Characterization of contour regularities based on the Levenshtein edit distance
Pattern Recognition Letters
An improved fast edit approach for two-string approximated mean computation applied to OCR
Pattern Recognition Letters
Hi-index | 0.00 |
The paper describes an attempt to use a machine learning approach to solve the problem of designing the set of prototypes to be used by an OCR system. The learning mechanism, based on a genetic algorithm, is exploited for providing the system with a set of reliable prototypes of the characters able to explain the variability encountered while dealing with specimen produced by different writers. In this framework, a new genetic algorithm with a variable population size is proposed, as well as a shape description scheme devised to improve the efficacy and the efficiency of the genetic search. Preliminary experiments show that the proposed approach is a promising step towards the automatic construction of the set of prototypes to be used for the recognition.