Ensemble classifier based on optimized extreme learning machine for motor imagery classification.

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Effective classification is crucial for brain-computer interface (BCI) systems.
• Motor imagery electroencephalogram (EEG) classification accuracy and generalization remain challenges.

Purpose of the Study:

• To propose an extreme learning machine (ELM) based method to enhance motor imagery EEG classification accuracy.
• To develop an ensemble classifier using optimized ELMs for improved BCI performance.

Main Methods:

• An ensemble classifier was constructed using optimized ELMs.
• Particle swarm optimization (PSO) tuned ELM parameters (input weights, hidden biases).
• Majority voting fused results from multiple base classifiers to mitigate local optima.

Main Results:

• The proposed method achieved significantly higher classification accuracies than competing methods on two-class and four-class motor imagery EEG data.
• Superior average accuracies were obtained compared to existing methods.
• Improved performance was observed for subjects with initially lower classification accuracies.

Conclusions:

• The optimized ELM ensemble classifier demonstrates superior performance for motor imagery EEG classification.
• This method shows promise as a robust solution for accurate EEG signal processing in BCI applications.