Table 6. Classification performance for the second level with the TestSet dataset.

	Confusion Matrix					Accuracy (%)
K3b		Left-estimated	Right-estimated	Tongue-estimated	Foot-estimated	89.2±2.0
	Left-real	11	3	0	0
	Right-real	2	21	0	0
	Tongue-real	0	0	12	1
	Foot-real	0	0	1	14
L1b		Left-estimated	Right-estimated	Tongue-estimated	Foot-estimated	78.3±1.4
	Left-real	9	1	0	0
	Right-real	2	12	0	0
	Tongue-real	1	1	5	3
	Foot-real	0	0	2	10
K6b		Left-estimated	Right-estimated	Tongue-estimated	Foot-estimated	57.5±8.5
	Left-real	6	5	0	0
	Right-real	4	7	0	0
	Tongue-real	2	3	2	0
	Foot-real	1	1	1	8

The values are the average results for the test over 5-runs. The standard deviation of the accuracy is also given.