JPH0348985A - Handwritten character recognizing device - Google Patents

Abstract

PURPOSE:To improve a recognition rate by generating a stroke composed of a coordinate value, correcting and recognizing the change of a vector value composed of the sequence of an input point near each starting point and ending point. CONSTITUTION:A processing control part 2 recognizes a character from the sequence of position coordinates, which are inputted to a coordinate input part 1, and displays the character in a display part 3. A stroke data generating part 2b of the control part 2 generates the sequence of a directional value composed of two points from the sequence of the coordinate value included in one stroke. According to the number of all the directional values in the stroke, the directional value train of an arbitrarily set rate is checked and the change of the vector value composed of the sequence near each start point and end point is investigated. When there is the change, a storke data correcting part 2c deletes the coordinate value up to the change point from the data and corrects the data. Next, when a recognition processing part 2d makes recognition by the corrected stroke data, the recognition rate of the character can be improved without fail.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Field of Application) The present invention relates to a handwritten character recognition device that recognizes two-dimensional patterns such as handwritten characters and figures and uses a data processing device or the like manually.

(Prior art) As an input means for a data processing device in place of a keyboard,
Enter handwritten characters and shapes using a tablet, etc.
There is an online handwritten character recognition technology that detects stroke information consisting of a sequence of coordinate values input in real time and recognizes characters and figures.

In the case of this device, the character input means replaces the conventional means such as paper and pen or pencil, and input is performed by writing on the input surface of the tablet device using a pen equipped with a switch and a detection device at the tip. This tends to cause wobbling, and when writing particularly quickly, so-called ``splashes'' may occur, where the end of the stroke is input in an unexpected direction, and the text may not be recognized as a character.

(Problems to be Solved by the Invention) As described above, in conventional handwritten character recognition devices, characters may not be recognized or may not be recognized erroneously due to "splashes" near the start and end points of handwritten stroke data. This method has the disadvantage that characters are often recognized incorrectly, and the recognition rate of characters decreases.

The present invention was made in view of the above-mentioned circumstances, and it is a handwritten character that can improve the recognition rate of characters by correcting the "splashes" near each start point and end point of input stroke data. The purpose is to provide a recognition device.

[Structure of the Invention] (Means and Effects for Solving the Problems) The present invention generates stroke data consisting of coordinate values manually input by a position manual means, and generates stroke data from a series of input points near each start point and end point. This method corrects large changes in vector values, and recognizes character patterns using the corrected stroke data.

(Function) By doing the above, even if fast handwritten input is performed, the input characters will not be erroneously recognized, and the character recognition rate can be reliably improved.

(Example) An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 is a diagram showing a schematic circuit configuration thereof.

The handwritten character recognition device shown in the figure employs an online character recognition method, and basically consists of a coordinate input unit l such as a tablet that sequentially inputs the positional coordinates indicating the stroke of the character pattern of the handwritten characters, and A processing control unit 2 that recognizes characters from a series of positional coordinates input by hand through the input unit 1, and a display consisting of, for example, a liquid crystal display, which displays character information etc. recognized and input by the processing control unit 2. It consists of Section 3.

Here, the coordinate input section 1 and the display section 3 have the same dimensions and the same coordinate plane, and as shown in FIG. and configured. Therefore, the information displayed on the display section 3 can be visually recognized through the coordinate input section I consisting of a transparent tablet.

Note that the processing control section 2 includes an input control section 2 as shown in FIG.
a, a stroke data generation section 2b, a stroke data correction section 2e, a recognition processing section 2d, a display control section 2e, a stroke data buffer 2r1 output buffer 72g, a recognition dictionary 2h, and the like.

Here, the transparent tablet which is the coordinate input section 1 will be explained. The coordinate input section I can be realized, for example, as an electrostatic transparent tablet as shown in FIG. As shown in FIG. 4(a), this electrostatic type transparent tablet has a plurality of transparent strips 71t5i1c on one side,
A pair of transparent glass plates 1a each having glass plates 1d arranged in parallel at predetermined intervals.

The transparent electrode Lc and the transparent electrode Id are made to face each other in directions orthogonal to each other and are bonded to each other via an insulating protective film (not shown) to form a tablet surface. The tablet has a shift controller le as shown in Fig. 4(b).
Shift registers if', Ig operating under the control of
are connected, thereby the transparent electrodes 1c, l
A predetermined voltage is sequentially applied to d, and the entire area of the tablet surface is scanned as the intersection of the transparent electrodes 1c and ld to which the voltage is applied.

