High Order Volumetric Directional Pattern for Video-Based Face Recognition

Table 4

Recognition rates using only 100 frames of each video as changing the threshold and the proposed approach order on Honda/UCSD database.

Descriptor	Recognition Accuracy (%) Using libsvm Classifier
Order
Order (VDP)	87.86 %	87.09 %	87.99 %	87.99 %	84.96 %
Order	89.71	89.71	90.21 %	88.89 %	86.25 %
Order	87.89 %	87.83 %	90.86 %	88.99 %	89.34 %
Order	88.34 %	89.57 %	91.68	89.36	90.20
Descriptor	Recognition Accuracy (%) Using knn Classifier
Order
Order (VDP)	90.47 %	92.13 %	92.16 %	91.29 %	90.47 %
Order	91.79	92.56	93.08	92.58	91.66
Order	88.89 %	91.10 %	91.58 %	91.63 %	91.13 %
Order	88.99 %	91.50 %	91.84 %	91.74 %	91.61 %