Description-Aided recognition of Handprinted characters

Abstract

A description-aided character recognition scheme is presented for the recognition of handprinted characters. This scheme uses a hexagonal-cellular regular hexagon (CH) for curve following and feature extraction. Characters are described in terms of certain “grammar-like” rules. A decision tree is employed for implementing this scheme. Open-ended inventories for features and character descriptions enhance the versatility of the recognition algorithm.

Chapter 1 gives a brief survey of the development in character-recognition techniques. The adoption of the description-aided approach, the use of a new technique for curve following, and a model of character handprinting are discussed. A general view of this recognition scheme is presented.

The employment of a CH for curve following is described in Chapter 2. The CH makes measurements on the curve at locations which are separated by distances equal to its radii. This curve-following technique obviates the need for line thinning and provides a simple method for measuring the curvature of a line. The measurements made by the CH scan provide sufficient information for feature extraction.

Chapter 3 shows how the description of a character is accomplished by using “large features,” i.e., straight lines and curves. The relationships existing at the junctions between large features are depicted by “small features.” An appropriate set of “grammar-like” rules is used for this description.

The extraction of a feature is discussed in Chapter 4. The local measurements made by the CH scan point out the presence of small features. A large feature is extracted by the cumulative measurements made by the scan on a curve. Open-ended inventories are maintained for the inclusion of additional features.

The culmination of the recognition scheme is presented in Chapter 5. A decision tree is used for the coupling of curve following, feature extraction, and character recognition. A character is recognized by the feature-by-feature construction of its description. An open-ended character description is used for accommodating more members in a character class or for adding new characters to the general class.

The last chapter is devoted to the computer implementation of the recognition scheme using handprinted samples. These results are analyzed, and suggestions are made for further studies.