As the global population ages, the demand for intelligent monitoring systems that support independent living while ensuring safety and well-being increases significantly. Ambient Assisted Living (AAL) technologies leverage unobtrusive sensing technologies, such as Passive Infrared (PIR) motion sensors, to track daily activity patterns and detect deviations that may indicate potential health concerns. This study aims to develop an unobtrusive in-home monitoring framework that overcomes PIR sensor limitations and enhances occupancy prediction accuracy using Long Short-Term Memory (LSTM) networks.We conducted a study in a single-resident apartment setup, where strategically placed PIR sensors were deployed. The analysis of habitual patterns in elderly individuals was performed by a predictive modeling approach using LSTM networks to forecast occupancy behavior over segments of 15, 30, 45, and 60 minutes.The produced LSTM-based predictive model on the 15-minute segments reached 93.0% training accuracy, with a validation accuracy of 91.4%, indicating strong predictive performance for short-term occupancy forecasting. The habitual disaccordance metric, used to quantify deviations between predicted and actual occupancy patterns, demonstrated its ability to successfully identify changes in daily routines and potential emergency situations.These findings suggest that a motion-sensor-based system utilizing LSTM models to analyze habitual patterns and behavioral discrepancies can be a valuable tool for detecting routine changes or emergencies, ultimately improving elderly care.Clinical Relevance- This study highlights the potential of motion-sensor systems powered by LSTM-based predictive models to enhance in-home monitoring for elderly individuals, enabling early detection of routine changes and potential emergencies while supporting independent living and ensuring timely intervention that can significantly reduce the risk of adverse health outcomes.
Analyzing Habitual Patterns and Behavioral Discrepancies in Ambient Assisted Living: An LSTM-Based Predictive Model / Miladinović, A.; Biscontin, A.; Kresevic, S.; Vascotto, M.; Gomiseli, C.; Accardo, A.; Ajčević, M.. - 2025:(2025), pp. 1-5. ( 47th Annual International Conference of the IEEE Engineering in Medicine and Biology Society Copenhagen, Danimarca 14 - 18 Luglio 2025) [10.1109/EMBC58623.2025.11254663].
Analyzing Habitual Patterns and Behavioral Discrepancies in Ambient Assisted Living: An LSTM-Based Predictive Model
Miladinović A.
Primo
;Biscontin A.Secondo
;Kresevic S.;Accardo A.;AjčevićM.Ultimo
2025-01-01
Abstract
As the global population ages, the demand for intelligent monitoring systems that support independent living while ensuring safety and well-being increases significantly. Ambient Assisted Living (AAL) technologies leverage unobtrusive sensing technologies, such as Passive Infrared (PIR) motion sensors, to track daily activity patterns and detect deviations that may indicate potential health concerns. This study aims to develop an unobtrusive in-home monitoring framework that overcomes PIR sensor limitations and enhances occupancy prediction accuracy using Long Short-Term Memory (LSTM) networks.We conducted a study in a single-resident apartment setup, where strategically placed PIR sensors were deployed. The analysis of habitual patterns in elderly individuals was performed by a predictive modeling approach using LSTM networks to forecast occupancy behavior over segments of 15, 30, 45, and 60 minutes.The produced LSTM-based predictive model on the 15-minute segments reached 93.0% training accuracy, with a validation accuracy of 91.4%, indicating strong predictive performance for short-term occupancy forecasting. The habitual disaccordance metric, used to quantify deviations between predicted and actual occupancy patterns, demonstrated its ability to successfully identify changes in daily routines and potential emergency situations.These findings suggest that a motion-sensor-based system utilizing LSTM models to analyze habitual patterns and behavioral discrepancies can be a valuable tool for detecting routine changes or emergencies, ultimately improving elderly care.Clinical Relevance- This study highlights the potential of motion-sensor systems powered by LSTM-based predictive models to enhance in-home monitoring for elderly individuals, enabling early detection of routine changes and potential emergencies while supporting independent living and ensuring timely intervention that can significantly reduce the risk of adverse health outcomes.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
