Recognition of handwritten characters by parts with multiple orientations

Recently, a regional decomposition method and a hierarchical model without consideration of orientation have been proposed to investigate the recognition rates of patterns. Based on the probability of occurrence of the patterns, new models with multiple orientations at various positions have been proposed in this paper, to produce a more complete character recognizer. New algorithms are developed to evaluate the recognition rates due to changes in orientation. Also, mathematical analysis is made to discover some important properties of character recognition versus the sample shapes, rectangular and square, which may facilitate their recognition and analysis. Numerical experiments have also been conducted on 89 patterns of the most frequently used alphanumeric handprints. This study displays a deeper, inherent similarity, and distinctness among different patterns and characters, which include part symmetry and part resemblance in different possible positions. The results of this paper should be useful to pattern analysis and recognition, understanding character with missing parts, poor scanning, noise, and various kinds of distortion.