An Improved EMD-Based Dissimilarity Metric for Unsupervised Linear Subspace Learning

Table 4

The recognition rates on noise AR face database over “F1” and “F2” features; recognition rates (dimensionality).

Face database	Algorithms	G4/P9	G5/P8	G6/P7	G7/P6	G8/P5
Noise	Baseline (F1)	0.2194 (19800)	0.2394 (19800)	0.2600 (19800)	0.2719 (19800)	0.2817 (19800)
	LPP (F1)	0.4148 (75)	0.4679 (100)	0.5160 (115)	0.5133 (140)	0.5383 (145)
	Algorithm 1 (F1)	0.5463 (65)	0.6246 (90)	0.6781 (120)	0.7128 (80)	0.7267 (90)
Noise	Baseline (F2)	0.4687 (240)	0.5323 (240)	0.5645 (240)	0.5864 (240)	0.6143 (240)
	LPP (F2)	0.4344 (80)	0.4935 (95)	0.5162 (100)	0.5594 (110)	0.5937 (110)
	Algorithm 1 (F2)	0.4685 (85)	0.5231 (85)	0.5395 (95)	0.5753 (115)	0.6250 (125)