JPH06301819A - Online handwriting recognition device - Google Patents

Abstract

(57) [Summary] [Purpose] Recognize continuous characters and handwritten characters with stroke order variation, and identify by characters with characteristic curved parts such as hiragana and stroke bending methods such as "shi" and "re". Easy and fast recognition of recognized characters. [Structure] An input coordinate data string input to a handwriting input circuit 4 is divided into segments by a segment dividing unit 1, and a feature amount such as a position, a length, a direction, and a curvature is obtained. Based on this, the partial handwriting extraction means 2 extracts the partial handwriting. continue,
The character identification means 3 uses the character dictionary 10 for the partial handwriting.
Characters are identified by determining the identification conditions such as the presence / absence of partial handwriting and the relationship between the partial handwritings described in. The curve portion can be extracted as a partial handwriting by using the curve information such as the curvature of the segment.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention provides free handwritten characters including stroke continuation, omission, and stroke order error in stroke count,
The present invention relates to an online handwritten character recognition device that flexibly recognizes changes in stroke order and the like.

[0002]

2. Description of the Related Art Conventionally, it is determined which basic stroke stored in a dictionary matches an input stroke, and a set of input basic strokes for one character is regarded as a basic stroke stored in the dictionary. There are online handwritten character recognition devices that perform character recognition by comparison. (Reference: Nikkei Electronics 1983.12.5 pp1
15-133) However, in this method, in order to recognize characters deformed by continuous characters or broken characters or characters with incorrect stroke order, various basic strokes can be registered in the dictionary or a large number of characters can be stored for one character. It is necessary to describe in the dictionary the set of basic strokes of.

In order to solve this drawback, Japanese Patent Laid-Open No. 2-56
As described in No. 689, the input coordinate data string is converted into linear element data, the partial handwriting which is a part of the handwriting is determined based on the linear element data according to the content of the character dictionary, and based on this determination result. There is an on-line handwritten character recognition device that recognizes a character by using a handwriting.

[0004]

In the conventional character recognition method based on the determination of the partial handwriting, since the partial handwriting is determined based on the straight line element data, it is difficult to determine the curved portion, and the characteristic of the curved portion such as hiragana is characteristic. There was a problem that it was difficult to recognize certain characters. For the same reason, recognition is possible for characters that have almost the same stroke structure, such as "shi" and "re", "2", "ko", and "tsuru", and are identified by the way the stroke is bent. There was a problem that it was difficult.

Further, since the partial handwriting is determined according to the contents of the character dictionary written separately for each character, the same partial handwriting is repeatedly determined for one input character, and the recognition speed is slow. It also had the problem of becoming.

[0006]

According to the present invention, an input coordinate data string is divided into appropriate sections (hereinafter referred to as "segments"), and feature quantities such as position, length, direction and curvature of the segment are calculated, Partial handwriting is extracted based on. The recognition of characters is performed by determining an identification condition such as the presence / absence of partial handwriting described in the dictionary and the relationship between the partial handwritings with respect to the partial handwriting.

Since the segment which is the basis of the determination of the partial handwriting also holds information as a curve, it is possible to determine the curved part handwriting. Further, by using the curve information of the segment, it is possible to identify the character identified by the way the stroke is bent.

Further, the determination of the partial handwriting, which was performed according to the character dictionary in the above-mentioned conventional example, can be performed regardless of the contents of the character dictionary, and the partial handwriting which is performed by duplicating some characters can be performed. Since the determination can be made only once, the recognition speed is increased.

[0009]

【Example】

(Embodiment 1) FIG. 1 is a block diagram of an online handwritten character recognition apparatus of the present invention. The handwriting input circuit 4 is a coordinate input circuit using a tablet and a pen. The segment dividing means 1 is composed of the segment dividing circuit 5 and the segment feature amount calculating circuit 6, the partial handwriting extracting means 2 is composed of the partial handwriting extraction control circuit 7 and the partial handwriting extracting circuit 8, and the character identifying means 3 is composed of the character identifying circuit 9 and the character. It is composed of the dictionary 10.

The segment dividing means 1 receives the input coordinate data string from the handwritten character input circuit 4, divides it into segments, calculates the characteristic amount of each segment, and outputs it. The input coordinate data string obtained by the handwriting input circuit 4 is divided by the segment dividing circuit 5 using the polygonal line approximation into segments having one or more coordinate data as elements. Then, the segment feature amount calculation circuit 6 calculates and outputs the feature amount of each segment.

