Lw-CNN-Based Myoelectric Signal Recognition and Real-Time Control of Robotic Arm for Upper-Limb Rehabilitation

Table 2

Offline testing accuracies with use of Lw-CNN model under different analysis window sizes and slippage rates.

Offline CNN-accuracy (%)
	W50–O5^2	W50–O10	W50–O15	W150–O5	W150–O10	W150–O15	W192–O5	W192–O10	W192–O15
S1	42	37	53	76	79	78	79	83	85
S2	32	41	40	79	82	84	81	84	88
S3	39	46	57	80	78	86	83	86	92
S4	22	37	30	83	85	89	85	87	91
S5	41	48	63	73	81	82	82	86	87
S6	43	40	45	80	90	93	89	90	94
S7	32	51	50	84	82	90	89	93	95
S8	29	38	48	80	82	87	84	86	86
Mean	35	42.25	48.25	79.38	82.38	86.13	84	86.88	89.75
SE^1	7.45	5.39	10.22	3.54	3.74	4.76	3.59	3.23	3.77

¹Standard error. ²w50–O5 denotes the analysis window of 50 ms, slippage 5 ms.