Specifically, a predetermined voltage is selectively applied to one of the transparent electrode ICs in order, and during a period when the predetermined voltage is applied to one of the transparent electrode ICs, the other transparent electrode 1d perpendicular to the predetermined voltage is applied. This is done by applying a predetermined voltage to all of them in sequence. That is, by sequentially applying a predetermined voltage to all of the other transparent electrodes 1d and shifting the transparent electrodes 1c to which a predetermined voltage is applied every time the transparent electrodes 1d are scanned, the matrix constituted by the transparent type title, ld is The entire plane (tablet surface) is sequentially scanned.

The counter 1h operates in synchronization with such scanning, and its count value represents the position coordinates on the matrix plane.

Therefore, the matrix plane (
The ben 11, which indicates the tablet surface), detects the voltage applied by scanning the transparent electrodes 1c and ld described above at the indicated position by electrostatic coupling. The scanning signal detected by this pen 11 is amplified via a preamplifier 1j, and then its noise component is filtered out via a filter (FIL) lk, and is applied to the counter lh as an indicated position detection signal. It will be done.

The counter lh receives this indicated position detection signal and outputs the counted value at that time as information on the indicated position on the tablet surface indicated by the pen II.

According to the coordinate input unit 1 configured in this way, the pen 11
By specifying an arbitrary position on the tablet surface, the pen 11 detects a scanning signal (applied voltage) that scans the transparent electrode 1cSld at the specified position. Then, the calculated value of the counter 1h at the time of detection of the scanning signal is detected as information on the positional coordinates corresponding to the indicated position. and the pen 11
When a character pattern is written on the tablet surface using the pen II, information indicating the writing stroke is input as a connection of positions indicated by the pen II, that is, a connection of indicated position coordinates (series of position coordinates). It turns out.

Information on the character pattern input by hand through the coordinate input unit l in this manner, that is, a sequence of position coordinates, is provided to the display unit 3 via the processing control unit 2. According to the series of position coordinates, the same position coordinates on the display unit 3 are driven to be displayed as described later. Through this display control, the character pattern input on the tablet surface is simultaneously displayed as the trajectory of the handwriting, making it appear as if the character information is being input as if writing characters on paper using a pen. ing.

Here, the procedure for character recognition based on handwritten input will be explained according to the processing procedure shown in FIG.

First, a written input is performed using the pen 11 on the coordinate input section l, and it is determined whether the switch of the pen tip is turned on or not based on the count value of the counter 1h (step 5a).
.

If the pen 11 is ON, the coordinate value data is considered to be data within one stroke, and the detected coordinate value data is stored in the stroke data buffer 2f. (Step 5b).

Thereafter, the display control unit 2e simultaneously displays the coordinate value data as information on the writing pattern on the display unit 3 (step 5c).

Furthermore, if the pen 11 is OFF, it is assumed that one stroke has ended and the next stroke is started, and a delimiter code is stored in the stroke data buffer 2f (step 5).
d).

At this time, if the start position of the stroke is within a certain value (distance M) from the previously stored stroke position data, this stroke is also determined to be a stroke that constitutes the same character. In step 5e), the coordinate value data of the stroke following the previous stroke is stored in the stroke data buffer 2r.

On the other hand, if the interval between strokes is more than a certain value, it is determined that the stroke is the beginning of writing a new character (step 5e), and the previous coordinate value string data is used to create the stroke data of one character sentence. . The stroke data correction unit 2c examines the start point and end point of each input stroke data as shown in FIG. 6(a), examines remarkable changes in the series of coordinate data of the stroke data, ), the changed portions as indicated by the broken lines in the circles indicated by A and B are deleted as correction processing (step 5r).

Hereinafter, this correction of stroke data will be explained in detail using FIG. 8.

First, at the beginning of the operation, a series of direction values between two points as shown in FIG. 7 is generated from a series of coordinate values included in one stroke (step 8a).

In the series of direction values generated in this way, first, from the starting point, check the direction value sequence that has a set arbitrary proportion, for example, 5%, of the number of all direction values in the stroke,
Check whether there is a value that changes by 90 degrees or more in the direction value sequence (step 8b).

Step 8c) If there is a change, the coordinate values up to the point where the change occurred are deleted from the stroke data (Step 8d).

The same process as above is executed for the end point of the stroke, and the points after the point where the change occurs are deleted (steps 8e to 8e).
Step 8g).

