The Ensemble Machine Learning-Based Classification of Motor Imagery Tasks in Brain-Computer Interface

Table 2

Comparison with previous studies.

Study	Feature extraction	Classifier	Classes/subject(s)	Accuracy (%)
[18]	Rényi min-entropy	RF	4/subject independent	80.55
[21]	Subbands PSDs	DNN	2/subject independent	82.48
[37]	Tangent space mapping	SVM	2/1-subject	97.80
[38]	Common spatial pattern	Backpropagation; Neural network	2/subject independent	80.73
[39]	Regularized common spatial pattern	SVM	2/subject independent	91.9
[40]	Fisher ratio of time domain parameters	SVM	2/subject independent	89.13
[41]	Common spatial pattern	SVM	2/subject independent	85.01
[42]	Stacked autoencoders (SAE)	CNN	2/subject independent	82.00
[43]	Inverse problem through beamforming	CNN	2/subject independent	90.50
[44]	Granger causality channel selection and common spatial pattern	Linear SVM	2/subject independent	88.46
Proposed	WPD	RF and RSM	2/subject dependent	98.69
	WPD	k-NN and RoF	2/subject independent	94.83