Epilepsy Signal Recognition Using Online Transfer TSK Fuzzy Classifier Underlying Classification Error and Joint...

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Epilepsy diagnosis relies heavily on electroencephalogram (EEG) signal analysis.
• Existing transfer learning models struggle with subject-specific variations in neural responses.
• Accurate recognition of epilepsy from EEG is crucial for timely intervention.

Purpose of the Study:

• To propose an online transfer TSK fuzzy classifier (O-T-TSK-FC) for robust epilepsy recognition from EEG signals.
• To enhance transfer learning by integrating subject-specific data and a novel regularization method.
• To improve computational efficiency without compromising classification performance.

Main Methods:

• Development of an objective function to integrate subject-specific target domain data.
• Introduction of a new regularization technique based on error consensus and probability density estimation.
• Application of clustering to partition source domains for computational efficiency.

Main Results:

• The O-T-TSK-FC demonstrated superior performance compared to existing benchmarking algorithms.
• Experiments were conducted across six different online transfer learning scenarios using Bonn University EEG dataset.
• The proposed classifier exhibited robust performance in recognizing epilepsy signals.

Conclusions:

• The O-T-TSK-FC offers an effective and robust solution for online epilepsy recognition.
• The integration of subject-specific data and novel regularization significantly improves transfer learning for EEG analysis.
• The method provides a promising advancement in automated epilepsy detection systems.