Table 3.

Three-class accuracy of evaluated CNN training strategies using the CWT-extracted vectors for the considered strategies: without spatial dropping (CWT), spatial dropping by removing all sensorimotor electrodes out (CWT*), spatial dropping by excluding nonrelevant electrodes (CWT**)

Subjects	[43]	[44]	CWT	$\kappa$	CWT*	$\kappa$	CWT**	$\kappa$
A03T	86.1	76.7	83.6 ± 6.6	0.75	84.6 ± 8.3	0.76	82.0 ± 5.6	0.73
A07T	83.6	82.8	73.5 ± 9.5	0.59	73.9 ± 7.5	0.60	75.0 ± 7.4	0.63
A08T	63.1	78.2	73.0 ± 9.0	0.59	76.0 ± 9.5	0.64	75.1 ± 10.4	0.63
A01T	80.8	84.1	73.0 ± 6.6	0.59	73.4 ± 6.1	0.60	75.7 ± 5.5	0.63
A09T	54.3	75.2	72.2 ± 5.6	0.58	69.4 ± 5.7	0.58	73.0 ± 7.6	0.60
A04T	54.0	62.4	69.7 ± 5.4	0.47	69.7 ± 6.9	0.47	69.9 ± 5.6	0.47
A06T	51.3	51.8	68.3 ± 6.7	0.49	68.8 ± 8.5	0.52	68.4 ± 6.8	0.52
A02T	71.2	57.7	65.0 ± 5.5	0.44	58.9 ± 6.5	0.46	62.4 ± 8.8	0.46
A05T	52.2	48.1	59.1 ± 4.2	0.37	58.5 ± 5.6	0.36	57.5 ± 6.0	0.34
Average	66.2	68.5	70.8 ± 6.5	0.54	70.3 ± 7.0	0.55	71.2 ± 7.0	0.56