Fall Detection using Motion History Image and Shape Deformation
Jantima Thummala, Suree Pumrin
Abstract
Computer vision takes more roles in daily life both at work and home. Especially, in healthcare area, it assists and relieves the burden of caregivers at either nursing homes or home environments. This paper presents a fall event detection applying computer image processing techniques as an assistant tool for a caregiver to look after the elderly. The system works as a video surveillance that detects activities of an elderly one and notifies a caregiver at certain events. Our system classifies activity into two categories: (1) a normal daily activity such as walking, sitting, and running, (2) an abnormal activity such as falling. The fall detection separates into two steps: first, the system applies a Motion History Image (MHI) to quantify human motion for a moment in terms of its percentage. Next, it analyzes human shape deformation after detecting a large motion (MHI >65%). Shape deformation consists of a ratio and speed of change in shape. The experimental results demonstrate that our system can accurately achieve fall detection by 95.16%.