Answering Machine Detection.

[Guest Hal Netkin]

I'm Using a Dialogic D/4PCI card. The operating System is Windows 2000 Professional.

 

When Autodialing out, the system doesn't detect an answering machine about half of the time. When no answering machine is detected, the message played is the one intended for when a person answers and starts prematurely.

 

Please advise.

 

Hal Netkin

halnet@pacbell.net

[SupportTeam]

Answering machine detection is not 100% reliable, but we usually see much better success rate then 50%.

 

Are you seeing this on home answering machines or are you dialing into phone company provided voicemail boxes?

[Guest Hal Netkin]

Dialing to ordinary home phones. This problem is recent. The system worked far more reliable in the past. Is there any kind of adjustment that can be made?

 

BTW: How does the Dialogic card know the difference between a person answering vs a person answering?

 

Hal Netkin

[Guest Hal Netkin]

In the last message, I meant to say:

BTW: How does the Dialogic card know the difference between a person answering vs an answering Machine?

 

Hal Netkin

[SupportTeam]

This problem is recent. The system worked far more reliable in the past.

Have you changed the phone lines used by this system? Anything else got changed?

How does the Dialogic card know the difference between a person answering vs an answering Machine?

They're not about to tell anybody but it's most likely from looking at the slightly different frequencies that are present on the line when an answering machine answers the call.

[Guest Hal Netkin]

Yes, I changed phone providers recently from SBC to ATT. Are there any adjustment in Dialogic's hardware and/or software or VG's software that can be made?

 

Hal Netkin

[SupportTeam]

Nothing that you can really change on Dialogic's side - but this is a good example of how each phone companies lines are usually of slightly different quality - and the lower quality of transmission the harder it is for Dialogic to correctly distinguish between an answering machine and a live person on outgoing calls...

[Guest Hal Netkin]

Thanks anyway. I guess I'll have to put up with the problem.

 

As an electrical engineer, it would seem to me that there could be a way where the system detects the tone that all answering machines emit which lets the caller know when to start his message. It could work as follows: The VG message starts whether or not a person or answering machine answers. If a person answers, the VG message starts and ends normally. But if an answering machine answers, the VG message starts over again from the beggining upon sensing the answering machine's tone. Of course when the message goes to completion, the tone sensor is disabled so as not to allow any key pad dial tones by the recipient to inadvertantly start the message over.

 

Seems to me that Katelina could design such a system in its own VG software. Or does any company offer such.

 

Thanks again,

 

Hal Netkin

[SupportTeam]

the VG message starts over again from the beggining upon sensing the answering machine's tone.

But each answering machine's tone is different... most of them are a single frequency or two frequency tone - to detect that the system would need to be able to detect the presence of a 'pure tone' on the line - guess you could do this by doing a continuous spectral analysis on the line and if you see a continuous distinct peak (or two peaks) then that could be used to indicate that the tone is heard... but the Dialogic card cannot record while playing at the same time (!) so the only way to set it to detect the tone is to have it set up to listen for a specific frequency - and that is not good enough as each answering machine's tone is different...

[Guest Hal Netkin]

I wasn't thinking of a Dialogic system necessarily. Somebody might have to design a card.

 

It wouldn't matter that each answering machine has a different tone frequency as long as they were in a broad frequency range. The card wouldn't be looking for a specific tone frequency. It would only have to detect a tone generated pulse whose amplitude is constant for about 100 millieseconds or more -- something that is common with almost all answering machines. It would be rare for such a tone to be generated by someone speaking or in the middle of an outgoing answering machine's message.

 

The VG message would start whether or not an answering machine or a person answered. If it did not hear a tone at the end of an outgoing message, it would mean that a person answered and the VG message would keep going to completion. But if the tone (which would mix with the VG message and detected by the system) at the end of the outgoing message came on in the middle of the VG message, the system would detect it and start the VG message from the beginning.

 

Just a retired engineer's brain storming.

 

Best,

 

Hal Netkin

 

P.S. The answering machine detection has mystereously returned to normal.

[AbsoluteControls]

In one of your posts you asked whether your customer was using VG to dial in to a regular home answering machine or to a phone company provided voicemail box. Is Dialogic's detection less acurate for the voicemail. I have been testing my D300-JCT E1 and it has been misidentifying voicemail as a live person every time.

[SupportTeam]

Sometimes the phone company's voicemail is set up to be of extremely good quality - and it looks like high quality recordings are harder to detect for what they are (an answering machine) then the low-quality recordings of the home answering machines out there.

 

Are you using the D300-JCT with "VoiceGuide for ISDN" ?

In that case please forward th trace of the outgoing call to support@voiceguide.com and we can have a look at what is happening during this outgoing call.

[SupportTeam]

There has been an issue with one of the releases of "VoiceGuide for ISDN" which has problems with Answering machine detection. This bug has now been fixed.