JPS5975374A - Character recognizer - Google Patents

Abstract

PURPOSE:To recognize a character in real time and at a high speed, by detecting the writing start and finish of a character and the moving order of a stylus pen within the writing of a character and then extracting and dividing the features of a character segment on the basis of the length and direction of the segment. CONSTITUTION:The paper 2 is put on a tablet 1, and characters are written on the paper 2 with a stylus pen 3. Thus a coordinate reading circuit 4 detects coordinates X, Y and Z of the pen 3 with a fixed time interval. The coordinates X and Y show the position of the pen 3, and the coordinates Z shows 1 and 0 when the pen 3 presses the paper 2 and the pen 3 is detached from the paper 2 respectively. Therefore a character section detecting circuit 5 detects the writing start and finish of a character and store them in a character coordinate memory 6. A normalizing circuit 9 normalizes the size of the character given from the memory 6. A Z coordinate differentiating circuit 12 detects the change of strokes, and then the direction and length of the segment are detected by X and Y coordinate differential circuits 15 and 18, a segment direction detecting circuit 19 and a segment length detecting circuit 21. Then the features of segments are extracted 33 from the emerging order of the segments together with the features of strokes. These features of segments and strokes are combined to recognize characters through a recognizing circuit 30.

Description

[Detailed Description of the Invention] +1) Background Art The present invention is an online character recognition system in which when characters are written using a pen on a prescribed tablet, the coordinates of the pen are imported into a data processing device at regular time intervals and the characters are recognized. Regarding equipment. Most of the conventional online character recognition devices connect a tablet to a general-purpose computer and use pre-prepared general-purpose computer software to recognize characters, which makes the data processing device expensive and takes a long time to recognize characters. The drawback was that it took time. On the other hand, it is possible to use a microcomputer in the data processing section, but it was thought that there would be problems with processing performance.

(2) Disclosure of the Invention According to the character recognition device according to the present invention, it is possible to detect the beginning and end of writing a character and the order in which the pen moves between them, and also to know the length and direction of the line segments of the character. By extracting and classifying the features of line segments, real-time character recognition becomes possible within an extremely short time with relatively simple data processing.

(3) Best Mode for Carrying Out the Invention FIG. 1 is a diagram showing the configuration of the present invention. A sheet of paper 2 is placed on a tablet 1 and characters are written with a stylus pen 3. A coordinate reading circuit 4 detects the coordinates X, Y, and Z of the stylus pen 3 at regular time intervals. The coordinates X and Y are numerical values indicating the position of the stylus pen 3, and the coordinate Z is the position where the stylus pen 3 is located on the tablet 1 (
Alternatively, the coordinates are indicated by a signal that distinguishes whether or not the top of paper 2) is pressed.

The tablet 1, stylus pen 3, and coordinate reading circuit 4 are the same as conventional ones, and for example, a commercially available one is DT-4000 manufactured by Daini Seikokin.

The character delimiter detection circuit 5 detects the coordinate x detected by the coordinate reading circuit 4.
, y, and z to detect from the beginning to the end of one character. For example, if the stylus pen 3 is pressing on the tablet 1, the coordinate Z' = 1,
This will be explained assuming that the coordinate Z=0 when the stylus pen 3 does not press the tablet 1.

The time when 2 = 0 changes to Z = 1 is considered as the start of writing one character, and the time when it changes from Z-1 to Z20 and the time in the z-00 state has passed for a predetermined time or more is considered as 1. This marks the end of the writing. Another method is to determine whether the X and Y coordinate values when Z = 1 are within a predetermined size range of the character, if the size of the character is determined in advance. The case may be the end of a character.

The character coordinate memory 6 captures and stores the coordinates x, y, and z from the start of writing to the end of one character.

The x, Y maximum/minimum detection circuit 7 determines the maximum and minimum values of the X and Y coordinates from the coordinates of one character stored in the character coordinate memory 6, and sends them to the character size calculation circuit 8.

The character size calculation circuit 8 calculates the maximum value of the X coordinate -Xmax.

Minimum value Ymin, maximum value of Y coordinate Ymax, Y
Find the font size from min.

Xmax-Xmin=A Ymax-Ymin=BA
, B are the respective character sizes in the X and Y directions.

The normalization circuit 9 converts the coordinates in order to make the written characters constant. Since the thickness of characters varies depending on whether they are written by different people or even when written by the same person, recognition processing is simplified by keeping the size of the characters constant.

For example, let xt, yt, zt be the coordinates of an arbitrary point in the imported character coordinate memory 6 that converts the character size to a constant size C in both the x and y directions, and then use the normalization circuit 9 to perform the coordinate transformation. The coordinates after this are x t', y t'.

If zt' Xt' = (Xt - Xmi n) -.

