JPS6031683A - Handwritten character recognition device - Google Patents

Abstract

PURPOSE:To allow more precise, higher-speed recognition of handwritten characters by calculating characteristic elements of character patterns of the same writer after dividing character frames, using a registered inclination. CONSTITUTION:Character samples of a writer are read beforehand, and an inclination of horizontal and vertical character lines are extracted 13 and registered. Thereafter, for the character patterns of the same writer, his handwritten characters can be recognized by dividing and deciding 5 a character frame into partial areas and calculating characteristic elements, using the registered inclination. In such a manner, a high-speed and stable handwritten character recognition device can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field) The present invention relates to a high-speed and highly accurate handwritten character recognition device.

(Background Art) In conventional handwritten character recognition devices, in order to absorb variations in features due to variations in the slope of character lines due to different writers, the variations in features are absorbed by using a plurality of dictionary masks. However, this device has the disadvantage that the time required to compare the extracted features with the dictionary during identification increases in proportion to the number of dictionary masks, resulting in a reduction in the processing speed of the device.

(Problems to be solved by the invention) In order to eliminate such conventional drawbacks, the present invention reads a sample of the author's characters in advance, extracts and registers the amount of inclination of character lines in the horizontal and vertical directions, and
Subsequently, for character patterns written by the same author, handwritten characters were recognized by dividing the character frame into partial regions using the registered slope amount and calculating characteristic elements.The purpose was to recognize handwritten characters at high speed. The objective is to provide a stable handwritten character recognition device.

(Structure and operation of the invention) FIG. 1 is a block diagram showing an embodiment of the handwritten character recognition device of the present invention. In the figure, 1 is a photoelectric conversion unit, 2 is a turn register, and 3 is a line width calculation unit. part, 4 is a character frame detection part,
5 is a character frame division determination unit, 6 is a sub pattern extraction unit, 7 is a feature matrix extraction unit, 8 is an identification unit, 9 is a control unit, 1o
1 is a memory section, 11 is a character output section, 12 is a stroke extraction section, 13 is a tilt extraction section, and 14 is a keyboard section.
The operation will be explained in the 0th order. First, (1) slant amount extraction and registration using character samples for each writer, and then (2) recognition operation.

Extracting and registering the amount of inclination using character samples for each author in (1) is performed in the following steps.

The author presses the author registration control key and the author registration number key on the keyboard section 14. The keyboard unit 14 detects the depression of the key and sends a registration control signal and a registration number signal to the control unit 9. Next, the oiler places the form containing the writer's sample of writing into the reading mechanism. In this example, the character sample is the kanji character ``木''.

The characters on the form set in the reading mechanism are read by the photoelectric conversion unit l.
It is converted into a binary quantized dinotal electrical signal and stored in the cutan renoster 2.

At the same time, the line width of the input pattern is calculated in the line width calculation section 3. The sub pattern extraction unit 6 performs a vertical whole blood scan on the mother turn renostar and extracts a vertical sub pattern (vsp) from the relationship between the continuous length of black pits and the line width calculated by the calculation unit 3. . Similarly, horizontal sub-pattern (H4F) by horizontal scan
The right diagonal subpattern (
H8F), the left diagonal sub-pattern (LSP) is extracted by scanning the left diagonal at 45 degrees. Figure 2 shows examples of the original pattern and each zapno e-turn (a) Hahara-gu-turn, (b)
(c) is the horizontal subpattern (H4F), (d) is the right diagonal subpattern (H4F), and (c) is the horizontal subpattern (H4F).
8F) and (e) are left diagonal sub-patterns (LSP).

The stroke extraction unit 12 performs a horizontal or vertical scan of the entire blood for each sub-gut turn register, and extracts from the white point to the black point (
), detect the change point from black point to white point, and calculate the number of change points in the previous scan of one column (or row), the coordinates of the change point, and the number of change points in the current column (or row). Extract the stroke from the relationship of the change point coordinates, and calculate the coordinates in the two-dimensional coordinate system defined by the noe turn reno star of both end points of the stroke in each extracted sapnozo turn register (the origin is at the bottom left of the pattern register). It is sent to the slope extraction section 13. The slope extractor 13 calculates the slope by referring to the coordinates of both end points of the stroke of the sub-i9 turn Renosta H8P and VSP extracted in the o-ri extractor 12.

That is, the coordinates of both end points of the stroke extracted from the horizontal sub-pattern H8P are (HXSn, HYSn) (HXFn,
HYEn) However, n=1. ..., N, N is the number of strokes, and the slope θ□ is calculated using equation (1). (However ■
(2) XEn>HXSn) In the formula (1), HLGn represents the length of the relevant stroke, which is determined by approximate formula (2).

HLGn= MAX (IHXEn-HXSnl,
l HYE, -HYSnl 1+(MIN HXEn
-HXS, l, 'l I-rYEn-HYSnl
)]/2(2) Equation (2) is an approximation equation in which the distance between two points is the sum of the smaller one of the differences in horizontal and vertical coordinates between the two points and the other one. Similarly, θ7 is calculated using equation (3). However, ■YEm>vYsm.

