| Author |
|
thomas.s
Intermediate
Joined: February 10 2010
Location: Sweden
Posts: 3
|
| Posted: February 10 2010 at 10:16am | IP Logged
|
|
|
Hi,
I am in the process of evaluating if we can replace Dialogic HMP with LanScape VoIP Media Engine.
I can't find any information about if you have an implementation for detecting answering machines. Do you have this feature or do you plan on implementing this?
Best Regards,
Thomas Sköldenborg
|
| Back to Top |
|
| |
support
Administrator
Joined: January 26 2005
Location: United States
Posts: 1666
|
| Posted: February 10 2010 at 11:53am | IP Logged
|
|
|
Hello Thomas,
Thank you for posting to this support forum.
I am very interested in working with your group in order to allow you to use the LS VOIP media engine (LME) instead of the Dialogic HMP software.
The media engine does not currently detect answering machines. What I would like to suggest is that we work together to define this capability and get it added to the product.
I have had other requests from other customers for this capability and I think it is time that we add this capability.
If you can, post to this thread some of your thoughts/requirements for answering machine detection. I have reviewed the Dialogic HMP specifications and I am sure we can get to a solution that will allow you to deploy your VOIP applications using the LS media engine.
I expect that answering machine detection in the media engine has to support in-band signal processing and RFC2833 tone detection. Right now we fully support in-band DTMF and RFC2833 DTMF generation and detection. Answering machine detection should be a logical extension of what is already implemented in the product.
Any information you would like to discuss is welcome here.
Thank you Thomas,
Randal
|
| Back to Top |
|
| |
thomas.s
Intermediate
Joined: February 10 2010
Location: Sweden
Posts: 3
|
| Posted: February 11 2010 at 4:29am | IP Logged
|
|
|
Hi Randal,
Thank you for going head-on with this challenge. I myself am not certain how answering machine detection is implemented in HMP but there is some information about how it is implemented in Asterisk here
http://www.voip-info.org/wiki/index.php?page=Asterisk+cmd+AM D
There is also some indication to what might be going on in Dialogics implementation on this page with tuning-information for the Dialogic-system:
http://www.dialogic.com/support/helpweb/dxall/tn924.aspx
Dialogic and Asterisk seem to base the decision of answering machine or not on just the voice-part of the initial greeting, i.e. not that much on the tones that might be heard after the greeting message has been played.
One requirement is that negative detection should be very fast, i.e. detection that it is not a machine. A human should not have to wait longer than say 2-3 seconds after he/she has stopped talking before the message is played.
Another requirement is of course that detection is correct most of the time. It is better if the system is wrong and treats a machine as a human than the other way around. We don't want humans being treated as machines.
I think I read that you already have voice activity detection implemented? I guess the answering machine detection should be a logical extension to this system as it seems to be in the Dialogic case.
Best Regards,
Thomas Sköldenborg - Enera International
|
| Back to Top |
|
| |
support
Administrator
Joined: January 26 2005
Location: United States
Posts: 1666
|
| Posted: February 11 2010 at 8:38am | IP Logged
|
|
|
Hello Thomas,
Thank you for the additional information.
Early this morning I reviewed the Asterisk and Dialogic HMP “answering machine detection” and I understand both approaches.
Yes, you are correct - the media engine has voice activity detection (VAD) on transmitted audio. This VAD is optionally use when building soft phone applications with the LME. It allows transmitted RTP media for voice calls to be reduced in an effort to preserve/reduce Tx bandwidth. Unfortunately that VAD control won’t help us here.
If you have basic signal processing knowledge, you could make an attempt your self to handle simple answering machine detection at the application level. This could be a short term solution.
The media engine will allow your VOIP application to access all received media for all VOIP phone calls – on a per line basis. In your application, you can tell the media engine that your app wants to access the received audio for a call. Your VOIP application can then take the received sample blocks for the call and signal process the received audio in any manner required. If you have a reasonable signal processing knowledge, you could implement a simple answering machine strategy right now.
Without going into too much detail or divulging any private information, what exactly is your group’s plans regarding deploying your VOIP applications and LME?
Is it possible for our two groups to work jointly to add this capability to the LME product?
We can also discuss this issue privately if needed.
Thank you,
Randal
|
| Back to Top |
|
| |
Active Topics 
Member List 
Search 
Help 
Register 
Login
LanScape Support Forum -> LanScape VOIP Media Engine™ - Pre-Sales Technical Support
Topic: Answering Machine Detection
