TABLE 2.

Classification accuracies of MI and MA tasks based on single-domain features and multi-domain hybrid features based on EEG or fNIRS.

Studies	Tasks	Classification Algorithms	Modalities	Features	Accuracies
Shin et al., 2016	MI	Linear Discriminant Analysis (LDA)	EEG	Power Spectrum	73.10%
				Common Spatial Pattern	63.39%
				Hybrid Features	73.28%
			fNIRS	Mean Value of HbO Channel Wise	82.76%
				Mean Value of Hb Channel Wise	79.66%
				Modified Common Spatial Pattern	78.74%
				Hybrid Features	86.84%
	MA		EEG	Power Spectrum	82.64%
				Common Spatial Pattern	77.24%
				Hybrid Features	84.6%
			fNIRS	Mean Value of HbO Channel Wise	82.76%
				Mean Value of Hb Channel Wise	79.66%
				Modified Common Spatial Pattern	78.74%
				Hybrid Features	86.84%
Li et al., 2021	MI	Deep Forest	EEG	Time-frequency	62.00%
				Common Spatial Pattern	72.00%
				Fusion	75.00%
Saadati et al., 2020	MI	Deep Neural Networks (DNN)	fNIRS	HbO_Mean	70.00%
				Hb_Mean	-
				Hybrid Features	80.00%
Yin et al., 2015	MI	Extreme Learning Machines	EEG	Power	70.00%
				Instantaneous Amplitude	72.00%
				Instantaneous Phase	81.00%
				Instantaneous Frequency	79.00%
				Hybrid Features	88.00%
			fNIRS	HbO_Mean	70.00%
				Hb_Mean	72.00%
				Total HbO and Hb	81.00%
				Differences between HbO and Hb	79.00%
				Hybrid Features	88.00%
Zhang et al., 2021	MI	LDA	EEG	Power Spectral Density	74.00%
				Common Spatial Pattern	75.90%
				Wavelet Transform	83.20%
				Hybrid Features	84.70%
Our work	MI	Random Forest	EEG	Statistic	58.01 ± 4.33%
				DE	54.17 ± 3.79%
				Hybrid Features	65.87 ± 3.78%
			fNIRS	HbO_statistic	89.26 ± 2.87%
				HbO_PSD	85.74 ± 5.20%
				Hb_statistic	90.00 ± 1.99%
				Hb_PSD	81.51 ± 3.57%
				Hybrid Features	92.19 ± 2.95%
	MA		EEG	Statisitc	76.42 ± 6.73%
				DE	75.06 ± 6.68%
				Hybrid Features	80.75 ± 7.60%
			fNIRS	HbO_statistic	92.51 ± 2.27%
				HbO_PSD	85.74 ± 5.20%
				Hb_statistic	90.00 ± 1.99%
				Hb_PSD	92.85 ± 2.23%
				Hybrid Features	94.88 ± 2.35%