Note that M in the above is the number of strokes extracted from the vertical sub-pattern. Also, when the number of strokes is O, the slope is also O.
shall be. Further, the stroke length VLGm is calculated using an approximate equation similar to equation (2).

The slope extraction unit 13 sends the slope of each sub-gutter calculated using the above equations (1) to (3) to the control unit 9. The control section 9 stores the slope data for each sub-pattern sent from the slope extraction section 13 at a predetermined address in the memory section 10 corresponding to the registration number sent from the keyboard section 14.

Next, the recognition operation (2) is performed according to the following procedure.

Okurake presses the author registration number key and the reading control key of the form to be read on the keyboard section 14, and sets the form to be read in the reading mechanism. The keyboard unit 14 detects the depression of the key and sends a reading control key depression signal and a registration number signal to the control unit 9.

When the control unit 9 receives the read control signal and the registration number signal, it reads out the author's slope data from the memory unit 10 at a predetermined address corresponding to the registration number, sends it to the feature matrix extraction unit 7, and then sends it to the identification unit. Category name from 8 (identification result)
Wait for it to be output.

The characters on the form set in the reading mechanism are converted into binary quantized digital electrical signals by the photoelectric conversion section 1 and stored in the pattern register 2. At the same time,
The line width calculation section 3 calculates the line width of the input pattern.

The sub-pattern extraction unit 6 performs a vertical scan on the whole blood and extracts a vertical sub-pattern (vsp) based on the relationship between the continuous length of black bits and the line width calculated by the calculation unit 3. Extract.

Similarly, by horizontal scanning, horizontal subpattern (H4F
) Extract the left diagonal sub-turns (LSP) by scanning diagonally 45 degrees to the right and sub-turns diagonally to the right.

The character frame detection unit 4 detects a character frame circumscribing the character pattern in the 9-turn reno star 2, and sends the result to the character frame division determination unit 5. The character frame division determination unit 5 divides the inside of the detected character frame into a JXI area (J and I are integers, in this embodiment J-I
-5) Determine the Y-axis to be divided and the coordinates of the dividing point on the Y-axis. Here, the Y axis indicates the horizontal direction of the character frame, and the Y axis indicates the vertical direction.

The division points are determined as follows. First, characters and turns are projected onto the Y-axis and the Y-axis to find the distribution of the number of black bits, respectively, and the center of gravity of the distribution is determined by dividing the first moment value of the distribution by the total number of point bits.

By repeating the procedure of dividing the distribution into two by the further determined center of gravity and dividing the center of gravity of each distribution, a plurality of coordinates of the center of gravity are obtained. The division coordinates necessary for the desired number of divisions are selected from this barycenter coordinate series, and this and the barycenter coordinates Xc, Yc of the first drawn character pattern are output to the feature matrix extraction section 7.

The feature matrix extraction section 7 extracts the division point coordinates X1' X2 j...XJ- determined by the character frame division determination section 5.
4 and Y, l Y2. . . . Yl-4, the center of gravity coordinates Xo, Yc of the letters and buttons, and the amount of inclination θ1 sent from the control unit 9. .

VSP, H4F, H8F, LS by θ7
The character frame area on each subpattern register of P is divided into JXI partial areas.

The method for determining divided regions will be explained below.

First, dividing point coordinates x, , x2...X, -1 and Y
l.

Y2. - Yl, , and Xc, Yc inclination amount θ□, θ7
(
4) Define (J+l-2) straight lines shown by the equation.

(However, Y<Y...<'Y'r-1+ Xl(X2・
・(x,,)2 When any point A (Xe'ye) within the character frame satisfies the inequality of equation (5), point A is included in the partial area (Li).

θ, ((Xe-Xc) + Yi-1<Ye≦01, (
xe-Xc)+Y1kθv (ye-YC) +
X5-1 (x, ≦θv (y, -Y(1)+Xj(5
) That is, the character frame is divided into a plurality of partial areas using the straight line in equation (4) as a boundary. FIG. 3 shows an example of division when I-J=5.

The feature matrix extraction unit 7 counts the number of black dots Bjj in each area of each sub-e turn register in which the character frame is divided into partial areas using the method described above, and calculates the line width W calculated by the line width calculation unit 3.
Using (6), calculate the features that determine the character line length and create a JXIX four-dimensional feature matrix.

L45 = Bij/W (6) After that, ■SP feature matrix) IJ is the length ΔY of the character frame in the Y-axis direction, H8P feature matrix is the length ΔX in the X-axis direction, and R3P and LSP feature matrix are ( Δχ1Δ
Normalize each with Y)/2 and finally JXIX4
Create a dimensional feature matrix. The identification unit 8 applies the distance (D) defined by equation (7) between the dictionary mask (f,) and the extracted feature matrix (f,), and calculates the distance so that D becomes the minimum value. The category name of the dictionary mask is sent to the control unit 9. The control unit 9 is configured to output the received category name to a character name output 11.

