JP3290110B2 - Handwritten character recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten character recognition device, and more particularly, to a handwritten character recognition device for recognizing an online handwritten character on the basis of a non-radical part of a character.

[0002]

2. Description of the Related Art As a conventional online handwritten character recognition device, for example, Japanese Patent Publication No. 60-3238,
Japanese Patent Application Laid-Open No. 69887 and the like are known.

The on-line handwritten character recognition system disclosed in Japanese Patent Publication No. 60-3238 uses a top-down description of the character shape of a character such as a kanji. That is, one kanji is defined as a set of one or more components, a component is defined as a set of one or more elements, and the elements are identified.
Defined as a set of one or more strokes.

[0004] Such a recognition method based on the top-down description of the character shape of a character is intended to easily cope with an increase in the character type of the identification character and to be able to recognize a deformed character.

In the online handwritten character recognition device disclosed in Japanese Patent Laid-Open No. 2-69888, when comparing and judging a character, the characteristic data of the input character and the characteristic data of the character stored in the characteristic dictionary are compared. It is intended to speed up character recognition by comparing and judging one character at a time and judging the pass / fail of each part of the characters stored in the dictionary, such as bias or side.

[0006]

However, in the case of the former of the prior art described above, it is necessary to prepare more basic strokes and elements in order to deal with deformation of characters such as continuation characters. There is a problem that the processing time is increased. Also, in the latter case of the above-mentioned conventional technique, there is a problem that even if a character has the same partiality, such as a partiality, a determination is made from the beginning every time, and the efficiency of the recognition processing cannot be said to be high.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a handwritten character recognition device capable of coping with deformation of characters such as continuous characters and performing high-speed character recognition.

[0008]

SUMMARY OF THE INVENTION A handwritten character recognition apparatus of the present invention stores a partial pattern forming a character having the same large classification information as the large classification information for large classification of a character for a plurality of characters. Storage means, a selection means for selecting a character having the same large classification information as the large classification information of the input character read in the stroke order as a candidate character from the characters stored in the storage means, and a part for forming the input character Collating means for sequentially and only once collating a pattern and a corresponding partial pattern in a partial pattern forming a candidate character stored in a storage means; and a collated partial pattern for a partial pattern forming an input character for each collation Of the candidate character having the partial pattern to be compared whose difference exceeds a predetermined threshold from the candidate characters selected by the selection means. It includes a character selection means, the, and recognizes an input character based on the remaining candidate characters by the candidate character selection means.

In the handwritten character recognition apparatus of the present invention, a character having the same large classification information as the large classification information of the input character read in the stroke order is selected as a candidate character from the characters stored in the storage means by the selection means. . The partial pattern forming the input character and the corresponding partial pattern in the partial pattern forming the candidate character stored in the storage means are sequentially and only once collated by the collation means, and the part forming the input character for each collation A difference between the pattern to be compared with the pattern is determined, and candidate characters having the pattern to be compared with the difference exceeding a predetermined threshold are removed from candidate characters selected by the selection unit by the candidate character selection unit. Go. Therefore, the selection based on the large classification information is made by the selection means, the recognition characters are narrowed down, and the narrowed down recognition characters are further narrowed down by collation of the partial patterns, so that the character recognition is performed quickly.

[0010] In the handwritten character recognition device, the partial pattern stored in the storage means is composed of segments. In addition to the segments which are always present in the character configuration, the partial patterns include segments which may appear as curved portions of continuous characters or scribbles. The partial pattern forming the input character is compared with the partial pattern consisting of the segment which is always present in the character configuration stored in the storage means and the segment which may appear as a continuation character or a scribble curve portion, and character configuration is performed. It is characterized in that a difference between a segment always present above and a corresponding partial pattern in a partial pattern forming an input character is obtained.

In the handwritten character recognition device, a partial pattern including a segment which may appear as a continuation character or a scribble curve portion in addition to a segment which always exists in the character configuration is stored in the storage means as a partial pattern. When matching with a partial pattern forming an input character, the partial pattern of the input character is compared with a partial pattern consisting of a segment that always exists in the character configuration and a segment that may appear as a continuation character or a scribble curve part. The difference between the segment that is always present in the character configuration and the segment that is always present in the character configuration stored in the storage means is required. Therefore, it is possible to cope with both the case where the input characters are in the regular style and the case where the input characters are the continuous characters.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a handwritten character recognition device according to the present invention will be described with reference to embodiments. FIG. 1 is a block diagram functionally showing the configuration of a handwritten character recognition device according to one embodiment of the present invention.

In the present specification, the term "partial pattern" is used hereinafter to mean the entire pattern or a part of the radical including the paraleg of the character, and the pattern of the radical forming the character, It includes a pattern, a pattern forming a part of a radical, and a pattern forming a part of a side, and each pattern is composed of a segment.

