Ein japanisches Eisenbahnunternehmen entwickelt ein KI-gestütztes Notfallwarnsystem, um ungewöhnliche Geräusche in Zügen, wie zum Beispiel Schreie, zu erkennen. Wenn an mehreren Stellen in einem Zug installierte Mikrofone Schreie, Schlägereien oder Hilferufe erkennen, wird sofort eine Warnung an das Zugpersonal gesendet.
https://www.asahi.com/ajw/articles/14628841
7 Comments
SUBMISSION STATEMENT: OSAKA–West Japan Railway Co. (JR West) is developing an emergency alert system that can detect unusual noises on trains, such as screaming, and accelerate the process of securing passengers’ safety. The project was started after a series of attacks on trains, including a stabbing rampage on a Shinkansen.
“We need to give train passengers a sense of security through various ways,” said Kazuaki Hasegawa, the president of JR West. “If (the system) works well, we would like to introduce it in our trains.” **Under the planned system, when microphones installed in multiple locations on a train detect screams, fights, cries for help and other anomalies, an alert will be immediately sent to crew members in their compartments**, according to JR West.
**If the system is successful, passengers experiencing an assault will be able to simply scream for help to initiate a response by train crew in the future.**
The system will allow for a much faster response time than through the conventional method, in which passengers press an emergency intercom.
JR West and one of its group companies are now in the process of analyzing sounds through artificial intelligence. The company has not determined when the system will be introduced on trains. JR West said the system will not automatically record and store voice data.
This seems like a good idea on paper, however we need to make sure that this AI doesn’t fuck up
That raises the question if people can trigger the system with random screams.
interesting, but what if there is a bug or virus on the AI. hope AI wont fck up
Japanese people are well-behaved enough that it might work. Not in the US.
This seems like a good use for AI. It’s probably extremely hard to either reliably detect or directly monitor for these sorts of situations and using AI to identify when something might need attention seems like good way to assist with security or medical emergencies.
so screamspotter instead of shotspotter? [been tried already](https://en.wikipedia.org/wiki/ShotSpotter).
>A 2024 audit from the New York City Comptroller found the New York City Police Department (NYPD) officers’ time was wasted 87 percent of the time by going to false alarms or unconfirmed sounds. It also stated that NYPD was over-representing the benefits of the system in reducing response times. Comptroller Brad Lander said “NYPD is wasting precious time and money on this technology and needs to do a better job managing its resources. Chasing down car backfires and construction noise does not make us safer.”