Spatial and Time Domain Feature of ERP Speller System Extracted via Convolutional Neural Network

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Understanding event-related potentials (ERPs) and the impact of illiteracy remains a challenge.
• The P300 peak, commonly used in brain-computer interfaces (BCIs), is not consistently observed across all subjects.
• Convolutional Neural Networks (CNNs) offer advanced capabilities for analyzing complex spatio-temporal features in neural data.

Purpose of the Study:

• To investigate the distinct spatio-temporal features of ERPs in illiterate versus non-illiterate individuals.
• To identify reliable ERP features for BCI applications, particularly those present in both subject groups.
• To explore the utility of CNNs in analyzing ERP data and differentiating neural patterns.

Main Methods:

• A convolutional neural network with two convolutional layers was trained to analyze spatial and temporal ERP features.
• ERP data from illiterate and non-illiterate subjects were compared.
• Neural activity correlations and peak latencies were analyzed.

Main Results:

• Non-illiterate subjects exhibited high correlation between occipital and parietal lobes, while illiterate subjects showed correlation between frontal and central lobes.
• Non-illiterates displayed ERP peaks at P300, P500, and P700; illiterates primarily showed peaks around P700.
• The P700 peak was prominent in both illiterate and non-illiterate subjects.

Conclusions:

• The P700 peak is a robust feature of event-related potentials, consistently present in both illiterate and non-illiterate individuals.
• This finding suggests P700 as a potential key feature for developing more inclusive BCIs.
• CNNs effectively capture spatio-temporal dynamics of ERPs, aiding in the understanding of neural correlates of literacy.