4 ms·
I'm not an expert, or even a layman in audio encoding, but it seems to me that a solution could be found that couples an open source capture program with their
by aorloff 11y ago
I'm not an expert, or even a layman in audio encoding, but it seems to me that a solution could be found that couples an open source capture program with their backend. The open source (audited and only updated from the open source branch) capture program filters the audio stream down to something that only contains a gunshot signature, and nothing else. If the motivation is there a solution is certainly possible.
- toomuchtodo 11y agoIt could be done, similar to this lightning detection project: http://www.lightningmaps.org/realtime http://www.lightningmaps.org/realtime http://www.blitzortung.org/Webpages/index.php?lang=en&page=3 http://www.blitzortung.org/Webpages/index.php?lang=en&page=3 TL;DR They do triangulated time of flight for EM emissions from lightning strikes.
- CHY872 11y agoThing is, with lightning it is dead easy because you're talking about what is essentially a straight line through the atmosphere. In a city, you get more issues: double datapoints from the gun and the bullet, multipath effects, etc etc etc.
- joosters 11y agoBut as the article says, they need the original audio in order to improve the gunshot detection - for example, if a gunshot is reported elsewhere but the detector doesn't classify it, they can grab the audio from the record to train the detector and hopefully improve it. If you never capture the raw audio, you won't be able to train the system to better detect the gunshots and to reduce false positives.