Assistive technologies are strategic to implement solutions for the well-being and active ageing of elderlies and people with neurological diseases. In this scenario, the monitoring of users' postural status is valuable also as an effective way to predict the risk of falls. This paper presents a customized solution for real-time postural sway monitoring, exploiting an embedded sensing node. The proposed approach allows for a reliable continuous monitoring of the user postural status, also thanks to the use of a Machine Learning approach for the classification of different classes of postural dynamics. Moreover, a MQTT based data-to-cloud approach has been implemented enabling the realization of a distributed sensor network aimed at the monitoring of frail users. The following indexes have been used to assess performances of the Machine Learning based classification algorithm: Accuracy, Precision, Recall and F1. In case of the test dataset, a 99.9% value has been obtained for the whole set of above-mentioned indexes.
A Customized Sensing Solution for Real-Time Postural Sway Monitoring, Exploiting MQTT Protocol for Data-to-Cloud Delivery
Ando, Bruno;Manenti, Mattia;Pistorio, Antonio
2024-01-01
Abstract
Assistive technologies are strategic to implement solutions for the well-being and active ageing of elderlies and people with neurological diseases. In this scenario, the monitoring of users' postural status is valuable also as an effective way to predict the risk of falls. This paper presents a customized solution for real-time postural sway monitoring, exploiting an embedded sensing node. The proposed approach allows for a reliable continuous monitoring of the user postural status, also thanks to the use of a Machine Learning approach for the classification of different classes of postural dynamics. Moreover, a MQTT based data-to-cloud approach has been implemented enabling the realization of a distributed sensor network aimed at the monitoring of frail users. The following indexes have been used to assess performances of the Machine Learning based classification algorithm: Accuracy, Precision, Recall and F1. In case of the test dataset, a 99.9% value has been obtained for the whole set of above-mentioned indexes.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.