Evolutionary Learning of Classifiers for Disk Discrimination

Abstract

A novel disk discrimination approach with two levels of classification is proposed. The optical pickup head is controlled to emit the digital versatile disk laser followed by the compact disk laser during the process of being pulled in from the neutral position toward the optical disk. Four features are measured and analyzed based on the reflected signals from the optical disk. Two levels of classifications are designed for the disk discrimination using these four features. A supervised learning approach based on the evolutionary ellipsoid classification algorithm is utilized to learn the classifiers and optimize the classifier parameterizations. Six different disk types have been successfully discriminated using the proposed approach.