Hybrid Deep Learning Approaches for sEMG Signal-Based Lower Limb Activity Recognition

Table 5

Comparison of the proposed method’s performance with studies using same dataset (in %).

Approach	Subject	Accuracy
		Walking	Sitting	Standing
EMD [41]	Healthy	64	67	69
MEMD [41]	Healthy	73	79	82
NA-EMD [41]	Healthy	79	83	83
MP-ANN [42]	Knee deformity	88	94	92
LRCN [43]	Healthy	98.2	97.7	98.4
	Knee deformity	92.8	92.3	92.2
ICA-EBM [36]	Healthy	96.0	96.2	96.2
	Knee deformity	86.6	86.4	85.5
WD-EEMD [34]	Healthy	85.11	88.70	93.50
	Knee deformity	98.86	96.38	96.77
V-1D-CNN [37]	Healthy	96.86	99.57	99.98
	Knee deformity	99.06	96.04	97.24
CNN-GRU (proposed)	Healthy	99.09	100	100
	Knee deformity	99.60	98.30	97.95