JPH0253830B2 - - Google Patents

Description

[Detailed Description of the Invention] (A) Technical Field of the Invention The present invention relates to a kanji recognition device, in particular, when performing recognition processing for handwritten kanji. The present invention relates to a kanji recognition device that extracts a sequence and matches it with a redundant sequence of each line segment stored in a dictionary along the sequence.

(B) Technical background and problems Recognition processing for kanji, especially handwritten kanji, is currently at a very difficult stage. As one such recognition process, a method is being considered that focuses on the background part of the kanji and makes a global judgment, but the two-dimensional characteristics of the background (or the space inside the character) are considered. If you try to capture this, the processing tends to become extremely complicated. For this reason, consideration has been given to focusing on the use of contour line segments, which has traditionally been applied to handwritten katakana.

(C) Purpose and Structure of the Invention The purpose of the present invention is to solve the above-mentioned problems, and the kanji recognition device of the present invention scans the kanji characters to be recognized and extracts features, In a kanji recognition device that determines the category of the kanji character to be recognized by comparing it with the contents of a dictionary in which features are stored, the kanji character to be recognized is obtained by at least a horizontal search from the right to the left. The left contour of the character obtained by searching from the horizontal left to the right, the upper contour of the character obtained by searching from the top to the bottom, and the top contour of the character obtained by searching from the bottom to the top. Based on the lower contour of the character obtained by and a series of contour underline segments along the lower contour of the character, and each of the contour line segment series is stored in the dictionary with redundancy in the series, and the recognition target kanji character is extracted. The present invention is characterized in that each of the contour line segment series obtained from the above is compared with each of the contour line segment series read out from the dictionary along the series. This will be explained below with reference to the drawings.

(D) Embodiments of the Invention FIG. 1 is an explanatory diagram for explaining the series of contour line segments according to the present invention, FIG. 2 is an explanatory diagram for explaining the series stored in the dictionary according to the present invention, and FIG. 1 shows the configuration of an embodiment of the present invention.

When the kanji 1 illustrated in Figure 1A is given, for example, the handwritten kanji ``shi'', scan 2 in the direction of the arrow shown in the figure.
(i) Extract the point A that first reaches the black area from the left white area of the background.

(ii) Next, extract the point α that reaches the white area from the black area.

(iii) Next, extract point B that reaches the black area from the white area.

(iv) Next, extract the point b that reaches the white area from the black area.

(v) Finally, extract the point b that reaches the right white area of the background from the black area.

do it like this.

Then, for each point corresponding to the point A in each scan, left contour line segments are extracted in series in the vertical direction, for example, as shown in L1, L2, . . . in FIG. 1B. Note that at this time, if the horizontal positions of the points corresponding to the above-mentioned point A obtained in correspondence with the two vertically arranged scans are separated by a threshold value or more, the line segment is considered to be discontinuous.
In addition, cases in which the start and end of each line segment are sealed by a black area of a character (black circles in the figure) and cases in which they are not sealed (white circles in the figure) are extracted separately.

Similarly to the above, the right line segment of the contour is created by connecting the points corresponding to the point n shown in FIG.
Extract as follows. Of course, in this extraction,
The scan from the right side to the left may be performed again.

Furthermore, if necessary, in order to code the white area between the strokes of the character,
The center point between point a and point B shown in the figure, the center point between point b and point N, etc. are extracted, and these respective center points are connected in the vertical direction to form the contour midline segment according to the present invention. You can extract and extract the series.

Extracted contour left line segment L as shown in Figure 1B
1, L2, . . . and right contour line segments R1, R2, . It goes without saying that this series represents the outline features of the kanji to be recognized 1.

In the case of the present invention, the contour line segment series obtained in this manner is sequentially compared with the contour line segment series in the dictionary in the direction of arrow 3. Figure 2 B shows the character “Shi”
The left contour line segment series 6 in the dictionary corresponding to is shown. In the figure, numerals 7 and 8 indicate either OK, numerals 9 and 10 indicate abbreviation, and numeral 11 indicates a series end point. In series 6 in the dictionary,
The reason why redundancy is given as in symbols 7 and 8 and symbols 9 and 10 can be considered to be to cope with deformation of characters.

