Jumlah Transisi pada Ciri Transisi dalam Pengenalan Pola Tulisan Tangan Aksara Jawa Nglegeno dengan Multiclass Support Vector Machines

Feature extraction is one of the most improtant step on characters recognition system. Transition features is one from many features used on characters recognition system. This paper report a research on handwritten basic Jawanesse characters recognition system to found the proper numbers of transitions used on transition features. To recognize the characters,the Multiclass Support Vector Machines were used. The Directed Acyclic Graph (DAG) SVM were used for multiclass classification strategy and to map each input vector to a higher dimention space, the Gaussian Radial Basis Function (RBF) kernel with parameter 1were used. It can be shown, for basicJawanesse characters recognition system, the optimal numbers of transitions used for transition features is 4 (a half of maximum numbers of transition on all patterns).