Human Gait Analysis: A Sequential Framework of Lightweight Deep Learning and Improved Moth-Flame Optimization Algorithm

Table 1

Proposed classification results of human gait recognition on the CASIA B dataset.

Method	Class	0°	18°	36°	54°	72°	90°	108°	126°	144°	162°	180°	Mean
LightweightDeep-ELM	NM	97.1	98.2	95	93.8	98.1	97.5	98.3	97.2	98	94	98.6	96.89
	BG	94.2	95.3	91	92.7	93	89.7	94.8	93.1	92.8	91.8	95.4	93.07
	CL	78.8	83.5	82.1	86.3	78.5	90.2	85.9	82	81.3	88.3	83.4	83.66
LightweightDeep-SVM	NM	96.2	97.5	96.1	92.8	97.9	96.2	98	97.5	97.1	93.2	98	96.40
	BG	92.5	93.6	93	92.2	94.1	87.6	94	93.5	92.1	92.4	93.4	92.58
	CL	79	82.8	81.5	86	79.1	87.4	83.9	82.3	80.7	87.5	82.1	82.93
LightweightDeep-FKNN	NM	93.2	94.5	92.9	93.1	94.6	91.8	93.5	94.8	94.5	93	94.1	93.63
	BG	87	89.1	90.5	89.6	91.4	82.6	90.8	89.4	87	89.3	90.7	88.85
	CL	73.5	78.4	77.2	81.6	75.3	82.2	80	77.5	76.1	82.4	78.5	78.42
LightweightDeep-EBT	NM	92.8	93.3	90.4	91.5	95	92.6	92.4	93.1	94	93.7	92.9	92.88
	BG	88.5	87.4	90.1	90.5	91.8	82	90.2	90.4	86.1	89.8	91.9	88.97
	CL	72.6	80.1	77	80.4	75	81.6	80.6	76.2	78.5	82	78.1	78.37
LightweightDeep-DT	NM	87.4	88.9	90.1	91.6	93	90.5	88	91.2	91.3	87.4	90.5	89.99
	BG	81.5	82.6	87.5	83.8	90.4	80.6	90	87.9	85.3	84	90.1	85.79
	CL	69.8	72.1	77	78.5	72.7	80	78.3	72.5	72.9	80.1	80.2	75.82

Bold values indicate the best values.