Real-time fall detection based on vision for low-power edge platforms

Falling detection is essential for elder care and surveillance, yet most vision methods treat it as static pose classification or simple temporal matching, ignoring instability dynamics.