JPS62229384A - On-line character recognizing device - Google Patents

Abstract

PURPOSE:To execute the processing at high speed by providing a picture number corresponding parameter setting part to set, for the number of the strokes of an input character, a weighting parameter added to a parameter in case a threshold and a candidate sequence are determined at the time of selecting a candidate character. CONSTITUTION:If a candidate character is classified for the number of the strokes, a normal distribution is generally approximately is obtained with about 10 pictures as a center, and therefore, a weighting parameter stored into a picture number corresponding parameter setting part 9 is weighted and added so that the threshold and the parameter of a middle classifying part 6 to select the candidate character, a part pattern Q value matching part 7 to determine a candidate sequence and a part pattern stroke code distribution matching part 8 can be set to a suitable value in accordance with the number of the stroke of an input character. Then, the optimum stroke corresponding processing is not necessary and the processing can be shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an online character recognition device that identifies written characters in real time.

(Prior Art) The following method is generally used as a recognition method in an online character recognition device. The data string of the handwritten stroke (the handwritten part from Benon to Peno) is
It is expressed in Y coordinate or r, θ polar coordinates, and feature points are extracted from these data (hereinafter referred to as human cover turns). Character recognition was performed by matching information on the extracted feature points with information on a pattern (hereinafter referred to as a W2i pattern) in which feature points were extracted and registered in advance using the same method.

Hereinafter, a conventional online character recognition method will be explained with reference to FIGS. 1O to 12.

First, FIG. 10 shows an example in which feature points are expressed by the start and end points of each stroke. The manually written j-th stroke data string Dj is 1DJ = ((X+-Vt)J) t-1, 2***
nJ'' (1) Qja+j If the number of stroke data is the starting point (Xs, VJj is (X+, Vt), +, the ending point (Xs, 3'e) = (xo,,yo,)J
is required.

j Here, the registered patterns obtained and registered using the same method are assumed to be (Xs, 3/s)J'' and (Xs, ya)J'.

When a simple matching method is adopted, when the number of character strokes is N, the distance d between the input cover pattern and the registered pattern is
is ÷(lxe, -X,,@l+lye, -ye,"l)
= (2) Calculated according to the horn. Then, matching with each registered pattern is performed, and eventually the distance is 1! The one with the smaller ld is output as the recognition result.

(Problems to be Solved by the Invention) However, the conventional online character recognition method using such a matching method has the following problems.

According to this method, a handwritten item with a shape similar to the registered pattern is recognized with a recognition rate of nearly 100%. However, in general, there are many cases in which the handwritten pattern does not have the same shape as the registered pattern. Figure 11 (
A) “River” whose shape is deformed with respect to the registered pattern
An example of a human cover turn is shown, and FIG. 11(b) shows a registered pattern corresponding to this. In this case, "2" and "OK" are shifted up and down, and their respective sizes also vary greatly.

Therefore, the distance lsd between the human cover pattern and the registered pattern becomes large, increasing the possibility of misrecognition.

As described above, according to the conventional method, there is a problem in that when a handwritten input is performed in a shape that is modified from the registered pattern, the misrecognition rate increases.

Furthermore, in matching simple handwritten strokes based on equation (2), the distance @d becomes large when the stroke order of human characters differs. Therefore, in reality, processing is required to optimally associate strokes to be matched.

That is, it is necessary to determine which stroke of the registered pattern should be matched with each manually written stroke. This process consumes a lot of time.

For example, let us consider the optimal correspondence when "shi" is written manually as shown in FIG. 12(a). Figure 12 (b
) is a diagram showing the corresponding registration pattern at this time, and Figure 12 (
C) is an explanatory diagram of optimal stroke matching processing. One way to make the correspondence is to first
Calculate the distance dll between the stroke (■ in Figure 12 (a)) and the first stroke (0 in Figure 12 (b)) of the registered pattern using formula 22, and then calculate the distance 1 from the registered pattern ■1.
11id+z, the distance d13 from the registered pattern ■1 is calculated, and the minimum stroke of d l l t d 12 + d 13 is associated. In this example, dll is the minimum, and ■-■6 is associated as the optimal stroke. Similarly, the second stroke of the human cover turn (■ in Fig. 12(a))
), the strokes of the registered pattern that have already been matched are removed (in this case, the strokes are matched with the remaining strokes), and the respective optimal strokes are matched.

