Data acquisition control method and system for a hand held reader

• US 4,897,880 A
• Filed: 03/06/1989
• Issued: 01/30/1990
• Est. Priority Date: 02/17/1987
• Status: Expired due to Fees

• Current Assignee: Soricon Corp
• Original Assignee: Soricon Corp
• Inventors: Joffe, David, Wilber, Scott A., Van Duzee, Brian, Goldstein, Andrew, Wieland, Chris M., Haas, Mark
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Couso, Jose L.
• Application Number:
  US07/319,930
• Time in Patent Office: 330 Days
• Field of Search: 382/9, 382/10, 382/13, 382/16, 382/30, 382/34, 382/48, 382/59, 235/454, 235/461, 235/485
• US Class Current: 382/177
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/166 Normalisation of pattern dimensions
• CPC Class Codes: G06V 30/248 involving plural approaches, e.g. verification by template match; Resolving confusion among similar patterns, e.g. "O" versus "Q" G06V30/242 takes precedence

Abstract

A data acquisition control method and system for alphanumeric character recognition. An unknown pattern is normalized, the normalized pattern is then divided into top and bottom parts. A center of mass is calculated for the two parts and is utilized for vertically aligning the unknown pattern. The vertically aligned pattern is width normalized and feature vectors are obtained. The feature vectors of the unknown pattern are then compared with feature vectors of known patterns stored in memory and the characters are identified.

Claims

1. Apparatus for pattern recognition comprising:

(a) means for scanning an unknown pattern to produce a first matrix of column and row signals representing the scanned image of the unknown pattern,

(b) means for producing from the first matrix a reduced size standard matrix of constant height and width containing an image only of a single character of the pattern, said producing means comprising;

(i) means for normalizing the height of the scanned image,

(ii) means for correcting the scanned image for skew by dividing the matrix with normalized height image into top and bottom parts, determining the center of mass separately for each of the top and bottom parts, and horizontally shifting rows of the image until the mass centers are vertically aligned,

(iii) means for normalizing the width of the horizontally shifted image by modifying the image by replicating and deleting vertical lines of the image until the image just fits within the standard matrix,

(c) means for creating from the standard matrix with normalized height and width image a multiple-word feature vector characteristic of the image by subdividing the image into smaller units and generating a word characteristic of the image part in each smaller unit,

(d) means for comparing the feature vector thus obtained or signals derived therefrom with corresponding signals obtained from known characters and stored in memory to assist in identifying a character of the unknown pattern.

2. A method of pattern recognition,, comprising the steps:

(a) scanning the unknown pattern to produce a first matrix of column and row signals representing the scanned image of the unknown pattern,

(c) producing from the first matrix a reduced size standard matrix of constant height and width containing an image only of a single character of the pattern by the following steps;

(i) normalizing the height of the scanned image,

(ii) dividing the matrix with image obtained in the previous step into top and bottom parts, determining the center of mass separately for each of the top and bottom parts, and horizontally shifting rows of the image until the mass centers are vertically aligned,

(iii) normalizing the width of the shifted image obtained in the previous step by modifying the image with respect to the selected matrix by replicating and deleting vertical lines of the image until the image just fits within the standard matrix,

(c) then from the standard matrix with image obtained following step (b), creating a multiple-word feature vector characteristic of the image by subdividing the image into smaller units and generating a cord characteristic of the image part in each smaller unit,

(d) thereafter comparing the feature vector obtained in step (c) or signals derived therefrom with feature vectors or signals derived therefrom obtained from known characters and stored in memory to assist in identifying a character of the unknown pattern.

3. A method as set forth in claim 2 wherein step (c) is carried out by forming separate blocks each representing a separate one of said smaller units, each blocks containing different multiple rows and columns of the matrix.