JPS58214972A - On-line handwritten character recognizer - Google Patents

Abstract

PURPOSE:To eliminate an operation for indication of capitals and small letters, by deciding a capital when the size detection result of the input character is larger than a fixed value and then a small letter when the size detection result is smaller than the fixed value and then delivering a character code corresponding to the decided character. CONSTITUTION:A written character is fed to a pre-processing part 4, and the size of the character is detected. A comparator 10 compares the detected size value with the set value to decide the character as a capital or small letter when the size value is larger or smaller than the set value respectively. A code converting circuit 8 receives a character code corresponding to the result obtained by giving pattern recognition to the written character and a size decision result signal through an input and regardless of the size of the written character. Then the character code corresponding to the written character is delivered as it is when a capital is decided by the decision of size. While a character code receives a prescribed conversion based on the character code received at the input and is delivered in the case of a small letter.

Description

DETAILED DESCRIPTION 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 reduces the burden on a scribe and prevents a decrease in character entry speed.

The prior art and its problems will be explained with reference to FIGS. 1 and 2. In Fig. 1, 1 is a tablet input device (coordinate input device);
2 is a character entry sheet, 3 is an input pen, 4 is a preprocessing unit,
5 is a feature extraction circuit, 6 is a matching circuit, 7 is a minimum value selection circuit, 8 is a code conversion circuit, 9 is an output terminal, 10 is UP
/DOi1%'N detection circuit, 11 is a stroke number detection circuit, 12 is a standard pattern memory, 13 is a position comparison circuit,
Reference numeral 14 represents an instruction frame position memory, and reference numeral 15 represents an uppercase/lowercase character flag memory. A scribe uses an input pen 3 to write characters in each character entry frame 2-1 of a character entry sheet 2 placed on a tablet 1.

At this time, the tablet 1
Coordinate position information is output from output lines 1-3 at regular intervals (sampling period).

In addition, the input pen 3 has a built-in switch that detects whether or not the input pen 3 is pressed against the character entry sheet 2, and the output of this switch is also used as Z-axis information.
It is output from output lines 1-3 at every sampling period along with the position information of the Y coordinate.

These X, Y, and Z axis information are supplied to the preprocessing section 4 as handwriting information. The preprocessing unit 4 first looks at the two-axis information, selects and imports only the data (hereinafter referred to as handwriting points) that the input pen 3 presses onto the character entry sheet 2, and performs the following normalization process. .

The series of handwriting points includes redundant points. This is because the pen tip does not move at a constant speed when writing characters, and the spatial distance between adjacent handwriting points may be very close. For this purpose, redundant handwriting points are removed. The removal method is to resample handwriting points that are a certain distance away from the starting point of a stroke (one line segment drawn from when the input pen 3 presses on the character entry sheet 2 until it leaves the character writing sheet 2, a series of handwriting points). Then, the process of again setting handwriting points that are a certain distance away from this re-marking point as re-sampling points is performed until the end point of the stroke. In other words, the handwriting point series of each stroke is converted from a time-space series to a distance-space series (
(Hereinafter, this process will be referred to as resampling process.)

Next, the preprocessing unit 4 normalizes the position and size of the data for one character that has been resampled. In which character entry frame 2-1 of the character entry sheet 2 should the characters be written? Since the XY coordinate values of the handwriting points will vary even if this is done, position normalization is a process of coordinate transformation for each character so that the center of gravity of the character remains constant. Also, since the size of the characters written in the character entry frame 2-1 differs depending on the scribe, it is recommended to perform coordinate transformation of each resampling point so that the size of the characters written is constant. It is.

This is done by ensuring that the average distance of each resampling point to the center of gravity of the character is constant.

The pre-processed character data is processed by the feature extraction circuit 5 to facilitate subsequent processing.
Expressed in a form with reduced amount of information. For example, for an input character ■ consisting of M strokes, the mth stroke entered is Im.

H= (I 1 , I 2 + ..., Ih1)
It is expressed in the order in which the strokes are written.