This process takes less time when the number of strokes is small, but for characters with a few strokes (characters with a few strokes), the number of processes M increases at the ratio of , where N is the number of strokes, so it takes a huge amount of time. This results in a large amount of processing. In order to compensate for this increase in processing amount, it is necessary to use a high-speed ALU or the like, so conventional online character recognition devices cannot be miniaturized and have the disadvantage of increasing costs.

As described above, the conventional method is sensitive to changes in the shape of the human cover drawn by the handwriting, and is prone to misrecognition, and also requires processing to perform optimal stroke correspondence in order to relax the stroke order, resulting in a large amount of processing. There was a drawback.

An object of the present invention is to solve the above-mentioned problems of the prior art, and to provide an excellent online character recognition device that is resistant to deformation of input patterns caused by handwriting and requires less processing. do.

(Means for solving the problem) The present invention removes unnecessary data from a coordinate data string obtained by handwriting input on a tablet, and linearizes the data by a surface processing unit that performs linearization processing and a preprocessing unit. A feature point extraction unit that extracts feature points representing characteristics of strokes that make up a written character from a coordinate data string, and a stroke coding unit that codes each stroke based on the positional relationship of the feature points extracted by the feature point extraction unit. In order to solve the problems of the prior art described above, the present invention is aimed at an online character recognition device that performs character recognition by comparing the output data of the stroke coding section with registered pattern data registered in advance. A large classification section that receives the output of the coding section and performs major classification based on the number of strokes of the written characters; The center of gravity of each partial pattern is calculated from the information on the segments connecting the feature points extracted by the point extraction unit, and based on the calculated center of gravity data,! Find the inter-sub-pattern vector whose starting point and end point are the centroid of the sub-pattern and the centroid of another sub-pattern, respectively.
A middle classification section that performs matching with vectors between subpatterns created from two patterns and performs middle classification according to the matching result: For candidate characters narrowed down by the middle classification section, each stroke of each segment of the partial pattern A partial pattern Q value representing the characteristics of the partial pattern extracted from the information is obtained, and this partial pattern Q value is matched with a partial pattern Q value previously created from registered patterns in the same manner. , a partial pattern Q value matching unit that ranks candidate characters according to the value obtained by weighted addition with the matching result by the middle classification unit; , for each partial pattern, calculate the distribution of stroke codes extracted by the stroke coding section, and match this stroke code distribution with a partial pattern stroke code distribution previously created from registered patterns in the same way, A partial pattern stroke code distribution matching unit that performs final ranking of candidate characters according to the value obtained by weighted addition of this matching result, the matching result by the intermediate classification unit, and the partial pattern Q value matching: Stroke For each number, threshold values that serve as criteria for candidate selection in the medium classification section and partial pattern Q value matching section, and weighting parameters used for weighted addition in the partial pattern stroke code distribution matching section are set in advance, and the weighting parameters used for weighted addition in the partial pattern stroke code distribution matching section are set in advance. A stroke number corresponding parameter setting section is provided for appropriately setting each parameter according to the number of strokes of a human character pattern.

(Function) In the present invention, since the online character recognition device is configured as described above, each technical means functions as follows.

The preprocessing unit linearizes the data by performing noise removal processing, moving average processing, smoothing processing, etc. on the character coordinate data string manually written on the tablet. For example, the feature point extraction unit may be configured to
, the characteristic point is extracted by a method such as finding the sign in the y direction and using the change point of the sign state as the characteristic point. The stroke encoding unit classifies strokes based on the extracted feature point information, for example, by the sign of each segment, the angle of the segments, and the rotation angle between the segments, and encodes the strokes.

