Table 4.

Accuracy and macro-averaged metrics of the proposed deep neural network classifiers for SHAP prehensile patterns.

	Accuracy	Macro-precision	Macro-recall	Macro-F1
DB5—8 channels : intact participants
100 ms	92.49 ± 1.26	69.16 ± 4.13	61.21 ± 6.18	64.71 ± 5.00
200 ms	93.71 ± 1.11	72.81 ± 4.26	68.59 ± 5.72	70.52 ± 4.79
400 ms	95.53 ± 1.24	80.56 ± 5.58	77.42 ± 6.24	78.76 ± 5.91
800 ms	97.29 ± 0.86	87.17 ± 2.90	86.27 ± 4.69	86.61 ± 3.94
1,000 ms	97.78 ± 0.75	89.14 ± 3.58	88.61 ± 3.95	88.93 ± 3.48
DB5—16 channels : intact participants
100 ms	94.66 ± 1.05	79.40 ± 3.23	74.14 ± 5.50	76.63 ± 4.22
200 ms	95.86 ± 0.89	82.74 ± 2.78	80.96 ± 4.37	81.79 ± 3.55
400 ms	97.28 ± 0.81	89.49 ± 2.34	87.10 ± 4.01	88.19 ± 3.05
800 ms	98.38 ± 0.84	93.17 ± 3.59	92.10 ± 3.95	92.60 ± 3.80
1,000 ms	98.82 ± 0.58	94.93 ± 1.91	94.48 ± 2.55	94.67 ± 2.18
DB7—12 channels : intact participants
100 ms	92.69 ± 4.19	86.42 ± 3.96	74.08 ± 9.32	79.40 ± 7.32
200 ms	93.34 ± 4.41	88.01 ± 4.02	76.81 ± 10.05	81.66 ± 7.86
400 ms	94.22 ± 4.03	88.90 ± 4.00	79.92 ± 8.91	83.77 ± 7.28
800 ms	95.46 ± 4.16	90.30 ± 4.38	85.18 ± 8.90	87.59 ± 7.03
1,000 ms	95.78 ± 3.99	90.94 ± 3.90	86.00 ± 8.35	88.31 ± 6.79
DB7—12 channels : amputee #1
100 ms	88.29	68.02	56.73	59.82
200 ms	88.88	69.94	56.66	58.09
400 ms	90.20	76.42	63.87	65.71
800 ms	91.14	77.48	67.39	69.22
1,000 ms	92.53	83.22	71.71	74.97
DB7—12 channels : amputee #2
100 ms	97.13	87.43	80.08	81.83
200 ms	97.55	87.11	83.99	83.38
400 ms	98.04	91.95	84.44	85.47
800 ms	98.66	94.93	88.92	89.92
1,000 ms	99.00	96.24	91.27	93.11