JPH04337890A - Device for recognizing on-line handwritten character - Google Patents

Abstract

PURPOSE:To improve a character recognition rate in an on-line handwritten character recognizing device. CONSTITUTION:When a user inputs a character on a tablet 1, coordinate data of a pen and ON/OFF information of the pen are transmitted to a preceding processing part 2 and based on this, respective input strokes are expressed by a direction code through the use of an eight direction code reference coordinate system so that an input pattern is generated in the on-line handwritten character recognizing device. A stroke recognizing part 5 collates the input pattern with a basic stroke dictionary and a character recognizing part 7 recognizes the inputted character based on this so as to output it to a display part 9. When the inputted character is correctly recognized, a writing habit detecting part 10 corrects the direction of the eight direction code reference coordinate system which is added in the preceding processing part 2 and becomes reference.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an online handwritten character recognition device, and more particularly to an online handwritten character recognition device that can improve the character recognition rate.

0002

[Prior art] Characters input from a tablet are recognized stroke by stroke using the coordinate points of the pen and the time series of ON/OFF of the pen, and a character dictionary in which characters are written in advance with strokes along the handwriting. For example, word processors, personal computers, and the like are examples of devices that are configured to input characters by comparing them with the characters.

[0003] For details of such a handwritten character recognition device, see, for example, "Online Handwritten Kanji/Hiragana Recognition".
(IEICE technical research report PRL8237).

0004

[Problems to be Solved by the Invention] In the conventional online handwritten character recognition apparatus, each scribe has a unique writing habit, which may reduce the recognition rate.

[0005] This invention was made to solve the above-mentioned problems, and an object thereof is to provide an online handwritten character recognition device that can learn the writing habits of scribes and improve the recognition rate. .

[0006]

[Means for Solving the Problems] An online handwritten character recognition device according to the present invention has a character input means for inputting handwritten characters for each stroke, and a predetermined reference stroke direction. A stroke direction specifying means for specifying the direction of a stroke, a recognized character outputting means for outputting a recognized character corresponding to an inputted character based on an inputted stroke, and the recognized character outputting means includes a predetermined reference stroke. A character stock means including a stock of a plurality of characters written using strokes is used to output a recognized character, and when an input character is correctly recognized, the direction of the input stroke is set as a predetermined reference stroke direction. and a correction means for correcting the predetermined reference stroke direction so that the difference is eliminated by comparison.

[0007]

[Operation] When the input handwritten character and the recognized character match, the direction of the input stroke is compared with the direction of a predetermined reference stroke, and the character recognition standard is used to eliminate the difference. The direction of the reference stroke is modified. Therefore, the direction of the reference stroke, which is the reference for character recognition, is corrected according to the writing habits of the scribe.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the main parts of the present invention. Referring to FIG. 1, the online handwritten character recognition device according to the present invention includes a tablet 1 on which a user inputs characters. When a user inputs characters on the tablet 1, pen coordinate data and pen ON/OFF information are sent to the preprocessing section 2. In the stroke end determination section 3,
Determine the end of one stroke input. Upon completion of one stroke, the preprocessing unit 2 extracts feature amounts from the coordinate data and creates a pattern of the input stroke. Specifically, as shown in FIG. 2, each stroke is represented by a direction code using an 8-direction code reference coordinate system, which will be explained later.
Create input patterns by integrating co-directional codes. The stroke recognition section 5 compares the input pattern with data recorded in the basic stroke dictionary section 6. When the character end determining section 4 determines the end of character input, the character recognition section 7 compares the time series of the input strokes with the data in the character dictionary and recognizes the input characters. The recognized characters are output to the display section 9, allowing the user to determine whether the input characters have been correctly recognized. If a correct recognition result is obtained for the input, the writing habit detection section 10 detects the character dictionary section 8.
Detects writing habits based on the input reference information recorded in .

Next, the operation of detecting writing habits of the online handwritten character recognition device according to the present invention will be explained with reference to FIG. Note that FIG. 4 shows the reference coordinate system of the 8-direction code included in the preprocessing unit 2, FIG. 5 shows examples of input characters and their stroke order, and FIG. 6 shows the reference coordinate system of the 8-direction code shown in FIG. The coordinate system when is corrected is shown.

