Subspace-Based Face Recognition in Analog VLSI (W77)
Gonzalo Carvajal, Waldo Valenzuela, and Miguel Figueroa
Dimensionality reduction using PCA and LDA Manhattan distances 12x12 images, 39 projections ORL database
1

distances 1 Test data Input Image y x dimensionality n 2

reduction

m

Da tab a se

LTA
k

Face ID

...

1 Classification Performance
Classi fication Performance

0.8

0.8

0.6

0.6

0.4 PCA+dist (SW) PCA with LMS+ dist (HW) GHA+dist (HW) 5 10 15 20 25 30 Number of Principal Componentes 35 40

0.4

0.2

0.2

LDA+dist+LTA (SW) LDA+dist+LTA (HW) LDA+dist (HW)+RBF (SW) 5 10 15 20 25 30 Number of LDA Projections 35 40

0

0

83% using PCA (software: 84%)

85% using LDA (software: 87%)