REDES NEURAIS CLASSE MODULAR APLICADAS NO RECONHECIMENTO DE CARACTERES MANUSCRITOS CLASS MODULAR NEURAL NETWORKS APPLIED IN RECOGNITION OF CHARACTER MANUSCRIPT
Clariane Silva Menezes, Leandro Luiz de Almeida, Mário Augusto Pazoti, Almir Olivette Artero · 2014
The handwritten character recognition is still a major challenge in the field of computer vision, primarily due to the diversity of styles that people can write, which makes it difficult to generalize the problem. In addition, there is also the difficulty in defining the descriptors that best characterize the character and build high performance OCR systems. This paper presents a system for recognizing handwritten characters offline, using Artificial Neural Networks Modular Class with classic backpropagation training algorithm, besides the methods used for feature extraction. Although training of neural classifiers require long processing and recognition of 62 classes of characters, few studies have considered the results obtained from the experiments are shown very promising, achieving hit rates above 90%.