A multi-label deep residual shrinkage network for high-density surface electromyography decomposition in real-time

Table 5 Decomposition accuracy of ML-DCNN for different window sizes and step sizes

Window size (data points)	Step size (data points)	Precision	Sensitivity	F1-score	Correctly identified MU counts
20	10	0.71 ± 0.24	0.49 ± 0.32	0.53 ± 0.31	2.54 ± 3.96
	20	0.72 ± 0.31	0.59 ± 0.40	0.60 ± 0.39	4.75 ± 6.55
	30	0.67 ± 0.36	0.61 ± 0.42	0.62 ± 0.41	6.33 ± 8.57
	40	0.78 ± 0.32	0.74 ± 0.34	0.75 ± 0.34	7.71 ± 9.41
	50	0.87 ± 0.24	0.85 ± 0.28	0.85 ± 0.28	9.63 ± 9.70
60	10	0.91 ± 0.09	0.78 ± 0.22	0.81 ± 0.19	5.79 ± 5.66
	20	0.88 ± 0.14	0.74 ± 0.31	0.77 ± 0.28	5.42 ± 5.79
	30	0.86 ± 0.18	0.75 ± 0.31	0.77 ± 0.29	6.13 ± 6.86
	40	0.87 ± 0.21	0.80 ± 0.33	0.81 ± 0.31	7.60 ± 7.60
	50	0.89 ± 0.20	0.82 ± 0.31	0.83 ± 0.30	8.81 ± 8.64
100	10	0.92 ± 0.08	0.86 ± 0.20	0.87 ± 0.18	7.85 ± 6.05
	20	0.91 ± 0.11	0.89 ± 0.13	0.89 ± 0.13	7.77 ± 6.69
	30	0.93 ± 0.12	0.91 ± 0.17	0.91 ± 0.17	10.04 ± 7.49
	40	0.95 ± 0.10	0.95 ± 0.12	0.95 ± 0.12	11.77 ± 7.99
	50	0.98 ± 0.07	0.98 ± 0.06	0.97 ± 0.07	13.06 ± 7.91
140	10	0.95 ± 0.07	0.91 ± 0.19	0.91 ± 0.16	11.56 ± 7.51
	20	0.98 ± 0.05	0.99 ± 0.03	0.98 ± 0.04	14.15 ± 7.87
	30	0.98 ± 0.06	0.99 ± 0.04	0.98 ± 0.05	13.79 ± 8.14
	40	0.99 ± 0.02	0.99 ± 0.03	0.99 ± 0.03	14.17 ± 8.38
	50	1.00 ± 0.00	1.00 ± 0.00	1.00 ± 0.00	14.71 ± 8.23

Values are expressed as mean ± standard deviation across 16 participants and 3 contraction intensities

ISSN: 1743-0003