On-line recognition method and apparatus for a handwritten pattern

• US 4,653,107 A
• Filed: 12/24/1984
• Issued: 03/24/1987
• Est. Priority Date: 12/26/1983
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Shojima, Hiroshi, Kuzunuki, Soshiro, Hirasawa, Kotaro
• Primary Examiner(s): Birmiel, Howard A.
• Application Number:
  US06/686,001
• Time in Patent Office: 820 Days
• Field of Search: 382/3, 382/13, 382/22
• US Class Current: 382/189
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06V 30/32 Digital ink

Abstract

Coordinates of a handwritten pattern drawn on a tablet are sequentially sampled by a pattern recognition unit to prepare pattern coordinate data. Based on an area encircled by segments created by the sampled pattern coordinate data of one stroke and a line connecting a start point and an end point of the one-stroke coordinate data, the sampled pattern coordinate data of the one stroke is converted to a straight line and/or curved line segments. The converted segments are quantized and normalized. The segments of the normalized input pattern are rearranged so that the input pattern is drawn in a predetermined sequence. Differences between direction angles for the rearranged segments are calculated. Those differences are compared with differences of the direction angles of the dictionary patterns read from a memory to calculate a difference therebetween. The matching of the input pattern and the dictionary pattern is determined in accordance with the difference. If the matching fails, the first or last inputted segment of the input pattern is deleted or the sampled pattern coordinate data of the next stroke is added, to continue the recognition process.

Claims

1. An on-line recognition method for a handwritten pattern comprising the steps of:

sampling successively coordinate information of strokes of a handwritten pattern;

segmentalizing the coordinate information of one stroke into line segments consisting of at least either of straight line segments and curved line seg- ments;

quantizing the line segments so that the direction of each line segment is classified by one of predetermined angles;

normalizing an input pattern constituted by the quantized line segments to determine presence and absence of connections between the line segments;

rearranging the order of the quantized line segments in accordance with a predetermined connection sequence, on the basis of the determination result of the preceeding step;

reading out a dictionary pattern already stored in a memory; and

calculating degree of difference between the input pattern and the readout dictionary pattern by comparing the both on the basis of the quantized angles and the rearranged order of the quantized line segments, to thereby recognize the input pattern.

2. An on-line recognition method according to claim 1, further comprising, after said difference degree calculating step, a step of altering the number of the line segments to continue the recognition processing when non-matching is detected by the comparison in said difference degree calculating step.

3. An on-line recognition method according to claim 2, wherein in said segment number altering step, the first or last inputted segment is deleted first to continue the pattern recognition processing.

4. An on-line recognition method according to claim 2, wherein in said segment number altering step, the coordinate information of a next stroke is added to the coordinate information of the current stroke to continue the pattern recognition processing.

5. An on-line recognition method according to claim 1, wherein in said segmentalizing step, segmentalization is carried out by calculating an area encircled by a fold line sequentially connecting the coordinate points on the coordinate information of one stroke and a line connecting the start coordinate point and the end coordinate point of the coordinate information.

6. An on-line recognition method according to claim 5, wherein said segmentalizing step includes the substeps of:

comparing the calculated area with a predetermined value;

determining as a desired line segment the line connecting the start coordinate point and the end coordinate point of the coordinate information when the calculated area is smaller than the predetermined value; and

dividing the fold line into two fold lines when the calculated area is not smaller than the predetermined value, to continue the segmentalizing step by comparing with the predetermined value an area encircled by the divided fold line and a line connecting the start coordinate point and the end coordinate point of the divided fold line.

7. An on-line recognition method according to claim 5, wherein in said segmentalizing step, two contiguous line segments are regarded as a curve to be compared with a curved portion of a dictionary pattern, when an absolute difference of angle made by the two segments is smaller than a predetermined angle and a sign thereof is equal to that of the difference of angle made by immediately preceding two line segments and a length ratio thereof is within a predetermined range.

8. An on-line recognition method according to claim 1, wherein in said difference degree calculating step, the difference of the quantized angles of each two contiguous rearranged line segments is compared with each corresponding difference of the readout dictionary pattern.

9. A recognition method for a handwritten pattern comprising the steps of:

(a) sampling successively coordinate information of strokes of a handwritten pattern;

(b) segmentalizing the sampled coordinate information by carrying out the following substeps;

(i) calculating an area encircled by a fold line formed by the sampled coordinate information and a line connecting the start coordinate point and the end coordinate point of the fold line;

(ii) comparing the calculated area with a predetermined value;

(iii) determining as a segnentalized straight line of the fold line the line connecting the start coordinate point and the end coordinate point of the fold line when the calculated area is smaller than a predetermined value, and returning to the substep (i); and

(iv) dividing the fold line into two fold lines and returning to the substep (i); and

(c) after the segmentalizing step, carrying out a pattern recognition processing on the basis of the line segments.

10. A recognition method for a handwritten pattern comprising the steps of:

(a) sampling successively coordinate information of strokes of a handwritten pattern;

(b) segmentalizing the sampled coordinate information;

(c) comparing an input pattern formed by line segments with a dictionary pattern already stored; and

(d) when no-matching is detected by the comparison, deleting either one of the first and last line segments from the inputted line segments and then continuing a pattern recognition processing on the basis of the inputted line segments in which one line segment has been deleted.

