Hybrid Deep-Learning Framework Based on Gaussian Fusion of Multiple Spatiotemporal Networks for Walking Gait Phase Recognition

Table 2

Summary of classification performance for different training functions.

Speed (m/s)	Training function	Classification result
		Accuracy (%)	Macro-F1 (%)	Macro-AUC
0.78	Bagging	90.5	67.7	0.97
	AdaBoosting	93.1	82.9	0.92
	CNN + LSTM	95.2	87.8	0.98
	CNN + GRU	94.7	87.3	0.97
	CNN + RNN	93.6	86.6	0.95
	GFM-Net	97.0	89.7	0.99
1.0	Bagging	91.4	68.7	0.95
	AdaBoosting	92.4	82.7	0.91
	CNN + LSTM	97.2	90.6	0.99
	CNN + GRU	96.6	89.3	0.98
	CNN + RNN	95.6	88.5	0.97
	GFM-Net	97.5	91.2	0.99
1.25	Bagging	89.2	66.3	0.95
	AdaBoosting	92.3	82.0	0.92
	CNN + LSTM	95.7	85.2	0.99
	CNN + GRU	95.3	84.6	0.99
	CNN + RNN	94.2	83.2	0.98
	GFM-Net	96.7	86.5	1.0