This coded stroke data is used for the following classification and ranking of each means, which is a feature of the present invention.

The major classification section performs major classification based on the number of strokes of written characters. For this reason, for each number of strokes, characters that can have that number of strokes are stored in advance, for example, in a character dictionary.

The intermediate classification section classifies candidate characters roughly classified according to the number of strokes by matching vectors between partial patterns, and further narrows down the candidate characters. Specifically, a vector between partial patterns is obtained for each candidate character, and this is compared with, for example, an inter-pattern vector in a character dictionary, and the difference is obtained as the distance between both vectors. This distance is compared with a predetermined threshold value set in the stroke number corresponding parameter section, and if it is larger than the threshold value, it is removed from the candidate characters, and if it is less than the threshold value, it is left as a candidate character.

The partial pattern Q value matching section first calculates the partial pattern Q value for the medium-classified candidate characters, compares this with the partial pattern Q value determined in advance from the registered pattern, and calculates the difference as the distance between both Q values. and perform matching. For this reason, the partial pattern Q value obtained in advance from the registered pattern is stored in, for example, a partial pattern dictionary. Next, the partial pattern Q value matching unit performs intermediate classification according to the distance obtained by weighted addition of the above matching result, that is, the distance between both Q values, and the matching result of the intermediate classification unit, that is, the distance between both vectors. The selected candidate characters are ranked and narrowed down using a predetermined threshold set in the stroke number corresponding parameter section.

The partial pattern stroke code distribution matching section calculates the partial pattern stroke code distribution for the top candidate characters narrowed down and ranked as described above, compares this with the stroke code distribution previously calculated from the registered pattern, and compares both distributions. Find the distance between them and perform matching. Then, the distance obtained by this matching and the distance obtained by matching in the intermediate classification part,
The distance obtained by matching in the partial pattern Q value matching section is weighted and added using the weighting parameter set in the stroke number corresponding parameter setting section, and the final ranking is performed based on the obtained distance. This is the recognition result.

In addition, the stroke number corresponding parameter setting section is the intermediate classification section,
It works to set appropriate thresholds and weighting parameters according to the number of strokes for selecting and ranking candidates in the partial pattern Q value matching section and the partial pattern stroke code distribution matching section. Generally, when characters are classified by the number of strokes, the number of candidate characters for each number of strokes is approximately normally distributed around 10 strokes, and the candidate characters vary greatly depending on the number of strokes. Therefore, it is efficient to change the threshold value in the intermediate classification section to a suitable value depending on the number of strokes. In addition, written characters are diverse, ranging from 1-stroke hiragana, katakana, alphabets, numbers, etc. to 25-stroke kanji. It is effective to change the weighting parameters used in the chord distribution matching section to suitable values depending on the number of strokes. Based on the above, appropriate threshold values and weighting parameters are set for the stroke number corresponding parameters in accordance with the stroke number.

Therefore, by the functions of each means of the present invention, it is possible to cope with fluctuations due to variations in character shapes input by hand, eliminate the need for optimal matching processing of strokes due to differences in stroke order, reduce the amount of processing, and reduce the amount of processing compared to the above-mentioned prior art. Problems are resolved.

(Embodiment) Hereinafter, an online character recognition device according to an embodiment of the present invention will be described in detail.

FIG. 1 is a block diagram showing the configuration of the online character recognition device of this embodiment. As shown in the figure, this online character recognition device includes a tablet 1, a preprocessing section 2, a feature point extraction section 3, a stroke coding section 4, a major classification section 5, a medium classification section 6, and a partial pattern Q value matching section. 7. It is composed of a partial pattern stroke code distribution matching section 8 and a stroke number corresponding parameter setting section 9. Note that IO in the figure indicates an output terminal to a display device or the like (not shown).