The partial handwriting extracting means 2 receives the segment feature amount from the segment dividing means 1, extracts a predetermined partial handwriting, and outputs it. The partial handwriting extraction control circuit 7 operates the partial handwriting extraction circuit 8 by specifying the extraction condition and the selection condition together with the feature amount of each segment received from the segment dividing means 1 to extract the determined partial handwriting. The extraction condition is a condition such as what kind of partial handwriting is extracted, and is a condition such as "downward straight handwriting". On the other hand, the selection condition is a condition of which partial handwriting is selected and output when a plurality of partial handwritings that match the extraction condition are matched, and is a condition such as "longest handwriting". The partial handwriting extraction circuit 8
Receive segment feature amount, extraction condition, selection condition,
Partial handwriting is extracted by combining the segments that meet the extraction condition based on the feature amount of the segment, and the partial handwriting that satisfies the selection condition is output from the extracted partial handwriting. The partial handwriting extraction circuit 8 may output 0 or a plurality of partial handwritings. All the partial handwritings extracted by the partial handwriting extracting means 2 are sent to the character identifying means 3.

The character identifying means 3 receives the partial handwriting from the partial handwriting extracting means 2, identifies the character according to the description in the character dictionary 10, and outputs the identified character. The character dictionary 10 describes identification conditions such as the existence of partial handwriting and the relationship between the partial handwriting necessary for identifying a character. The character identification circuit 9 determines the identification condition described in the character dictionary 10 for the partial handwriting received from the partial handwriting extraction unit 2, and outputs the identified character when the condition is satisfied.

FIG. 2 is a diagram showing the segment feature amount of the present invention. The following seven are used as the segment feature amount.

Characteristic 1. Start point coordinates (parameter P21) 2. End point coordinates (parameter P22) 3. Length (parameters P23, L) 4.8 Direction from the oriented start point to the end point (parameter P24) 5. The ratio of the maximum value (parameter P25, h) of the distance from the line segment connecting the start point and the end point to each element and the length of the line segment h / L (curvature) 6. When moving from the start point to the end point on the line segment, each element moves to the left or right Which is the distribution (left / right distribution) 7. Whether the direction difference between the previous segment and the current segment is 1 or less in 8 directions, and whether it can be connected to the previous segment to form a straight line handwriting in the direction of the previous segment ( Possibility of connection) The feature quantity 5 (curvature) is that the larger the ratio of the maximum value of the distance from the line segment connecting the start point and the end point to the element and the length of the line segment, the more the line segment and the element are separated, and the segment is more curved. Showing that A feature quantity, the feature quantity 6 (right and left distribution) is a feature amount indicating whether bending left or segment bending right.

FIG. 3 is a flowchart showing the operation of the partial handwriting extraction circuit 8. The partial handwriting extraction circuit 8 performs the step S31,
In S32, the segments are scanned one by one in the input order. In step S33, it is determined whether the segment extracted in step S31 satisfies the extraction condition. If the condition is met, the partial handwriting currently being extracted is combined in step S34. In step S34, if there is no partial handwriting currently being extracted, it is stored as a new partial handwriting and extraction is started. When scanning of all the segments is completed in step S32, the process proceeds to step S35 to select and output a partial handwriting satisfying the selection condition from the stored partial handwriting.

FIG. 4 is a flow chart showing the operation of the partial handwriting extraction circuit 8 when instructed to extract the "longest clockwise curve handwriting". The extraction condition is “clockwise handwriting” and the selection condition is “longest handwriting”. Figure 3
Step S33, which was determining the extraction condition in step S, is specifically expanded to steps S43, S44, and S45. Steps S43 and S44 determine the condition that the segment is curved, that is, curved handwriting, and step S45 determines the condition that the segment is curved right, that is, clockwise writing. The partial handwriting extraction circuit 8 scans the segments one by one in the order of input in steps S41 and S42. For the segment extracted in step S41, it is determined in step S43 whether the feature amount 5 (curvature) is large, step S44.
Then, it is checked whether the feature amount 6 (left-right distribution) is biased to the left or right, and if either is satisfied, it is determined that this segment is curved, and step S
Proceed to 45. In step S45, it is checked whether the feature amount 6 (horizontal distribution) is biased to the left, that is, whether this segment is bent to the right. If it is satisfied, the extraction condition is satisfied, so that the partial handwriting being extracted is combined in step S46. In step S46, if there is no handwriting currently being extracted, it is stored as a new partial handwriting. Step S42
When scanning of all the segments is completed in step S47, the process proceeds to step S47, and the longest partial handwriting is selected from the extracted partial handwriting and is output.

