JPS62260287A - On-line recognition system for handwritten character - Google Patents

Abstract

PURPOSE:To recognize even a handwritten character which projects from a character entry frame by finding the center and size of the handwritten character and segmenting the character when the handwritten character is inputted on a tablet. CONSTITUTION:When a character begins to be handwritten on the board surface of the tablet 1, the start and end points of each input stroke are decided by a coordinate position maximum/minimum decision circuit 2 in the form of X-Y coordinates. Thus, the X-Y coordinate positions of the start and end points of each stroke are decided and a character center position calculating circuit 3 calculates the center position of each stroke. This center position becomes data for determining the size of the character based on input strokes and a character area decision circuit 4 decides the area that the character occupies. When the final stroke is finished, the character area is determined and the input character is segmented 5; and the area is estimated rectangularly because Japanese characters has the property of a square.

Description

[Detailed Description of the Invention] [Summary] This is an online handwritten character recognition method that does not cause typos or unrecognizable characters even if the handwritten characters input from the input tablet surface extend beyond the character writing area. A method for correct recognition without reducing the recognition rate is shown.

[Industrial application field]

The present invention relates to a method for recognizing characters handwritten on the surface of an input tablet. More specifically, when handwriting characters into an input tablet, there is a condition that the characters must be written within a certain frame. This relates to a system in which handwritten characters can be recognized even if the constraints are relaxed and some protrusion occurs.

The online handwritten character recognition device is a device that instantly recognizes characters written on the keyboard and throat, and is particularly expected to be used in the field of Japanese input for word processors, personal computers, etc.

Even today, when Japanese word processors and personal computers are widespread, there are many people who are not accustomed to keyboard operations or who do not like keyboard operations, and online handwriting, which does not require much skill, is a character input method aimed at these people. Research into character recognition devices is becoming more active, and several types of devices have already been commercialized.

This method is also being considered for its convenience of allowing data to be entered into a computer at the same time it is written on a slip, or for learning proper kanji education.

However, I am acutely aware of the inconveniences of the current method.
The problem is that the frame for writing handwritten characters is limited, and since Japanese kanji in particular are characterized by a large number of strokes, it is extremely inconvenient to use a system that restricts writing within a certain frame.

Therefore, there is a strong demand for a system that allows for a flexible writing frame and that allows handwritten characters to be recognized even if the writing frame is slightly exceeded.

[Prior Art] Conventional online handwritten character recognition systems are subject to various restrictions upon input.

The most basic of these is the restriction that characters must be "written within a certain frame" on the input tablet surface.

This restriction is extremely inconvenient when it comes to Japanese characters, especially kanji, as many characters have a large number of strokes.

If characters are entered by handwriting while ignoring these restrictions, typographical errors will be recognized or cannot be recognized, resulting in a decrease in character recognition rate and an input error.

FIG. 3 shows the configuration of a conventional general online handwritten character recognition system.

31 is an input tablet notebook for inputting handwritten characters, and a character entry area 31a is divided on the input tablet.

When handwritten character input data is sent from the input tablet 31, a preprocessing unit 32 performs preprocessing such as normalizing the character size and character position.

33 is a character feature extraction unit that extracts the features of the strokes that make up the characters (the handwriting of the pen from when the pen is turned on to when it is turned off on the input tablet) from the preprocessed data, and extracts these characteristics from the preprocessed data. In the matching section 34, matching is performed with a dictionary 35 of characters.

Reference numeral 36 is a similar character identification unit, which is a system for identifying similar characters for characters that cannot be recognized by matching with the character dictionary 35 alone, such as similar character types such as "±" and "±". It is.

In this conventional method, a combination of strokes written in the character writing area 31a is determined to be one character, character cutting and other preprocessing are performed, and character recognition processing is performed.

Therefore, when writing handwritten characters, it is difficult to write characters with a large number of strokes, and if the character accidentally goes outside the character writing area 31a, it will be processed later as a different character, causing problems such as typos or unrecognizable characters, resulting in a decrease in the recognition rate. be.

[Problem that the invention seeks to solve]

In this way, the conventional handwritten character input method has a restriction that characters must be entered so that they fit one by one in the character entry area displayed on the input tab left, and if this restriction is ignored, The present invention, which causes input errors, was created in consideration of these points.
When entering handwritten characters from the input tab left, you are no longer strictly restricted to the character entry area as in the past.
The purpose is to provide a system with a high degree of freedom that can be recognized even if a small amount of overflow occurs.

[Means for solving problems]

FIG. 1 shows a principle block diagram of the online handwritten character recognition method of the present invention.

In FIG. 1, numeral 1 is an input tablet for inputting handwritten characters, and numeral 2 is a coordinate position maximum/minimum value determining circuit for determining the coordinate positions of the start and end points of the stroke of the character input from the input tablet 1.

3, when a character is input stroke by stroke from the coordinate position of the stroke determined by the coordinate position maximum/minimum value determination circuit 2,
This is a character center position calculation circuit that calculates the center position each time.

4 is a character area determination circuit that receives handwritten character data from the character center position calculation circuit 3 and the coordinate position maximum/minimum value determination circuit 2 and determines the size of the handwritten character input to the input tablet 1.

Reference numeral 5 indicates output data of the character area determination circuit 4, which is a character cutting signal which is pre-processing for character recognition.

