Iris recognition is a prosperous biometric method, but some technical difficulties still exist especially when applied in embedded systems. Support Vector Machine (SVM) has drawn great interests recently as one of the best classifiers in machine learning. In this paper, we develop an iris recognition system using SVM to classify the acquired features series. Even though the SVM outperforms most of other classifiers, it works slowly, which may hinder its application in embedded systems, where resources are usually limited. To make the SVM more applicable in embedded systems, we make several optimizations, including Active Learning, Kernel Selection and Negative Samples Reuse Strategy. Experimental data show that the method presented is amenable: the speed is 5 times faster and the correct recognition rate is almost the same as the basic SVM. This work makes iris recognition more feasible in embedded systems. Also, the optimized SVM can be widely applied in other similar fields.
[1]
B. Mandelbrot.
Fractal Geometry of Nature
,
1984
.
[2]
John Daugman,et al.
High Confidence Visual Recognition of Persons by a Test of Statistical Independence
,
1993,
IEEE Trans. Pattern Anal. Mach. Intell..
[3]
Daphne Koller,et al.
Support Vector Machine Active Learning with Applications to Text Classification
,
2000,
J. Mach. Learn. Res..
[4]
Vladimir Vapnik,et al.
Statistical learning theory
,
1998
.
[5]
T. Tan,et al.
Iris Recognition Based on Multichannel Gabor Filtering
,
2002
.
[6]
Bernhard Schölkopf,et al.
Training Invariant Support Vector Machines
,
2002,
Machine Learning.
[7]
Samy Bengio,et al.
SVMTorch: Support Vector Machines for Large-Scale Regression Problems
,
2001,
J. Mach. Learn. Res..