Lightweight Real-time Fall Detection using Bidirectional Recurrent Neural Network

As the world's population is aging, the home care systems for elderly people have been getting high attention. According to the National Council on Aging, every 11 seconds, an older adult is treated in the emergency room for a fall, and every 19 minutes, an older adult dies from a fall. The number of single households is also increasing with an aging society. In a single household, there is no one to help the elderly when they fall. This could lead to serious problems such as disability or death. In this paper, we propose a lightweight real-time system for fall detection, distinguished from other activities of daily living (ADL). The entire system is divided into a preprocessing and prediction part. With the system, falls and ADLs can be distinguished with more than 92% accuracy which is higher than the existing approach even without any additional resampling method.