The above configuration is a method for recognizing characters handwritten on the input tablet l.

[Effect]

When characters begin to be handwritten on the input tablet surface, the coordinate position maximum and minimum value determination circuit 2 is activated for each input stroke.
Accordingly, the start point and end point of the stroke are determined as XY coordinate positions.

Once the XY coordinate positions of the start and end points of the stroke are determined, the character center position calculation circuit 3 calculates the center position for each stroke.This center position is calculated based on the input stroke.
This data is used to determine the size of the character (the size of the character and the spread of the character from the center position), and based on this data, the character area determination circuit 3 determines the area occupied by the character.

The area occupied by a character is the area occupied by the size of the input character when the input character is finished, based on the input stroke. When the final stroke is input, the area of the character is determined and the input character is Used as a cutout signal.

This area is expected to be square, because of the principle that Japanese regular typefaces are designed as squares.

〔Example〕

FIG. 2 shows an embodiment of the present invention, in which an input tablet 1
The user sets a character entry area frame 1a on the board surface, which serves as a guide for the size of handwritten characters. However, this character entry area frame 1a does not strictly restrict the entry of characters, and if characters of extremely different sizes are input, it may affect the time and accuracy of recognition processing for one input character. Therefore, it is only set as a guideline.

The coordinate position maximum/minimum value determination circuit 2 uses a Xm1n+Y@in coordinate format coordinate determination table 2 araax, Ytrrax coordinate determination unit 2b, which represents the starting point and end point position of a stroke, and a character stroke determined by these coordinate determination units. A circuit including a memory circuit 2c that stores data on coordinate positions is used.

The character center position calculation circuit 3 is a circuit composed of an X-direction center position calculation section 3a and a Y-direction center position calculation section 3b for strokes of characters that are input one after another.

The character area determination circuit 4 determines the area of the size occupied by the handwritten character based on the data of the coordinate position and center position of the stroke. This data is used as a character cutting signal.

FIG. 4 illustrates the method for obtaining the above-mentioned character cutting signal. When the upper horizontal stroke of the kanji character ``earth'' is written on the surface of the input tablet 1, the Y shown in FIG.
The coordinate positions of a+in, Yzax, Xm1n, and Xa+ax are determined, and based on these coordinate positions, the center position 42 of the character is determined.
A is required.

The spread of characters 43 in the left and right direction from this center position 42A
If the length of A is r, then according to the principle that Chinese characters are designed within a square, the length of the vertical spread of the characters is also r, which determines the character area 41A.

Next, a vertical stroke of "±" is written, and it extends beyond the character area 41A determined by the initial stroke, as shown in FIG. Find the position, the spread of the character 43B is r + X, and from the center position 42B of the character in the vertical and horizontal directions
A character area 41B having a size of is obtained.

After that, the bottom horizontal stroke of "±" is written as shown in FIG.
+x, a character center position 42C and a character area 41C are obtained.

This completes the input of the kanji "±", and one side of the character area of "±" written on the input tablet is determined to be 2 (r + x + x) by the character area determination circuit 4, and in this way, the character is cut out. I get a signal.

In parallel with this process, stroke data of handwritten characters input from the input tablet 1 is input to the stroke data buffer 6.

- When half of the characters have been input, the stroke similarity calculation circuit 7 reads the stroke data stored in the stroke data buffer 6 under the control of the control circuit 8 and performs preprocessing using the character cutout signal described above.

The stroke data for which the preprocessing has been completed is stored in the stroke information storage section 9, and the recognition section 10 stores the stroke data in the character dictionary 1.
1, handwritten input characters are recognized. The processing method after this pre-processing is performed in the same manner as the conventional method.

〔Effect of the invention〕

According to the present invention, when inputting handwritten characters from the input tablet surface, even if the characters are written outside the character entry frame, the character recognition function is not affected, and it is possible to input handwritten characters with a large number of strokes. This method is extremely useful in practical terms.

[Brief Description of the Drawings] Fig. 1 is a principle block diagram of the online handwritten character recognition method of the present invention, Fig. 2 is a diagram of an embodiment of the present invention, Fig. 3 is an explanatory diagram of the conventional method, Fig. 4 is FIG. 3 is an explanatory diagram of a method for obtaining a cutout signal for a character area. In FIG. 1, 2 is an input tablet, 2 is a coordinate value 1 maximum/minimum value determination circuit, 3 is a character center position calculation circuit, and 4 is a character area determination circuit. 1 input 7-ray-1- Half-invention ~ Tori 7'l17η7m 1st illustration 7'l17η7m j Tomoe Lugo Figure 4

Claims (1)

[Scope of Claims] A coordinate position maximum/minimum value determination circuit (2) that determines the coordinate position for each stroke of a handwritten character input from the input tablet (1); and the coordinate position maximum/minimum value determination circuit. A character center position calculation circuit (3) calculates the center position of the coordinate position for each stroke of the character determined in (2), and a character center position calculation circuit (3) calculates the center position of the coordinate position determined by the coordinate position maximum/minimum value determination circuit (2). A character area determination circuit (4) is provided that determines the size of the handwritten character from the center position determined by the character center position calculation circuit (3), and the character area determination circuit (4) determines the size of the character determined by the character area determination circuit (4). An online handwritten character recognition method characterized by cutting out and recognizing handwritten characters input from the input tablet (1) above using data from the input tablet.