Character recognition method utilizing compass directions and torsion points as features

• US 5,659,633 A
• Filed: 04/25/1994
• Issued: 08/19/1997
• Est. Priority Date: 11/29/1991
• Status: Expired due to Term

• Current Assignee: Advanced Recognition Technologies, Inc.
• Original Assignee: Advanced Recognition Technologies, Inc.
• Inventors: Isaak, Shmuel, Ilan, Gabriel
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/233,086
• Time in Patent Office: 1,212 Days
• Field of Search: 382/186, 382/187, 382/185, 382/242, 382/197, 382/119, 382/123, 382/198, 382/241
• US Class Current: 382/186
• CPC Class Codes: G06V 30/347 Sampling; Contour coding; Stroke extraction
• CPC Class Codes: G06V 30/373 using a special pattern or subpattern alphabet

Abstract

A method of recognizing a character is described which includes the following operations: (a) measuring the angle of the tangent to the character at each of a plurality of sample points on the character, (b) assigning those points having a tangent direction within a first, small predefined angle from the vertical axis with values of Up or Down compass directions, and those points having a tangent direction within the first, small predefined angle from the horizontal axis with values of Left of Right compass directions and (c) selecting special points. The special points are those sample points: i) at the beginning and end of the character; ii) whose compass direction is different than that of the previously selected special point; iii) whose compass direction is the same as the previously selected special point only if: a) there is at least one sample point, among the sample points between the two sample points with the same compass direction, which has no compass direction and b) at least one of the sample points with no compass direction is a "torsion" point defined as a point having a tangent direction which is a significant local extremum among those sample points with no compass direction between the two sample points having the same compass direction; and iv) which are torsion points.

Claims

1. A method of recognizing a character comprising the operations:

(a) measuring the angle of the tangent to the character at each of a plurality of sample points on the character;

(b) assigning those points having a tangent direction within a first, small predefined angle from the vertical axis with values of Up (u) or Down (D) compass directions, and those points having a tangent direction within the first, small predefined angle from the horizontal axis with values of Left (L) or Right (R) compass directions;

(c) selecting as special points at least those sample points

i) at the beginning and end of said character;

ii) whose compass direction is different than that of the previously selected special point;

iii) whose compass direction is the same as the previous special point only if;

a) there is at least one sample point, among the sample points between the two sample points with the same compass direction, which has no compass direction, and

b) at least one of said sample points with no compass direction is a "torsion" point defined as a point having a tangent direction which is a significant local extremum among those sample points with no compass direction between the two sample points having the same compass direction;

iv) which are torsion points;

(d) arranging said special points according to predefined rules;

(e) and utilizing the resulting arrangement of special points for recognizing the character.

2. The method according to claim 1, wherein said first predefined angle is not more than 15°

from the respective axis.

3. The method according to claim 2, wherein said first predefined angle is about 11°

from the respective axis.

4. The method according to claim 1, wherein the tangent angle at each of said plurality of sample points is defined by the angle between an imaginary chord, connecting a preceding sample point with a succeeding sample point both spaced a predefined number of sample points away on either side of the respective sample point, and the vertical axis of the handwriting device.

5. The method according to claim 4, wherein the length of said chord is also measured, and if below a predetermined minimum length, the sample point is assigned an identification indicating an ambiguous tangent direction.

6. The method according to claim 1, wherein the method is used for the real-time recognition of handwritten characters as they are written as a series of handwritten strokes, and wherein the arrangement of special points resulting from predefined sequence in operation (d) is a time-ordered sequence.

7. The method according to claim 6, further including determine the location of the start and end points of each handwritten stroke, and utilizing also the location of said start and end points as start special points and end special points for recognizing the character according to the operations (d) and (e).

8. The method according to claim 6, wherein said time-ordered sequence of special points is utilized for recognizing the character by matching it with stored sequences of special points of a plurality of stored reference characters.

9. The method according to claim 8, wherein said operation (d) includes examining said time-ordered sequence of changes in the compass directions of the handwritten character for the presence of a 180°

change between first and second adjacent special points in the sequence; and

upon detecting the presence of such a 180°

change, injecting a further special point, identified with the attribute "breaking", into said sequence between said adjacent special points, and applying to said further special point a compass direction changed by 90°

from the compass direction of the first adjacent point, towards the compass direction of the second adjacent point.

10. The method according to claim 9, wherein:

all special points not within the first predefined angle from the horizontal and vertical axes are rounded to the nearest axes but are identified with the attribute of "diagonal" (D); and

the sequence of the special points of the handwritten character to be identified, and the sequences of the special points in the library, are matched according to the following rules;

(a) if the corresponding compass directions in the sequence match, the two sequences are incremented;

(b) if the corresponding compass directions in the two sequences are opposite to each other, and if both special points are identified with the attribute "breaking", the two sequences are incremented;

(c) if the two corresponding compass directions are opposite to one another, and one is identified with the attribute "breaking", while the other is identified with no attribute, the two sequences are incremented;

(d) if one of the corresponding points is identified with the attribute "diagonal", and the other is not, then;

(i) if the two points are the first ones in their respective sequences, none of them is the last point of its sequence, and the tangent directions of the two special points are not opposite to one another, only the sequence containing the "diagonal" attribute is incremented;

(ii) if the two points are not the first ones in the sequence, none of them is the last point of its sequence and the tangent directions of the two special points are opposite to one another, the sequence with the "diagonal" attribute is incremented by "2";

(iii) if one point is diagonal, and the other is the last of its sequence, then the sequence containing the diagonal attribute is incremented;

(e) continue matching the corresponding special points in the two sequences for the complete sequences, and determine a "successful" match to the present if the two sequences are thus completed; and

(f) terminate the matching of the two sequences whenever a match as above is not made between the two sequences, and determine the match to be an "unsuccessful" match.

11. The method according to claim 6, wherein said time-ordered sequences of special points are stored in a library which also includes the number of strokes of each of a plurality of stored reference characters;

said method further including a preliminary screening operation of;

(i) comparing the number of strokes of the handwritten character to be recognized with the number of strokes of the stored reference characters in the library, and

(ii) eliminating from further matching those reference characters not having the same number of strokes as in the handwritten character to be recognized.

12. The method according to claim 11, further including normalizing the coordinates of all special points of the handwritten character to be recognized to the smallest rectangle which encompasses all said special points.

13. The method according to claim 12, wherein, if more than one reference character of the library matches the sequence of special points of the handwritten character to be recognized, a decision operation is performed to determine the reference character having the best match with the handwritten character to be recognized, by:

(i) calculating the sum of the square distances between corresponding points of the handwritten character to be recognized and each of said matched reference characters; and

(ii) identifying the handwritten character as the reference character having the smallest sum so calculated.