A Multisensor Data-Fusion Approach for ADL and Fall classification

Abstract

This paper deals with an advanced approach for the monitoring of elders and people with neurological pathologies (e.g., Alzheimer). The system that adopts a multisensor approach is able to recognize critical events, such as falls or prolonged inactivity, to monitor the user posture, and to notify the alerts to caregivers. In particular, this paper focuses on smart algorithms developed for the activities of daily living (ADL) classification, which use the information provided by inertial sensors embedded in the user device. In particular, a novel multisensor data-fusion approach, combining data from an accelerometer and a gyroscope, is presented. Apart from alert management, the information provided by this system is useful to track the evolution of the user pathology, also during rehabilitation tasks. Results obtained during tests with users demonstrate suitable performances of the adopted paradigm, in terms of sensitivity and specificity, in performing falls and ADL classification tasks. The average value of the sensitivity index for the classes of falls and ADL considered through this paper is 0.81, while the average value of the specificity index is 0.98.