FIG. 5 is a conceptual diagram showing a part of the character dictionary 10. In the character dictionary 10, identification conditions such as presence / absence of partial handwriting, positional relationship between partial handwriting, and input order are described in a tree. The character dictionary 10 has a plurality of trees as shown in the figure, and each has a root 51. Characters are divided into two sets depending on the success or failure of each identification condition, and the identification conditions are arranged so that the two sets formed by them are similar character sets. For example, "a", "me", "o", "no", and "ke" and "i" are considered to be a set of similar characters, but in the figure, the above two sets are set under the discrimination condition of "there is a long clockwise curve handwriting". Is made.

In the character dictionary 10, the minimum handwriting required to identify a character is described as an identification condition, and not all strokes forming a character are described as an identification condition. For example, the stroke at the upper right of "o" is an unstable stroke in which the position, direction, and length are completely different depending on the writer, but there is no discriminating condition regarding the stroke in the character dictionary 10. By identifying using partial handwritings other than the strokes, the influence of unstable strokes by the writer is avoided.

FIG. 6 is a flowchart showing the operation of the character identification circuit 9. The character identification circuit 9 reads the character dictionary 10 in step S61, and in step S63, the root 51 of the tree.
The individual identification conditions are read one by one according to the determination result of the identification conditions. When the read identification condition is a condition regarding the presence or absence of partial handwriting (step S65),
The partial handwriting extracted from the partial handwriting extracting unit 2 is searched for a partial handwriting satisfying the identification condition (step S
66) If it exists, the partial handwriting is stored, and the process proceeds to step S67 to continue the determination of the identification condition. If it does not exist, the process returns to step S62. If the identification condition is not the condition regarding the presence or absence of the partial handwriting in step S65, the process proceeds to step S67. Subsequently, when the identification condition is a condition regarding the relationship between the partial handwritings (step S67),
In step S68, a target partial handwriting is selected from the partial handwritings retrieved and stored in step S66 so far and it is determined whether the identification condition is satisfied. If the condition between the partial handwritings is satisfied, the process returns to step S63, the next identification condition is read, the determination of the condition is continued, and if the condition is not satisfied, the process returns to step S62. If the identification condition is not the relationship between the partial handwritings in step S67, step S67
Returning to 63, the next identification condition is read. When the end of the tree of the character dictionary is reached, there is no further identification condition in step S64, so the identified character is output (step S69) and the process returns to step S62. Step S62
Now, find the root of the next tree and follow the identification conditions. If the character dictionary ends and there is no next tree,
End character identification.

FIG. 7 to FIG. 10 are examples showing processing steps using the input of the continuous character "A". 71, 9 in the figure
Reference numerals 1, 101 denote input coordinate points, and 72, 92, 102 denote segment boundaries.

FIG. 7 is a diagram showing a processing example of the segment dividing circuit 5, in which the segment 73 is obtained by dividing the input handwriting by performing polygonal line approximation. As a polygonal line approximation method, reference is made to “Online handwritten kanji recognition by two types of feature point analysis” (The Institute of Electronics, Information and Communication Engineers, D-II Vol.
J-72-D-II No.9 pp.1432-1440) is an example.

FIG. 8 is a diagram showing a processing example of the segment feature amount calculation circuit 6, showing an example of the obtained feature amount. The segment shown in the figure is a part of the stroke that curves from the upward direction to the right direction.
This is a segment whose end point is point B. In the coordinate system where the origin is at the lower left of the character, the x axis is to the right, and the y axis is to the up,
Feature 1 (starting point coordinates) is the coordinate of point A (18,84), feature 2
(End point coordinate) is the coordinate of point B (69, 106). Further, the feature amount 3 (length) is the distance 55 of the point AB, and the feature amount 4 (direction) is 8
Turn to the upper right. The feature amount 5 (degree of curvature) is 4.1%, which indicates that this segment is slightly curved. Feature 6 (horizontal distribution) is 84% on the left side and 16% on the right side with respect to the line segment from the start point to the end point, and it is distributed to the left side so that this segment is a right turn. Is shown. The feature amount 7 (connectability) is above, the direction of the immediately preceding segment is up, and the direction of the current segment is upper right, so the difference is 8 or less in 1 direction,
It shows that it can be an upward straight line handwriting.