In addition, each stroke 11 to IM is a series of N+1 broken line approximation points that divide the width of one stroke from the start point (starting handwriting point) to the end point (finishing handwriting point) into N equal parts. express. In other words, the n1th stroke 1m is the approximate point Pm1. Pm2. -, P
Im=(Pml, Pm2
, -, PmN+1). Here, the approximate point P of the broken line
+nn is PIITI = (X m n , Y m r
It is the XY locus a value indicated by .

The input characters described by the feature extraction circuit 5 in this manner are supplied to one input terminal of the matching circuit 6. At the other input terminal of the matching circuit 6, the average pattern of characters input by a large number of scribes, whose features have been extracted, is processed for each character to be recognized.
Supplied from standard pattern memory 2.

Φθ Here, the standard pattern S for the character θ is ・θ
θ θ θ8 w(S
1. S2. ..., 8M). However, M is the number of strokes of the character θ and θ Sm is θ θ θ θS m
This is the m-th stroke expressed as e=(Pm1.Pmz, -, PmN+1).

θ P m n is θ θ θ Pmn! (Xmn, ymn) is the nth XY coordinate value of a broken line approximate point that divides the line segment of the mth stroke into N equal parts.

The matching circuit 6 calculates the distance D(θ) between the input character I and the standard pattern S having the number of strokes equal to the number of strokes M of this pattern as follows.

Here, Im is the mth stroke θ of the input character
・θ-ri, Sm is the m-th stroke of the standard pattern S,
ds (S'm, Im) represents the distance between the m-th stroke Sm, Im of both patterns, and n==1. N+1 is the number of broken line approximation points of the stroke, θ Pmn, Pmn is the nth broken line approximate point of the second stroke of both patterns, dP (Pmn, Pmn
) indicates the distance between the n-th broken line approximate points of the m-th stroke of both patterns dP(P”mn, Pmn) = (X”mn-Xmn)%
-(7+koymJ9).

Combining the above equations (7) to (9), matching circuit 6
is calculated as the distance D(θ) between both patterns. However, Xmn, ) (n+n is the nth stroke of the mth stroke
The X coordinate value, yo-., y-. of the th broken line approximate point is also the Y coordinate value.

Here, if there are L standard patterns equal to the number of strokes M of the input character I, the matching circuit 6 sequentially calculates the inter-pattern distance D between the input character ■ and the L standard patterns.
(θ) and supplies the result to the minimum value selection circuit 7.

The number of strokes of the input character ■ is determined by detecting the UP and DOWN of the input pen 3 by the UP/DOWN detection circuit 10 based on the Z-axis information from the tablet 1, and then calculating the number of strokes from U l) to DOWN. The change is determined by counting over one character by the stroke number detection circuit 1.

The output of the stroke number detection circuit 11 controls the standard pattern memory 12 to select a standard pattern S equal to the number of strokes M of the input character l and supplies it to the matching circuit 6.

The minimum value selection circuit 7 takes the minimum value of the L pieces of inter-pattern distances D(dl) to D(θL) sequentially supplied from the standard pattern memory 12 and outputs it to the code conversion circuit 8. .

The problem with the conventional online 111 character recognition method as described above is, for example, the uppercase letter ``ki'' and the lowercase letter [ya J
This is an input method and recognition method for uppercase and lowercase letters, such as r-J r-J. These lowercase letters have exactly the same shape as uppercase letters, only differing in size, and The reason for the difference in character size is that in conventional technology, due to normalization in the preprocessing section, the recognition section ends up with exactly the same characters, making it impossible to determine whether they are uppercase or lowercase. be.

For example, if the kana ``tsu'' is written in the corner of the character entry frame 2-1 as shown in (a) of Figure 2, and if it is written in the corner of the character entry frame 2-1 as shown in (C). The preprocessing result is exactly the same pattern with WO as the center of gravity, such as (bl, (dl). In this way, when the uppercase and lowercase character patterns have the same shape, , exists not only in Japanese katakana and hiragana, but also in English characters.

