Abstract
In recent years, connected home healthcare, which involves multiple technologies such as wearable sensors, audio and video technology, and pervasive computing, has drawn attention for its ability to improve quality of life for elderly people. One of its most important services is fall detection. Falls represent a significant threat to the health and independence of adults older than 65. However, commercial fall detection devices are expensive and charge a monthly fee for their use. A more cost-effective, adaptable, and reliable fall detection system is necessary to detect falls and send alarms to an appropriate authority. This article introduces the MEFD framework that detects falls by elderly people and allows family members and caregivers to help by immediately locating them. In the proposed framework, real-time data retrieved from an accelerometer sensor on a smartphone are processed and analyzed by an online fall detection system running on the smartphone itself. The system sends an indoor sound alert to family members through a wireless access point at home or an outdoor SMS alert to a hospital or caregiver through a mobile network base station. A hybrid deep learning model is used to detect falls. The model is trained offline using a public dataset called MobiAct. Experimental results show that the proposed framework can detect falls from real-time streaming data with high accuracy compared to state-of-the-art approaches.