JPH0150955B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to an online handwritten character recognition device, and in particular to a similar character identification method for identifying katakana and kanji that are similar in shape to katakana through simple processing. One problem in handwritten character recognition is how to identify similar characters. In particular, some kanji with a small number of strokes have almost the same shape as katakana, as shown in Table 1, and are very difficult to identify. Conventionally, in order to identify these similar characters, the first
As shown in the figure, write text frame 1-1 on entry sheet 1.
An identification frame 1-21 is provided to identify similar characters, and in the case of katakana, write the katakana character (for example, ka) in the character entry frame 1-1 as shown in Figure 2a, and on the other hand, If you want to enter a kanji that has a similar shape to katakana (for example, power), write that kanji in the character entry frame, and also write it in the identification frame 1-21.
Identification mark (mark) indicating that it is a kanji character 1-
Enter 22. Then, the recognition unit selects the identification frame 1.
- Determine whether it is Katakana or Kanji depending on whether identification mark 1-22 is written in 21 or not.

[Table] Measures were taken. In addition, as mentioned above, instead of providing an identification frame 1-21 for each character entry frame 1-1, one identification frame 1-21 is provided on the entry sheet 1, and each character entry frame 1-2 is If so, write identification mark 1-22 in identification frame 1-21. Alternatively, a separate keyboard for specifying the character type (katakana, kanji, etc.) was installed, and the keyboard was operated to specify the character type each time a character was entered. However, in addition to writing characters, these methods require writing identification marks 1-22 and operating a keyboard, which not only burdens the user but also slows down the speed at which characters can be entered. The disadvantage is that it also causes a decline in SUMMARY OF THE INVENTION An object of the present invention is to provide a similar character identification method that eliminates the drawbacks of the prior art described above, reduces the burden on the user, and does not reduce the speed of character entry. In the present invention, in order to achieve the above object, katakana characters are written in a predetermined area set in a character writing frame, and the circumscribed rectangle of the written character, the center position of the circumscribed rectangle, or Kanji and katakana whose shapes are similar to katakana are identified depending on the position of the center of gravity of the written character in the entry frame. Hereinafter, embodiments of the present invention will be described based on FIG. First, before writing characters, the input pen 2 is used to inform the recognition system of the position of the entry sheet 1.
-1, press the position set marks 1-3 and 1-4 on the entry sheet 1 shown in FIG. At this time, the tablet 2 supplies the corresponding coordinate position information to the frame position memory 10 from 2-2. The frame position memory 10 stores position set marks 1-
Based on the coordinate values corresponding to 3, 1-4, know the placement position of entry sheet 1 and calculate the position of each character entry frame in entry sheet 1 (the character entry frame arrangement information of entry sheet 1 is (stored in the frame position memory 10) and store the new position of each character entry frame. This process is performed every time the entry sheet 1 is replaced. Next, characters are written in the character entry frame with the input pen 2-1. At this time, as shown in Figure 4, a katakana entry area 1-2 is provided in each character entry frame 1-1, and as shown in Figure 5a, when writing katakana, this area is used. Write in Katakana entry area 1-2. In addition, in the case of kanji, as shown in FIG. 5b, the katakana entry area 1-2 is ignored and the character is written on the side of the character entry frame. The characters written in this way are written using the input pen 2-1.
The coordinate position series and the information on the input pen 2-1 pressed onto the entry sheet are supplied to the preprocessing circuit 3 as handwriting information. The preprocessing circuit 3 removes noise components included in the XY coordinate position series (X, Y value series);
Then, smoothing, redundant handwriting points, etc. are performed. The pre-processed XY coordinate position series is supplied to the feature extraction circuit 4, where the features of the input character and information compression are performed to facilitate subsequent recognition processing. As a feature of the input character, the number of strokes of the input character is detected based on the crimping information of the input pen 2-1.
In addition, in order to compress information, each stroke of the input character (equivalent to the number of strokes, the line segment drawn before the input pen presses and leaves) is compared with a predetermined standard stroke, and the most similar standard stroke is It is expressed in code, and is written using input characters and a code string of this standard stroke. The input character information (standard stroke code string) described in this way is supplied to the similarity calculation circuit 6, and is sequentially compared with the standard character stroke code string stored in the standard pattern memory 5 in advance to find similarities. degree is calculated. At this time, the similarity calculation is performed for standard characters having the same number of strokes as the input character obtained by the feature extraction circuit 4. Then, the stroke codes of both strokes are compared for each stroke order, and if the codes match, a constant value α is set, and if they do not match, a value β is set in advance between the stroke codes (different between stroke codes. The value is larger for those that exist.β<
α) and find the sum of all strokes as the similarity. The similarity value for each standard character pattern is sequentially supplied to the identification circuit 7, and the code corresponding to the standard character pattern is supplied from 5-1 to the identification circuit 7. The identification circuit 7 detects the highest similarity value from the codes and similarity values for these standard character patterns, and when the difference from the next highest similarity value is greater than a predetermined value, the highest similarity is detected. Outputs the standard character pattern code corresponding to the degree value as the recognition result. However, when characters as shown in Table 1 are written, the stroke code matches both katakana and kanji, and both show a high degree of similarity, making it impossible to uniquely identify the input character. In such a case, it is determined whether the input characters are katakana or kanji, using information obtained from the above-described method of writing katakana and kanji. The process is as follows. Every time a character is written, the handwriting information of the written character is preprocessed by the preprocessing circuit 3 and then supplied to the circumscribed rectangle detection circuit 8. The circumscribing rectangle detection circuit 8 detects a rectangle 8-1 circumscribing the input character as shown in FIG. In other words, from the XY coordinate position series for one character,
The minimum value (point a in FIG. 5) and maximum value (point b in FIG. 5) of each axis are detected. Then, the position information of this circumscribed rectangle 8-1 (values of the XY coordinate positions of points a and b) is supplied to the frame position memory 10 and also to the position comparison circuit 9. The frame position memory 10 detects in which frame the input character is written based on the position information of the circumscribed rectangle 8-1, and stores a code (writing position code) corresponding to the character writing frame 1-1. At the same time as outputting, katakana entry area 1-2 in character entry frame 1-1
The minimum value Xc on the X axis and the maximum value Yc on the Y axis are supplied to the position comparison circuit 9. The values of Xc and Yc are determined by rearranging entry sheet 1, as mentioned above.
Point A (X _A , Y _A ) in Figure 5 for each character entry frame 1-1
_{In addition} to calculating and arranging the _coordinate position of point _{B (} X _{B , Y B )} , From the two equations (2), find X _C and Y _C. (k is a constant that determines the size of the katakana entry area 1-2.) The position comparison circuit 9 determines that the position of the circumscribed rectangle 8-1 is
Katakana entry area 1 in the character entry frame 1-1
It is determined whether it is within 2. The determination method is, for example, katakana entry area 1-2.
Let the minimum value of the X-axis be X _C and the maximum value of the Y-axis be Y _C , then the minimum X-axis value x _a of the circumscribed rectangle 8-1 (a in Figure 5)
X coordinate value of point), Y axis maximum value y _a of circumscribed rectangle 8-1
(Y coordinate value of point b in Figure 5) both satisfy the following two equations: x _a > X _C + l (3) y _b < Y _C + l (4) (l is a constant) Get up and do it. This determination result (whether or not the above equations (3) and (4) are satisfied, that is, whether or not the input character was written in the katakana writing area 1-2) is supplied to the identification circuit 7. , is used as identification information when input characters cannot be uniquely identified as katakana or kanji. That is, if the determination result is positive, the identification circuit 7 outputs a standard character pattern code corresponding to the katakana character. Moreover, in the negative case, a code of a standard character pattern corresponding to the Kanji character is output. As explained above, according to the present invention, by writing katakana characters in the katakana character entry area set in advance within the character entry frame, katakana and kanji with similar shapes can be easily identified, and scribes can On the other hand, the operation of specifying character types for katakana and kanji, which was a huge burden in the past, can be eliminated. Furthermore, it is possible to prevent the input speed from decreasing due to the conventional character type specification operations for katakana and kanji.

