This thesis presents the development of a smart shoe prototype for the detection and classification of motor events, with a particular focus on accidental falls. The proposed approach integrates machine learning techniques and knowledge representation to enable semantic analysis of data in embedded and Internet of Things (IoT) contexts. A dataset was selected from inertial sensors mounted on the shoe, and subsequently processed through preprocessing, semantic annotation, and classification using the MAFALDA algorithm (Matchmaking Features for mAchine Learning Data Analysis). The resulting architecture proves suitable for real-world scenarios with limited resources, enabling intelligent functionalities directly on-device.
La presente tesi descrive lo sviluppo di un prototipo di scarpa smart per il rilevamento e la classificazione di eventi motori, con particolare attenzione alla caduta accidentale. L’approccio proposto integra tecniche di apprendimento automatico e rappresentazione della conoscenza per abilitare un’analisi semantica dei dati in contesti embedded e Internet of Things (IoT). È stato selezionato un dataset a partire da sensori inerziali montati su una calzatura, poi sottoposto a un processo di preprocessing, annotazione semantica e classificazione tramite l’algoritmo MAFALDA (Matchmaking Features for mAchine Learning Data Analysis). L’architettura sviluppata si dimostra adatta a contesti reali caratterizzati da risorse limitate, abilitando funzionalità intelligenti direttamente on-device. .
ALGORITMI DI CLASSIFICAZIONE DI EVENTI SU BASE SEMANTICA PER UN PROTOTIPO DI SCARPA SMART
RICCHIUTI, CARLO
2024/2025
Abstract
This thesis presents the development of a smart shoe prototype for the detection and classification of motor events, with a particular focus on accidental falls. The proposed approach integrates machine learning techniques and knowledge representation to enable semantic analysis of data in embedded and Internet of Things (IoT) contexts. A dataset was selected from inertial sensors mounted on the shoe, and subsequently processed through preprocessing, semantic annotation, and classification using the MAFALDA algorithm (Matchmaking Features for mAchine Learning Data Analysis). The resulting architecture proves suitable for real-world scenarios with limited resources, enabling intelligent functionalities directly on-device.I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14253/87528