A simple and efficient method for global handwritten word recognition applied to Brazilian bankchecks

A simple and efficient method for the recognition of isolated handwritten words in Brazilian bankchecks is proposed. For this end, the global word recognition technique was adopted. It relies mainly on topological features to identify the shape of the digitized word images. A genetic algorithm (GA) was used to select those features that best represent each word class and the nearest neighbor method was used as a classification means. The experimental results obtained in our laboratory present a higher average classification rate than a previous work that used the same database.