Abstract: A method is provided that includes capturing an input handwritten character with parameter representation for each stroke and applying a polygonal approximation thereto; assuming each polygonal line segment approximated to be vector that reaches an end point from a start point, and obtaining an angle between an axis that becomes a reference and each line segment as a polygonal line segment angle sequence; obtaining an exterior angle sequence of vertices of the line segments; making a sum of exterior angles of the same sign, where the same sign of plus or minus in the exterior angle sequence continues, to be a winding angle sequence; extracting a global feature according to each obtained sequence and a localized or quasi-localized feature in each curved portion divided corresponding to the winding angle sequence, hierarchically and divisionally; and performing character recognition by comparing the extracted result with a template of an object character.

Abstract: In a case in which handwritten character recognition such as online handwritten character recognition is implemented, simple and high performance character recognition by which transformation such as remarkable transformation and rotation is able to be treated and which recognizes an input that is written in a single stroke such as kaisho-writing, is provided.