APPLICATION OF SVM AND SOFT FEATURES TO AZERBAIJANI TEXT RECOGNITION

Abstract

The purpose of this study is to establish more accurate and less timeconsuming recognition system for Azerbaijani text recognition. The main problem of investigating and developing recognition systems is the extraction of features, in view of the fact that, most of current recognition systems use features, which are unintelligible for human mind and proposed for operating by computers. For eliminating abovementioned problem, in this paper was offered “soft” features, extracted on base of human-mind techniques. On the side of validating SVM approach and “soft” features provided in this paper, experiments were executed using various feature classes offered for Azerbaijani handprinted characters and different methods.