Resumen de Words recognition using methods of word shape coding
Maite García-Ordás, Laura Fernández Robles, Óscar García-Olalla Olivera, Diego García-Ordás, Enrique Alegre Gutiérrez
In this work, a holistic word recognition and a word shape coding methods have been evaluated and improved. In the first method, the feature vector was composed of six scalar features and three features vectors extracted from normalized profiles.
Four statistical moments for each profile vector were added in order to improve the results. In the second method, the features vector was made up by the superior and inferior extrema points with a special codification and the number of cuts of the word with the central line. We improved it by adding cuts with two lines instead of the central one. Results were improved from 73.33% to 80% of success rate for the first method and up to 93.33% for the second one not only obtaining lower error rates but also lower draws.