FIG. 2 is a flowchart showing an outline of the operation of the apparatus shown in FIG. The major classification section 5 executes step 104, and the intermediate classification section 6 executes steps 105 to 1.
08, the partial pattern Q value matching section 7 executes steps 109 to 112, and the partial pattern stroke code distribution matching section 8 executes steps 113 to II&. Note that the stroke number corresponding parameter 9 is set for each part of 6 to 8 in accordance with the number of strokes of the input character.

FIG. 3 is a diagram showing the structure of a character dictionary included in the apparatus of this embodiment, and this dictionary allows characters to be selected based on the number of strokes. Further, FIG. 4 is a diagram showing the configuration of a partial pattern dictionary included in the apparatus of this embodiment.

Hereinafter, the operation of the apparatus of this embodiment will be explained in detail step by step for each part.

First, the tablet 1 is used for manually writing characters, and when characters are input by hand, a writing data string ((Xt, L)j-1, 2...nJ, is extracted and sent to the preprocessing section 2.

The preprocessing unit 2 linearizes the data as shown in FIG. 5(b) by performing noise removal processing, moving average processing, and smoothing processing on the received writing data string, and then the feature point extraction unit 3 Output to.

Next, the operation of the feature point extraction section 3 will be described. There are several methods for feature point extraction processing performed by the feature point extraction unit 3, but here we will use a linearized data string ((ma + - Vi
), l-1, 2""nJ) X between the data of J.

A method of calculating the sign (positive, negative, and 0 signs) in the y direction and extracting points of change in the sign state as feature points will be described.

The x and y direction sine XS, YS, between the data is YSz
=Sign(Yi-3'+-t), +, 0. −
Expressed as Compare the signs in the X direction and X direction between each piece of data obtained in this way with the signs between the previous data, and if they are the same, they are not registered as feature points, and if they are different, the state is considered to have changed and the feature point is Register as. Figure 5 (C)
In addition to the points obtained in this way, the characteristic points are shown including the starting point and the ending point. Generally, this process is sometimes called linear approximation. The space between these feature points will hereinafter be referred to as a segment, and the feature points will be expressed as ((X+, Yi), i-1, 2...fj)j.

The feature point information obtained as described above is output to the stroke coding section 4 and the medium classification section 6.

Next, the operation of the stroke coding section 4 will be described. The stroke encoding section 4 encodes each stroke based on the feature point information obtained by the feature point extraction section 3. There are many methods for this encoding, but in general, the encoding is performed by classifying the X and Y signs of each segment, the angle of the segment, and the rotation angle between the segments. FIG. 6 is an explanatory diagram of this encoding process, in which θ1. θ2. θ3 indicates the angle of the segment (the angle formed in the +xX direction), and θ1゜θ
2 indicates the rotation angle between adjacent segments.

The encoded stroke data is output to the major classification section 5EL and the partial pattern stroke code distribution matching section 8.

Next, the operation of the major classification section 5 will be described. The major classification section 5 receives the output from the stroke encoding section 4 and performs major classification of the target characters based on the number of strokes. Therefore, for each number of strokes, characters that correspond to that number of strokes are prepared in advance in a character dictionary as shown in FIG. For example, suppose that the number of strokes of a character pattern drawn manually is 10 strokes. In this case, among the characters stored in the character dictionary, the character ``唖'', which can be 10 strokes as shown in Figure 3, is
, "Ko", "Ai", etc. are selected as candidate characters.

Next, the operation of the intermediate classification section 6 will be described. The middle classification section 6 is
The candidate characters obtained by major classification according to the number of strokes in the major classification section 5 are further divided into medium classifications using vectors between partial patterns, which will be described below. Here, a partial pattern refers to a part of one character that is written as a series of characters, and a vector between partial patterns refers to the center of gravity of one partial pattern and the center of gravity of another partial pattern as the starting point and ending point, respectively. is the vector.

First, an example of a method for calculating vectors between partial patterns will be described. X for each segment in the partial pattern.

