Enhancement of SSVEPs Classification in BCI-Based Wearable Instrumentation Through Machine Learning Techniques

• Published in: IEEE sensors journal 2022-05, Vol.22 (9), p.9087-9094
• Main Authors: Apicella, Andrea, Arpaia, Pasquale, De Benedetto, Egidio, Donato, Nicola, Duraccio, Luigi, Giugliano, Salvatore, Prevete, Roberto
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial neural networks; Augmented reality; BCI; brain-computer interface; Classification; Classification algorithms; Convolutional neural networks; EEG; Electroencephalography; Helmet mounted displays; Human-computer interface; industry 4.0; instrumentation; Instruments; Machine learning; neural networks; Performance enhancement; real-time systems; Sensors; SSVEP; Support vector machines; wearable; Wearable technology
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2022.3161743

This work addresses the adoption of Machine Learning classifiers and Convolutional Neural Networks to improve the performance of highly wearable, single-channel instrumentation for Brain-Computer Interfaces. The proposed measurement system is based on the classification of Steady- State Visually Evoked Potentials (SSVEPs). In particular, Head-Mounted Displays for Augmented Reality are used to generate and display the flickering stimuli for the SSVEPs elicitation. Four experiments were conducted by employing, in turn, a different Head-Mounted Display. For each experiment, two different algorithms were applied and compared with the state-of-the-art-techniques. Furthermore, the impact of different Augmented Reality technologies in the elicitation and classification of SSVEPs was also explored. The experimental metrological characterization demonstrates (i) that the proposed Machine Learning-based processing strategies providea significantenhancement of theSSVEPclassification accuracy with respect to the state of the art, and (ii) that choosing an adequate Head-Mounted Display is crucial to obtain acceptable performance. Finally, it is also shown that the adoption of inter-subjective validation strategies such as the Leave-One-Subject-Out Cross Validation successfully leads to an increase in the inter-individual 1- \sigma reproducibility: this, in turn, anticipates an easier development of ready-to-use systems.