JPS592192A - Character recognizing system - Google Patents

Abstract

PURPOSE:To enable character recognition by matching that absorbs two-dimensional expansion and contraction, by performing firstly horizontal one-directional normalization of expansion and contraction from projecting information on the horizontal axis of a character pattern, and then performing expansion and contraction matching in the vertical direction. CONSTITUTION:An input pattern is read from an input character pattern memory 1 as a signal 101, projecting information histogram on the horizontal axis is obtained and outputted as a projecting information signal 102. A mapping function generating device 3 obtains a mapping function of the projecting information signal 102 and a standard projecting information signal 104 and outputs as a mapping functon signal 103. A two- dimensional expansion and contraction matching device 5 reads the signal 101 and performs one-directional expansion and contraction normalization by using the number of mapping for each read character read as a signal 103, and reads the standard character pattern signal 106 of read character corresponding to above-mentioned mapping function stored in a standard character function memory 6, performs expansion and contraction matching in the vertical direction to find out difference, and outputs the difference between input character pattern and each read character as a signal 105. A discriminating device 7 reads it as a signal 105, and outputs, for instance, the result of output of character which is leaset in difference as a signal 107.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for recognizing characters composed of many strokes, such as kanji, hiragana, alphanumeric characters, etc.

The development of optical character recognition technology has been remarkable in recent years, and products that recognize alphanumeric characters, both handwritten and printed characters, have been commercialized and put into practical use. Also kanji,
For those that recognize Japanese characters including hiragana,
As far as single fonts for printed characters are concerned, the development of prototype machines has already been announced.

However, in order to recognize handwritten characters such as kanji, hiragana, hiragana, and alphanumeric characters, approaches such as extending the recognition method for handwritten alphanumeric characters to recognize handwritten kanji, and approaches from printed kanji recognition are being taken. However, nothing effective has yet been achieved. One reason for this is that kanji have more complex shapes than alphanumeric characters, making it difficult to select feature information.Another reason is that kanji have many deformations, making it difficult to obtain stable feature information. For example, it is difficult.

On the other hand, although experiments have been attempted on small amounts of data by combining the brush effect and the template matching method, good results have not yet been obtained.

Now, there are two possible causes of fluctuations in handwritten characters: (1) positional deviation (expansion and contraction), and (2) rotation. This causes distortion in the entire character pattern.

Of these, the basic cause of the positional deviation (1) is that the positional deviation occurs in two directions because it is a character slam or two-dimensional information. In contrast, if we focus on speech recognition processing, since it is basically one-dimensional information in the time axis direction, the DP matching method is used to perform expansion/contraction matching in the time axis direction to solve the problem of positional deviation. . (
For example, Japanese Patent Publication No. 49-45943 ``Batan Similarity Calculation Device'').

The present invention provides a high-precision character recognition method that achieves stretch matching in two different directions on a two-dimensional plane in a pseudo manner, and absorbs two-dimensional positional deviations that have been considered difficult in the past.

The present invention will be explained in detail below using the figures.
Let us take the horizontal direction and the vertical direction as two different directions.

The reason for this is that it is easy to explain and is often used when dealing with two-dimensional batons.
The same effect can be obtained by using two different directions.

Figure 1 is a diagram for intuitively explaining two-dimensional expansion/contraction matching. (a) A one-way normalized character button in which the input character button is stretched and normalized in the horizontal direction so that the expansion and contraction matching between the upward projection information and the standard character button (a) is optimally matched with the projection information on the horizontal axis, and (dl indicates a two-way normalized character button that has been vertically expanded and contracted to optimize vertical expansion/contraction matching between the one-way normalized character button (e) and the standard character button (a); The present invention realizes, when matching the standard character BUTTON (at) and the input character BUTTON (b), as if matching the standard character BUTTON (a) and the two-way normalized character BUTTON (d). As a result, characters can be recognized while absorbing two-dimensional stroke positional deviations.

