Visual and haptic feedback in detecting motor imagery within a wearable brain–computer interface. (January 2023)

• Record Type: Journal Article
• Authors: Arpaia, Pasquale; Coyle, Damien; Donnarumma, Francesco; Esposito, Antonio; Natalizio, Angela; Parvis, Marco
• Is Part Of: Measurement. Volume 206(2023)
• Publication Date: 2023-01
• Subjects: Brain–computer interface -- Motor imagery -- Electroencephalography -- Extended reality -- Haptic -- Neurofeedback; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.112304
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: This paper presents a wearable brain–computer interface relying on neurofeedback in extended reality for the enhancement of motor imagery training. Visual and vibrotactile feedback modalities were evaluated when presented either singularly or simultaneously. Only three acquisition channels and state-of-the-art vibrotactile chest-based feedback were employed. Experimental validation was carried out with eight subjects participating in two or three sessions on different days, with 360 trials per subject per session. Neurofeedback led to statistically significant improvement in performance over the two/three sessions, thus demonstrating for the first time functionality of a motor imagery-based instrument even by using an utmost wearable electroencephalograph and a commercial gaming vibrotactile suit. In the best cases, classification accuracy exceeded 80 % with more than 20 % improvement with respect to the initial performance. No feedback modality was generally preferable across the cohort study, but it is concluded that the best feedback modality may be subject-dependent. Highlights: Wearable brain–computer interface with vibrotactile chest based feedback Non-invasive measures are exploited along with an extended reality scenario Different feedback modalities are exploited Machine learning is adopted both for offline and online signal processing The validation was carried out by using data from 8 subjects in a total of 20 session