Incremental Face Recognition based on Binary Gabor Features and Recursive SVM Classifiers

다국어 초록 (Multilingual Abstract)

Unlike batch-based face recognition which suffers from appearance variations over
time, adaptive nature of the incremental learning is advantageous for real world applications
in which one confronts with a sequence of data. Unfortunately, a number of
issues such as real-time efficiency and robustness to local variation for high dimensional
face images still remain to be resolved for various applications. In this paper,
we focus on a realtime training algorithm for face recognition related applications.
Particularly, we propose a data independent binary Gabor feature extraction and derive
a recursive formulation of support vector machine. The dimension of our Gabor representation
is reduced considering various orientations at different grid positions and the
reduced Gabor features are trained by a recursive support vector machine. The overall
accuracy is consequently improved via an ensemble of those combination. Our experimental
results show that the proposed algorithm significantly outperforms several
popular face recognition methods with a dramatic reduction in computational speed.
We demonstrate the effectiveness of the proposed algorithm and the feasibility of its
application in a large-scale web services and mobile devices with limited resources.
We believe that the proposed framework can be easily extended to other computer
vision and pattern recognition techniques.
KeyWords