When the y component is (dx+, dy+), the length dl of each segment is d i r = dx + + d'J r-...(5
). Also, the center coordinates of each segment are normalized by dividing by the character width HX, HY (x+”, y
t”), the centroid coordinates of the partial pattern (x,, y=
) is given by. The centroids of the external partial patterns are determined by the above method, and vectors between partial patterns are determined by setting the centroids of one partial pattern and another partial pattern as starting points and ending points, respectively. Note that here, the inter-partial pattern vectors are considered separately for the X direction and the X direction.

FIG. 7 shows the partial pattern “a” of the character “唖”, the centroid of each of the centroids of “vomit” and the vector between the partial patterns “; (mouth, t),”;
An example of Ca, t) is shown below.

The intermediate classification unit 6 narrows down the candidate characters selected by the major classification unit 5 using the vectors between sub-patterns described above and performs intermediate classification. Considering this case, the procedure for intermediate classification for this human-powered character will be explained below.

Since the hand-written character "Ai" has 10 strokes, first refer to the 10 stroke portion of the character dictionary as shown in FIG. Then, the character "唖" is placed first, and in that column, the partial patterns that make up "唖", the handwriting of the partial patterns, and the stroke number information of each partial pattern (hereinafter referred to as cut position) are displayed. ), and vector values between each partial pattern calculated in advance from the registered pattern are shown. Similar information is arranged for the characters "Ko" and "Ai" in the following order, and the intermediate classification unit 6 determines whether or not they should be candidates according to this order of characters, and performs intermediate classification as follows.

First, assuming that the handwritten character is "唖", the matching distance d vec of vectors between partial patterns is determined. As stated in the character dictionary in Figure 3, the cut position of "Muta" is (3, 7), that is, the "mouth" is formed from the 1st stroke to the 3rd stroke, and the "mouth" is formed from the 4th stroke to the 1st θ.
A "break" is formed by the stroke, but in this example, the incoming cover turn is "Ai", so if we consider the vector between partial patterns for "Ai" at this cut position, the result will be as shown in FIG. Vector length 1 between sub-patterns of the input pattern obtained as shown in FIG.

The difference between the partial pattern vector X (mouth, ri, y (o, ken)) written in the dictionary is the matching distance d VeC, and , A) I+l?” (B, Ken) - 7 (B, Kinzuki) Calculated.

Generally, when the number of partial patterns of a written character is multiple,
Normalize with the number of partial patterns BPN and find the matching distance!
11dvec is calculated according to:

Here, a certain threshold VECREJ is set and the calculated d v
It is determined whether ec is greater than VECREJ. When dvIIc > VECREJ, the referenced character (
In this case, it is assumed that the character is not "唖"), and the vectors between the partial patterns of the next character are matched. d Vae≦V
In the case of ECREJ, calculation and matching of the partial pattern Q value, which will be explained next, is performed on the assumption that it seems to be "dumb".

The threshold value VECREJ is set in advance in the stroke count corresponding parameter setting unit 9 for each stroke count, and the value is set according to the stroke count of the input pattern at the time of recognition.

Since the number of candidate characters differs depending on the number of strokes, for example, in a stroke number with a small number of candidate characters, even if the VECREJ value is increased, the number of candidates remaining in the medium classification by the vector matching between partial patterns will be about the same number as candidates for other stroke numbers. Therefore, it is better to set a small VE(:REJ value) for strokes with a small number of candidates, and to set a large VE(:REJ value for strokes with a large number of candidates).

Next, the operation of the partial pattern Q value matching section 7 will be described. The partial pattern Q value matching unit 7 calculates the partial pattern Q value for the candidate characters remaining after the intermediate classification using the vectors between partial patterns in the intermediate classification unit 6, and calculates the partial pattern Q value in the partial pattern dictionary shown in FIG. Perform matching. The partial pattern Q values of this partial pattern dictionary are created and stored in advance from registered patterns. Here, the partial pattern Q value refers to a characteristic parameter representing the length, direction, and position of each segment. In online character recognition, the X, Y direction, + or - direction is obtained as important information regarding the movement of the writing pen, and this partial pattern Q value is an effective use of this information.

