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

Table 1

Summary of classification performance of different models at unsynchronized speed.

Model	Speed	0.78 m/s				1.0 m/s				1.25 m/s
	Phase	HS	FF	HO	SW	HS	FF	HO	SW	HS	FF	HO	SW
Bagging	Precision (%)	0	84.7	85.9	98.1	0	86.5	87.5	97.2	0	84.3	80.4	97.3
	Recall (%)	0	96.2	85.1	94.6	0	97.5	88.4	93.4	0	97.6	81.0	91.3
	F1 (%)	0	90.1	85.5	96.3	0	91.7	88.0	95.3	0	90.5	80.7	94.2
AdaBoosting	Precision (%)	48.8	91.8	92.2	96.9	42.1	90.8	94.2	97.2	40.7	90.4	96.0	96.3
	Recall (%)	52.6	90.4	93.4	97.0	53.7	91.7	91.9	94.7	54.5	92.1	93.8	94.0
	F1 (%)	50.7	91.1	92.8	97.0	50.7	91.3	93.0	95.9	46.6	91.3	94.9	95.1
GFM-Net	Precision (%)	79.1	94.5	98.2	99.2	80.6	95.8	99.0	98.9	80.9	95.0	97.9	97.9
	Recall (%)	56.4	98.3	97.0	98.4	63.1	98.4	97.1	98.9	40.1	98.6	96.8	98.4
	F1 (%)	65.9	96.4	97.6	98.8	70.8	97.0	98.1	98.9	53.6	96.8	97.3	98.1