In this way, the stroke data correction unit 2c deletes the unexpected input points that occurred at the start point and the end point, and after the obtained corrected stroke data is stored again in the stroke data buffer 2f, this stroke data is used. Subsequently, the process returns to the process shown in FIG. 5, and the recognition processing section 2d performs character recognition.

That is, the recognition processing unit 2d analyzes the data in the stroke data buffer 2'' as shown in FIGS. 9 and 10, first examines the delimiter data, and determines the number of strokes.Next, for each stroke, From each coordinate string data that makes up the stroke, find the direction value between two points and create direction string data.Furthermore, search this direction string data, and if the same direction value continues, The direction value string data is compressed by representing the direction value.

Here, the above-mentioned direction values are quantized into eight directions (direction values) and given, for example, as shown in FIG. 10(b).

According to this quantization, for example, the number "3" is
), it is converted into a direction value sequence r252357J.

Such processing is executed for each stroke data. Then, information on the direction value sequence obtained and the recognition dictionary 2
The recognition processing unit 2d collates h and the input character is recognized. As shown in FIG. 9, the recognition dictionary 2h used for this recognition process expresses the characteristics of the strokes constituting the characters to be recognized as direction value sequences, and classifies them according to the number of strokes constituting each character. It is configured in association with each character code.

The recognition processing unit 2d reads the direction string information D2 and the character code D3 corresponding to the number of strokes from the recognition dictionary 2h, stores them in a temporary buffer (not shown) provided in the recognition processing unit 2d, and reads out the created direction value string. The data is compared with the direction column information of this - time buffer to see if there is a match, and if there is a match, the character code is extracted and written to the output buffer 2g. At this point, the characters are recognized. Even when there is no exact match, if there is a match within the allowable range, that character code is extracted (step 5g).

Based on the character code obtained by the recognition processing section 2d and stored in the output buffer 2g, the display control section 2e displays a corresponding character pattern on the display section 3.

In the manner described above, characters can be recognized from input strokes.

It should be noted that the present invention is not limited to the embodiments described above; for example, the coordinate manual unit 1 may be a pressure-sensitive tablet or an opaque tablet. Similarly, in the present invention, the coordinate input means and the display means have been described as being integrated, but a coordinate input means such as a mouse or the like that can input the position of the amount of movement may be used. The display section 3 may be configured by a plasma display or the like.

Also, in the stroke data correction means, points recognized as "splashes" are deleted, but the direction value may be corrected by going back to the direction of movement of the pen.

Further, the modification range can also be changed within a range that does not cause any trouble to the recognition processing section 2d. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, stroke data consisting of coordinate values input by the position input means is generated, and vector values consisting of a series of input points near each start point and end point are generated. By correcting large changes and recognizing character patterns using the corrected stroke data, the entered characters will not be misrecognized even when inputting quickly by hand, ensuring a reliable character recognition rate. It is possible to provide a handwritten character recognition device that can improve the performance of handwritten characters.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing the configuration of a processing control section according to an embodiment of the present invention, Fig. 2 is a block diagram showing a schematic circuit configuration, Fig. 3 is a perspective view showing the structure, and Fig. 4 is a coordinate input section. Figure 5 is a flowchart showing the overall flow of character recognition.
FIG. 6 is a diagram showing the contents of correction of stroke data, FIG. 7 is a diagram showing direction value string data created from coordinate value string data, FIG. 8 is a flowchart showing stroke data correction processing, and FIG. Diagram showing an example of the configuration of a recognition dictionary, No. 10
The figure is a diagram showing an example of expressing the direction of a writing stroke. DESCRIPTION OF SYMBOLS 1... Transparent tablet (coordinate input part), 2... Processing control part, 2a... Input control part, 2b... Stroke data generation part, 2C... Stroke data correction part, 2
d... Recognition processing unit, 2e... Display control unit, 2''...
- Stroke data buffer, 2g...output buffer, 2h...recognition dictionary, 3...liquid crystal display (display section).

Claims (1)

[Claims] A position input means for inputting coordinate values corresponding to a designated position; a generation means for generating stroke data consisting of the coordinate values input by the position input means; and a stroke generated by the generation means. A correction means for correcting a large change in a vector value consisting of a series of input points near each start point and end point of data, and a recognition means for recognizing a character pattern using stroke data corrected by the correction means. A handwritten character recognition device characterized by: