Hiroyasu Takahashi

A spelling correction method and its application to an OCR system

Abstract This paper describes a method of spelling correction consisting of two steps: selection of candidate words, and approximate string matching between the input word and each candidate word. Each word is classified and multi-indexed according to combinations of a constant number of characters in the word. Candidate words are selected fast and accurately, regardless of error types, as long as the number of errors is below a threshold. We applied this method to the post-processing of a printed alphanumeric OCR on a personal computer, thus making our OCR more reliable and user-friendly.

DRS: a workstation-based document recognition system for text entry

Document recognition system (DRS), a workstation-based prototype document analysis system that uses optical character recognition (OCR), is described. The system provides functions for image capture, block segmentation, page structure analysis, and character recognition with contextual postprocessing, as well as a user interface for error correction. All the functions except image capture and character recognition have been implemented by means of software for the Japanese edition of OS/2.<<ETX>>

Handwritten numeral verification method using distribution maps of structural features

Character recognition methods can be categorized into two major approaches. One is pattern matching, which is little affected by topological changes such as breaks in strokes. The other is structural analysis, which tolerates distorted characters only if the topological features of their undistorted versions are kept. We developed a new recognition method for hand-written numerals by combining the merits of the two approaches. The recognition process consists of three steps: (1) an input character is recognized by a patternmatching method, which reduces the number of possible categories to 1.5 on the average, (2) the character is yenfled to be true, false, or uncertain by a structural analysis method that we have newly developed, and (3) special heuristic verification logics are applied to uncertain characters. In the second step, the new structural analysis method uses the positions and directions of terminal points extracted from thinned character images as a main feature. The extracted terminal points are labeled according to a structural-feature distribution map prepared for each category. The generated labels are matched with template label sets constructed by statistical analysis. The characteristics of the method are as follows: (1) it copes with distortion of hand-written characters by using distribution maps for the positions and directions of feature points, and (2) distribution maps can be automatically generated from statistical data in learning samples and easily tuned interactively. The merits of combining the two methods are as follows: (1) the advantages of both pattern matching and structural analysis are obtained, (2) the probabilities of steps 2 and 3 needing to be executed are 22% and 9% respectively, which hardly affect the total processing time, and (3) as a result of steps 1 and 2, only a small number of special logics are required. In a test using unconstrained hand-written characters of low quality, the recognition rate and substitution rate were 95.2% and 0.42% respectively. A recognition speed of 80 characters/second was achieved on a small hardware system.

Character string detection algorithm using horizontal boundaries, and its application to a part number entry system

This paper proposes a new algorithm for detecting character strings in an image containing illustrations and characters. It also describes a part number entry system that utilizes this algorithm. The algorithm detects character strings by investigating the horizontal boundaries of rectangles representing characters strings. It can be performed a high speed, and can detect characters touching an illustration. Using this algorithm, the part number entry system extracts areas of part numbers scattered among illustrations and then recognizes the. This is a software program implemented on a personal computer, and is composed of four subprograms: detection of character strings, character recognition, post-processing, and flexible user-interface for error correction.

Neural network architectures for rotated character recognition

Explores a neural network (NN) approach that is analogous to the human straightforward pattern matching, where some rotation is taking place in high level neurons close to symbols. The main objective is to develop ideas to simulate the rotation and verify them by using a large number of handwritten characters. The author proposes a feedforward NN where the links between input and hidden units are locally connected and weights are symmetrically shared. In the recognition process the total input values to hidden units are rotated according to the number of possible orientations and the activation values of output units are calculated for each orientation to find the best output.<<ETX>>