Now, even if the character size is different, it is converted to a constant size. Furthermore, since Xt s Yt is an absolute coordinate, the coordinate y! which has been coordinate-converted by the normalization circuit 9.
t and ft are relative coordinates with Xm1n and Ymin as the origin.

Hereinafter, the coordinates after coordinate transformation by the normalization circuit 9 will be referred to as normalized coordinates. The normalization circuit 9 includes a Z coordinate register 10, an X coordinate register 13 . The normalized coordinates converted by the normalization circuit 9 are stored in the Y coordinate register 16.

x't, Qt are stored. The coordinates xt, yt, zt captured at regular time intervals are stored in the character coordinate memory 6, and these coordinates are sequentially converted into normalized coordinates X't, Y't, z't by the normalization circuit 9. is,
X coordinate register 13. The Y coordinate register 16 is stored in the Z coordinate register.

Note that the Z coordinate register 10. Immediately before storing the normalized coordinates converted by the normalization circuit 9, the X-coordinate register 13, Y-coordinate register 16 stores the respective contents in the two coordinate registers 11. The X coordinate register 14 and the Y coordinate register 17 are set to 7, and then the normalized coordinates are stored. This results in
The z, x, X coordinate registers 11, 14.17 contain z, x
, X coordinates, and the normalized coordinates captured immediately before the normalized coordinates stored in the register 10.13.16 are stored.

The X coordinate differentiating circuit 12 examines the change in the X coordinate from the contents of the Z coordinate registers IO and 11. For example, from Z=1 to Z−
A series of 0 or Z=1, or from Z=0 to Z=1,
Alternatively, it is detected whether Z=0 is continuous. If the contents of the X coordinate register 11 are 1 and the contents of the X coordinate register 10 are 0,
If it is 0, it means that Z=0 has changed from 2-1.

Z, X, Y coordinate registers 10.13.16 and 11°14
The coordinates represented by , , and l7 will be explained as two points.

The X-coordinate difference circuit 15 calculates the amount of change in the X-coordinate between two sampling points, and the X-coordinate register 13.14
Let the content of be X't+ X't-x, then a-shield-
Calculate X't-1.

Similarly, the Y-coordinate difference circuit 18 calculates the amount of change in the Y-coordinate between two points, and if the contents of the Y-coordinate register 16 and 17 are Y't+ Y't-x, then b = Y't-
Calculate Y't-1.

The moving direction of the stylus pen 8 and the length of the line segment between two points are determined and recognized from the above α and b.

The line segment direction detection circuit 19 determines the direction of the line segment between the two sampled points, that is, the direction of movement of the pen.If the direction of the line segment is θ, then θ is determined by tznθ=−.

Here, in order to simplify the recognition method, θ is divided into eight types as shown in FIG. 2, for example. It would be possible to classify more accurately, such as 16 types, 32 types, etc., but recognition would be complicated, so in the embodiment, 8 types are used.

The division value of θ from the line segment direction detection circuit 19 is sent to the line segment direction register 20 and the direction differentiation circuit 24, and is stored in the line segment direction register 20.

The line segment direction detection register 20 sends the contents of the line segment direction detection register 20 to the direction differentiation circuit 24 before storing the segment value of θ from the line segment direction detection circuit 19. Changes in the θ segment values can be known from the θ segment values sent from the segment direction detection circuit 19 and the line segment direction register.

Next, a method for determining the length l of a line segment between two points will be explained.

The line segment length detection circuit 21 determines the length l of a line segment between two points, and calculates it using the following equation.

l Heaven 7 The comparator circuit 22 uses the line segment length l as a predetermined constant e.
(e is determined from the font size, etc.) Values above are valid, and values less than 0 are invalid. This is a minute amount of movement of the stylus pen 3 due to hand vibration, etc., and is treated as noise and removed. If this l is less than 0, 2 coordinate register 11, X coordinate register 14
, leave the contents of the Y-coordinate register 17 unchanged, set the normalized coordinates of the next sampling point to the X-coordinate register 10, X-coordinate register 13, and Y-coordinate register 16, and calculate Calculate l.

When the length l of the line segment is 8 or more, the length l of the line segment is sent to the distance addition circuit 28. Directional differentiation circuit 24. indicates that the line segment direction is the same based on the division value of the line segment direction θ sent from the line segment direction detection circuit 19 and the division value of the line segment direction θ obtained at the previous sampling point from the line segment direction register 20. Determine whether it is or is different. If the division values of θ are the same, a signal is sent to the distance addition circuit 28. The distance addition circuit 28 adds the length l of the line segment sent from the comparison circuit 22 and stores it in the distance addition circuit 23. If the division value of the line segment direction θ sent from the line segment direction detection circuit 19 is different from the division value of the line segment direction θ sent from the line segment direction register 20, different signals are sent to the distance addition circuit 23. The added value of the line segment length l that was added to the distance addition circuit 28 is sent to the line segment length classification circuit 26, and the line segment length l sent from the comparison circuit 22 is newly sent to the distance addition circuit 23. Seven times.

