Sparse measures with swarm-based pliable hidden Markov model and deep learning for EEG classification.

Area of Science:

• Biomedical Signal Processing
• Machine Learning
• Computational Neuroscience

Background:

• Electroencephalography (EEG) signals are complex, requiring advanced models for accurate feature extraction and classification.
• Standard techniques often lose crucial structural information present in EEG data.
• A versatile and novel approach is needed for effective EEG signal modeling and classification.

Purpose of the Study:

• To propose a novel and versatile approach for EEG signal modeling and classification.
• To integrate sparse representation with Swarm Intelligence (SI) techniques and Hidden Markov Models (HMM).
• To develop and compare a deep learning methodology using Convolutional Neural Networks (CNN) with the proposed approach.

Main Methods:

• Initial analysis of sparse representation measures for EEG signals.
• Classification using a novel convergence of sparse representation measures with SI-based HMM (PSO, DE, WOA, BSA).
• Development of a deep learning methodology utilizing Convolutional Neural Networks (CNN).

Main Results:

• The deep learning methodology combined with sparse representation achieved the highest classification accuracy of 98.94%.
• The SI-based HMM method combined with sparse representation achieved a high classification accuracy of 95.70%.
• The proposed methods demonstrated superior performance compared to standard pattern recognition classifiers.

Conclusions:

• Sparse representation is a highly effective technique for enhancing EEG signal classification.
• The integration of sparse representation with deep learning (CNN) offers state-of-the-art performance in EEG analysis.
• The proposed SI-based HMM approach provides a robust alternative for complex biomedical signal classification.