[Progress of classification algorithms for motor imagery electroencephalogram signals]

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Motor imagery (MI) involves the intention of movement without physical action, a key area in neuroscience.
• Classifying motor imagery electroencephalography (MI-EEG) signals is crucial for understanding brain activity and developing brain-computer interfaces.
• Extracting relevant features from MI-EEG signals is essential for accurate classification.

Purpose of the Study:

• To review and evaluate recent progress in classification algorithms for MI-EEG signals.
• To summarize and assess various classifiers and machine learning strategies used in MI-EEG analysis.
• To offer insights for developing higher-performance MI-EEG classification algorithms.

Main Methods:

• Review of traditional machine learning classifiers, deep learning, and Riemannian geometry classifiers.
• Analysis of machine learning strategies including ensemble learning, adaptive learning, and transfer learning.
• Evaluation of the effectiveness of different approaches in improving classification accuracy for MI-EEG signals.

Main Results:

• Traditional classifiers have seen improvements, while deep learning and Riemannian geometry are increasingly applied.
• Ensemble, adaptive, and transfer learning strategies have been successfully employed to boost classification performance.
• Significant advancements have been made in distinguishing different motor imagery tasks using improved algorithms.

Conclusions:

• The field of MI-EEG classification is rapidly evolving with novel classifiers and strategies.
• Further research into these advanced techniques can lead to more robust and accurate MI-EEG analysis.
• This review provides a foundation for future developments in high-performance MI-EEG classification algorithms.