As a result, when the same division value of θ continues,
Since the total length of the line segments is obtained, it indicates the length of the line segments until the segment value of θ changes.

The line segment length division circuit 26 classifies the added value of the line segment length l sent from the distance addition circuit 28 into a plurality of values, thereby simplifying the recognition process.

For example, in the example, the font size is converted to C, so the added value of the length l of the line segment is less than one of C. It is divided.

The line segment feature extraction circuit 33 extracts a line segment from the segment value of the line segment direction θ sent from the line segment direction register 2o and the segment value of the added value of the line segment length l sent from the line segment length segment circuit 26. Extract the features of

The division value of the line segment direction θ is θ4. Let la be the division value of the added value of the length l of the line segment. As for the number of divisions, θα is 8 divisions as shown in FIG. 2, and la is 4 divisions as described above.

For example, the number "5" as shown in Figure 8 is "5" on the first stroke.
, and the second stroke in the order of "-".

Figure 4 shows the character pattern obtained by converting the coordinates of the stylus pen 3 when writing characters as shown in Figure 3, determining θ and l from the normalized coordinates, and dividing them into segment values θa and lα. It shows.

Figure 4 (a) is the "5" part of the first car, and (b)
is the "-" part of the second car. The separation between the first car and the second car is done as follows. Z=1 to 2=0
When it changes to Z, it is considered that the first car has finished, and then Z
It is assumed that the second car starts when Z = O changes to Z = 1. Similarly, the second car and the third car, the eighth car and the fourth car
This is what separates the eyes from each other, and the same applies hereafter.

Characteristics of each stroke are determined from the appearance order of segment values θa and la for each stroke, and characters written with the stylus pen 3 are recognized by combining these characteristics.

However, when determining the characteristics of each stroke from the order of appearance of the segment values θα and la, there is no problem if the character consists of only straight lines, but if the character consists of curved lines as shown in Figures 8 and 4, Sometimes, there are so many combinations of segmentation values θα and la that it is not possible to easily extract features for each surface. Even in the first car of the number "5", depending on the person writing it, θα, l as shown in Figure 5 (a) or (b).
Since α appears and is different from the case shown in FIG. 4, it becomes very complicated when extracting the features of the first two eyes. Therefore, in the present invention, the line segment feature extraction circuit 38 detects straight lines and curves based on θa+la and then extracts the features of the image, so that the features of the image can be easily extracted. In the embodiment, the lines are divided into 8 types of straight lines shown in FIG. 2 and 8 types of curves shown in FIG. 6.

The division of straight lines and curves (hereinafter referred to as feature division of line segments) output by the line segment feature extraction circuit 88 is represented by P (IJ).

(2) indicates the division θα of the straight line direction (FIG. 2) and the division of the curve (see FIG. 6), and J indicates the division h of the length of the line segment. In the case of a curve, J may be expressed by dividing it by the size of the curve, but in the embodiment, J is simplified by not dividing it and J=0.

For example, the extraction of P (IJ) in Figure 4 (a) is P (62)
, P (80).

The value becomes P(90), and the value shown in FIG. 4(b) becomes P(O8).

Also, Fig. 5(a) shows P(62), P'(80),
P(90), (b) is P(62), p(s
O), P(90), P(AO), and Fig. 4(
α), the first two eyes "5" of the number "5" in FIGS. 5(a) and (b) can be extracted as being very similar, making it easy to extract features for each subsequent stroke.

Feature extraction for each image will be described below as image feature extraction.

Next, image feature extraction will be explained.

One stroke is from Z=1 to Z-0, and features are extracted for each stroke.

The read-only storage device consists of a read-only storage device 27 and a register 28, and uses the contents of the register 28 and the contents IJ of the line segment feature classification P(IJ) output from the line segment feature extraction circuit 33 as an address. The data is read from 27 and set in register 28. It is read out from the read-only storage device 27 and sent to the register 28 at the timing of the output signal from the AND gate 29 between the output signal of the line segment feature extraction circuit 33, which is output when the feature of the line segment changes, and the timing signal CKI. Since it is set, it will be read when the direction of the line segment changes. Furthermore, register 28
Since the content of is also read from the read-only storage device 27 as an address, the state in which the line segment direction has changed is set in the register 28.