FIG. 9 is a diagram showing a processing example of the partial handwriting extraction means 8, and is an example in which all the linear handwritings in eight directions and the longest clockwise handwriting are extracted as the partial handwriting. 9 in the figure
3 is the extracted handwriting. Since the partial handwritings can share and overlap the segment, all the partial handwritings that satisfy the extraction condition are extracted.

FIG. 10 is a diagram showing a processing example of the character identification circuit 9, which is an example in which the identification condition described in the character dictionary 10 is determined. In the figure, reference numerals 103, 104, and 105 are partial handwritings used for the determination of the identification condition. When the character dictionary 10 shown in FIG. 5 is traced for "a", there are 1. a clockwise curve handwriting in which the handwriting entry is downward, and 2. a sufficiently long downward line handwriting in the center. , 3. And if there is a right-handed straight handwriting in the center of the top,
"Ah".

The character identification circuit 9 describes three partial handwritings as a partial handwriting necessary for identifying the hiragana "A", a clockwise curve handwriting 103, and a downward straight line handwriting 10.
4. The right handwriting 105 is selected to determine the identification condition. As a result, the input character shown in FIG. 8 is identified as the hiragana “A”.

As described above, according to the present invention, a curved handwriting such as a clockwise handwriting can be extracted as a partial handwriting, so that a character such as a hiragana characterizing the curved portion can be recognized.

Since the conventional extraction of the curve handwriting based on the straight line element data is performed by tracking while the angular change of the continuous segment is within the predetermined value,
For example, a portion with a large angle change, such as the lower left tip of "a", could not be continuously extracted. In order to solve this, the segment was made small enough so that the angle change of the segment was within the default value, but if the segment is made small, the number of segments in the input character increases and the processing time becomes slow. On the other hand, according to the present invention, since it is possible to extract the curve handwriting by using a feature other than the angle change of continuous segments, it is possible to continuously extract a portion with a large angle change, and the processing can be performed by increasing the number of segments. You can avoid slowing down.

Further, a recognition candidate character can be easily obtained by using a set of similar characters created in the character dictionary. For example, in FIG. 5, "a", "me", "o", "no", and "ke" and "i" each form a set of similar characters, so that for the input character identified as "a", "me" and "me"""
“No” can be a recognition candidate. In fact, handwritten characters such as “a” and “me” are often similar, and by using similar characters as candidate characters, the recognition rate including candidates can be easily increased.

In the embodiment, the segment division is performed based on the polygonal line approximation. However, the input stroke is divided into fixed lengths, the horizontal and vertical equal intervals are divided, or the fixed strokes are divided. A method or the like may be used.

Further, the segment characteristic amount calculation circuit 6 calculates the seven characteristic amounts shown in FIG. 2, but in addition to this, the characteristic amount such as the time required for writing a segment and whether or not it can be a curve handwriting is calculated. It is also possible to ask for.

(Embodiment 2) In the previous embodiment, the partial handwriting extraction means 2 extracts all predetermined partial handwriting irrespective of the character to be identified, and the character identification means 3 discriminates against these partial handwriting. We determined the identification conditions such as the presence or absence of partial handwriting required for. However, in this method, since the predetermined partial handwriting is extracted, the partial handwriting most suitable for identification cannot always be extracted. Further, partial handwriting that is not used for character recognition and is unnecessarily extracted may be generated to reduce recognition speed. In the present embodiment, the character identifying means 3 extracts the partial handwriting by using the partial handwriting extracting means 2 as needed, thereby enabling character identification without extracting the partial handwriting in advance, The partial handwriting suitable for identification is extracted, and the partial handwriting to be extracted is reduced. Further, at this time, by extracting not only all the strokes of the character as shown in FIG. 5A but only the partial handwriting necessary for identifying the character, it is possible to further reduce the partial handwriting to be extracted.

FIG. 11 shows such an online handwritten character recognition apparatus of this embodiment. The difference from the previous embodiment of FIG. 1 is that the partial handwriting extraction means 113 has a segment division means 11
Instead of directly receiving the segment feature amount output from No. 1 and performing the process, the character identifying unit 112 receives the segment feature amount and causes the partial handwriting extraction unit 113 to process the feature amount as necessary. Each circuit other than the character identification circuit 117, that is, the handwriting input circuit 11
4, segment division circuit 115, segment feature amount calculation circuit 116, character dictionary 118, partial handwriting extraction circuit 119
The operation of is the same as that of the previous embodiment. Character identification circuit 117
Upon receiving the segment feature amount, the sub-feature extraction circuit 119 operates in accordance with the contents of the character dictionary 118 to extract a sub-handwriting required for identifying a character. At this time, the character identification circuit 117 causes the character dictionary 118 to
The partial handwriting extraction circuit 119 is operated by using the identification conditions such as the position, direction, and length of the partial handwriting described in 1) as extraction conditions and selection conditions. The partial handwriting extraction circuit 119 outputs the partial handwriting that satisfies the given extraction condition and selection condition. The character identification circuit 117 outputs the character dictionary 1 during the output.
Characters are identified by determining whether there are partial handwritings that satisfy the identification condition 18 or whether the relationship between the output partial handwritings satisfies the identification condition.