Figures 2 (a) to 11) are diagrams for explaining means for realizing two-dimensional expansion/contraction matching and their effects; 21 in Figure 2 (a) is a standard character button; 22 is an input character button, 23 in (C) is standard projection information on the horizontal axis of the standard character button 21, and in the same figure (24 in dl is input projection information on the horizontal axis of the input character button 22, tel in the figure shows that the standard projection information 23 and the input projection information 24 are expanded and contracted in the horizontal direction to obtain the mapping function 25; This is a diagram showing that a horizontal one-way stretch/contract normalization button 26 of the input character button 22 is obtained, and (g) of the same figure shows the difference between the one-way stretch/contraction normalization button 26 and the standard character button 21. This is a schematic diagram showing that the mapping function at that time is equal to C when expansion/contraction matching is performed in the vertical direction.

In the above explanation, the character button used in stretch matching consists of an MxN matrix, where M is the number of pixels in the horizontal direction and N is the number of pixels in the vertical direction. It is assumed that the projection information on the horizontal axis is described as a one-dimensional vector, that is, a series of M-rans of scalar quantities. Further, the horizontal one-way expansion/contraction normalization process of the input character pattern (indicated by fl/(b)) is to sequentially expand/contract and normalize each of the N M-dimensional vectors.

In contrast to the understatement shown in the above explanation, the present invention can achieve two-way expansion/contraction alignment.
By repeating this process multiple times, the aim is to achieve two-dimensional expansion/contraction matching.

On the other hand, Fig. 2 (h) shows the horizontal one-way expansion/contraction normalization process ρ.
This is a diagram to show the effect. In the figure, the black parts 21 and 22 are the parts that are aligned by vertical expansion and contraction, and the white blank parts are the parts that are not aligned even if they are expanded and contracted in the vertical direction. ing. By comparing Figure 2 (h) and Figure 2 (gl) above, it is found that unidirectional horizontal expansion/contraction normalization processing is performed on the input character button 22 before vertical expansion/contraction matching processing. The effect is shown.

Figure 2 (1) is a temporary diagram to explain the case where two-dimensional baton information itself is used instead of projection information for the horizontal one-way stretching/contraction normalization process, and the black parts 21 and 22 in the figure are This is the part that has been aligned by expanding and contracting in the horizontal direction.The white blank area shows the part that is not aligned even if it is expanded and contracted in the horizontal direction, but the alignment is extremely unstable due to positional displacement of the stroke. I understand. This one-way horizontal stretching/contraction normalization process greatly affects the accuracy of the next vertical stretching/contraction matching process, and stable matching is required to obtain the mapping function. Even if this information is missing, it is necessary to use projection information that absorbs the positional deviation of the stroke.

FIGS. 3(a), (b), (e), and (d) are diagrams for explaining an example of the DP matching method used in speech recognition as one-way expansion/contraction matching processing.

The standard baton Ao is an M-dimensional vector Ao e Ao +・
...Ao sequence, the input pattern /A is an M-dimensional vector A, A2. ..., is made up of a series of ANs. Also, any vector of standard batons A oj
and any vector A of the human-powered slam: Let the distance be d(
l.

j). When simple matching is performed, the degree of difference D (A, Ao) between the input baton A and the standard baton Ao can be obtained, for example, by the following formula.

This formula is shown in Figure 3 (on the mapping function j = i of al, J A and Ao are made to correspond and the degree of difference between the two batons is determined. So, A1 and Aoj
If it is possible to make these patterns correspond to each other, it is possible to partially expand or contract the input batten A to match it with the standard batten Ao in order to find the degree of difference between the two putters.

The DP matching method is a method for matching the input data by partially expanding or contracting it. For example, in Fig. 3 (1b), f (N, N) is found from the following initial values and recurrence formula. In particular, matching can be achieved by making Aj and Aoj correspond on the mapping function j=ψ(1).

g(1,1)=d(1,1)('yD"d(1-j)+*x(l' (i 1
, j) −9(11−j 1)+g(l −1tj−
2) ) However, d(1,j)=cx:+ (1≦O1 or j
≦0).