To deal with this, conventionally a method has been adopted in which a scribe supplies information on whether a letter is an uppercase or a lowercase letter via the tablet 1 at the time of writing.

For example, as shown in FIG. 1, the tablet 1 is provided with an uppercase character designation frame 1-1 and a lowercase character designation frame 1-2, and by pressing either of the designation frames with the input pen 3, subsequent input characters can be changed. Declare uppercase or lowercase letters.

Then, in the recognition section, the position comparison circuit 13 determines the coordinate values corresponding to the indication frame 1-1, 1-2 (indication frame position memory 1
4) as a reference for comparison, input pen 3
It is detected which designation frame is pressed or whether it is not pressed at all, and an uppercase/lowercase character flag memory 15 is set in accordance with the detected combination.

Here, the output F of the uppercase/lowercase character flak memory 15 is
For uppercase letters, F=1; for lowercase letters, F=0. Also, the character code for the uppercase letter “tsu” is A3C4, and the lowercase letter “tsu” is A3C4.
Assume that the character code for tsuj is A3C5, and the standard pattern for ``tsu'' is given an uppercase character code A3C4.

And if you look at the case where "tsu" is written with input pen 3,
The character code A3C4 is outputted from the minimum value selection circuit 7. This character code is supplied to the code conversion circuit 8. When the instruction from the case/lowercase flag memory 15 is F-1 (uppercase instruction), the code conversion circuit 8 outputs the character code A3C4 as is, and when F'=0 (lowercase instruction), it outputs the character code A3C4, which corresponds to a lowercase letter. Outputs character code A3C5. That is, the code conversion circuit 8 internally has a character code correspondence table of lowercase letters corresponding to uppercase letters, and uses this to convert the uppercase character code to a lowercase character code when F=0 and outputs the converted character code.

However, the conventional method described above has the problem that in addition to writing the letters, it is also necessary to indicate whether the letters are uppercase or lowercase, which not only burdens the scribe but also slows down the input speed. .

The purpose of the present invention is to solve the above-mentioned problems in the prior art, reduce the burden on the reporter, prevent a decrease in character entry speed, and improve the recognition efficiency for English letters. An object of the present invention is to provide a character recognition device.

In order to achieve the above object, the present invention has a character size detection means for detecting the size of a written character, and compares this detected value with a set value, and when the size is larger than the set value, it is determined to be an uppercase letter, and when it is smaller than the set value, it is determined to be a lowercase letter. When the size determination result is an uppercase character by receiving the character code corresponding to the result of pattern recognition of the written character without depending on the character size of the written character and the size determination result signal as input, The apparatus is configured to include a code conversion means that directly outputs a character code corresponding to a written character, and when the size determination result is a lowercase character, performs a predetermined code conversion based on the input character code and outputs the result. ζ It's cold.

An embodiment of the present invention will be described below with reference to FIG. Third
In the figure, 4-1 is a resampling circuit, 4-2 is a position normalization circuit, 4-3 is a centroid normalization circuit, 4-4 is a size normalization circuit, and 4-5 is an average radius extraction circuit. , 16 is a comparison circuit, and the other parts are the same as in the case of FIG. Human power 1. The marriage information of the characters is on the output line l-3 of the tablet.
Here, as described above, the l and TK points behind the handwriting points of each stroke are removed, and the handwriting point series of each stroke is extracted from the spatiotemporal column. Convert to metric space series. That is, the line segment C from the start point to the end point of the stroke is resampled at constant distance intervals.

The m-th stroke Im of this resampled input character ■ is the resampled point □□□], Φ2..., (J)nE
Expressed as a series of (rr), Im= (Qml , Qm2 , −・−, Q+nB(
tri ). However, Qme is the 0th resampling point of the mth stroke, and E (e) is the number of resampling points of the mth stroke. Also, the resampling point Q
me indicates the XY coordinate value, Qme = (Xme, Yme)
It is.

