Table 3.

Accuracy and macro-averaged metrics of the proposed deep neural network classifiers for 41 hand movements.

	Accuracy	Macro-precision	Macro-recall	Macro-F1
DB5—8 channels : intact participants
100 ms	74.00 ± 2.10	42.79 ± 4.40	36.14 ± 4.80	38.68 ± 4.70
200 ms	77.97 ± 2.09	51.58 ± 4.73	46.13 ± 4.56	47.95 ± 4.69
400 ms	80.88 ± 1.99	56.44 ± 4.12	52.48 ± 4.55	53.85 ± 4.49
800 ms	87.04 ± 1.83	68.88 ± 4.08	67.56 ± 4.01	68.00 ± 4.06
1,000 ms	89.00 ± 2.05	73.32 ± 4.11	71.78 ± 4.67	72.35 ± 4.54
DB5—16 channels : intact participants
100 ms	81.37 ± 2.17	59.43 ± 5.13	55.10 ± 5.00	56.90 ± 5.03
200 ms	84.25 ± 2.02	65.21 ± 4.48	62.07 ± 4.56	63.38 ± 4.57
400 ms	87.21 ± 1.86	71.69 ± 3.68	68.54 ± 4.43	69.70 ± 4.31
800 ms	90.72 ± 1.62	77.45 ± 3.34	76.88 ± 3.84	76.88 ± 3.75
1,000 ms	93.87 ± 1.49	85.57 ± 2.46	84.00 ± 3.40	84.67 ± 3.20
DB7—12 channels : intact participants
100 ms	82.83 ± 4.90	73.96 ± 3.21	66.73 ± 4.89	69.78 ± 4.05
200 ms	85.08 ± 4.83	78.83 ± 2.88	70.67 ± 5.47	74.18 ± 4.38
400 ms	87.74 ± 4.94	81.81 ± 3.29	76.56 ± 5.31	78.90 ± 4.47
800 ms	90.61 ± 4.73	85.48 ± 3.55	82.30 ± 5.10	83.77 ± 4.50
1,000 ms	91.69 ± 4.68	87.03 ± 4.06	84.66 ± 4.78	85.74 ± 4.55
DB7—12 channels : amputee #1
100 ms	74.64	55.27	48.20	50.18
200 ms	78.23	63.91	53.79	56.05
400 ms	79.16	64.55	57.56	58.44
800 ms	81.54	69.97	62.02	63.08
1000 ms	82.42	72.84	65.10	65.66
DB7—12 channels : amputee #2
100 ms	87.49	63.62	59.23	60.26
200 ms	89.68	69.20	65.33	66.23
400 ms	89.01	64.25	62.39	61.66
800 ms	93.41	76.39	74.87	74.35
1,000 ms	94.07	75.78	76.55	74.90