FIG. 3(c) is a diagram showing an example of the DP matching method for obtaining the above recurrence formula.
, 5), the standard baton has the same 5 series (1, 2
, 3, 4, 5) and (1, j) is (1, 1), (2
, 2), (3, 4), (4, 5), and (5, 5).

Figure 3(d) shows that the recurrence formula calculation is performed within the range of I-△≦j≦10∆ in order to reduce the amount of calculation in the above recurrence formula calculation. In the law, this range is called the matching window.
We are actually trying to make the amount of calculation more efficient.

The above recurrence formula is only for calculating the degree of dissimilarity, but 1n(f(1-1, D, f(1-1, j-1), g(1
-1,j-2))=9(11,j(1-1ρ (however, j(1-1) is j, j-1, j-2), then h(l, By calculating the function h(i, J) as j)=j(1-1), the value of h(1, 1) can be changed to h(N, N) after the degree of dissimilarity is calculated.
The mapping function can be obtained by sequentially obtaining J1(, t, t) from J1(, t, t). For example, in Figure 3 (cl example) h(5,5) h(5,5)=5 , h(4,5)=4 , h(3
,4)=2.

Since h(2,2)=1, the mapping function (1,j) is (1,1), (2,,2), (3,4),
(4,5), (5,5) are found.

FIG. 4 is a diagram showing an example of expansion/contraction normalization processing,
)] 1) (1≦i≦16) is input (tan, Y (j)
(1≦j≦16) is the expansion/contraction normalization (tan), and j=ψ(1
) is a mapping function for stretch normalization. In this example, Y
(jl is determined by the following rule.

(1) j=ψ(i)〉ψ(i-1) and ψ(i)〈
When ψ(++1), Y(j)=X(1) (2) When j=ψ(1)=ψ(i-1)+2, Y
(j-1) =X(1)(3) j=ψ(1)=ψ
When (ti)<ψ(i+1), Y(j)=X(1) FIG. 5 is a block diagram showing an embodiment of a device for realizing the method of the present invention. 100 is an input character button signal, and 1 is an input character button storage section that stores the input character button. 2 is a projection information extracting means that reads the input button as a signal 101 from the input character noktan storage unit 2 and obtains a projection information histogram on the horizontal axis;
It is output as a projection information signal 102. 3 is a mapping function generating means which generates the projection information signal 102 and a standard projection information signal 104 for each character type to be read, which is stored in the standard projection information storage section 4 in the same format as the projection information signal 102.
A mapping function is obtained and output as a mapping function signal 103. 5 is a two-dimensional expansion/contraction matching means, which receives input character slam signal 1;
01 is read, one-way expansion/contraction normalization processing is performed using the mapping function for each character type read as the signal 103, and the mapping function corresponding to the mapping function stored in the standard character button storage section 6 is read. Standard character slam signal 1 for reading type
06 is read in, vertical expansion/contraction matching is performed to determine the degree of difference, and the degree of difference between the input character BUTTON and each type of character to be read is output as a signal 105. The identification means 7 reads the degree of dissimilarity from each character type to be read as a signal 105, and outputs, for example, simply the character type with the smallest degree of dissimilarity, or the smallest degree of dissimilarity and the second smallest degree of dissimilarity. When the difference is greater than a certain value, the character type with the smallest degree of difference is outputted as the output result, and in other cases, the recognition result is outputted as a signal 107 using a normal method in character recognition, such as outputting a blemish.

In the above description, the input button storage section 1 and the projection information extraction means 2 may be those commonly used in the button processing.

