Towards decoding motor imagery from EEG signal using optimized back propagation neural network with honey badger...

Area of Science:

• Neuroscience and Biomedical Engineering
• Signal Processing and Machine Learning

Background:

• Brain-Computer Interface (BCI) systems utilizing electroencephalography (EEG) are crucial for decoding motor imagery (MI) and have applications in medicine and assistive technologies.
• Existing EEG-based BCIs face challenges including signal noise, low decoding accuracy, and signal instability.

Purpose of the Study:

• To present a novel approach for classifying motor imagery (MI) from EEG signals, addressing limitations of current BCI systems.
• To improve the accuracy and stability of EEG signal decoding for BCI applications.

Main Methods:

• A synergistic approach combining Hilbert-Huang Transform (HHT) for pre-processing, Permutation Conditional Mutual Information Common Space Pattern (PCMICSP) for feature extraction, and an optimized back propagation neural network (BPNN) using the Honey Badger Algorithm (HBA) for classification.
• The Honey Badger Algorithm (HBA) was employed to optimize the weights and thresholds of the BPNN, incorporating chaotic mechanisms for refined optimization.

Main Results:

• Experimental analysis on the EEGMMIDB benchmark dataset demonstrated the efficiency of the proposed method.
• The technique achieved a maximum classification accuracy of 89.82% on EEG signals, outperforming other methods.
• The study considered both epileptic and non-epileptic EEG signal levels, highlighting the robustness of the developed mechanism.

Conclusions:

• The proposed HBA-optimized BPNN with HHT and PCMICSP offers a significant advancement in EEG-based BCI for motor imagery classification.
• The method demonstrates improved accuracy and stability, addressing key challenges in current BCI technology.
• This approach holds promise for enhancing the effectiveness of assistive technologies and rehabilitation tools.