Development of a Real-Time, Simple and High-Accuracy Fall Detection System for Elderly Using 3-DOF Accelerometers

• Published in: Arabian journal for science and engineering (2011) Vol. 44; no. 4; pp. 3329 - 3342
• Main Authors: Van Thanh, Pham, Tran, Duc-Tan, Nguyen, Dinh-Chinh, Duc Anh, Nguyen, Nhu Dinh, Dang, El-Rabaie, S., Sandrasegaran, Kumbesan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-04-2019; Springer Nature B.V
• Subjects: Accelerometers; Accuracy; Algorithms; Datasets; Engineering; Fall detection; Humanities and Social Sciences; multidisciplinary; Older people; Recognition; Research Article - Computer Engineering and Computer Science; Science; Sensitivity; Smartphones; SMS; Post-fall recognition; Upper fall threshold (UFT); VIP contacts; Vertical velocity; Lower fall threshold (LFT)
• ISSN: 2193-567X; 1319-8025; 2191-4281
• DOI: 10.1007/s13369-018-3496-4

Falls represent a major problem for the elderly people aged 60 or above. There are many monitoring systems which are currently available to detect the fall. However, there is a great need to propose a system which is of optimal effectiveness. In this paper, we propose to develop a low-cost fall detection system to precisely detect an event when an elderly person accidentally falls. The fall detection algorithm compares the acceleration with lower fall threshold and upper fall threshold values to accurately detect a fall event. The post-fall recognition module is the combination of posture recognition and vertical velocity estimation that has been added to our proposed method to enhance the performance and accuracy. In case of a fall, our device will transmit the location information to the contacts instantly via SMS and voice call. A smartphone application will ensure that the notifications are delivered to the elderly person’s relatives so that medical attention can be provided with minimal delay. The system was tested by volunteers and achieved 100% sensitivity and accuracy. This was confirmed by testing with public datasets and it also achieved the same percentage in sensitivity and accuracy as in our recorded datasets.