Reconhecimento de palavras manuscritas usando análise multivistas.

Abstract

This work presente a multiple classifier system applied to the handwritten word recognition (HWR) probiem. The goal is to investigate the use of perceptual models in the development of recognition systems. The handwritten words are analyzed considering different approximation leveis, in order to get a computational approach of the reading human process. The application proposed is the recognition of the Portuguese handwritten names of the months. The considered system is formed by the following architectures: 2 fixed sub-regions, 8 fixed sub-regions and N variable sub-regions. Each architecture is formed by a module of features extraction, based on perceptual models and specific for each type of segmentation, and an appropriate classifier. The experimental teste have performed on a database specifically built for this probiem, also described in this work.