D−f■τ7q了 (7) As explained above, in this embodiment, by reading the writer's character sample in advance, the amount of inclination of the character line of the writer is extracted and registered. For character patterns of the same author, the registered inclination amount is used to divide the character frame to recognize the characters of the same author, which has the advantage of being able to recognize accurately and quickly. .

(Effects of the Invention) In the present invention, the tendency of the inclination of character lines is registered in advance from the author's character sample, and this inclination amount is used to divide the character frame and extract characteristic elements. It is possible to perform highly accurate recognition, and the registered tilt amount can be used for characters written by the same author, so it can be used in a high-speed and highly accurate handwritten character recognition device.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the character recognition device of the present invention, FIG. 2 is a diagram showing an example of an original pattern and each sub pattern, and FIG. 3 is a diagram showing an example of area division. . 1... Photoelectric conversion unit, 2... and turn register, 3
. . . line calculation unit, 4 . . . character frame detection unit, 5-character frame division determination unit, 6 . , 9...Control unit, 1o...Memory unit, 11...Character output, 12...
- Stroke extraction part, 13 - Inclination extraction part, 14... Keyboard part. Patent Applicant Oki Electric Industry Co., Ltd. Patent Application Agent Megumi Yamamoto - Second Cause (a) (b) '(c) (d) (e) Figure 3 Procedural Amendment (Voluntary) 1982 Kazuo Wakasugi, Commissioner of the Japan Patent Office (May 25) 1, Indication of the case 1982 Patent Application No. 140518 2, Title of the invention Handwritten character recognition device 3, Person making the amendment Relationship with the case Patent applicant name (029) ) Oki Electric Industry Co., Ltd.\4, Agent 5, Claims column and Detailed Description of the Invention column of the specification to be amended, Drawing 6, Contents of the amendment - (1) Claims of the specification Correct the range as shown in the attached sheet. (2) "By the author" in line 13 of page 3 is corrected to "by the author." (3) On page 3, line 18, "by author" is corrected to "'4E's". (4) "Length and calculation section" on page 4, line 11 is corrected to "length and line width calculation section." (5) Formula (2) on page 6 of the same page is corrected as follows. r HLGTl= MAX (IHXEn-HXSnl
, IHYEn-HYSnl ) + [KIN (lHX
En HXSnl, 1HYEn HYSnl 't]
/2 (2) J(6) Change "Inclination data of each subpattern 11--" from line 10 to line 12 of page 7 to "Control sending the inclination of sub-patterns 3P and VSP to the control unit 9. Part 9 is the tone H3P sent out from the slope extraction part 13,
Supplement with VSP slope data 8J. (7) "Determine the center of gravity" in lines 10 to 11 of page 9. ] and correct it. (8) Page 9, lines 12 to 14 [Furthermore -m-
get. '' is further divided into two by the centroid determined, and the barycenter coordinates of each divided region are calculated.Then, the distribution is divided into multiple regions by the centroid determined so far, and the coordinates of the center of gravity of each region are calculated. By repeating the process of determining the center of gravity, multiple coordinates of the center of gravity are obtained.'' (9) Same page 10, line 7 c7) rYc slope amount'
. Correct the amount of inclination. (10) Correct “Character name output” on page 11, line 8 of the same page to “Character output J.” (11) Correct “tendency of slope” on line 1 O of page 12 of the same page to “amount of slope” do. (12) Correct Figure 1 of the same figure as shown in the attached sheet. 5. Photoelectrically converting the author's character sample and storing the obtained digital signal as an original pattern in a pattern register as an original pattern, and extracting a sub-pattern representing a stroke in a desired direction from the original pattern.5. It has a function that extracts and registers the weighted average of the stroke inclination obtained from the Mizugotan J1 Sada subpattern using the stroke length as a weight, as the inclination of the character line of the writer. For each partial area obtained by dividing each of the sub-patterns obtained from the original pattern by specifying the author according to the registered slope within the character frame, the number of black bits in the partial area is determined by the character line width and stroke. A feature matrix is created using the amount obtained by normalizing with the size of the character frame corresponding to the direction as the feature element, and the standard character mask is recognized by referring to a dictionary written in the same format as the feature matrix. A handwritten character recognition device characterized by:

Claims (1)

[Claims] The writer's character sample is photoelectrically converted and quantized, and the obtained dinotal signal is stored in a pattern register as an original pattern, and a sub-pattern representing a stroke in a desired direction is extracted from the original four turns. It has a function that extracts and registers the weighted average of the obtained stroke inclinations using the stroke length as the weight, and registers it as the inclination of the author's character. During normal reading, by specifying the author, the original or Each of the above subs obtained from the turn
For each partial area obtained by dividing the turn within the character frame according to the registered slope, the number of black pits in the partial area is normalized by the character line width and the size of the character frame corresponding to the stroke direction. 7. A handwritten character recognition device characterized in that a feature matrix is created using quantities obtained as feature elements, and recognition is performed by referring to a dictionary in which standard character masks are described in the same format as the feature matrix.