Re: [Freeswitch-users] Better results from mod_vmd
I just tried mod_avmd and mod_vmd and seem to have the same issue, my question is if someone has this working with Panasonic answering machine and Verizon Wireless (as well as other US wireless carriers) voicemail, what change if any should I apply to get it working. Thanks for the help On Thu, Aug 20, 2009 at 8:36 PM, Matthew Fong <mattdfong@gmail.com> wrote:
I changed
/*! Minimum time for a beep. */ #define MIN_TIME 8000 to 6500 and it seemed to work, but I'm not sure how many false positives I will get in a real-world environment. at 4000 it fired the event like 5 times in a session, but 6500 only once. Do you think I should expect a lot of false positives after changing this value?
--matt http://www.hellohunter.com
On Thu, Aug 20, 2009 at 4:54 PM, Eric des Courtis < eric.des.courtis@gmail.com> wrote:
Matt,
As is mod_vmd will not detect tones shorter then 138ms. However I could get that value down to ~30ms at best by making a few modifications to the algorithm.
Cheers.
Eric des Courtis
On Thu, Aug 20, 2009 at 7:51 PM, Eric des Courtis<eric.des.courtis@gmail.com> wrote:
Matt,
For your information the tones you gave me are exactly 738Hz. If you want to try that tone detection thing.
Cheers.
Eric des Courtis
On Thu, Aug 20, 2009 at 2:20 PM, Michael Collins<msc@freeswitch.org> wrote:
On Thu, Aug 20, 2009 at 11:06 AM, Steve Underwood <steveu@coppice.org> wrote:
On 08/20/2009 05:22 AM, Michael Collins wrote:
There is no noise on those 3 beeps. In fact, for something
that's
been through ulaw/alaw compression those beeps are very clean. They are quite short, though.
Heck yeah they're short! Steve, in your experience is there a practical way to detect a beep that short without chewing up system resources or having lots of false positives? -MC
The tone samples I just looked at are about 130ms long. The problem is the detector is trying to be a very open ended detector of anything narrowband enough to be a single tone, and of any duration beyond some small minimum. Its difficult to make such a thing voice immune unless you can also count on a very large signal to noise ratio. With a digital trunk you can probably rely on a large SNR, but what happens when people use analogue lines? There is a reason why DTMF detectors try hard to work down to about 10dB SNR. :-)
Steve
Thanks for the lesson uncle Steve! I'm guessing that the OP will need a new strategy. Possibly waiting for silence? Not sure what's the best way to go but I'm interested in hearing if someone has a solution.
-MC
_______________________________________________ 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
_______________________________________________ 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
Hi Abaci, what issues do you mean by "same issue"? Is this amount of false positives? Could you describe the tests you perform in more details please? The avmd module is being revised by myself at this moment and there is at least one bug in this module that influence the results FS-8855 fix calc of variance of tone's freq estimator and it has been fixed now in pull request #714 and will be merged soon after I do some more testing and rebase, to the end of this week. Apart from this, there are improvements coming, e.g. FS-8875 Enable faster beep detection cheers, Piotr
My problem was simple it doesn't detect voicemail beeps of wireless carriers voicemail (e.g. Verizon Wireless) the only beep it detected without issue was a beep from another FreeSWITCH Voicemail On Tue, Mar 15, 2016 at 9:13 PM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
what issues do you mean by "same issue"? Is this amount of false positives? Could you describe the tests you perform in more details please?
The avmd module is being revised by myself at this moment and there is at least one bug in this module that influence the results FS-8855 fix calc of variance of tone's freq estimator
and it has been fixed now in pull request #714 and will be merged soon after I do some more testing and rebase, to the end of this week. Apart from this, there are improvements coming, e.g.
FS-8875 Enable faster beep detection
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Abaci, is this attached Verizon voicemail the kind of beep that you mean? cheers, Piotr
yes On Wed, Mar 16, 2016 at 7:02 PM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci, is this attached Verizon voicemail the kind of beep that you mean?
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
if it helps I can email you a Verizon wireless number that I used to test so that you can test yourself On Wed, Mar 16, 2016 at 7:31 PM, Abaci B <abaci64@gmail.com> wrote:
yes
On Wed, Mar 16, 2016 at 7:02 PM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci, is this attached Verizon voicemail the kind of beep that you mean?
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Abaci, It is not necessary. I will take a look at this, but first we need to merge what is already awaiting. cheers, Piotr
Hi Piotr, I am glad that you are working on the avmd module! I tried myself a few months ago but found the code very difficult to follow and debug due to its overuse of macros. Do you plan to rewrite these macros as inline functions? For my part, I ended up rewriting all these macros as functions just to understand the code. I eventually wrote a new module from scratch as I noticed the bugs you mention and needed incompatible changes in the api. But had not time to clean it up and check the licensing matters yet. The real breakthrough is actually the one you mentioned in FS-8875: I have noticed real improvements in the cpu usage and detection accuracy by checking only the convergence of OMEGA instead of sin(OMEGA) and thus avoiding the sin calculation entirely! I will be glad to help in any way. Please keep me in touch when you publish your patch, I’ll be happy to have a look. Best Regards, Vallimamod .