FIG. 6 is a block diagram showing an example of the configuration of the mapping function generating means 3. The process here is to calculate the recurrence formula f(1, j) in the explanation of the DP matching method, find the trajectory h(i, j) obtained as a result of the recurrence formula calculation, and calculate h(1, j) to find the mapping function.

102 is the projection information signal, which is a series of scalar quantities A, .
...A, ..., A, and 104 is the standard projection information signal, which is a series Ao of scalar 1 for each character type to be read.
31 corresponds to e, Ao + . 32 is a recurrence formula calculation unit that calculates the above-mentioned recurrence formula %)).
゜mg*(f(1-1,j), f(amount-1,j-1),
f(1-1,j-2)) as the signal 341,
The calculation result 9(1,j) is output to the cumulative value storage section 33 as a signal 321.

34 is a minimum value selection unit, and cumulative value storage unit 33 to 9 (1-
1,j), f(1-1,j-1) and q(s-1,
j-2) as the signal 331, signal 332 and signal 333, m(f(1-1,j), f(1-1,j-
1), f(i-1, j-2)) as the signal 341, and h
(1, j) is output to the mapping locus storage unit 35 as a signal 342. When the recurrence formula calculation is completed, the mapping locus storage unit 3
5 outputs the mapping function as a signal 103.

FIG. 7 is a block diagram showing an example of the configuration of the two-dimensional expansion/contraction matching means 5. As shown in FIG. Reference numeral 51 denotes a one-way expansion/contraction normalization means, which generates a one-way expansion/contraction normalization means bang signal corresponding to each type of character to be read from the input character bang signal 101 and the mapping function signal 103 corresponding to each type of character to be read. 510 is output. 52
is a character slam expansion/contraction matching means that calculates the degree of difference corresponding to each type of characters to be read from the one-way expansion/contraction normalized character slam signal 511 and the standard character slam signal 106 corresponding to each type of characters to be read. Output as .

The expansion/contraction normalization means 51 reads the input character button 101 as a series of vectors A , A , ..., A , and uses the mapping function determined by the signal 103 for each vector according to the rules explained in FIG. 4. Vector A', A
2. ...outputs a 1M sequence as signal 510,
This is a one-way stretch normalized character bang. This one-way expansion/contraction normalized character pattern is obtained for each type of character to be read. That is, the one-way expansion/contraction normalized character button is sequentially outputted as a signal 510 to the mapping function corresponding to each character type to be read.

FIG. 8 is a block diagram illustrating an example of the configuration of the character punch expansion/contraction matching means. Reference numeral 521 denotes a vector distance calculation unit, which converts the one-way expansion/contraction normalization process button and standard character button corresponding to each character type to be read into vector series signals 510.
is read as the signal 106, and the DP matching method d(
1, j) is performed and output as a signal 5211. Reference numeral 522 denotes a recurrence formula calculation unit, which may be the same as the recurrence formula calculation unit 32 of the mapping function generation means 3, and is used to calculate the recurrence formula 9). as the signal 5211, and the moon (f(1-1,j), f(i-1,j-1)
, 9(1-1, j-2)) as a signal 5241, and output g(t, j) as a signal 5221 to the difference cumulative value storage unit 523.

524 is a difference minimum value selection unit, and a difference degree cumulative value storage unit 5
23 to f(i-x+j)-g(t-i, j-1) and g(i-1, j-2) as signals 5231 and 52
32, and the signal 5233 includes the bladder, m(g(i-
1,j), q(i-1゜j-1), g(1-1,j-2
)) is output as a signal 5241. When the recurrence formula calculation is completed, the cumulative difference value storage unit 523 stores the difference g(N
, N) as a signal 105. Through the above processing, the degree of difference corresponding to each character type to be read is sequentially outputted as a signal 105.

Figure 9 is a diagram showing another example of information that can be used as projection information.The number of times the character stamp crosses the character part by scanning it in the vertical direction is used as the projection information. , is similar to the projection information described above.

