Gunshot Tracking Cameras to be Deployed in LA 480
apok04 writes "Get out your tinfoil hats (and ski masks). A USC engineer uses his expertise with nerve cells to create a surveillance system that can recognize the sound of a nearby gunshot - and identify the shooter. In a unique pilot program, L.A. and Chicago will deploy test units in high-crime areas. The creator emphasizes that the system cannot recognize voices or words, but his previous research into speech recognition systems suggests otherwise."
Re:Bay Area Scam (Score:2, Informative)
Re:I thought they already have this? (Score:2, Informative)
"A microphone surveillance system now is using his insights to recognize - instantly, and with high accuracy - the sound of a gunshot within a two-block radius. The system can then locate, precisely, where the shot was fired, turn a camera to center the shooter in the camera viewfinder and make a 911 call to a central police station."
So, this system can locate exactly where a shot was fired, as well as turn local cameras in the area of the shot towards the shooter.
What about silencers/suppressors? (Score:3, Informative)
Fine, so it detects the sound. Minimize the volume of the sound, or change the profile of that sound, and the shot becomes less-likely to be detected. A suppressor would help in the former, but I'm not sure about the latter (any experts?).
Suppressors are not difficult to manufacture [yahoo.com], after all, although it's a felony to do so (or to possess one), in violation of the 1934 National Firearms Act...
Predictions:
1) monitoring devices get destroyed and/or hacked, and/or
2) suppressors increase in popularity, and/or
3) alternate means of killing (knives, swords, blowguns, etc.) increase in popularity
or,
4) nothing changes, except more shooters are detected
Anyway, just because the microphone's input is piped to a neural-net program which detects gunshots does not mean the input cannot *also* be outputted to a file, or to speakers on a computer, etc..
Re:what if (Score:3, Informative)
They look like a turret, and we're told they're bullet proof and even work with a silencer.
Then again they have cameras that give you $90 tickets for trying to go through a yellow light.
The difference... (Score:2, Informative)
This soulds a bit like a neural net. I know of neural nets taking a FFT and being able to tell one jet engine from another (eg. 747 vs 727) or a Toyota engine vs a VW or an accoustic return from a box vs a sphere.
Gunshot signatures could be quite easy to decipher since a pistol sounds different to a shotgun or rifle and a subsonic (eg. .45 APC) sounds different to a supersonic (eg. 9mm). However the sound does get filtered and some components are lost. Perhaps this is why the system only works for a couple of blocks.
As one currently working for a voice company (Score:5, Informative)
Also speech recognition knowledge is very different from speaker recognition (one cares about what the person says regardless of how they say it, the other cares about how they say it regardless of what it is). The mathematical models for both are very different.
Also the microphones are likely specialized in the wrong frequency/volume range to be useful for speaker authentication.
prior art (Score:4, Informative)
Redwood City CA has had this for eight years (Score:4, Informative)
Here's an evaluation. [ncjrs.org] Median location error is about 25 feet. That at least gets it down to two or three houses.
I met the designer of this system some years ago. The original prototype worked using microphones and hard-wired phone connections for each microphone. The signal from each microphone was transmitted using an analog FM carrier system over the phone line designed to trade frequency response for dynamic range. The system had terrible audio frequency response but huge dynamic range, so that pulse events like gunshots come through cleanly without overload. When you have enough dynamic range, gunfire is easy to recognize, because the leading edge of the pulse is so sharp. Few other sounds have that form.
The microphones are up on telephone poles and atop buildings, and they're omnidirectional. So they mostly pick up loud bangs, wind, and aircraft noise. The original pole units were entirely analog, phone line powered, and very dumb. The original central processing system was a PC with some data acquisition cards running LabView. Since then, it's become fancier, with better integration with mapping programs and transmission of gunfire locations to PDA-type devices. But it's not really very complicated.
tech review covered this a month ago. (Score:3, Informative)
Re:Redwood City CA has had this for eight years (Score:3, Informative)
A banger or a car backfire miss the Swoosh.
Re:Bay Area Scam? (DOJ NIJ Report) (Score:2, Informative)
I never read anything about corruption regarding ShotSpotter, nor did I find any mentions in news archives.
The article I found just mentions that there was significant debate in Redwood City before buying the system from Trilon for $85K. "Opponents, however, claim it is a boondoggle and that the money could be better spent elsewhere, such as on hiring more police officers." (SFChronicle, 3/18/97, "Redwood City Endorses Gunshot Locator System")
The National Institute of Justice [usdoj.gov] funded a study of the ShotSpotter system in Redwood City and Dallas.
The December 1999 report can be found on the NIJ website:
http://www.ncjrs.org/pdffiles1/nij/179274.pdf [ncjrs.org]
The report compared Alliant's SECURES system in Dallas to Trilon's ShotSpotter system in Redwood City.
It sounds like they had a lot of fun with this test in RWC:
Dallas chose not to allow the firing of blank rounds on random street corners:
If you're wondering why Redwood City would be picked, keep in mind that neighboring East Palo Alto had the highest per capita murder rate in the country after a string of drug murders in 1992. [paloaltoonline.com] (The homicide rate is lower now.)
The NIJ report page is pretty entertaining reading:
Re:Response Time (Score:4, Informative)
In English, we do not posess a grammatical gender to refer to the unknown: we have male, female, and neither, but not a possibly-either. The convention in English for nearly two millennia, and in her precursor languages (English is grammatically feminine, incidentally, much like a ship), has always been to use the masculine when referring to the unknown or the general. That is, the masculine gender serves double duty: it (amusingly, the masculine grammatical gender is itself grammatically neutral) refers to both males and other grammatically-male individuals, but also to those whose grammatical gender is unknown or general. This isn't sexist so much as a limitation of the language. Incidentally, the very word 'man' is actually a gender-unknown holdover from Old English; the word for a male man (a phrase which seems redundant now) was 'were' (like werewolf, and pronounced similarly); because 'man' could refer to either a man or a woman, words like 'wifman' (means wife-man), which became 'woman,' or 'leman' (a mistress: means love-man) could be formed.
Moreover, in this specific case the distribution of male vs. female shooting perpetrators can hardly be said to justify the use of the feminine. Quite the opposite, really.