81690 research outputs found
Sort by
Simulation and Detection of Abnormal Behaviors for the Elderly Living Alone in a Smart Home
No full textThis study presents a simulator-based approach to detect anomalies among elderly individuals living alone in smart homes. The proposed simulator generates six typical anomalies: being semi-bedridden, being housebound, wandering, forgetting to turn off appliances, falls while walking, and falls while standing. These anomalies, categorized into state, activity, and moving types, are statistically modeled based on the mini-mental state examination score, which governs their frequency and duration as a proxy for dementia progression. Unlike previous simulators, this framework can generate both short-term and long-term anomalies, ranging from seconds to years, and simulate outputs from ambient sensors such as passive infrared motion sensors and pressure sensors.
Experimental evaluations on simulation data demonstrate the realism of activity sequences and the practicality of the proposed anomaly detection methods. The generated activity sequences closely mimic real-world activity patterns, exhibiting variances and similarities, measured by edit distance, greater than random data. Detection classifiers trained on simulated data achieved a sensitivity of over 0.9 with fewer than one false alarm every 50 days for most anomalies. Sensitivity for falls while walking, initially 0.75, can be improved to 0.96 through oversampling. Newly proposed detection methods for being housebound, being semi-bedridden, and forgetting achieved high precision and recall. By leveraging simulated data, the framework avoids the high costs and challenges associated with collecting long-term real-world data while maintaining a high degree of similarity to real activity sequences.
This work highlights the potential of privacy-preserving ambient sensors, such as infrared motion sensors, pressure sensors, door sensors, and cost sensors, to detect long-term behavioral changes in elderly individuals. The simulator-based framework offers a cost-effective solution for enhancing care in smart homes, emphasizing its scalability and adaptability. Future research will focus on refining the simulation models, incorporating richer sensor data, and bridging the gap between simulated training and real-world deployment to further support early intervention and improved elderly care
Policy significance of comprehensive initiative for long-term care prevention and daily life support : Focusing on its positioning as one of community-based support initiatives included within long-term care insurance
No full textThis article considers positioning and framework, historical background, practical examples, acceptability of inclusion within long-term care insurance, a type of social insurance, and necessity of positioning it as a community-based support initiative rather than an insurance benefit regarding comprehensive initiative for long-term care prevention and daily life support, in order to clarify its policy significance by focusing on its positioning as one of community based support initiatives included within long-term care insurance, based not only on a legal perspective but also on a practical perspective