First, a method for calculating the partial pattern Q value will be explained.

Note that in the formula below, Σ indicates addition regarding all strokes and all segments. Further, HX and HY indicate character width.

■ X-direction position of the +X-direction component■ -X-direction position of the X-direction component■ +X-direction position of the y-direction component■-X-direction position of the y-direction component■X-direction position of the +X-direction component■-Y direction of the X-direction component Position■ +X-direction position of the y-direction component■ -X-direction position of the y-direction component■~■ In the case of the origin, it is the value of each direction position when the origin is set to the lower left, but in this case, the value near the origin is multiplied. When subjected to , it becomes 0, so in order to prevent it from becoming 0, change the origin and set the origin to the upper right, and write the values Q9 to Q16 of each direction position in the same way, and Q + "=
It is assumed that a total of 16 values of Q + s represent the length, direction, and position of each stroke segment of the target character.

The partial pattern Q value matching unit 7 calculates the partial pattern Q value explained above for the remaining blanks by classification based on the vector between partial patterns.
Suppose that ``Ai'' is manually written down and left as ``Ko'' or classified by vectors between partial patterns.In this case, the cut position of ``Ko'' is (3, 2, 5) as shown in Figure 3, and the partial The pattern is J+enter+shu, so cut the human cover turn at the cut position (3*2, 5) and cut each Q
+” to Q166. At this time, in this example, the 9th
As shown in the figure, "A" or "sai" corresponds to "2", "2" corresponds to "mu", and "24" corresponds to "su". The calculated partial pattern Q values QI' to Q16' are matched with the partial pattern Q values in the partial pattern dictionary shown in FIG. 4. In other words, first, calculate QI'~Q+ using "mu".
o' and Q1~Q+g of the partial pattern dictionary "O" are matched, then Qge~Q,6' calculated in "2" is matched with Q l -Q +a of the partial pattern dictionary "Mu", and then Match Q1'' to Q161 calculated in "Obu" with Q! to Qr6 of the partial pattern dictionary "table". The sum of the differences in these matchings is set as a matching distance of 11dap. In this case, the distance is 911 d n
p represents how close the introductory pattern "Ai" is to "Ko".

Generally, weighting is performed as shown in equation (17) using the number of strokes BSJ of each partial pattern, and this is set as the matching distance 1fl d a p.

Q+ (j); Q, value of the fifth partial pattern, and the weighting parameter Wv0 of equation (18) for each number of strokes in advance
°, wBp are determined and stored in the stroke number corresponding parameter setting section 9, and weighting is performed by the weighting parameter value set by the stroke number parameter setting section 9 according to the number of strokes of the input pattern at the time of recognition. Then, the distance dap obtained as above and d v, , e obtained by matching the inter-partial pattern vectors obtained in the previous step are weighted by Wv, , c and wlip, respectively, and the distance is added. Find d1.

di”Wvec”dvac”WBP”dBF m-
(to) The above operations are performed on all candidate characters remaining after classification by partial pattern vectors, and sorting by dl is performed.

Next, partial pattern stroke code distribution matching section 8
We will describe the operation of. The partial pattern stroke code distribution matching section 8 is the partial pattern Q value matching section 7.
In response to the output from the stroke coding unit 4, a partial badan stroke code distribution is obtained for the candidate characters narrowed down by the intermediate classification, and is created in advance from this distribution and the registered pattern and stored in the partial pattern dictionary shown in FIG. Matching with the partial pattern stroke code distribution is performed, and the top candidates are further ranked. The range of targets for this ranking is determined, for example, by the ratio of d to the first complementary distance fa dr obtained by sorting. That is, ranking is performed with candidate characters of dJ/dI≦ZRATE as the target range. Here, the threshold value ZRATE is determined in advance by the stroke number corresponding parameter setting unit 9 for each stroke number of the written character.
, and at the time of recognition, the value is set according to the number of strokes in the input pattern.

