JPH0425588B2 - - Google Patents

Description

[Detailed description of the invention]

(1) Technical Field of the Invention The present invention relates to a character recognition method, and particularly to a method for discriminating between hiragana and kanji in an optical character reading device that optically reads handwritten hiragana and kanji. (2) Prior art and problems Conventional optical character reading devices recognize hiragana and kanji using a single method, but hiragana and kanji are generally distinguished by complexity such as the number of strokes and the number of curved parts. There are clear differences as shown in the figure. Therefore, it seems likely that both were recognized by the same means.
There are problems in that good recognition results cannot be obtained because the differences cannot be clearly understood, and character post-processing cannot be performed accurately on the recognition results, such as word processing consisting only of kanji. Ta. (3) Purpose of the Invention The purpose of the present invention is to improve the recognition accuracy of hiragana and kanji by distinguishing hiragana and kanji input into an optical character reading device using separate recognition means, and to perform post-processing on the recognition results. The aim is to improve the accuracy of (4) Structure of the Invention According to the present invention, a character pattern input to an optical character reading device is calculated by determining the number of loops surrounded by the character, the number of connected components that are separated from each other, and the horizontal and vertical directions. The characters are classified into characters with a large number of strokes and characters with a small number of strokes according to the number of black runs, which are continuous parts of the characters scanned, and are further classified into characters with a large number of strokes and characters with a small number of strokes. A character recognition method is provided which is characterized by classifying characters with a small number of strokes into kanji and hiragana with a small number of strokes based on a line segment series. (5) Hereinafter, the present invention will be explained by way of examples with reference to the accompanying drawings. FIG. 1 is a circuit configuration diagram for implementing a character recognition method according to the present invention. When a character M consisting of hiragana and kanji is input to the circuit shown in Fig. 1, first, the multi-stroke number kanji TK and other characters are distinguished by the loop number connected component number part 1 and the average black run number part 2.
The contour line segment series section 3 further distinguishes between kanji SK and hiragana H, which have a small number of strokes. The loop number and connected component number section 1 includes a loop number counting circuit 11, a connected component number counting circuit 12, and a large stroke number/small stroke number determination circuit 13. this house,
Circuits 11 and 12 each have the function of counting the number of loops A and the number of connected components B defined in FIG. Loop A refers to a portion surrounded by the character portion of input character M, and in the example of FIG. 2, there are two loops. Connected components B refer to mutually separated parts of the input character M, and in the example of FIG. 2, there are five connected components as shown by broken lines. In addition, the circuit 13 for determining the number of strokes and the number of strokes is small.
is a circuit that determines the number of strokes of the input character M. This circuit 13 is also incorporated in the other average black run counting unit 2, and distinguishes between the kanji with a large number of strokes TK and other characters, and calculates the number of kanji with a corresponding number of strokes. Extract TK (Figure 1). The average black run number section 2 is an average black run number counting circuit 21
and a large stroke number/small stroke number determination circuit 22, which determines the degree of complexity of the input character M together with the loop number connected component number section 1 at the previous stage. Black run C refers to the continuous part of the character to be scanned when scanning in the x or y direction in Figure 2, and if the number of meshes in the x and y directions is M and N, then the average black run in each direction is The numbers n _x and n _y are respectively

【formula】

Represented by [Formula]. The contour line segment series section 3 includes a contour line segment series extraction circuit 31
and a low stroke count/hiragana determination circuit 22, which has a function of discriminating between low stroke count kanji SK and hiragana H. What is a contour line segment series?As shown in Fig. 3, input character M is scanned in the x direction (Fig. 3 1) and y direction (Fig. 3 2), and whether the character is open at the edge of the character (○ A series of four line segments (○ and ○, ○ and ●, ● and ○, ● and ●) that are formed by the combination of closed (●) marks. FIG. 3 1' shows a contour pattern and a series of contour line segments of the character in FIG. 3 1 viewed from the left side. The contour line segment series is formed as a continuous portion when the contour pattern is viewed from the left side. The state of both ends of a contour segment (open or closed) is determined by the fact that when looking above (in the case of the top end) or below (in the case of the bottom end) the contour line segment, if it collides with the contour, it is closed, and if it does not collide with the contour, it is open. be. By extracting these line segment sequences, they are inputted to a low stroke count/hiragana determination circuit 32 to discriminate between the low stroke count kanji SK and the hiragana H. Comparing kanji with a small number of strokes and hiragana, hiragana are often made up of curved strokes, while kanji are often made up of straight strokes. FIG. 4 shows a method for distinguishing between kanji and hiragana with a small number of strokes based on a series of contour line segments. Taking as an example the outline segments seen from the right side of two types of line segments a and b as shown in Figure 4, for a straight line like a, the outline segments are generated as one continuous line part. On the other hand, a curve like b is generated by dividing it into a plurality of contour line segments. Therefore, when the number of loops, the number of connected components, the number of black runs, etc. are the same, hiragana and kanji can be separated depending on the number of contour segments. The pattern of the character M input to the circuit having the above configuration is first classified into the Chinese character TK with a large number of strokes and other characters based on the loop number connected component number part 1 and the average black run number part 2. Since other characters have relatively few characteristics, the outline segment series section 3, which has a more detailed recognition function, uses SK characters for kanji with fewer strokes.
and Hiragana H. (6) Effects of the Invention As described above, according to the present invention, since hiragana and kanji input to an optical character reading device can be distinguished by separate recognition means, the recognition accuracy of hiragana and kanji can be improved. Moreover, it is possible to improve the accuracy of post-processing based on this.

[Brief explanation of drawings]

Fig. 1 is a circuit configuration diagram for implementing the character recognition method according to the present invention, Figs. 2 and 3 are conceptual explanatory diagrams of the circuit shown in Fig. 1, and Fig. 4 shows a small number of strokes by a series of contour line segments. FIG. 2 is a diagram illustrating a method for distinguishing between kanji and hiragana. 1... Loop number connected component number part, 2... Average black run number part, 3... Contour line segment series part, 11... Loop number counting circuit, 12... Connected component number counting circuit, 13
...High stroke number/low stroke number determination circuit, 21: Average black run number counting circuit, 22: High stroke number/low stroke number determination circuit, 31: Contour line segment series extraction circuit, 32: Low stroke number, Hiragana determination circuit.

Claims (1)

[Claims] 1 The pattern of characters input into an optical character reader is determined by the number of loops surrounded by the character, the number of connected components that separate the character from each other, and the number of consecutive parts of the character scanned in the horizontal and vertical directions. Characters with a large number of strokes are classified into characters with a large number of strokes and characters with a small number of strokes according to the number of black runs, and characters with a small number of strokes are classified into characters with a small number of strokes based on a contour line segment series, which is a set of predetermined line segments that constitute the outline of the character with a small number of strokes. A character recognition method that is characterized by classification into kanji and hiragana.