. 2016 Jul 21;16(7):1134. doi: 10.3390/s16071134

Table 5.

Recognition accuracies of previous method [16] using the combination of visible light and thermal images with feature-level fusion and score-level fusion approaches (unit: %).

Feature Extraction Method	First SVM Layer Kernel	Accuracy of Recognition System Using Combined Images
		Feature-Level Fusion			Score-Level Fusion
		EER	FAR	GAR	Second SVM Layer Kernel	Accuracy
		EER	FAR	GAR	Second SVM Layer Kernel	EER	FAR	GAR
HOG	Linear	17.892	10.00	70.80	Linear	19.955	10.00	63.40
			10.00	70.80			15.00	71.08
			15.00	78.70			19.960	80.050
			15.00	78.70			20.00	80.08
			17.90	82.116			25.00	84.96
			17.90	82.116	RBF	20.059	15.00	70.385
			20.00	84.32			20.00.	79.841
			20.00	84.32			20.060	79.941
			25.00	86.947			25.00	85.259
			25.00	86.947			30.00	88.901
	RBF	16.632	10.00	71.90	Linear	16.333	10.00	71.46
			10.00	71.90			15.00	81.80
			15.00	80.59			16.340	83.675
			15.00	80.59			20.00	87.64
			16.660	83.396			25.00	92.32
			16.660	83.396	RBF	16.277	10.00	71.368
			20.00	86.52			15.00	81.99
			20.00	86.52			16.280	83.726
			25.00	90.00			20.00	87.408
			25.00	90.00			25.00	92.21