Car crash detection with automatic SOS

A phone that detects a car crash from its sensors and sends an SOS to chosen contacts.

Sensor and behaviour ML
Samsung R&D Institute India
Nov 2022 – Aug 2024
Built
short entry

1 of 6Four signals at once

The phone reads its accelerometer, gyroscope, pressure and sound together. Normal driving barely moves them.

6 steps · 42 s

Four signals at once. The phone reads its accelerometer, gyroscope, pressure and sound together. Normal driving barely moves them.

Rash driving is not a crash. Hard braking and sharp turns push the readings up, but not to a crash's level. The model sorts normal, rash and extreme cases.

Look-alikes. A thrown phone, loud music or a film can spike one signal. One signal alone is not enough, and the model was tuned against each kind of false alarm.

A crash. In a crash, g-force spikes past any driving, sound and pressure catch the airbag, and the gyroscope catches a roll.

SOS. When the signals agree, the phone sends an SOS to the contacts the user chose.

Tested without crashing cars. Real crash data is rare, so crashes were simulated with a toy car carrying a phone. Everything that is not a crash was recorded for real.

Illustration drawn by code, not a recording: the sensor traces are drawn to show which signals move together in each case.

The phone reads its accelerometer, gyroscope, pressure and sound together. A statistical model with heuristics sorts normal driving, rash driving and the extreme cases that could be crashes, and was tuned against each kind of false alarm. Sound and pressure catch an airbag opening; the motion sensors catch a roll, a fall or toppling.

Real crash data is rare, so crashes were simulated with a toy car carrying a phone, and everything that is not a crash was collected for real: rash driving, thrown phones, loud music, films.

More in sensor and behaviour ml