Microprocessor-based recognition of handprinted characters from a tablet input

摘要

This paper describes a highly efficient system for the recognition of handprinted English characters using a tablet-based input (mouse, graphics tablet, light pen, etc.). A major system constraint was the limiting of computational/storage resources to an 8-bit microprocessor using integer arithmetic only and no more than 32 kbyte of combined RAM/ROM. A hierarchical control structure exploits character/stroke features of varying degrees of complexity to achieve this efficiency. A binary decision tree uses simple features to rapidly reduce the set of candidate characters to a very small set. More complex features are used on a need-only basis to verify the correctness of a potential match. A top level character parser allows input of either a single character or a sequence of distinct characters.