11. An on-line handwritten pattern recognition apparatus comprising:

a coordinate input unit for sampling serial coordinate information and pen-up and pen-down information of strokes of a handwritten pattern;

a pattern recognition unit for comparing the input information from said input unit for each input stroke starting from pen-down and ending at pen-up with standard pattern information stored in a memory to determine analogy therebetween;

a pattern generation unit for generating reshaped information of the pattern recognized by said pattern recognition unit;

said pattern recognition unit including;

segmentation means for converting the input information from said coordinate input unit to straight line and curved line segments;

quantization data generation means for quantizing the segments in accordance with angles made by lines connecting start points and end points of the respective segments;

normalization means for listing presence or absence of connection between the start point and the end point of each of the segments;

sequencing means for rearranging the quantized data into one-dimension input data sequence based on said list;

difference calculation means for rearranging the standard pattern data into one-dimension standard data sequence by said sequencing means and comparing the one-dimension standard data sequence with the one-dimension input data sequence to calculate the analogy therebetween; and

number of segments altering means for increasing or decreasing the number of segments to be compared with the standard pattern information when the difference is smaller than a predetermined value, to continue a pattern recognition processing.

12. An on-line handwritten pattern recognition apparatus according to claim 11 wherein said segmentation means includes:

calculation means for calculating an area encircled by segment pattern sequentially connecting points on the input stroke and a line connecting the start point and the end point of the segment pattern;

first decision means for determining whether the segment pattern is to be regarded as a straight line or not based on the area calculated by said calculation means; and

second decision means for determining whether the segments determined as the straight line by said first decision means are a portion of a curved line.

13. An on-line handwritten pattern recognition apparatus according to claim 11 wherein said sequencing means includes means for selecting one segment from said segments and arranging the quantized data in one direction starting from the quantized data of said selected segment, based on said list.

14. An on-line handwritten pattern recognition apparatus according to claim 11 wherein said number of segments altering means includes a start pointer and an end pointer for the segments to be compared, means for incrementing or decrementing said pointers, and segment updating means for replacing the segments by new segments pointed by said pointers.

15. An on-line handwritten pattern recognition apparatus for evaluating analogy between a handwritten pattern and each of stored dictionary patterns, and of a pattern recognized to be identical to the handwritten pattern is found in the dictionary patterns, outputting that dictionary pattern as a recognized pattern, and if such a pattern is not found, repeating the above process for a next input handwritten pattern, characterized by that:

the handwritten pattern is divided into a plurality of segments in accordance with a value of an area encircled by a straight line connecting two of a plurality of sample points on the handwritten pattern and the sample points interposed between said two points and tracing said handwritten pattern.

16. An on-line handwritten pattern recognition apparatus according to claim 15 wherein the handwritten pattern is divided into a plurality of segments by substituting said handwritten pattern by a segment when said area is smaller than a predetermined threshold, and if said area is larger than the predetermined threshold, the line connecting the sample points is repeatedly shortened until said area becomes smaller than said threshold.

17. An on-line handwritten pattern recognition apparatus according to claim 16 wherein the line connecting the sample points is shortened by dividing each stroke of the handwritten pattern into n segments (n≧

2, an integer), and if an angle difference of m contiguous segments (n≧

m≧

2, an integer) is smaller than a predetermined angle, reducing the m segments to l (m>

l, an integer) segments.

18. An on-line handwritten pattern recognition apparatus according to claim 15 wherein the handwritten pattern is divided in accordance with a ratio of the length of the line connecting the two of the sampling points on the handwritten pattern and a sum of lengths of the segments tracing the handwritten pattern.

19. An on-line handwritten pattern recognition apparatus according to claim 18 wherein the handwritten pattern is substituted by a segment when said ratio is within a predetermined range, and if said ratio is not within the predetermined range, the line connecting the sample points on the handwritten pattern is repeatedly shortened until said ratio comes into said range.

20. An on-line handwritten pattern recognition apparatus according to claim 19 wherein the line connecting the sample points is shortened by dividing each stroke of the handwritten pattern into n (n≧

2, an integer) segments, and if an angle difference of m (n≧

m≧

2, an integer) contiguous segments is smaller than a predetermined angle, reducing m segments to l (m>

l, an integer) segments.

21. An on-line handwritten pattern recognition apparatus for evaluating analogy between a handwritten pattern and each of stored dictionary patterns, and if a pattern recognized to be identical to the handwritten pattern is found in the dictionary patterns, outputting that dictionary pattern as a recognized pattern, and if such a pattern is not found, repeating the above process for a next input handwritten pattern, characterized by that:

combinations consisting of portions of or entire handwritten pattern are compared with the dictionary patterns, the combinations consist of segments of the handwritten patterns, and the combination having a larger number of segments is compared first.

22. An on-line handwritten pattern recognition apparatus according to claim 21 wherein said segments are time-serially numbered, the segment having the smallest or largest number is deleted and the remaining segments are used as one of said combination.

23. An on-line handwritten pattern recognition apparatus according to claim 22 wherein the number of segments to be deleted is limited to n (0≦

n<

number of input segments, an integer), and the opposite ends of the segment number sequence are alternately deleted.

24. An on-line handwritten pattern recognition apparatus according to claim 22 wherein a plurality of segments are deleted in one step when said segments include segments connected at an angle larger than a predetermined angle.

25. An on-line handwritten pattern recognition apparatus according to claim 22 wherein segments connected at an angle larger than a predetermined angle in said segments are deleted in one step.

26. An on-line handwritten pattern recognition apparatus according to claim 23 wherein a plurality of segments are deleted in one step when said segments include segments connected at an angle larger than a predetermined angle.

27. An on-line handwritten pattern recognition system according to claim 23 wherein segments connected at an angle larger than a predetermined angle in said segments are deleted in one step.