Improving EEG-Based Motor Imagery Classification for Real-Time Applications Using the QSA Method

Table 1

Accuracy of classifier in motor imagery based BCI: multiclass. The classes are (T1) left imagined hand movements, (T2) right imagined hand movements, (T3) imagined foot movements, (T4) imagined tongue movements, (T5) relax (baseline), (T6) word generation starting with a letter specific, and (T7) mental calculation.

Dataset	Activity	Subjects	Trials	Filter	Feature	Classifier	Accuracy	References
BCI competition III	T1, T2, T3, T4	5	240 to 360	Yes	AAR	Linear SVM	63%	[16]
						KNN	41.74%
						LDA	54.46%
						Mahalanobis distance	53.50%
	T1, T2, T6	3		Yes		Neuro-fuzzy algorithm S-dFasArt	89.04%	[23]
	T1, T2, T3, T4	3	240	Yes	CSP	CSP + SVM	79%	[24]
						CSP + SVM + KNN + LDA	69%
						PCA + ICA + SVM	63%
						CBN + SVM	91%
	T1, T2, T6	3	12	Yes	SCSSP/LDA	FBCSP-NBPW		[14]
						FBCSP-Lin
						SCSSP-NBPW
						SCSSP-Lin
	T1, T2, T3	5	280	Yes	CSP/ERD-ERS	BSSFO, SVM	75.46%	[25]
BCI competition IV	T1, T2, T3, T4	9	288	Yes	CSP	CSP + SVM	31%	[24]
						LDA + SVM	30%
						CSP + SVM	29%
						CBN + SVM	66%
	T1, T2, T3, T4	9	288	Yes	SCSSP/LDA	LDA	85%	[14]
	T1, T2, T3, T4	9	288	Yes	CSP/ERD-ERS	BSSFO, SVM	80.26%	[25]
Other data sets	T1, T2, T3, T4	9	240	Yes	AAR, Kalman Filter	MDA	—	[13]
	T1, T2, T5	9	120	Yes		SMR, two-dimensional linear classifier	85%	[26]
	T1, T2, T3, T4	8	288	Yes	ICA, CSP	FDA	33%–84%	[27]
	T1, T2, T3	3	480	Yes	(SampEn)	SVM (RBF Kernel)	69.93%	[28]
	T1, T2, T5	33		Yes	AAR	PCA		[12]
Epoc	T1, T2, T3	3	100	Yes		SVM	66.16%	[29]
	T1, T2	5		Yes	ICA, CWT	Simple logistic, Meta, MLP	80.40%	[30]
	T1, T2, T4	1	75	Yes		Neural networks (PSO)	91%	[31]
	T1, T2, T3, T4, T7	3	40	Yes	BP	HMM	77.50%	[32]
	T1, T2, T3	3	30	Yes	BP	LDA	95%	[33]