As described above, according to the present invention, from the projection information of the character pattern on the horizontal axis, first a one-way stretching/contracting normalization process in the horizontal direction is performed, and then a stretching/contracting matching process is performed in the vertical direction. It is possible to realize character recognition using alignment that absorbs natural expansion and contraction.

Contrary to the above process, by performing unidirectional stretching/contraction normalization processing in the vertical direction from the projection information on the vertical axis of the character pattern, and then performing stretching/contraction matching in the horizontal direction, similarly two-dimensional stretching/contraction is performed. It is possible to realize character recognition by matching that absorbs Further, the two different directions are not limited to the above-mentioned vertical direction and horizontal direction.

In general, many character recognition methods try to achieve the effectiveness of the recognition method by normalizing the position and size, smoothing character patterns, blurring, etc., but the character recognition method according to the present invention also Effects similar to those of other methods can also be obtained by performing preprocessing on the pattern.

Furthermore, various DP matching methods have been published so far, and are not limited to the method described in this specification.

[Brief explanation of the drawing]

Figure 1 (a), (bl, (c), (dl) is a diagram for intuitively explaining two-dimensional expansion/contraction matching, Figure 2
Figures (a) to (i) are diagrams for explaining the means for realizing two-dimensional expansion/contraction matching and their effects, and Figure 3 (at, (bl)
, (cl, (dl is a diagram showing an example of the DP matching method, FIG. 4 is a diagram showing an example of expansion/contraction normalization processing, and FIG. 5 is an example of an implementation of a device for realizing the method of the present invention. A block diagram showing an example, FIG. 6 is a block diagram showing an example of the configuration of the mapping function generation means, FIG. 7 is a block diagram showing an example of the configuration of the two-dimensional expansion/contraction matching means, and FIG. 8 is a block diagram showing an example of the configuration of the two-dimensional expansion/contraction matching means. FIG. 9 is a block diagram showing an example of the configuration of , and FIG. 9 is a diagram showing another example of information that can be used as projection information. Mapping function generator stage, 4 is a standard projection information storage section, 5 is a two-dimensional expansion/contraction matching means, 6 is a standard character button storage section, 7 is an identification means, 31 is a distance calculation section, 32 is a recurrence formula calculation section, 33 34 is a minimum value selection unit, 35 is a mapping locus storage unit, 51 is a one-way expansion/contraction normalization means, 52 is a character slam expansion/contraction matcher n21 is a vector distance calculation unit, 522 is a recurrence formula calculation unit, 523 indicates the cumulative difference value storage unit.
(bn(C)
(d) (e) O 2 μs (4) (S) No. 2 B (h) Yoshi 3 μ5 (cL) (b) Year 3 Figure (C) Total 4 Figure / / / 000 / / / l 000 / /
/ (x (L) Fang 5 Ro Fang 6 μs O 7 Symbol 8 En> (

Claims (1)

[Claims] Regarding the method of recognizing input character bangs expressed as two-dimensional mesh-like information, an input character bang storage means for storing the input character bangs, and two different predetermined directions for the input character bangs are provided. A projection information extraction means for extracting projection information that is a series of one-dimensional information for one direction, and a standard that stores standard projection information created in advance for each character type in the same format as the projection information of the input character slam. a projection information storage means, inputting the projection information of the input character button and the standard projection information;
a mapping function generating means for calculating a mapping function that stretches and contracts the two and maximizes the degree of matching; a standard character button storage means for storing two-dimensional mesh-like standard character buttons created in advance for each character type; Using the input character button, the mapping function, and the standard character pattern, the input character pattern is stretched and matched in the one direction by the mapping function, and then stretched and matched in a direction different from the one direction. and an identification means that outputs a recognition result based on the degree of difference between the input character stamp obtained as a result of the expansion and contraction matching and the standard character pattern of each type of character to be read. A character recognition method that is characterized by being able to perform alignment by absorbing dimensional expansion/contraction fluctuations.