In the case of a character with a small number of strokes, the amount of information in the vector between partial patterns and the Q value of the partial patterns is small. Therefore, in this case, the ranking determined by dl in equation (I8) tends to be uncertain, and the ZI'tATE value in the partial pattern stroke code distribution matching section 8 also needs to be large. On the other hand, for characters with many strokes,
For the same reason, it is preferable to make the ZRATE value small.

Here, a method for calculating the partial pattern stroke code distribution will be explained. - As an example, assume that the input cover turn is "Ai" and the character selected as the first candidate is "Ai". From the character dictionary in Figure 3, the cut position of the candidate character “Ai” is (3
, 4, 3), the partial pattern is found to be %+4+L, so cut the incoming cover turn "Ai" at this position. In this case, the partial patterns obtained by cutting are the same as the contents of the character dictionary, but the distribution of the number of stroke codes obtained by the stroke coding section 4 is calculated for each partial pattern. For example, if you look at the partial pattern of “゛°,” “/”.

7”, “\”, so 501 is one tree, S03 is one tree,
A stroke code distribution with one SO5 is required.

The partial pattern stroke code portion I5 calculated in this way is based on the partial pattern stroke code distribution of the partial pattern dictionary shown in Fig. 4, which was created by previously calculating and averaging several Rjl patterns using the same procedure. Matched. In the case of this example, first, the difference between the partial pattern stroke code distribution of "mu" calculated as above and the stroke code distribution of "n" in the partial pattern dictionary in FIG. ). In other words, Hitokabataan Partial Pattern Dictionary 5ol-1 book SO! -0,9 pieces 5O2 -0 pieces SO2
・-0,1 SO3-1 5O3-0,4 5
O4-0 5O4-0, 6 5O5-1
5O5-1 difference 0. I+ 0.1+ 0.6+ 0.6=1.
Four lines are obtained as d s (4). Similarly, “椹”
, “L” is matched and d s ($),
dSL) is obtained. Then, these are normalized by the number of stroke codes of each partial pattern BS, and the total distance gl d of the matching distance of each normalized partial pattern is
s is further calculated.

Next, for the distance @ d v, e obtained by vector matching between partial patterns, the distance by partial pattern Q value matching (iap, and the distance 1lid s by partial pattern stroke code distribution matching), weighting parameters WVe, WBp , W, the distance value is calculated as shown in equation (20).

Here, the weighting parameters wvclc, wBp, w,
is a parameter for which the optimum value of the weighting parameter is calculated in advance for each number of strokes and stored in the number-of-strokes corresponding parameter setting section 9, and the weighting is set by the number-of-strokes corresponding parameter setting section 9 according to the number of strokes of the human covert. is a parameter value.

D = W vee” d vec”W ap' d
ap” Ws・d ap ... (20) Find this distance for each candidate character to d, /d, ≦ZR8T,
The candidate characters are ranked according to the distances obtained, and the recognition results are output from the output terminal IO to a display device or the like.

Although not mentioned in the above embodiment, partial patterns "Fu" and "Dust" are used for those whose handwriting shapes are very similar in the ranked recognition results, such as "Dust" and "Dust".
A more reliable recognition result can be obtained by providing a detailed check section that recognizes "dust" if there is an intersection of the ○ marks as a difference between "a" and "dust" if there is no difference.

Furthermore, in the description of the above embodiment, an example was shown in which each parameter value is set according to the number of strokes of a human kataan, but it is also possible to classify each character into types and set parameters according to the types.

Furthermore, in addition to narrowing down the candidates by matching vectors between partial patterns in the intermediate classification unit 6, a process is added in which characters are not considered candidates when the distance is greater than a certain threshold using partial pattern stroke code distribution matching, thereby narrowing down candidate characters for post-processing. It is also possible to do so.

