The evolution of millimeter-wave radar technologies has opened new perspectives for indoor human monitoring in healthcare and assistive contexts, by enabling non-invasive, privacy-preserving solutions that are independent of lighting conditions. Within this framework, this thesis aims to design and experimentally validate an indoor monitoring system based on mmWave FMCW radar, intended for the estimation of human presence, position, and posture in indoor environments. The work systematically addresses the entire processing chain, from radar sensor configuration to the generation and analysis of three-dimensional point clouds in Cartesian coordinates. In particular, a modular and reproducible software pipeline is defined, encompassing data acquisition, geometric transformation of radar measurements, subject tracking, and the extraction of geometric and kinematic features. Specific attention is devoted to the analysis of the impact of the main sensor configuration parameters, examining their practical implications in terms of field of view, spatial resolution, tracking stability, and data quality. System validation is carried out through a structured experimental protocol based on controlled acquisitions in a realistic indoor environment and the use of multimodal ground truth annotations. Experimental results highlight a good capability of the system to detect human presence and estimate subject position, as well as to distinguish between static and dynamic postures in single-subject scenarios. At the same time, several critical issues emerge, particularly related to the management of complex scenarios, point cloud stability, and ambiguities in certain postural transitions. Overall, the work demonstrates the potential of mmWave FMCW radar as an enabling technology for non-invasive indoor monitoring, while also highlighting current limitations and outlining possible future developments oriented toward multimodal integration and improved algorithmic robustness.
L’evoluzione delle tecnologie radar a onde millimetriche ha aperto nuove prospettive per il monitoraggio indoor di soggetti umani in contesti assistivi e sanitari, grazie a soluzioni non invasive, privacy-preserving e indipendenti dalle condizioni di illuminazione. In tale ambito, la presente tesi si propone di progettare e validare sperimentalmente un sistema di monitoraggio indoor basato su radar mmWave FMCW, finalizzato alla stima della presenza, della posizione e della postura di soggetti umani in ambienti chiusi. Il lavoro affronta in modo sistematico l’intera catena di elaborazione, dalla configurazione del sensore radar alla generazione e all’analisi della point cloud tridimensionale in coordinate cartesiane. In particolare, viene definita una pipeline software modulare e riproducibile, comprendente l’acquisizione dei dati, la trasformazione geometrica delle misure radar, il tracciamento dei soggetti e l’estrazione di feature geometriche e cinematiche. Un’attenzione specifica è dedicata allo studio dell’impatto dei principali parametri di configurazione del sensore, analizzandone le implicazioni pratiche in termini di campo visivo, risoluzione spaziale, stabilità del tracciamento e qualità dei dati. La validazione del sistema è condotta mediante un protocollo sperimentale strutturato, basato su acquisizioni controllate in un ambiente indoor realistico e sull’impiego di annotazioni di ground truth multimodali. I risultati sperimentali evidenziano una buona capacità del sistema di rilevare la presenza e di stimare la posizione dei soggetti, nonché di distinguere posture statiche e dinamiche in scenari a singolo soggetto. Al contempo, emergono alcune criticità legate alla gestione di scenari complessi, alla stabilità della point cloud e all’ambiguità di alcune transizioni posturali. Nel complesso, il lavoro dimostra il potenziale dei radar mmWave FMCW come tecnologia abilitante per il monitoraggio indoor non invasivo, evidenziandone al contempo i limiti attuali e delineando possibili sviluppi futuri orientati all’integrazione multimodale e al miglioramento della robustezza algoritmica.
Progettazione e validazione sperimentale di un sistema di monitoraggio indoor basato su radar mmWave FMCW
EGIDATI, LEONARDO
2025/2026
Abstract
The evolution of millimeter-wave radar technologies has opened new perspectives for indoor human monitoring in healthcare and assistive contexts, by enabling non-invasive, privacy-preserving solutions that are independent of lighting conditions. Within this framework, this thesis aims to design and experimentally validate an indoor monitoring system based on mmWave FMCW radar, intended for the estimation of human presence, position, and posture in indoor environments. The work systematically addresses the entire processing chain, from radar sensor configuration to the generation and analysis of three-dimensional point clouds in Cartesian coordinates. In particular, a modular and reproducible software pipeline is defined, encompassing data acquisition, geometric transformation of radar measurements, subject tracking, and the extraction of geometric and kinematic features. Specific attention is devoted to the analysis of the impact of the main sensor configuration parameters, examining their practical implications in terms of field of view, spatial resolution, tracking stability, and data quality. System validation is carried out through a structured experimental protocol based on controlled acquisitions in a realistic indoor environment and the use of multimodal ground truth annotations. Experimental results highlight a good capability of the system to detect human presence and estimate subject position, as well as to distinguish between static and dynamic postures in single-subject scenarios. At the same time, several critical issues emerge, particularly related to the management of complex scenarios, point cloud stability, and ambiguities in certain postural transitions. Overall, the work demonstrates the potential of mmWave FMCW radar as an enabling technology for non-invasive indoor monitoring, while also highlighting current limitations and outlining possible future developments oriented toward multimodal integration and improved algorithmic robustness.| File | Dimensione | Formato | |
|---|---|---|---|
|
Egidati_Leonardo.pdf
Accesso riservato
Dimensione
3.97 MB
Formato
Adobe PDF
|
3.97 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/106273