On 16 Mar 2016, at 02:13, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
what issues do you mean by "same issue"? Is this amount of false positives? Could you describe the tests you perform in more details please?
The avmd module is being revised by myself at this moment and there is at least one bug in this module that influence the results FS-8855 fix calc of variance of tone's freq estimator
and it has been fixed now in pull request #714 and will be merged soon after I do some more testing and rebase, to the end of this week. Apart from this, there are improvements coming, e.g.
FS-8875 Enable faster beep detection
cheers, Piotr _________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Vallimamod, the macros must unfortunately stay as they are because we need to avoid unnecessary function calls on heavy code paths of execution. Actually, I've had to add more of them recently to enable change of resolution in arc cosine table creation (this is used if FASTMATH is defined). To ease debugging you may wish to compile module with -E flag, this will expand macros. I am going to complete FS-8875 in coming week. Most likely this will simply comes down to another option in options header (which if enabled will avoid computing of arcus cosine) plus some minor adjustments to make the rest of the code reflect that change. cheers, Piotr
If you are using FS as a dialer. The beep detect thing is there for amusement. If you are serious and making money from this ridiculous election remember FS is saving you countless thousands and donate or make an investment in the commercial AMD module. consulting@freeswitch.org On Friday, March 18, 2016, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Vallimamod,
the macros must unfortunately stay as they are because we need to avoid unnecessary function calls on heavy code paths of execution. Actually, I've had to add more of them recently to enable change of resolution in arc cosine table creation (this is used if FASTMATH is defined). To ease debugging you may wish to compile module with -E flag, this will expand macros.
I am going to complete FS-8875 in coming week. Most likely this will simply comes down to another option in options header (which if enabled will avoid computing of arcus cosine) plus some minor adjustments to make the rest of the code reflect that change.
cheers, Piotr
-- Anthony Minessale II ♬ @anthmfs ♬ @FreeSWITCH ♬ ☞ http://freeswitch.org/ ☞ http://cluecon.com/ ☞ http://twitter.com/FreeSWITCH ☞ irc.freenode.net #freeswitch ☞ *http://freeswitch.org/g+ <http://freeswitch.org/g+>* ClueCon Weekly Development Call ☎ sip:888@conference.freeswitch.org ☎ +19193869900 https://www.youtube.com/watch?v=9XXgW34t40s https://www.youtube.com/watch?v=NLaDpGQuZDA
Hi Abaci, back to your problems with Verizon voicemail - the cause of them is: 1. Audio is stereo 2. Audio is sampled at 44100Hz 3. DESA algorithm used in avmd is very sensitive to noise If you resample audio to 8000Hz and set avmd module to 8000Hz bit rate or leave it at 44100Hz but adjust avmd to 44100Hz while always properly handling number of channels (mono/stereo/...) then results are good. Avmd module in current form is unable to meet these requirements, but I am going to add the support for this. cheers, Piotr
I didn't try with stereo 44100Hz, I tried simple g711 calls to Verizon cell phones and it didn't trigger the beep, when calling a voicemail on another freeswitch box it did detect the beep. On Mon, Mar 21, 2016 at 9:27 AM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
back to your problems with Verizon voicemail - the cause of them is:
1. Audio is stereo 2. Audio is sampled at 44100Hz 3. DESA algorithm used in avmd is very sensitive to noise
If you resample audio to 8000Hz and set avmd module to 8000Hz bit rate or leave it at 44100Hz but adjust avmd to 44100Hz while always properly handling number of channels (mono/stereo/...) then results are good. Avmd module in current form is unable to meet these requirements, but I am going to add the support for this.
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Abaci, if the sample rate is correct and it is mono audio and there is not too much noise in the signal avmd should work. You may want to retry this after new changes have been merged into master branch today. I have verified avmd detects 1000Hz beep in the audio attached previously. You can send me the mobile number I can test in private email. cheers, Piotr On 21 March 2016 at 21:53, Abaci B <abaci64@gmail.com> wrote:
I didn't try with stereo 44100Hz, I tried simple g711 calls to Verizon cell phones and it didn't trigger the beep, when calling a voicemail on another freeswitch box it did detect the beep.
On Mon, Mar 21, 2016 at 9:27 AM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
back to your problems with Verizon voicemail - the cause of them is:
1. Audio is stereo 2. Audio is sampled at 44100Hz 3. DESA algorithm used in avmd is very sensitive to noise
If you resample audio to 8000Hz and set avmd module to 8000Hz bit rate or leave it at 44100Hz but adjust avmd to 44100Hz while always properly handling number of channels (mono/stereo/...) then results are good. Avmd module in current form is unable to meet these requirements, but I am going to add the support for this.
