given this thing was designed to run on an ARM processor.
I emailed Dave about it as well. Perhaps we need to incent him more ;)
Fred
Brian West sip:brian@freeswitch.org <sip%3Abrian@freeswitch.org>
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Brian West sip:brian@freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I think I see what exactly is going on.. its something thats very rare in this module but give me some time and I think we can nail it down. /b On Jul 14, 2008, at 1:38 PM, Frederick Jabre wrote:
That did help a bit. It will still spike to 60-100% occasionally but its a noticeable difference. Now its mostly between 3-20% CPU.
In lumenvox there is a way to tell the speech engine to only detect when the soundfile is finished playing. Is there a way to do that here? or is it set to interrupt-speaker mode all the time..?
Fred
Brian West sip:brian@freeswitch.org
Sounds good. Will continue testing from this end. I'll keep an eye open for the fix. Thanks. Sent from my iPhone On Jul 14, 2008, at 2:49 PM, Brian West <brian@freeswitch.org> wrote:
I think I see what exactly is going on.. its something thats very rare in this module but give me some time and I think we can nail it down. /b
On Jul 14, 2008, at 1:38 PM, Frederick Jabre wrote:
That did help a bit. It will still spike to 60-100% occasionally but its a noticeable difference. Now its mostly between 3-20% CPU.
In lumenvox there is a way to tell the speech engine to only detect when the soundfile is finished playing. Is there a way to do that here? or is it set to interrupt-speaker mode all the time..?
Fred
Brian West sip:brian@freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Dave's territory in that case..? Or is it something that could be patched on this end...? On Jul 14, 2008, at 5:23 PM, Brian West <brian@freeswitch.org> wrote:
I think the issue is now in the lib.
/b
On Jul 14, 2008, at 4:00 PM, Frederick Jabre wrote:
Sounds good. Will continue testing from this end. I'll keep an eye open for the fix.
Thanks.
Sent from my iPhone
Brian West sip:brian@freeswitch.org
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Open up your pocketsphinx.conf.xml and set listen_hits to 1 The issue comes from the yes or no grammar.. if you say yes or now too fast it was hitting below the listen_hits which would cause the issue you seen. Guess we can code around that also. /b On Jul 14, 2008, at 4:29 PM, Frederick Jabre wrote:
Dave's territory in that case..? Or is it something that could be patched on this end...?
Brian West sip:brian@freeswitch.org
I re-rolled the tarballs and uploaded new ones with the fix from David. I also patch cmd_ln.c in the build so it doesn't spew crap all over the place. On Jul 14, 2008, at 4:29 PM, Frederick Jabre wrote:
Dave's territory in that case..? Or is it something that could be patched on this end...?
Brian West sip:brian@freeswitch.org
I have a cepstral Allison voice for 16khz. I am using javasrcipt session.speak. The audio output is not nearly as good as what comes out directly from swift. I am guessing that speak is down sampling to 8khz. Is this true? Is there a way to get speak to not do this? Thanks
is the call 8khz? If you call with a 16khz codec like g722 it should choose 16k On Fri, Oct 31, 2008 at 4:08 PM, Frank @ Impact <frank@impactfax.com> wrote:
I have a cepstral Allison voice for 16khz. I am using javasrcipt session.speak.
The audio output is not nearly as good as what comes out directly from swift.
I am guessing that speak is down sampling to 8khz. Is this true?
Is there a way to get speak to not do this?
Thanks
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> iax:guest@conference.freeswitch.org/888 googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:213-799-1400
Hmm. Well the call is g711 -----Original Message----- From: freeswitch-users-bounces@lists.freeswitch.org [mailto:freeswitch-users-bounces@lists.freeswitch.org] On Behalf Of Anthony Minessale Sent: Friday, October 31, 2008 5:30 PM To: freeswitch-users@lists.freeswitch.org Subject: Re: [Freeswitch-users] session.speak sample rate is the call 8khz? If you call with a 16khz codec like g722 it should choose 16k On Fri, Oct 31, 2008 at 4:08 PM, Frank @ Impact <frank@impactfax.com> wrote: I have a cepstral Allison voice for 16khz. I am using javasrcipt session.speak. The audio output is not nearly as good as what comes out directly from swift. I am guessing that speak is down sampling to 8khz. Is this true? Is there a way to get speak to not do this? Thanks _______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org -- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ AIM: anthm MSN:anthony_minessale@hotmail.com <mailto:MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com <mailto:PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <mailto:sip%3A888@conference.freeswitch.org> iax:guest@conference.freeswitch.org/888 googletalk:conf+888@conference.freeswitch.org <mailto:googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:213-799-1400
Ok, That explains your question then. Since g711 is an 8khz codec, the system will downsample the audio to match and you will lose some of the clarity. On Sat, Nov 1, 2008 at 9:06 AM, Frank @ Impact <frank@impactfax.com> wrote:
Hmm. Well the call is g711
-----Original Message----- *From:* freeswitch-users-bounces@lists.freeswitch.org [mailto: freeswitch-users-bounces@lists.freeswitch.org] *On Behalf Of *Anthony Minessale *Sent:* Friday, October 31, 2008 5:30 PM *To:* freeswitch-users@lists.freeswitch.org *Subject:* Re: [Freeswitch-users] session.speak sample rate
is the call 8khz? If you call with a 16khz codec like g722 it should choose 16k
On Fri, Oct 31, 2008 at 4:08 PM, Frank @ Impact <frank@impactfax.com> wrote:
I have a cepstral Allison voice for 16khz. I am using javasrcipt session.speak.
The audio output is not nearly as good as what comes out directly from swift.
I am guessing that speak is down sampling to 8khz. Is this true?
Is there a way to get speak to not do this?
Thanks
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Anthony Minessale II
FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/
AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch
FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> iax:guest@conference.freeswitch.org/888 googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:213-799-1400
_______________________________________________ Freeswitch-users mailing list Freeswitch-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> iax:guest@conference.freeswitch.org/888 googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:213-799-1400
participants (4)
-
Anthony Minessale
-
Brian West
-
Frank ＠ Impact
-
Frederick Jabre