GM Patents System That Can Detect Drunk Drivers by the Way They Walk

General Motors has filed a patent for a new system capable of identifying intoxicated individuals based on their walking patterns.

The technology was developed in collaboration with researchers from National University of Colombia. It relies on a neural network trained to analyze surveillance video footage. The system tracks around 20 key points on a person’s body, evaluating movement speed, rhythm, and symmetry, then comparing those metrics to patterns typical of a sober individual.

According to the research team, the system achieves an accuracy rate of about 93%. However, false positives are possible—particularly in cases involving fatigue, illness, or musculoskeletal conditions that may affect a person’s gait. The technology has already been tested using real-world video footage, including controlled scenarios where participants voluntarily consumed alcohol.

Developers say the system could be integrated into security setups in parking garages, near bar entrances, or at gas stations. If cameras detect suspicious movement from someone approaching a vehicle, the system could automatically alert security personnel or even prevent the car from starting through a connected mobile app.