Personalization and Adaptation to the Medium and Context in a Fall Detection System

The main objective of this paper is to present a distributed processing architecture that explicitly integrates capabilities for its continuous adaptation to the medium, the context, and the user. This architecture is applied to a falling detection system through: 1) an optimization module that finds the optimal operation parameters for the detection algorithms of the system devices; 2) a distributed processing architecture that provides capabilities for remote firmware update of the smart sensors. The smart sensor also provides an estimation of activities of daily living (ADL), which results very useful in monitoring of the elderly and patients with chronic diseases. The developed experiments have demonstrated the feasibility of the system and specifically, the accuracy of the proposed algorithms and procedures (100% success for impact detection, 100% sensitivity and 95.68% specificity rates for fall detection, and 100% success for ADL level classification). Although the experiments have been developed with a cohort of young volunteers, the personalization and adaption mechanisms of the proposed architecture related to the concepts of "design for all" and "design space" will significantly ease the adaptation of the system for its application to the elderly.