ONE DIMENSIONAL WITH DYNAMIC FEATURES VECTOR FOR IRIS CLASSIFICATION USING TRADITIONAL SUPPORT VECTOR MACHINES
Mohd Zaid Abdullah · 2014
This paper proposes an iris classification for small and imbalance dataset employing traditional Support Vector Machine classifier. A technique from the combination of modified conventional moving average and histogram equalization are proposed to produce the figurative and interpretability iris texture. In doing so, the smooth effect of the iris texture will take place for feature extraction. This study also proposed onedimensional features with dynamic vector which are extracted by manipulating the global mean and intensity variation on the un-normalized iris image. The uniqueness of the proposed feature extraction technique is where eyelid, eyelashes and lighting effect will be merged together in the calculation in order to produce the feature vector. Therefore, this study has not considered the noise removal method, and this differs compared to most previous works, which have implemented noise removal method for eliminating the eyelid and eyelashes information. The images from CASIA Version 1 and Version 4-Interval are used to assess the proposed method. The results obtained on these data sets reveal the effectiveness of the suggested method.