First, consider a horizontal and vertical coordinate system as shown in FIG. 4 as a reference for detecting writing habits. Handwriting habits are detected by focusing on long strokes in the horizontal direction and long strokes in the vertical direction among the strokes that make up a character, and determining the inclination of the input stroke corresponding to these strokes.

That is, let us now consider the case where the user of the online handwritten character recognition device inputs strokes in the order shown in (B) in order to input "6" as shown in FIG. Assume that in the character dictionary, information serving as a reference for writing habit detection is written in the stroke string forming "6", indicating that the second stroke is the reference for the horizontal stroke. When "6" is recognized as a recognition candidate character among the input characters, a writing habit is detected in the following procedure.

That is, referring to FIG. 3, it is first determined whether there is a reference stroke (step S11,
(The following steps are omitted). When it is determined in S11 that there is a stroke that serves as a reference for writing habit detection, the slope of the corresponding input stroke is calculated (S13). In the example of Fig. 5, it is written that the second stroke is the reference in the horizontal direction, so find the slope of the straight line connecting the starting end and the ending end (the angle with respect to the horizontal reference axis 1) for the second stroke. .

[0014] During character input, cases in which the inclination of strokes in both the horizontal and vertical directions exceed a certain threshold are counted (S15 to S23). If each stroke is frequently tilted, the horizontal and vertical reference axes are tilted by the threshold angle (S25, S27). FIG. 6 is a diagram showing a reference coordinate system when the lateral reference axis 1 is tilted toward the upper right direction for characters that rise to the right.

[0015] When the reference axis of the reference coordinate system is corrected, the direction of the stroke of a character input thereafter is specified using the corrected reference coordinate system. In addition, a new threshold value is set, and writing habits are newly detected.

[0016] When it is determined in S11 that there is no reference stroke, or when it is determined in S25 that the frequency is not so frequent (NO in S11 and NO in S25),
No reference axis correction is performed.

[0017] In the above embodiment, writing habits were detected by focusing on straight line strokes in the vertical and horizontal directions, but the present invention is not limited to this, and it is possible to detect writing habits using strokes in any direction. Judgment of writing habits may also be made.

[0018]

As described above, according to the present invention, when an input handwritten character and a recognized character match, the direction of the input stroke is compared with the direction of a predetermined reference stroke, and the direction of the input stroke is compared with the direction of a predetermined reference stroke. The direction of the reference stroke, which is the reference for character recognition, is corrected so that the difference disappears. That is, the writing habits of the scribe are detected, and the reference for determining the stroke direction is corrected based on the writing habits.

As a result, it is possible to improve the character recognition rate of the user in the online handwritten character recognition device.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the main parts of an online handwritten character recognition device according to the present invention.

FIG. 2 is a diagram showing input stroke processing.

FIG. 3 is a flowchart for explaining the operation of detecting writing habits.

FIG. 4 is a reference coordinate system of an 8-direction code.

FIG. 5 is a diagram showing an example of inputting "six" and its stroke order.

FIG. 6 is a diagram showing a reference coordinate system of a corrected eight-direction code.

[Explanation of symbols]

1 Tablet 2 Pre-processing section 5 Stroke recognition section 6 Basic stroke dictionary section 7 Character recognition section 8 Character dictionary section 9 Display section 10　Writing habit detection section

Claims (1)

[Claims] 1. Character input means for inputting characters by hand for each stroke; stroke direction specifying means having a predetermined reference stroke direction and thereby specifying the direction of the input stroke; A recognized character output means outputs a recognized character corresponding to the input character based on the strokes inputted, and the recognized character output means writes a character with strokes in the stroke order, outputting the recognized character using a dictionary comparison means with a plurality of character dictionaries in which information serving as a reference for writing habit detection is added to strokes that are not very dependent on the scribe, and the input character is correctly recognized; If there is a stroke to which information serving as a reference for the writing habit detection is added in the character dictionary, the inclination of the input stroke corresponding to that stroke is compared with the inclination of a reference axis in a predetermined reference stroke direction. a writing habit detecting means for detecting the writing habit; and a writing habit detecting means for detecting the writing habit by detecting the writing habit; and a correction means for correcting a reference axis of direction.