Classification of finger vibrotactile input using scalp EEG

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) primarily focus on output pathways, with limited research on decoding sensory input.
• Scalp electroencephalography (EEG) offers a non-invasive method for neural signal acquisition.
• Understanding sensory input pathways is crucial for developing comprehensive BCIs.

Purpose of the Study:

• To investigate the feasibility of building a BCI system capable of decoding sensory input.
• To classify neural activity associated with vibrotactile stimuli delivered to different fingers using scalp EEG.

Main Methods:

• Utilized Local Fisher Discriminant Analysis (LFDA) and Gaussian Mixture Model (GMM) for neural activity classification.
• Recorded EEG data from participants receiving vibrotactile stimuli on individual fingertips.
• Designed two classification tasks: differentiating ipsilateral and contralateral finger stimuli.

Main Results:

• Achieved high decoding accuracies of 97.6% for ipsilateral finger differentiation and 99.3% for contralateral finger differentiation.
• Identified relevant event-related EEG features in both amplitude and power domains for accurate classification.

Conclusions:

• Scalp EEG can effectively decode vibrotactile sensory input to the fingers.
• The developed LFDA-GMM classifier demonstrates high performance in classifying neural activity related to tactile stimuli.
• This research supports the development of BCIs with integrated sensory input capabilities.