FIG. 12 is a flow chart showing the operation of the character identification circuit 117 of this embodiment. Character identification circuit 9 of the previous embodiment
6 searched for whether the partial handwriting satisfying the identification condition exists in the partial handwriting received from the partial handwriting extraction means 2 in step S66 of FIG. 6, the character identification circuit 117 of the present embodiment identifies in step S126. The partial handwriting satisfying the conditions is extracted by operating the partial handwriting extraction circuit 119 to check the existence.

FIGS. 13, 14 and 15 are diagrams showing the processing of the character identification circuit 117 for the hiragana "A". In the figure, 131, 141 and 151 are input coordinate points, 13
2, 142 and 152 indicate the boundaries of the segments. The character identification circuit 117 uses the partial handwriting extraction circuit 119 according to the contents of the character dictionary 118 to extract the partial handwriting necessary for identifying the character. The character dictionary 118 is the same as the character dictionary 10 shown in FIG.
Followed by: 1. There is a whole clockwise curve handwriting with the entry of the handwriting facing downward, 2. And there is a sufficiently long straight line handwriting pointing down in the center, and 3. And a rightward handwriting writing in the center of the top part. If there is a,
"Ah".

It is described as follows.

In accordance with the description in the character dictionary 118, the character identification circuit 117 first uses the partial handwriting extraction circuit 119 to extract the right handed curved handwriting 133 with the leading segment facing downward (FIG. 13). Subsequently, a downward long straight line handwriting 143 in the central portion is extracted (FIG. 14).
Regarding the conditions regarding the lengths of the two partial handwritings, the longest partial handwriting is given to the partial handwriting extraction circuit 119 as a selection condition, and it is determined whether or not the extracted partial handwriting has a sufficient length. Finally, the rightward straight line at the center of the upper part is extracted. However, since there is no description of the length of this line, the rightward straight line 1 at the center of the upper part is extracted using the partial handwriting extraction circuit 119.
53 is extracted and the existence of this partial handwriting is examined (Fig. 1
5). In this way, the input character is identified as hiragana "a". The partial handwritings 144, 154, 15
Reference numeral 5 indicates a partial handwriting that has already been extracted.

(Embodiment 3) Generally, in character recognition,
The recognition of similar characters, such as "shi" and "re", "2", "ko" and "tsu", which have almost the same stroke structure and can be identified only by the way the stroke is bent, is a big problem. ing. According to the present invention, the character identifying means identifies the similar character by using the feature amount regarding the curve of the segment such as the feature amount 5 (curvature degree) and the feature amount 6 (horizontal distribution) obtained by the segment dividing means. It is possible to do. For example, if you follow the character dictionary for "shi" and "re": 1. Long straight line handwriting 2.1. Handwriting with a starting point near the end point of the handwriting of 2.1 and pointing to the right or upper right 3.1. Both are described as "shi" if they are relatively curved, and "le" otherwise. In response to the above condition 3, the character identification circuit performs step S66 in FIG. 6 and step S126 in FIG.
At, it is determined whether the segment in the partial handwriting that satisfies the above conditions 1 and 2 is curved. Whether or not the segment is curved is determined by steps S43 and S4 in FIG.
Similar to 4, the segment characteristic amount 5 (curvature) is large, and the characteristic amount 6 (horizontal distribution) is biased to one of left and right. If it is determined that the partial handwriting is curved, it is identified as "shi", and if it is determined that it is not curved, it is identified as "re".

[0038]

According to the present invention, it is possible to use a curved line portion as a partial handwriting by using a segment which holds not only the linear element data used conventionally but also the curved line information. As a result, it becomes easy to recognize characters such as hiragana that are difficult to obtain by the conventional method based on linear element data, which are characteristic of the handwriting of the curved portion.

