Re: Glass breaks, thunderstorms and HA (was Re: Help with 7 Circuit Project?)

[Marc F Hult <MFHult@xxxxxxxxxxxxxxxxxxxxx>]

On Mon, 1 Aug 2005 01:44:50 -0400, "Robert Green"
<ROBERT_GREEN1963@xxxxxxxxx> wrote in message
<NuOdna_chuSLJXDfRVn-gw@xxxxxxx>:

>"Marc F Hult" <MFHult@xxxxxxxxxxxxxxxxxxxxx> wrote in message
>
>> Which is what I was referring to, namely,  using software to identify a
>> characteristic sound from a single microphone. Savoy's Cyberhouse's
>> Listen.exe can do this.
>
>Yes, I understand that.  It works like DragonDictate, except that instead
of
>putting a word on the screen in response to a "sound match" Savoy's program
>allows the controller to take an action.  I assume, based on my experience
>with Dragon, that it's not 100% reliable without training.


It can be made to be ~100% reliable in detection. Problem is with falsing.


>> There are also commercial, scientific, and freeware
>> programs to do this that might be adapted. Google on
>> "sound recognition"
>
>I think I'm going to spend a little more time modeling the problem.  I want
>to detect thunder mainly to prevent false alarms from my glass break
>detectors.  Since, as you point out, the audio "signature" of thunder is
>highly variable, it seems it would be pretty futile trying to make a sound
>match.  It would also require PC horsepower, and I'd like to see if a very
>simple hardwired circuit could also do the trick.  I think what you
>suggested earlier in the thread (some op-amps and gates) would work very
>nicely simply based on the amplitude of the sound.  It might also be
>necessary to filter out all but the lowest frequency sounds and, as you
>pointed out, compensate for possible cancellation of the sound waves.
>
>> http://www.merl.com/projects/SoundRecognition/
>
>> As I noted previously, the sound of lightning-induced thunder is complex
>> and variable.
>>
>> See: http://freesound.iua.upf.edu/tagsViewSingle.php?id=951
>
>Yes, agreed.  Sound pattern matching is not a good solution for a number of
>reasons.  But four microphones, a few IC's and stuff from my junkbox might
>work.  I have to decide whether to amplify the signals close to the
>microphone (probably best since it would be more likely to compensate for
>difference in output level due to line length) and how to test the circuit.



Seems to me that you could eliminate most falses by ANDing the output of the
glass-break sensor on the monitored window with the output of an ultra
sensitive (i.e., "too sensitive") glass-break on an upper floor away from
the monitored window.


Or as you suggest, the output of a microphone --> amp --> bandpass filter
--> comparator,  which could also be mounted on a remote window pane. (IF
upstairs window rattling, THEN ignore patio door glass-break.)

Marc
Marc_F_Hult