Through-Wall Radar Classification of Human Posture Using Convolutional Neural Networks

Table 4

Confusion matrices of experimental results for the drywall scenario.

(a)

				Actual
				Empty	Standing				Sitting
					Micro		Macro		Micro		Macro
					@ 2.5 m	@ 5 m	@ 2.5 m	@ 5 m	@ 2.5 m	@ 5 m	@ 2.5 m	@ 5 m
Predicted	Empty			200	1	3	2	7	2	5	3	6
	Standing	Micro	@2.5 m		198				2
			@5 m			196				1
		Macro	@2.5 m				194				3
			@5 m					192				2
	Sitting	Micro	@2.5 m		1				196
			@5 m			1				194
		Macro	@2.5 m				4				194
			@5 m					1				192	Total
			Total	200	200	200	200	200	200	200	200	200	1800
			Accuracy	100%	99%	98%	97%	96%	98%	97%	97%	96%	97.56%

(b)

		Actual
		Empty	Standing	Sitting
Predicted	Empty	200	13	16
	Standing	0	780	8
	Sitting	0	7	776	Total
	Total	200	800	800	1800
	Accuracy	100.00%	97.50%	97.00%	97.56%