[Asterisk-Users] Answering Machine Detection

[Chris Ziomkowski]

At 11:52 PM 10/29/2003 +0800, you wrote:
>Ray Burkholder wrote:
>
>>Might want to write a new
>>>"energy detector" algorithm in dsp.c though based on a wideband/low Q 
>>>resonator approach (move the pole way in towards the origin) as opposed 
>>>to narrow band goertzels (pole on the unit circle). More robust for this 
>>>type of work.
>>>
>>
>>Where does one go to learn this terminology and the math to implement it?
>Apparantly not to this source. :-) A Goertzel filter finds one output bin 
>of a DFT. Since the width of the bins in a DFT is directly related to the 
>number of samples you include in a processed block, the width of the 
>Goertzel is too. A Geortzel is as wide or as narrow as you want it to be. 
>Oh, and k does not need to be an integer, unless you are trying to 
>evaluate phase. That is a common misconception. There is a sliding window 
>version of a Goertzel filter, but this has the same characteristics as the 
>standard version, as it is just a trick for calculating a continuous 
>stream of Goertzels efficiently.

Now, now, Steve. Let's not get personal. You are correct. I am also 
correct. It depends on how you look at it. A Goerzel is only as wide as you 
want it to be if you have short periods and/or infinite representation. 
Goertzels are on the unit circle, and as such on the edge of stability. 
They never dampen their responses, so they will grow infinitely large if 
you keep ringing them.

Try using them on a 16 bit fixed point DSP and you'll understand how 
important this point really is.

>Regards,
>Steve
>
>
>_______________________________________________
>Asterisk-Users mailing list
>Asterisk-Users at lists.digium.com
>http://lists.digium.com/mailman/listinfo/asterisk-users