Error Potentials (ErrPs) are neurophysiological signals generated by users when they perceive errors in their actions and during interaction with brain-computer interfaces (BCIs), following an incorrect response of the BCI. This work focuses on identifying and detecting ErrP signals during the discrete control of a powered wheelchair. Nine healthy subjects voluntarily participated in the experiment. Electroencephalogram (EEG) signals were acquired from the subjects while they controlled the powered wheelchair using a joystick along a predefined path. Random errors were intentionally introduced during the control sessions to elicit ErrP responses. The EEG signals were analyzed to identify, to characterize the ErrPs, and ultimately to construct a classifier capable of detecting them. The results show a differentiation between the neural responses corresponding to correct and erroneous actions, confirming the presence of distinct ErrP signals following incorrect commands during the discrete control of the wheelchair. A classifier was successfully developed and trained to detect these ErrP signals on a trial-by-trial basis, showcasing promising accuracy in identifying real-time errors. Furthermore, individual variability in neural activity among subjects was acknowledged, highlighting the necessity for personalized calibration and optimization of system parameters. Future directions involve extending this research to more complex environments without predefined paths to simulate realistic scenarios and testing the system’s efficacy with individuals having motor impairments, who are the final end-users.
I potenziali di errore (ErrP) sono segnali neurofisiologici generati dagli utenti quando percepiscono errori nelle loro azioni e durante l’interazione con le interfacce cervello-computer (BCI), in seguito a una risposta errata della BCI. Questo lavoro si concentra sull’identificazione e la rilevazione dei segnali ErrP durante il controllo discreto di una sedia a rotelle motorizzata. Nove soggetti sani hanno partecipato volontariamente all’esperimento. I segnali dell’elettroencefalogramma (EEG) dei soggetti sono stati acquisiti mentre controllavano la sedia a rotelle motorizzata utilizzando un joystick lungo un percorso predefinito. Durante le sessioni di controllo sono stati introdotti intenzionalmente errori casuali per suscitare i potenziali di errore. I segnali EEG sono stati analizzati per identificare e caratterizzare gli ErrP e, infine, costruire un classificatore in grado di rilevarli. I risultati mostrano una differenziazione tra le risposte neurali corrispondenti ad azioni corrette ed errate, confermando la presenza di segnali ErrP distinti in seguito a comandi errati durante il controllo discreto della sedia a rotelle. È stato sviluppato con successo un classificatore in grado di rilevare questi segnali ErrP per ogni comando dato alla sedia a rotella, dimostrando una promettente accuratezza nell’identificazione degli errori in tempo reale. Inoltre, è stata riconosciuta la variabilità individuale dell’attività neurale tra i soggetti, evidenziando la necessità di una calibrazione soggetto-specifica e dell’ottimizzazione dei parametri del sistema. Le direzioni future prevedono l’estensione di questa ricerca ad ambienti più complessi e privi di percorsi predefiniti, per simulare scenari realistici, e la verifica dell’efficacia del sistema con persone con disabilità motorie, che saranno gli utenti finali.
Analysis of EEG Signals for the Detection of Erroneous Commands During the Control of a Powered Wheelchair
CATTABRIGA, MICHELA
2022/2023
Abstract
Error Potentials (ErrPs) are neurophysiological signals generated by users when they perceive errors in their actions and during interaction with brain-computer interfaces (BCIs), following an incorrect response of the BCI. This work focuses on identifying and detecting ErrP signals during the discrete control of a powered wheelchair. Nine healthy subjects voluntarily participated in the experiment. Electroencephalogram (EEG) signals were acquired from the subjects while they controlled the powered wheelchair using a joystick along a predefined path. Random errors were intentionally introduced during the control sessions to elicit ErrP responses. The EEG signals were analyzed to identify, to characterize the ErrPs, and ultimately to construct a classifier capable of detecting them. The results show a differentiation between the neural responses corresponding to correct and erroneous actions, confirming the presence of distinct ErrP signals following incorrect commands during the discrete control of the wheelchair. A classifier was successfully developed and trained to detect these ErrP signals on a trial-by-trial basis, showcasing promising accuracy in identifying real-time errors. Furthermore, individual variability in neural activity among subjects was acknowledged, highlighting the necessity for personalized calibration and optimization of system parameters. Future directions involve extending this research to more complex environments without predefined paths to simulate realistic scenarios and testing the system’s efficacy with individuals having motor impairments, who are the final end-users.File | Dimensione | Formato | |
---|---|---|---|
Cattabriga_Michela.pdf
accesso aperto
Dimensione
10.75 MB
Formato
Adobe PDF
|
10.75 MB | Adobe PDF | Visualizza/Apri |
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/20.500.12608/60578