Table 1.

Average classification accuracies (%) of the Dataset IVa of BCI Competition III and Dataset IIa of BCI Competition IV for CSP, LTCSP, and LTCCSP with increasing occurrence frequencies of outliers.

Freq		Dataset IVa of BCI Competition III					Dataset IIa of BCI Competition IV
	Subject	aa	al	av	aw	ay	1	2	3	4	5	6	7	8	9	Mean
0	CSP	75.0	100.0	69.9	91.5	74.2	84.0	54.5	95.8	75.3	58.0	68.1	80.6	92.4	93.7	79.5
	LTCSP	63.4	96.4	61.7	75.9	49.6	88.8	58.5	95.1	73.9	70.0	72.3	81.3	95.8	92.3	76.8
	LTCCSP	77.7	100.0	73.0	92.9	78.2	90.9	60.2	96.5	78.2	72.9	70.2	81.3	95.1	93.7	82.9*
0.1	CSP	70.5	100.0	68.2	86.1	76.5	86.9	56.2	93.5	65.9	56.1	61.7	66.0	92.2	92.7	76.6
	LTCSP	65.2	98.8	61.8	77.5	49.1	87.8	56.8	94.5	69.4*	69.5*	71.6*	80.4*	97.5*	92.4	76.6
	LTCCSP	74.0*	100.0	71.9*	91.6*	78.5*	91.7*	61.1*	95.5*	72.0*	64.7*	66.4*	71.9*	95.2*	93.7*	80.6*
0.2	CSP	69.6	98.9	68.0	89.2	76.3	86.3	56.0	93.4	64.8	51.7	61.8	62.6	93.1	92.5	76.0
	LTCSP	64.6	97.7	62.1	75.2	49.4	88.3*	59.1*	94.8*	68.9*	69.1*	71.3*	80.2*	97.3*	92.5	76.5
	LTCCSP	73.6*	99.8*	70.2*	92.8*	79.9*	90.6*	61.9*	95.7*	69.5*	59.1*	67.5*	69.4*	94.7*	93.6*	79.9*
0.3	CSP	67.6	99.8	69.3	86.6	77.7	87.1	55.9	93.0	64.0	51.9	62.2	60.1	94.2	92.2	75.8
	LTCSP	63.2	98.9	63.0	74.2	49.2	87.6	58.6	94.1*	67.5*	69.1*	70.1*	81.3*	97.4*	92.5	76.2
	LTCCSP	71.2*	100.0	73.3*	91.9*	81.4*	89.8*	60.0*	94.9*	68.0*	58.9*	67.2*	65.5*	95.6*	93.6*	79.4*
0.4	CSP	58.7	99.6	61.7	82.4	74.0	85.3	53.7	93.1	65.1	53.9	61.0	58.6	93.3	92.5	73.8
	LTCSP	64.6*	97.7	63.1*	72.0*	49.0	87.9*	58.1*	94.3*	67.8*	67.4*	69.5*	81.5*	97.0*	92.5	75.9
	LTCCSP	67.3*	99.8	67.8*	86.6*	77.7*	89.2*	59.3*	95.1*	67.7*	56.8*	65.8*	64.3*	95.0*	93.8*	77.6*

*Paired t-test P < 0.05 between two concerned methods, that is, LTCSP versus CSP and LTCCSP versus CSP. The bold values indicate the best performance among the three methods.