Concept-to-implementation of New Threshold-based Fall Detection Sensor
Kaleb Clarke, Thilanga Ariyarathna, Sarla Kumari · TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON) · 2021
Fall-detection is a growing industry, and one which is increasingly subject to attention from medical professionals and academics. Fall-detection is commonly applicable to aged-care practices, where it is used to detect if a human subject has fallen over as a result of a trip, stumble or misstep. This study has researched, designed, and delivered an accurate and functional fall-detection application for elderly persons. Medical-alert products have been on the market since the 1990s, but most of these systems require user-interaction before respondents are dispatched. The proposed application uses accelerometers to measure the movements of its subject, and runs a unique, threshold-based algorithm to identify the pattern of an actual fall. The algorithm distinguishes falls from daily activities such as sitting, standing, lying down and walking short distances. The main attributes considered to differentiate these activities from actual falls are: peak acceleration, instantaneous jolt, and post-fall movement. The development of this program involved significant experimentation to establish statistical parameters related to fall data. The final application has a sensitivity of 100% and a selectivity of 97.5%, which offers a significant degree of protection against fall-related injuries. This paper gives insight into pertinent literature, presents the implementation details, and discloses experimental results which signify the validity.