Table 4.

Performance comparison on CHASEDB.

Methods		Acc (%)		Spe (%)		Sen (%)		AUC (%)
		Base	Impro	Base	Impro	Base	Impro	Base	Impro
Proposed	5-way 3-shot	96.76	96.91	97.83	97.92	80.82	81.86	96.64	97.10
	4-way 3-shot	96.85	96.96	97.97	97.99	80.14	81.60	96.92	97.12
	3-way 3-shot	96.85	96.91	98.05	97.87	79.08	82.60	97.48	97.85
	3-way 4-shot	96.90	96.96	98.05	98.02	79.63	81.18	96.93	97.07
	3-way 5-shot	96.70	96.82	97.70	97.72	81.83	83.44	97.12	97.24
Machine learning	Fraz et al. (14)	94.69		97.11		72.24		97.12
	Orlando et al. (17)	-		97.12		72.77		95.24
	Li et al. (18)	95.81		97.93		75.07		97.16
	Mo and Zhang (23)	95.99		98.16		76.61		98.12
	Jiang et al. (24)	96.68		97.45		86.40		98.10
	Zhou et al. (25)	95.20		97.51		75.53		-
	Yan et al. (26)	96.10		98.09		76.33		97.81
	Filipe et al. (27)	96.53		98.64		77.79		98.55
	Park et al. (28)	97.36		-		-		98.59

The bold values indicate the maximum value of the corresponding metric in the manuscript.