Further, according to the present invention, it is possible to identify a character depending on how the stroke is bent by using the segment curve information, which is difficult to recognize by the conventional method based on the straight line element data having no curve information. "Le",
It is easy to recognize characters such as "2", "ko", and "Otsu" that have almost the same stroke structure and are identified by the way the stroke is bent.

Further, according to the present invention, since the partial handwriting can be extracted without using the character dictionary, the determination of the partial handwriting which is conventionally performed for one input character in duplicate is performed only once. Is possible and recognition speeds up. further,
In the character dictionary, the minimum handwriting required to identify a character is described as an identification condition, and the extracted handwriting can be less than or equal to the strokes that make up the character. It is faster than if you do it.

Further, in the present invention, the recognition candidate character can be easily obtained by using the set of similar characters created in the character dictionary.

According to the present invention, the partial handwriting is extracted from the input coordinate data string, so that a continuous character string in which two or more strokes continue, a stroke number variation due to omission of the stroke, or a stroke order variation character string, etc. It can be recognized without changing. As for the stroke number variation due to the continuation of strokes, the position and direction of the leading segment of the partial handwriting are specified in the extraction condition of the partial handwriting, so that the partial handwriting can be extracted and recognized from the middle of the input stroke. Also, for characters with strokes that are easily omitted, or for characters with strokes whose positions, directions, and lengths are completely different depending on the writer, by using partial handwriting other than the unstable strokes for identification, It is possible to perform recognition without being affected by omission of strokes and unstable strokes by the writer. Further, with respect to the stroke order variation, since the partial handwriting is extracted regardless of the stroke input order, it is possible to recognize the stroke without changing the character dictionary.

Further, according to the present invention, since the partial handwriting necessary for identifying a character is extracted from the input stroke, there is no influence of noise input other than the strokes forming the character as compared with the method of processing all the input strokes. Can be recognized.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an online handwritten character recognition device of the present invention.

FIG. 2 is a diagram showing segment feature amounts of the present invention.

FIG. 3 is a flowchart showing the operation of the partial handwriting extraction circuit of the present invention.

FIG. 4 is a flowchart showing the operation of the partial handwriting extraction circuit of the present invention.

FIG. 5 is a conceptual diagram showing a part of a character dictionary of the present invention.

FIG. 6 is a flowchart showing the operation of the character identification circuit of the present invention.

FIG. 7 is a diagram showing a processing example of a segment division circuit of the present invention.

FIG. 8 is a diagram showing a processing example of a segment feature amount calculation circuit of the present invention.

FIG. 9 is a diagram showing a processing example of a partial handwriting extraction means of the present invention.

FIG. 10 is a diagram showing a processing example of a character identification circuit of the present invention.

FIG. 11 is a configuration diagram of an online handwritten character recognition device of the present invention.

FIG. 12 is a flowchart showing the operation of the character identification circuit of the present invention.

FIG. 13 is a diagram showing a processing example of a character identification circuit of the present invention.

FIG. 14 is a diagram showing a processing example of a character identification circuit of the present invention.

FIG. 15 is a diagram showing a processing example of a character identification circuit of the present invention.

[Explanation of symbols]

 1 segment division means 2 partial handwriting extraction means 3 character identification means 4 handwritten character input circuit 5 segment division circuit 6 segment feature amount calculation circuit 7 partial handwriting extraction control circuit 8 partial handwriting extraction circuit 9 character identification circuit 10 character dictionary 111 segment division means 112 character identification means 113 partial handwriting extraction means 114 handwritten character input circuit 115 segment division circuit 116 segment feature amount calculation circuit 117 character identification circuit 118 character dictionary 119 partial handwriting extraction circuit

Claims (4)

[Claims] 1. A segment dividing unit that divides an input coordinate data string of filled-in handwritten characters into segments and calculates feature amounts such as position, length, and curvature of the segments, and a stroke of a character based on the segments. Alternatively, it has a partial handwriting extracting means for extracting a partial handwriting which is a part of a stroke, and a character identifying means for recognizing a character or a part of the character by collating the partial handwriting with a condition described in a dictionary. Online handwritten character recognition device. 2. The on-line handwritten character recognition apparatus according to claim 1, wherein the partial handwriting extraction means extracts curved part handwriting. 3. The on-line handwritten character recognition device according to claim 1, wherein the character identification means performs collation of partial handwriting by using a feature amount of a segment. 4. The on-line handwritten character recognition according to claim 1, wherein character recognition is performed by extracting only the minimum partial handwriting necessary for the character identification means to identify a character using the partial handwriting extraction means. apparatus.