Classification of Motor Imagery EEG signals using high resolution time-frequency representations and convolutional...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor imagery (MI) based Brain Computer Interface (BCI) systems are crucial for neuro-rehabilitation in individuals with motor disabilities and brain injuries.
• Classifying left and right hand MI tasks is essential for developing effective BCI systems.

Purpose of the Study:

• To classify left and right hand motor imagery (MI) tasks using Electroencephalogram (EEG) signals.
• To improve the accuracy of MI-based BCI systems for neuro-rehabilitation.

Main Methods:

• Utilized Complex Morlet Wavelets (CMW) with frequency-dependent widths for high-resolution time-frequency representations (TFR) of EEG signals from channels C3 and C4.
• Developed a novel method for selecting the number of cycles relative to the center frequency of CMW for feature extraction.
• Employed a Convolutional Neural Network (CNN) to classify the generated TFRs into left or right hand MI tasks.

Main Results:

• Achieved a classification accuracy of 82.2% on the BCI Competition IV dataset 2a.
• Demonstrated that the proposed TFR generation method yields higher classification accuracy compared to baseline and existing algorithms.

Conclusions:

• The proposed framework effectively classifies left and right hand MI tasks using advanced EEG signal processing and CNN.
• The novel TFR generation technique shows significant potential for enhancing the performance of MI-based BCI systems in neuro-rehabilitation applications.