Electroencephalogram based communication system for locked in state person using mentally spelled tasks with...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Increasing population leads to a rise in individuals with disabilities, particularly those with Locked-in State (LIS) resulting from Spinal Cord Injury (SCI).
• Effective assistive technologies are crucial for improving the quality of life for individuals with severe motor impairments.

Purpose of the Study:

• To design a four-state moving robot controlled by brain-computer interface (BCI) signals.
• To develop and validate an optimized neural network model for interpreting electroencephalogram (EEG) signals for robot control.
• To compare the performance of the proposed model against conventional neural network models.

Main Methods:

• Acquisition of four-imagery task signals using a three-electrode system (T1, T3, FP1).
• Feature extraction using Continuous Wavelet Transform (CWT).
• Training and evaluation of an optimized neural network model, compared with conventional Feed Forward Neural Network, Time Delay Neural Network, and Elman Neural Networks.

Main Results:

• The proposed optimized neural network model achieved a classification accuracy of 93.86% and an offline recognition accuracy of 97.50%.
• Information Transfer Rate (ITR) analysis showed the optimized model outperformed conventional models, reaching 21.67 bits per sec.
• The study indicated superior performance of male subjects compared to female subjects.

Conclusions:

• Continuous Wavelet Transform (CWT) features combined with an optimized neural network model offer superior performance for robot control in LIS patients.
• The developed system demonstrates significant potential for enhancing communication and mobility for individuals with severe disabilities.
• Further research may explore gender-based performance differences and refine the BCI system for broader applications.