(Effects of the Invention) As described in detail above, according to the present invention, by extracting and matching the feature parameters for each partial pattern, the variation in shape created by hand writing, which is a drawback of the pattern matching method, can be eliminated. In particular, since fluctuations in distance values due to positional deviations between partial patterns constituting characters are absorbed, erroneous recognition caused by these fluctuations can be prevented. Also,
By narrowing down the candidates by matching vectors between partial patterns and by matching using the partial parameter Q value representing the information of each segment of each stroke for each partial pattern, the optimum stroke matching process becomes unnecessary and the process is shortened. As described above, the present invention can provide an excellent online character recognition device that is resistant to variations in writing shapes and that requires less processing.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an online character recognition device according to an embodiment of the present invention, FIG. 2 is an operation flowchart of the device shown in FIG. 1, FIG. 3 is a diagram showing an example of the configuration of a character dictionary, and FIG. The figure shows an example of the structure of a partial pattern dictionary, FIG. 5 is an explanatory diagram of preprocessing, FIG. 6 is an explanatory diagram of stroke encoding processing, FIG. 7 is an explanatory diagram of vectors between partial patterns, and FIG. FIG. 9 is an explanatory diagram of calculation of vectors between partial patterns. FIG. 9 is an explanatory diagram of partial pattern Q value calculation. FIG. FIG. 12 is an explanatory diagram of stroke association processing. l...Tablet 2--Surface processing section 3-Feature point extraction section 4...S]-Loaf coding section 5...Main classification section 6...Medium classification section 7...Partial pattern Q value Matching unit 8...Partial pattern stroke code distribution Matching unit 9...Number of strokes corresponding parameter setting unit i o-...Output terminal

Claims (1)

[Scope of Claims] A preprocessing unit that removes unnecessary data from a coordinate data string obtained by handwritten input on a tablet and performs a linearization process, and a written character from the coordinate data string linearized by the preprocessing unit. A feature point extraction unit that extracts feature points representing the characteristics of the strokes that make up the stroke, and a stroke coding unit that codes each stroke based on the positional relationship of the feature points extracted by the feature point extraction unit. In an online character recognition device that performs character recognition by comparing the output data of the section with registered pattern data that has been registered in advance, there is a large classification section that receives the output of the stroke coding section and roughly categorizes the written characters according to the number of strokes. For the candidate characters selected by the major classification section, the part that is written as a series in handwriting is defined as a partial pattern, and the center of gravity of each partial pattern is calculated from the information on the segments connecting the feature points extracted by the feature point extraction section. , Based on the calculated center of gravity data, find a vector between partial patterns whose starting and ending points are the center of gravity of one partial pattern and the center of gravity of another partial pattern, respectively, and create a vector between these partial patterns and a registered pattern in the same manner in advance. The intermediate classification section performs matching with the vectors between partial patterns that have been prepared, and performs intermediate classification according to the matching results.The intermediate classification section performs intermediate classification based on the matching results. Find a partial pattern Q value that represents the characteristics of the partial pattern, match this partial pattern Q value with a partial pattern Q value created in advance from registered patterns in the same way, and use this matching result and the intermediate classification section. The partial pattern Q value matching section ranks the candidate characters according to the value obtained by weighted addition with the matching result of Next, the stroke code distribution extracted by the stroke coding section is calculated, and this stroke code distribution is matched with a partial pattern stroke code distribution previously created from registered patterns in the same manner. , a partial pattern stroke code distribution matching unit that performs final ranking of candidate characters according to the value obtained by weighted addition of the matching result by the medium classification unit and the partial pattern Q value matching, and for each number of strokes, The threshold values used as criteria for candidate selection in the medium classification section and partial pattern Q value matching section and the weighting parameters used for weighted addition in the partial pattern stroke code distribution matching section are set in advance, and the weighting parameters used for weighted addition in the partial pattern stroke code distribution matching section are set in advance. An online character recognition device comprising: a stroke number corresponding parameter setting unit that appropriately sets each parameter according to the number of strokes.