The contour line segment series as explained above using the contour left line segment series is, for example, (i) the contour left line segment series,
Four types are extracted: (ii) contour right line segment series, (iii) contour upper line segment series, and (iv) contour lower line segment series. Furthermore, two types of line segment series existing between character strokes are extracted: (v) a contour center line segment series during horizontal direction scanning, and (vi) a contour center line segment series during vertical direction scanning. Then, the same verification as above is performed for each.

FIG. 3 shows the configuration of an embodiment of the present invention.
In the figure, reference numeral 12 is a horizontal line segment extraction circuit, 13 is a vertical line segment extraction circuit, and 14 and 15 are characteristic line segment series (line segment series) connection circuits, which are the line segment L1 shown in FIG. 1B. , L2, ..., 16 is a contour left line segment series buffer, 17 is a contour right line segment series buffer, 18 is a contour middle line segment series buffer, 19 is a contour line segment series buffer, and 20 is a contour line segment series buffer. 21 is a contour center line segment series buffer; 22 is a dictionary; 23 to 28 are matching judgment circuits; and 29 is a determining circuit.

The illustrated characteristic line segment connection circuit 14 includes the line segments L1, L2, . . . and R1, R
2, . . . are checked and stored in the corresponding buffers 16, 17, 18, respectively. The feature line segment connection circuit 15 corresponds to vertical scanning and operates in the same manner as the circuit 14.

The contents of each buffer 16 to 21 can be found in the dictionary 22.
Each line segment series read out from the matching judgment circuit 23
to 28. In this case, the second
It may be considered that the verification is performed sequentially in the direction of arrow 3 shown in Figure A.

The determination results from each of the comparison determination circuits 23 to 28 are led to a determination circuit 29, which comprehensively examines them and determines the category of the Chinese character to be recognized.

(E) Effects of the Invention As explained above, according to the present invention, a series of character outline segments can be used for recognition of handwritten kanji characters, etc. It is easy to extract two-dimensional features of characters to be recognized, and by preparing a dictionary with some redundancy, it is possible to obtain a sufficient recognition rate.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram for explaining the series of contour line segments according to the present invention, FIG. 2 is an explanatory diagram for explaining the series stored in the dictionary according to the present invention, and FIG. 3 is an explanatory diagram for explaining the configuration of an embodiment of the present invention. shows. In the figure, 1 is the kanji to be recognized, 2 is the scanning line, 4,
5 and 6 are contour line segment series, 12 and 13 are line segment extraction circuits, 14 and 15 are each line segment series concatenation circuits, 16 to 21 are line segment series buffers, respectively.
2 is a dictionary; 23 to 28 are respective matching judgment circuits;
29 represents a decision circuit.

Claims (1)

[Claims] 1 Scan the kanji characters to be recognized and extract the features,
In a kanji recognition device that determines the category of the kanji character to be recognized by comparing it with the contents of a dictionary that stores features corresponding to standard kanji characters, the kanji character to be recognized is at least horizontally moved from right to left. The left contour of the character obtained by searching backwards, the right contour of the character obtained by searching horizontally from left to right, the top contour of the character obtained by searching from top to bottom, and the top contour of the character obtained by searching from bottom to top. Based on the lower contour of the character obtained by searching in the direction, a series of contour left line segments along the left contour of the character, a series of right contour line segments along the right contour of the character, and a series of contour right line segments along the left contour of the character, and the upper contour of the character. A series of contour upper line segments that follow the character lower contour and a series of contour lower line segments that follow the character lower contour are extracted, and each of the contour line segment series is stored in the dictionary with redundancy added to the series. , a kanji recognition device characterized in that each of the contour segment series obtained from the kanji character to be recognized is compared with each of the contour segment series read from the dictionary along the series. .