[Brief explanation of drawings]

Fig. 1 is a schematic diagram showing a conventional entry sheet, Fig. 2 a and b are pattern diagrams showing a conventional character type designation system for katakana and kanji, and Fig. 3 is an embodiment of the device according to the present invention. FIG. 4 is a schematic diagram showing the entry sheet of the present invention, and FIG. 5 a, b
is a pattern diagram explaining the katakana and kanji entry method of the present invention. 1... Entry sheet, 2... Tablet, 2-1... Input pen, 3... Preprocessing circuit, 4... Feature extraction circuit, 5
... Standard pattern memory, 6... Similarity calculation circuit, 7
...Identification circuit, 8. Circumscribed rectangle detection circuit, 9. Position comparison circuit, 10. Frame position memory.

Claims (1)

[Claims] 1. A sheet with a handwritten character entry frame printed on it is placed on the top of the tablet, characters are written one by one with an input pen in the character entry frame of the sheet, and based on the handwriting information of the input characters output from the tablet. ,
In an online handwritten character recognition device that recognizes written input characters and outputs the recognition result and position information of the frame in which the input characters are written, katakana characters are written in a predetermined area within the preset character entry frame. The online handwritten character recognition device recognizes whether the shape of a katakana character is similar to that of a katakana character, depending on the position of the circumscribed rectangle of the written character, the center position of the circumscribed rectangle, or the position of the center of gravity in the frame of the written character. A similar character identification method characterized by distinguishing between kanji and kanji.