ENHANCING CLASSIFICATION ACCURACY OF EXO-ANKLE USING HYBRID FNIRS-EMG BCI

Abstract

Brain-computer interfaces (BCIs) have grown as novel technologies that utilize brain activity for communication and control, typically via non-invasive techniques like electroencephalogram (EEG). Although significant advancements have been made, dependence on single-modality systems frequently impacts classification accuracy, limiting their practical use. This study addresses the challenge by utilizing electromyography (EMG) and functional near-infrared spectroscopy (fNIRS) features to improve classification accuracy for motor cortex activity. The aim was to integrate data obtained from several modalities to enhance performance compared to single-modality techniques. EMG and fNIRS data have been acquired during the same motor tasks for four separate classes. To determine the optimal features combination, the Particle Swarm Optimization (PSO) technique has been applied with linear discriminant analysis (LDA), achieving a classification accuracy of 98%. The findings indicated considerable enhancements in classification accuracy by integrating EMG and fNIRS features. This research advances the non-invasive brain-computer technique by presenting a new method for combining features of hybrid modality, allowing the development of more reliable and precise systems for motor rehabilitation and assistive technologies.