For example, if you write the number ``5'' as shown in Figure 8, the direction segment θα of the line segment as shown in Figure 4 (α), (b),
Suppose that the line segment length division lα is obtained in the first stroke and the second stroke, and as mentioned above, the line segment feature division (IJ) is obtained in the first stroke.
Suppose that P(62), P(80), P(90) are obtained in the first picture, and P(O3) is obtained in the second picture.

FIG. 7 shows a part of a state transition diagram for image feature extraction, which consists of the read-only storage device 27 and the register 28.

SOO-S20 in the circle means states SOO-S20, and 00-20 are read from the read-only storage device 27 and stored in the register 28. Also, P near the arrow (
P(IJ) such as 00) indicates the feature classification P(IJ) of the line segment output from the line segment feature extraction circuit 83.

For example, when the feature segment P(62) of the line segment is extracted in the initial state SOO, this means that the state changes to SO2. In other words, the register 28 has the content 00, and when P(62) is extracted, the read-only storage device 28 uses the numbers indicated by -100 and 62 as addresses.
The content read from 7 is 02 corresponding to the state of SO2, and 02 is stored in register 28. Next, when the feature segment p(so) of the line segment is extracted, the read-only storage device 27 is read using the numerical values indicated by the contents 02 and 80 of the register 28 as addresses.
8 is read and 08 is stored in register 28. Similarly, when the feature segment P(:90) of the line segment is extracted, 15 corresponding to the state S15 is read from the read-only storage device 27 and stored in the register 28. When the first stroke is completed, the feature of the first stroke is 1 stored in the register 28.
It is 515 corresponding to 5.

Next, the second stroke is θα-0 as shown in the fourth figure (b). lα=
3 is obtained, the feature classification P'(IJ) of the line segment
P(O3) is extracted, and the feature of the second image is S20.

Note that the diagram inside the circle in FIG. 7 shows a schematic diagram of the configuration of line segments corresponding to state S, and S15 means the shape of "5". Also, SO6 has a “2” shape, and S
12 is in the shape of a "2".

Furthermore, the characters written by combining the characteristics of each stroke are recognized. The recognition circuit 80 in FIG. 1 recognizes the relative position of the coordinates of the starting point or ending point of each line sent from the relative position detection circuit 25 in combination with the image characteristics.

The recognition circuit 30 combines image features for each stroke using a signal sent from the Z coordinate differentiating circuit 12 when the coordinate Z changes, that is, from Z=1 to 2=0 or from Z=O to Z=1. . For example, in the case of FIG. 4(α)(b) described above, S15 and 520 are combined to result in the recognition of the number "5", and the signal CH is output. If it is a character with 8 strokes, the result of the combination of 1 stroke and 2 strokes is combined with the special i of 3 strokes, and the result becomes the recognition result.

(4. Brief explanation of the drawings) Figure 1 shows the configuration of the present invention, Figure 2 shows an example of line segment direction division, Figure 8 shows an example of characters to be recognized, Figures 4 (α), (b )
5(a) and 5(b) are diagrams showing the division values for each stroke of a character pattern in the order of their appearance, FIG. 6 is a diagram showing divisions of curved lines of characters, etc., and FIG. FIG. 3 is a diagram showing state transition of feature extraction.

l is tablet 2 is paper 8 is a stylus pen 4 is the coordinate reading circuit 5 is character delimiter detection circuit 6 is character coordinate memory 7 is the XY maximum maximum minimum detection circuit 8 is a character size calculation circuit 9 is a normalization circuit 10 is 2 coordinate register 11 is 2 coordinate register 12 is the Z coordinate differential circuit 13 is the X coordinate register ■4 is the X coordinate register 15 is the X coordinate difference circuit 16 is the Y coordinate register 17 is the Y coordinate register 18 is the Y coordinate difference circuit 19 is a line segment direction detection circuit 20 is the line segment direction register 21 is a line segment length detection circuit 22 is a comparison circuit 23 is a distance addition circuit 24 is a directional differentiation circuit 25 is a relative position detection circuit 26 is a line segment length division circuit 27 is a read-only storage device 28 is a register 29 is and gate 30 is recognition circuit

Claims (2)

[Claims] (1) By capturing the coordinates X, Y, and Z of the pen at regular intervals, a character recognition device that recognizes characters, symbols, etc. written with a pen detects the beginning and end of writing one character. means for normalizing the size of characters; means for detecting changes in stroke from changes in coordinate 2;
means for detecting the direction and length of a line segment from coordinates X and Y;
A means for extracting line segment features from the order of appearance of line segment directions and lengths, a means for extracting features for each stroke from the appearance order of line segment features, and recognition of characters, etc. from a combination of stroke features. A character recognition device characterized by comprising means for: (2) A means for detecting the length and direction of a line segment from the coordinates 2. The character recognition device according to claim 1, further comprising means for extracting and classifying minute features.