A Fusion-Based Technique With Hybrid Swarm Algorithm and Deep Learning for Biosignal Classification.

Area of Science:

• Neuroscience and Neural Engineering
• Biomedical Signal Processing
• Machine Learning Applications

Background:

• Electroencephalography (EEG) signals are crucial for understanding brain electrical activity.
• EEG analysis is vital in neuroscience, neural engineering, and commercial applications.
• Machine learning integration enhances EEG analysis for neural classification.

Purpose of the Study:

• To propose a Fusion Hybrid Model (FHM) for efficient EEG biosignal feature extraction.
• To develop a Hybrid Differential Particle Artificial Bee (HDPAB) algorithm for optimal feature selection.
• To achieve robust EEG signal classification for practical applications.

Main Methods:

• Feature extraction using a Fusion Hybrid Model (FHM) with Singular Value Decomposition (SVD).
• Feature selection employing a Hybrid Differential Particle Artificial Bee (HDPAB) algorithm.
• Classification using a Zero Inflated Poisson Mixture Regression Model (ZIPMRM) and deep learning, compared to standard methods.

Main Results:

• The proposed methodology achieved 98.79% classification accuracy for epileptic datasets.
• The methodology achieved 98.35% classification accuracy for schizophrenia datasets.
• The FHM and HDPAB algorithms demonstrated effective feature extraction and selection for EEG analysis.

Conclusions:

• The developed FHM and HDPAB algorithms provide efficient feature extraction and selection for EEG signals.
• Robust classification of EEG signals is achievable, reducing reliance on expert interpretation.
• The proposed approach shows significant potential for diagnosing neurological disorders like epilepsy and schizophrenia.