During the last decade, the use of inertial sensors in human motion analysis has occupied a central role in the biomedical and electronic engineering scientific literature. The portability and relative low cost of such equipment opened up to a number of domestic, clynical, research and rehabilitation scenarios, leading to a growing interest in software methodologies to obtain accurate data from inertial and magnetic information. In the present dissertation a thorough study is reported regarding a series of common application of inertial sensors in human motion analysis, considering purely inertial measurements. Regardless of the complexity of the techniques that may be involved in the compensation of magnetic distorsions, the present work aims to investigate and assess the accuracy limits of purely inertial setups, in order to give an estimate of the reliability of such kind of minimal setup when other information sources are obstacled for relatively long epochs of time (if not for the whole duration of the measurement session). The present experimental studies concern a comparative study on Extended and Unscented Kalman filtering for sensor fusion in displacement and orientation estimation of human center of mass (CoM) during treadmill walking, center of pressure (CoP) and lower body joint angles estimation during unperturbed posture, pedestrian dead reckoning. A custom calibration method for navigation and local reference frames alignment in magnetometer-free setups is also reported. Results present accuracies within 2 deg (orientation) and 1.5 mm (displacement) for CoM and lower body joints angles estimations, high correlation between estimated and measured CoP, predictability in dead reckoning error drift and improvement in 90 deg turns estimation accuracy using a custom methodology.
Negli ultimi dieci anni, l'uso di sensoristica inerziale nell'ambito dell'analisi del movimento ha occupato un ruolo centrale nella letteratura scientifica relativamente agli ambiti di ingegneria elettronica e biomedica. La portabilità ed il basso costo di questo tipo di strumentazione hanno generanto un cresente interesse nello sviluppo di metodologie software per ottenere informazione accurata da dati inerziali e magnetici. Nel presente lavoro è riportato uno studio su comuni scenari di applicazione di tale strumentazione in analisi del movimento, considerando tuttavia misure puramente inerziali. Indipendentemente dalla complessità delle tecniche adottabili nella compensazione dei disturbi magnetici, il presente lavoro mira ad investigare e verificare i limiti di accuratezza ottenibili adottando assetti di misura puramente inerziali, al fine di esprimere una stima dell'attendibilità di questo tipo di configurazione, imprescindibile qualora altre sorgenti di informazione siano ostacolate per periodi di misura relativamente lunghi. Gli studi presentati riguardano uno studio comparativo sulle varianti Extended ed Unscented del filtro di Kalman nella stima di orientazione e spostamento del centro di massa (CoM) durante cammino su tapis-roulant, stima del centro di pressione (CoP) e degli angoli delle articolazioni dell'arto inferiore durante postura non perturbata, pedestrian dead-reckoning. Viene presentata inoltre una metodologia personalizzata per l'allineamento dei sistemi di navigazione e locale. I risultati evidenziano accuratezze entro 2 deg (orientazione) e 1.5 mm (spostamento) nella stima del CoM e angoli dell'arto inferiore, alte correlazioni riguardanti il CoP stimato e misurato, predicibilità nella deriva d'errore nel pedestrian dead-reckoning e miglioramenti nella stima delle rapide variazioni d'orientazione (90 deg) tramite una tecnica sperimentale qui proposta.
Optimal Filtering and Magnetometer-free Sensor Fusion: Applications to Human Motion Inertial Sensing / Cardarelli, Stefano. - (2020 Mar 20).
Optimal Filtering and Magnetometer-free Sensor Fusion: Applications to Human Motion Inertial Sensing
CARDARELLI, STEFANO
2020-03-20
Abstract
During the last decade, the use of inertial sensors in human motion analysis has occupied a central role in the biomedical and electronic engineering scientific literature. The portability and relative low cost of such equipment opened up to a number of domestic, clynical, research and rehabilitation scenarios, leading to a growing interest in software methodologies to obtain accurate data from inertial and magnetic information. In the present dissertation a thorough study is reported regarding a series of common application of inertial sensors in human motion analysis, considering purely inertial measurements. Regardless of the complexity of the techniques that may be involved in the compensation of magnetic distorsions, the present work aims to investigate and assess the accuracy limits of purely inertial setups, in order to give an estimate of the reliability of such kind of minimal setup when other information sources are obstacled for relatively long epochs of time (if not for the whole duration of the measurement session). The present experimental studies concern a comparative study on Extended and Unscented Kalman filtering for sensor fusion in displacement and orientation estimation of human center of mass (CoM) during treadmill walking, center of pressure (CoP) and lower body joint angles estimation during unperturbed posture, pedestrian dead reckoning. A custom calibration method for navigation and local reference frames alignment in magnetometer-free setups is also reported. Results present accuracies within 2 deg (orientation) and 1.5 mm (displacement) for CoM and lower body joints angles estimations, high correlation between estimated and measured CoP, predictability in dead reckoning error drift and improvement in 90 deg turns estimation accuracy using a custom methodology.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Cardarelli.pdf
accesso aperto
Descrizione: Human motion analysis
Tipologia:
Tesi di dottorato
Licenza d'uso:
Creative commons
Dimensione
7.65 MB
Formato
Adobe PDF
|
7.65 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.