The human character data resampled in this way is supplied to the center of gravity extraction circuit 4-3, and the center of gravity Wo of the input character is extracted. This center of gravity position WO is the total resampling point Qme for − characters (In” 1 to M, e: l, F!)
), the average value Xo of the X Zaura value, the average value YO of the Y coordinate value
The Y coordinate value is ↓), so it is expressed as Vvo=(Xo, Yo).

Next, in the position normalization circuit 4-2, C1 this center of gravity position W
The coordinate values of each resampling point Qme are transformed so that O is the origin of the new XY coordinate axes.

In other words , -XQ -Y.

Qme = (Xme, Yme)), the XY coordinate value XmeYm of each resampling point Qme
Subtract Xo and Yo from e. x me here
= X me −X.

yme = Yme'-Y.

Then, the center of gravity position is X=O, y=O.

Next, the average radius extraction circuit 4-5 calculates the size of the input character, here the average radius R of the input character, for the human character data that has been subjected to the position normalization process as described above. Here, U=ΣE(,,,) is the number of resampling points of input character 1n=1, M is the number of strokes of input character,
1xme1. l ymel is the absolute value of the X-axis value and Y-axis value of the 0th resampling point of the mth string whose origin is the center of gravity WO of the input character. This means that the average radius R is obtained as the average value of the distances of each resampling point QrTIe from the center of gravity W01. This average radius R
The larger the character is written, the larger the value becomes, and is a malameter that corresponds to the size of human-powered characters.

The size normalization circuit 4-4 calculates this average radius R.

The coordinate values of each resampling point Qme are transformed so that Qme becomes the set value RO. X of the resampling point Qme after this conversion process
, the Y-axis value is X ll1e 'i''Y ITle, then the normalization of the size is to normalize (divide) each XY coordinate value by the average radius of the input character, as shown below.

This average radius R of the input characters is also input to one input terminal of the comparison circuit 16 and compared with a set value ath input to the other input terminal. In other words, the comparison circuit 16 determines whether the input character is larger than the set value. The determination result is supplied to the code conversion circuit 8, and the operation of the code conversion circuit 8 is controlled.

The output signal of the size normalization circuit 4-4 is supplied to the feature extraction circuit 5, where the input character is expressed in a compressed form as explained in the prior art section, and the matching circuit 6 converts the input character into a standard pattern 1. Perform matching calculation (calculation of distance between patterns) with

The standard pattern memory 12 contains letters A, B,
Regarding uppercase letters such as C, ..., kana characters, etc., the average of patterns of uppercase letters such as A, I, tsu, ...tsu, written by many scribes and subjected to the above-mentioned preprocessing and feature extraction. The pattern is stored together with the character code corresponding to that character. Also, at this time, the characters are classified and stored according to the number of strokes.

Here, as shown in Figures 4 (1) and 9 (2), write capital letters 1" A J in the character entry frame 2-1 with the first stroke ■l [△" and the second stroke 12 [' - Assuming that the handwriting information of output lines 1-3 is written using two strokes of J, the handwriting information of output lines 1-3 is subjected to pre-processing and feature extraction, and the matching circuit 6 performs matching calculation ( (calculation of inter-pattern distance) is performed and the results are sequentially supplied to the minimum value selection circuit 7.

Then, in the minimum value selection circuit 7, the inter-pattern distance D
The minimum value of (θ) is detected and the standard pattern character code corresponding to the minimum value is supplied to the code conversion circuit 8.

Here, if you set the character code corresponding to the standard pattern and its letters (uppercase letters) as shown on the left side of Table 1, and set the character codes of lowercase letters corresponding to the uppercase letters as shown on the right side of Table 1. do.

In the minimum value selection circuit 7, when the distance between patterns for the capital letter rAJ becomes the minimum, the minimum value selection circuit 7
The character code A3C1 given to the capital letter rAJ is supplied to the code conversion circuit 8. Figures 4 (1) and 2 (
2) The average radius R(1) and R(2) extracted by the average radius extraction circuit 4-5 for the writing of the capital letter "Aj" are as follows.
A comparison circuit 16 compares whether or not it is larger than a set value th.