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Abaci, after FS-8875 was merged yesterday in PR 775 I have tested avmd on your Verizon voicemail (contains 1000hz beep) and on another Voipfone voicemail (contains beep of lower frequency, 724.1914 is detected). Detection of tone works just fine while not giving many false positives in the same time. But you should note the avmd module is in beta stage so not everything is working as expected yet. In case of Verizon sometimes it gives false alarm on silence, this can be handled at this moment by adjusting avmd session parameters AVMD_SAMLPE_TO_SKIP_N increase BEEP_TIME increase SAMPLES_CONSECUTIVE_STREAK increase VARIANCE_THRESHOLD may need to decrease or starting avmd after 3 - 4 first ring tones. Regardless of this it will be treated in more generic way - by adding of estimation of amplitude. Related to this is a need for adjusting parameters on a basis of expected characteristics of incoming signal, e.g. SNR. And it should be possible to be set the settings on a per call basis. So you may wish to checkout master branch, build, test, and let us know the results. cheers, Piotr
was it merged into 1.6 or just master? On Wed, Mar 30, 2016 at 11:09 AM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
after FS-8875 was merged yesterday in PR 775 I have tested avmd on your Verizon voicemail (contains 1000hz beep) and on another Voipfone voicemail (contains beep of lower frequency, 724.1914 is detected). Detection of tone works just fine while not giving many false positives in the same time. But you should note the avmd module is in beta stage so not everything is working as expected yet. In case of Verizon sometimes it gives false alarm on silence, this can be handled at this moment by adjusting avmd session parameters
AVMD_SAMLPE_TO_SKIP_N increase BEEP_TIME increase SAMPLES_CONSECUTIVE_STREAK increase VARIANCE_THRESHOLD may need to decrease
or starting avmd after 3 - 4 first ring tones. Regardless of this it will be treated in more generic way - by adding of estimation of amplitude. Related to this is a need for adjusting parameters on a basis of expected characteristics of incoming signal, e.g. SNR. And it should be possible to be set the settings on a per call basis.
So you may wish to checkout master branch, build, test, and let us know the results.
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
Hi Abaci, It was merged into master branch only as everything now goes into master while last commits on 1.6.6 are dated on January. cheers, Piotr
Just in master so far. It will hit 1.6 branch when we do the next release On Wednesday, March 30, 2016, Abaci B <abaci64@gmail.com> wrote:
was it merged into 1.6 or just master?
On Wed, Mar 30, 2016 at 11:09 AM, Piotr Gregor <piotrek.gregor@gmail.com <javascript:_e(%7B%7D,'cvml','piotrek.gregor@gmail.com');>> wrote:
Hi Abaci,
after FS-8875 was merged yesterday in PR 775 I have tested avmd on your Verizon voicemail (contains 1000hz beep) and on another Voipfone voicemail (contains beep of lower frequency, 724.1914 is detected). Detection of tone works just fine while not giving many false positives in the same time. But you should note the avmd module is in beta stage so not everything is working as expected yet. In case of Verizon sometimes it gives false alarm on silence, this can be handled at this moment by adjusting avmd session parameters
AVMD_SAMLPE_TO_SKIP_N increase BEEP_TIME increase SAMPLES_CONSECUTIVE_STREAK increase VARIANCE_THRESHOLD may need to decrease
or starting avmd after 3 - 4 first ring tones. Regardless of this it will be treated in more generic way - by adding of estimation of amplitude. Related to this is a need for adjusting parameters on a basis of expected characteristics of incoming signal, e.g. SNR. And it should be possible to be set the settings on a per call basis.
So you may wish to checkout master branch, build, test, and let us know the results.
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org <javascript:_e(%7B%7D,'cvml','consulting@freeswitch.org');> http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org <javascript:_e(%7B%7D,'cvml','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
Just tested mod_avmd on master and mod_avmd was able to detect the beep of a Verizon voicemail. Will do more testing and report back. On Wed, Mar 30, 2016 at 8:39 PM, Michael Jerris <mike@jerris.com> wrote:
Just in master so far. It will hit 1.6 branch when we do the next release
On Wednesday, March 30, 2016, Abaci B <abaci64@gmail.com> wrote:
was it merged into 1.6 or just master?
On Wed, Mar 30, 2016 at 11:09 AM, Piotr Gregor <piotrek.gregor@gmail.com> wrote:
Hi Abaci,
after FS-8875 was merged yesterday in PR 775 I have tested avmd on your Verizon voicemail (contains 1000hz beep) and on another Voipfone voicemail (contains beep of lower frequency, 724.1914 is detected). Detection of tone works just fine while not giving many false positives in the same time. But you should note the avmd module is in beta stage so not everything is working as expected yet. In case of Verizon sometimes it gives false alarm on silence, this can be handled at this moment by adjusting avmd session parameters
AVMD_SAMLPE_TO_SKIP_N increase BEEP_TIME increase SAMPLES_CONSECUTIVE_STREAK increase VARIANCE_THRESHOLD may need to decrease
or starting avmd after 3 - 4 first ring tones. Regardless of this it will be treated in more generic way - by adding of estimation of amplitude. Related to this is a need for adjusting parameters on a basis of expected characteristics of incoming signal, e.g. SNR. And it should be possible to be set the settings on a per call basis.
So you may wish to checkout master branch, build, test, and let us know the results.
cheers, Piotr
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
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
participants (5)
-
Abaci B
-
Anthony Minessale
-
Michael Jerris
-
Piotr Gregor
-
Vallimamod Abdullah