Re: Technology 'secures' gunfire [CNN]

[Jonathan Wienke <JonWienk@ix.netcom.com>]

At 09:06 PM 10/31/97 -0600, Jim Choate wrote:
>Forwarded message:
>
>>                    TECHNOLOGY 'SECURES' GUNFIRE IN THE CITY
>>                                        
>>      Secures October 31, 1997
>>      Web posted at: 4:44 p.m. EST (2144 GMT)
>>      
>>      ARLINGTON, Virginia (CNN) -- If you heard gunshots ring out in your
>>      neighborhood, you might be able to tell the general direction they
>>      came from. And if you happened to glance at your watch, you could
>>      say about what time. In maybe a minute, if you were so inclined, you
>>      could call the police to report it.
>>      
>>      Now police have an electronic witness that can provide similar
>>      assistance: a device called SECURES that pinpoints the time and
>>      location of gunshots.

This would be a network of microphones and processing stations which could
perform a reverse-GPS location analysis of sounds picked up by 3 or more
microphones.  (Sounds common to 2 microphones could be localized with a
lower degree of accuracy if directional microphone arrays are used.)  Yet
another instance of Big Brother technology that is of limited value to the
police.  Of course, this means that you will have the police responding to
every backfiring car, which will dampen their enthusiasm for responding
unless full-auto fire or a prolonged gunfight is overheard.  Of course, if
you have a silenced weapon and some cherry bombs with cigarette time-delay
fuses, you can use this system to docoy the police into the wrong
neighborhood.  Or if you confine yourself to single-shot assassinations
near busy streets, it will probably be written off as a vehicle backfire,
especially if you are doing a drive-by with a suppressed shotgun.  (Not
possible to silence completely, but certainly possible to quiet to the
point that it wouldn't attract undue notice along a busy street.)

In order for this system to be worth anything, it would have to be able to:
1.  Use voice recognition techniques to classify the type of weapon
(primarily useful for machine guns--it could evaluate the frequency
characteristics, rate of fire, etc. to distinguish between an AK-47 and an
UZI) sufficiently well to distinguish between small-arms fire, fireworks
(cherry bombs, M-80's, etc) and vehicle backfires.

2.  Perform "scream analysis" to distinguish the typical screams of
children at play from those of gunshot victims.

3.  Monitor conversations throughout the coverage area.  A suspicious sound
preceded by a male voice saying "Give me your money, bitch" would be much
more interesting than one preceded by a revving engine.  This would have
the added benefits of allowing LEO's to track fugitives via the sound of
their footsteps, their breathing, vehicle engine sounds, etc., as well as
gathering voiceprint data from crimes in progress.

I think point 3 is the scariest.  A properly designed system could do
voiceprint analysis of almost every word spoken in public, tie the
conversations to the identities of the speakers, and archive the time,
location, content and participants of every spoken conversation for long
periods of time in a database that could be searched by keyword, speaker
identity, time, and/or location.  The following searches could be done:

1.  "I want a list of everyone who uttered the words 'buy' and 'crack' in
the same sentence between 2100 and 0330 hours within 500 feet of 123 Maple
Drive between August 7 and December 5."

2.  "I want a list of all participants in conversations with Citizen-Unit
754-35-9710 which included the phrase 'BATF agent' in the last 6 months.

3.  "I want a map of Citizen-Unit 754-35-9710's movements for the last 2
weeks."

4.  "I want to see the movement history of all '96 Ford Escorts with a
misfire on cylinder #1."

5.  "I want the identities of everyone involved in the assault that
happened at the corner of Maple and Main at 1752 hours yesterday.

The Big Brother potential of such a system should be obvious.  What is
really scary is that such a system could be built mostly with currently
existing hardware, and at most a few man-years of software development.  If
each node in the network performs its own speech to text conversion and
archiving, and coordinates with a central voiceprint ID server, (which
could also provide the sync signal that the nodes would use to
cross-reference between nodes to locate sounds) each node could consist of
a Pentium 200 with some specialized audio signal processing cards and 15-20
GB of storage.  The only really new thing required would be an .AVI-style
format for storing MPEG audio, a text transcript of said audio (which would
need to include keywords for gunshots, passing vehicles, and other events
of interest), and location coordinates (updated on a second to second
basis) which could be indexed for reasonably efficient searching.


Jonathan Wienke

PGP Key Fingerprints:
7484 2FB7 7588 ACD1  3A8F 778A 7407 2928
3312 6597 8258 9A9E D9FA  4878 C245 D245 EAA7 0DCC

"If ye love wealth greater than liberty, the tranquility of servitude
greater than the animating contest for freedom, go home from us in peace.
We seek not your counsel, nor your arms. Crouch down and lick the hand that
feeds you. May your chains set lightly upon you; and may posterity forget
that ye were our countrymen."
-- Samuel Adams

"Stupidity is the one arena of of human achievement where most people
fulfill their potential."
-- Jonathan Wienke

RSA export-o-matic:
print pack"C*",split/\D+/,`echo "16iII*o\U@{$/=$z;[(pop,pop,unpack"H*",<>
)]}\EsMsKsN0[lN*1lK[d2%Sa2/d0<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<J]dsJxp"|dc`

PGP signature