Authentication with a one-dimensional CNN model using EEG-based brain-computer interface

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-computer interface (BCI) technology leverages electroencephalogram (EEG) signals for direct human-environment interaction.
• Motor imagery (MI) classification via EEG is crucial for assisting stroke survivors and individuals with motor impairments.
• Current EEG classification faces challenges due to noise, signal interference, and limited generalization, hindering practical applications.

Purpose of the Study:

• To develop a robust and efficient deep learning methodology for classifying EEG signals corresponding to motor imagery tasks (right hand, left hand, feet, sedentary).
• To introduce a lightweight 1-D Convolutional Neural Network (1-D CNN) model that minimizes parameters while maximizing classification accuracy.
• To explore innovative applications of the four-class output for secure authentication systems and assistive technology for individuals with disabilities.

Main Methods:

• Implementation of a one-dimensional Convolutional Neural Network (1-D CNN) architecture tailored for EEG signal processing.
• Training and validation of the 1-D CNN model on motor imagery datasets encompassing four distinct tasks.
• Evaluation of the model's performance based on classification accuracy and parameter efficiency.

Main Results:

• The proposed 1-D CNN model achieved a high classification accuracy of 91.75% for the four motor imagery tasks.
• The model is characterized as lightweight, featuring a reduced number of parameters compared to existing deep learning approaches.
• Demonstrated the feasibility of utilizing the four output classes for novel applications, including secure password authentication and assistive control.

Conclusions:

• The developed 1-D CNN model offers a practical and accurate solution for EEG-based motor imagery classification.
• The lightweight nature of the model enhances its suitability for real-time BCI applications and deployment on resource-constrained devices.
• The innovative application of multi-class EEG signal classification presents a promising avenue for enhanced security systems and assistive technologies for individuals with motor impairments.