Here, Rth × Rm...In the case of Figure (1), Rth
> R(2)...If it is determined that the case shown in Figure (2) is the case, the code conversion circuit 8 outputs the input character code A3C1 (character code for 1-A-1) as it is in the case of Figure (11). In the case of Figure (2), the input character code is A3C1 with a small "A" written in it.
(Based on the character code for 1'AJ, the character code A3B1 of the corresponding alphabetic lowercase letter 1a-(
(built-in) and output it.

The same is true when the katakana ``tsu'' is written as shown in Figures (3) and 1 (4) in Figure 4. These conditions are summarized in Table 2.

Table 2 (Note) Large: means to write in a larger size than the character entry frame.Small: means to write in a smaller size relative to the character entry frame.

In the above example, the position is normalized so that the center of gravity of the input character is crossed from the origin l, and the size of the input character is set to the average distance (average radius) between the center of gravity of the input character and the re-drifting point. ), and the size is normalized based on this average radius.However, the present invention is not limited to this, and instead of the above, position normalization is performed based on the center of the circumscribed rectangle of the input character. The position is set as the origin, and the size of the input character is set as the diagonal length of the circumscribed rectangle, and the size is normalized based on this diagonal length so that it becomes a constant diagonal length. You can also do that.

As explained above, according to the present invention (1), the size of an input character is detected, and when the detection result is larger than a certain value, it is judged as an uppercase letter, and when it is smaller than a certain value, it is judged as a lowercase letter. By having a configuration that outputs a character code corresponding to the input character, it is no longer necessary to input the input character by pressing the input pen into the uppercase and lowercase letter indication frame IJ7, In addition to eliminating the burden on the scribe, it is now possible to prevent a drop in character input speed, and alphabetic characters, 1, and characters whose characters are mainly composed of curved lines, are excluded from recognition. Traditionally,
This has the effect of preventing lowercase letters from being deformed in various ways by scribes, which lowers the recognition rate because they are mainly curved.

[Brief explanation of drawings]

Fig. 1 is a block diagram of a conventional example, Fig. 2 is a diagram showing input characters and preprocessing results, Fig. 3 is a block diagram of an embodiment of the present invention, and Fig. 4 is a diagram showing an example of character writing according to the present invention. ~It is a figure. Explanation of symbols 1...Tablet 2...Character entry sheet 2
-1...Character entry frame 3...Manual pen 4...Preprocessing unit 4-4...Size normalization circuit 4-5...Average radius extraction circuit 5...Feature extraction circuit 7.・Minimum value selection circuit 8・
...Code conversion circuit 12...Standard pattern memory 16...Comparison circuit agent Patent attorney Usui 1) Toshiyuki 1 ・ 2P spear (D (2) 4 Closed (3) (4) 418

Claims (1)

[Scope of Claims] (Li) An online handwritten character recognition device that writes characters on a tablet with an input pen and recognizes the written characters based on daughter trace information of the written characters output from the tablet. a font size detection means for detecting the size of the font, a font size determination means for comparing the detected value with a set value and determining that it is an upper case letter when it is larger than the set value, and a lower case letter when it is smaller than the set value, and a font size judgment means that does not depend on the font size of written characters. When the size judgment result is an uppercase letter, the character code corresponding to the pattern recognition result of the written character is output as is, and the size judgment result is a lowercase letter. An online handwritten character recognition device characterized by comprising: code conversion means for converting a predetermined code based on a character code received as an input and outputting the resultant code. (2. Claim 1) The above-mentioned character size detection means detects the center of gravity of the written character obtained based on the re-sampling points extracted from the line segments of each stroke of the written character at fixed distance intervals, and the above-mentioned re-sampling points. An online handwritten character recognition device characterized in that it is a character size detection means that determines the size of the handwritten character as the average value of the distance from the marked point. (3) In the device according to claim 1, the character An online handwritten character recognition device characterized in that the size detection means is a character size detection means that determines the length of a diagonal line of a rectangle circumscribing the handwritten character as the size of the handwritten character.