Handwriting recognizer with estimation of reference lines

• US 5,774,582 A
• Filed: 04/24/1995
• Issued: 06/30/1998
• Est. Priority Date: 01/23/1995
• Status: Expired due to Fees

• Current Assignee: Advanced Recognition Technologies, Inc.
• Original Assignee: Advanced Recognition Technologies, Inc.
• Inventors: Gat, Yoram, Kadosh, Arie
• Primary Examiner(s): Bost, Dwayne
• Assistant Examiner(s): JOHNSON, BRIAN
• Application Number:
  US08/428,806
• Time in Patent Office: 1,163 Days
• Field of Search: 382/184, 382/186, 382/187, 382/188, 382/189, 382/177, 382/178, 382/179, 382/174, 395/792
• US Class Current: 382/186
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/146 Aligning or centring of the image pick-up or image-field

Abstract

The system of the present invention has a reference character database, which stores a multiplicity of reference characters, and a feature extractor which extracts shape information for each input character. For each reference character, the database lists shape information and reference line measurements indicating how reference lines pass through it. For example, the reference line measurements can be percentages of a height of a bounding box which bounds each reference and/or input character.

In some embodiments, the input characters are collected into a stroke buffer and in other embodiments, they are collected into a line buffer. In all embodiments, the reference line estimation is performed for a series of input characters, whether as a full line of characters or as a collection of strokes.

Claims

1. A handwritten character recognizing system which includes estimation of reference lines, the system comprising:

a stroke buffer for storing a plurality of consecutive strokes of input characters, wherein each stroke has a multiplicity of sample points, said characters having reference lines associated therewith, each said reference line being of a type, each said type from a collection of a plurality of reference line types;

a feature extractor for forming input characters from said strokes and for producing shape information for each input character formed;

a reference lines estimator, connected to said stroke buffer, for producing a plurality of sets of reference lines all of which are parallel to a central line estimated from a central sample point of each stroke, each said set of reference lines containing one and only one reference line of every said type of said collection;

a reference character database comprising a multiplicity of reference characters, wherein, for each reference character, said database lists its shape information and reference line measurements indicating how reference lines pass through it;

a character classifier for matching said reference characters to said input characters in accordance with said shape information, for matching a set of said reference line measurements to each set of reference lines and for producing scores thereby; and

a decision unit for classifying said input characters based on said scores produced by said character classifier.

2. A system according to claim 1 and wherein said reference lines estimator comprises means for receiving said strokes, means for estimating a central line from a set of central sample points of said strokes, and means for producing, from said central line, said plurality of possible reference lines which are parallel to but at a distance from said central line.

3. A system according to claim 1 and wherein said reference line measurements are percentages of a height of a bounding box which bounds each reference character.

4. A handwritten character recognizing system which includes estimation of reference lines, the system comprising:

a line buffer for storing a line having a plurality of input characters therein;

a feature extractor for producing shape information for each input character in said line buffer;

a reference character database comprising a multiplicity of reference characters, wherein, for each reference character, said database lists its shape information and reference line measurements indicating how reference lines pass through it;

a character classifier for matching said reference characters to said input characters in accordance with said shape information and for producing scores thereby;

a top score filter for selecting those reference characters having scores above a predetermined threshold and for receiving said reference line measurements for each selected reference character,

a reference lines estimator, connected to said top score filter, for generating reference lines from said reference line measurements of said selected reference characters, wherein one set of reference lines is produced for each combination of selected reference characters, and for producing a score for each set of reference lines; and

a decision unit for classifying said line of input characters based on the best score, for said line, produced by said reference lines estimator.

5. A system according to claim 4 and wherein said reference line measurements are percentages of a height of a bounding box which bounds each reference character.

6. A method for recognizing handwritten characters which includes estimation of reference lines, the method comprising the steps of:

in a stroke buffer, storing a plurality of consecutive strokes of input characters, wherein each stroke has a multiplicity of sample points, said characters, having reference lines associated therewith, each said reference line being of a type, each said type from a collection of a plurality of reference line types;

forming input characters from said strokes and producing shape information for each input character formed;

producing a plurality of sets of reference lines all of which are parallel to a central line estimated from a central sample point of each stroke;

storing, in a reference character database, a multiplicity of reference characters, wherein, for each reference character, said database lists its shape information and reference line measurements indicating how reference lines pass through it;

matching said reference characters to said input characters in accordance with said shape information;

matching a set of said reference line measurements to each set of reference lines and for producing scores thereby, each said set of reference lines containing one and only one reference line of every said type of said collection; and

classifying said input characters based on said scores produced by said step of matching.

7. A method according to claim 6 and wherein said step of producing comprises the steps of receiving said strokes, estimating a central line from a set of central sample points of said strokes, and producing, from said central line, said plurality of possible reference lines which are parallel to but at a distance from said central line.

8. A method according to claim 6 and wherein said reference line measurements are percentages of a height of a bounding box which bounds each reference character.

9. A method for recognizing handwritten characters which includes estimation of reference lines, the method comprising the steps of:

storing, in a line buffer, a line having a plurality of input characters therein;

producing shape information for each input character in said line buffer;

storing, in a reference character database, a multiplicity of reference characters, wherein, for each reference character, said database lists its shape information and reference line measurements indicating how reference lines pass through it;

matching said reference characters to said input characters in accordance with said shape information and producing scores thereby;

selecting those reference characters having scores above a predetermined threshold and receiving said reference line measurements for each selected reference character;

generating reference lines from said reference line measurements of said selected reference characters, wherein one set of reference lines is produced for each combination of selected reference characters, and producing a score for each set of reference lines; and

classifying said line of input characters based on the best score, for said line, produced by said reference lines estimator.

10. A method according to claim 9 and wherein said reference line measurements are percentages of a height of a bounding box which bounds each reference character.