Accuracy in EEG Data Systems Acquisition for Motor Imagery BCI Home Assistive Applications
Year: 2026
Type: conference-paper
Abstract: The translation of Brain-Computer Interfaces (BCIs) from laboratory environments to real-world home assistive applications is heavily influenced by the choice of Electroencephalography (EEG) acquisition hardware. In this study, we systematically benchmark the performance of a wireless consumer-grade system (Emotiv EPOC X, 14 channels, saline electrodes) against a gold-standard research-grade system (BrainVision, 16 channels, wet gel electrodes) for a binary (left vs. right hand) motor imagery (MI) task. Utilizing a rigorous multi-domain feature extraction pipeline (combining 170 spectral features across four in-band frequency ranges) and 4 machine learning classifiers (SVM, Random Forest, XGBoost, LightGBM), we evaluate the devices across classification accuracy, signal-to-noise ratio (SNR), event-related desynchronization (ERD) lateralization, and Information Transfer Rate (ITR). Our results demonstrate that BrainVision achieved a peak classification accuracy of 95.8% (LightGBM) compared to 92.5% for Emotiv (XGBoost), alongside a +4.83 dB advantage in ASR-based SNR at motor-relevant channels (−0.28 vs. −5.11 dB), a nearly 1.9× higher lateralization index (+77.6 vs. +41.9), and a 1.2× higher Information Transfer Rate (8.18 vs. 6.71 bits/min). However, when transitioning BCIs to daily assistive use for patients with severe motor impairments, the lightweight, wireless, and rapid-setup nature of the consumer-grade headset provides a profound usability advantage over tethered, rigid laboratory equipment. These empirical and ergonomic findings, supported by established BCI literature, offer crucial insights and practical recommendations for balancing statistical precision with patient comfort during BCI hardware selection.
Institutions: University of Catania, University of Pisa
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: hybrid
OpenAlex ID: W7203751822