From an online character input device such as a tablet, handwritten characters are read by the character input unit 1 in the order of strokes and are temporarily stored in a storage device. For a character input to the character input unit 1, when a segment in the same direction continues in the same stroke, a straight line is used, and a segment in a range including a curved portion within a predetermined range is used as a straight line. The pre-processing unit 2 performs pre-processing such as linear approximation, normalization, and noise removal processing.

The straight line processed and approximated by the preprocessing unit 2 is converted into a large classification code based on the large classification code shown in Table 1 according to the direction in which the straight line is drawn, and based on the segment code shown in Table 1. Segment coding. Here, FIG. 2 is a diagram provided for explanation of the large classification coding and the segment coding in cooperation with Table 1, and based on the direction of the approximated straight line processed and approximated by the preprocessing unit 2, the segmentation processing is performed. The unit 3 performs a segment coding process. In FIG. 2, the large classification code and the segment code are shown superimposed. In Table 1, right payment means a character portion written diagonally from left to right, and left payment means a character portion written diagonally from right to left.

[0016]

[Table 1]

For example, as a result of preprocessing, when the direction of the straight line drawn in the horizontal direction falls within the range of 350 ° to 70 °, the line is coded into the large classification code {I} by the large classification coding, It is coded into segment code {0} by segment coding.

A character code such as a JIS code in which the number of input characters processed in the segmentation processing unit 3 and the large classification code of the first and last segments are the same is extracted from the character code group in the character dictionary 9. The large characters are roughly classified and extracted by the large classification unit 4 as candidate characters.
Here, the large classification code of the first and last segments is classified into {I} when the segment code is {0} or {1}, and the segment code is {2},
Alternatively, when {3}, the large classification code is classified as {II}, when the segment code is {4}, the large classification code is classified as {I}, and when the segment code is {5},
The major classification code is classified into {I} or {II}. When the segment code is {6}, the major classification code is {II}.
are categorized.

Here, the character dictionary 9 includes (a) a character code such as a JIS code of an input character. (B) The minimum number of segments that may appear for an input character; used for major classification. (C) the maximum number of segments that may appear for input characters; used for major classification. (D) Type of direction of the first segment of the input character; used for major classification. (E) Type of the direction of the last segment of the input character; used for major classification. (F) The number of partial patterns constituting the input character; used for collation of the partial patterns. (G) Partial pattern dictionary numbers stored in the order in which they were written; used for position matching of partial patterns. (H) positional relationship information of the partial patterns; after the partial pattern matching, radicals constituting the characters and side by side positions to distinguish characters having different partial pattern positions such as, for example, character {member} and character {song} Used for matching. Are stored.

Here, the minimum number of segments that may appear for an input character and the maximum number of segments that may appear for an input character are stored in the character dictionary 9 in the case of a block style. This is because the number of segments differs depending on the case of the continuation character, and there is a width.

The number of segments in the input character is between the minimum number of segments that can appear for the input character and the maximum number of segments that can appear for the input character, and The major classification code of the segment is the same as the major classification code of the first segment of the character stored in the character dictionary 9, and the direction of the last segment of the input character is the last segment of the character stored in the character dictionary 9. The character that is the same as the large classification code is roughly narrowed down from the character dictionary 9 by the large classification unit 4 as a candidate character. Here, the large classification unit 4 corresponds to a selection unit that selects a candidate character based on an input character.

The partial pattern constituting the candidate character roughly detected from the character dictionary 9 by the large classification unit 4 is collated by the partial pattern collation unit 5. The matching of the partial pattern in the matching unit 5 is performed by searching the character dictionary 9 for the number of the dictionary 8 of the partial pattern stored for the candidate character narrowed down by the large classification unit 4.

The partial pattern dictionary 8 includes (a) the number of the partial pattern dictionary, (b) the number of stroke orders (On) that may appear for the partial pattern, (c) the direction code sequence of the segment of the stroke order 1, and Information on the positional relationship between segments,
(D) the direction code sequence of the segment in the stroke order 2 and the positional relationship information between each segment; (e) the direction code sequence of the segment in the stroke order 3 and the positional relationship information between each segment.
Are stored. Here, the partial pattern dictionary 8 and the character dictionary 9 correspond to storage means. Further, as described above, the partial pattern is formed from a radical and a segment constituting a part of the radical.

By referring to the number of the partial pattern dictionary constituting the candidate character narrowed down from the character dictionary 9, the partial pattern dictionary 8 having the same number as the number is searched and stored in the searched partial pattern dictionary 8. The direction of each segment is compared with the direction of the segment of the input character, and the relative positional relationship between segments stored in the searched partial pattern dictionary 8 and the relative positional relationship between input characters are compared. The comparison is performed, the collation and the comparison result are weighted, and a penalty is calculated based on the weighted collation and the comparison result. The number of segments of the input character having undergone the collation processing is temporarily stored for each partial pattern in the character dictionary.

The calculated penalty is compared with a predetermined threshold value to perform a threshold process for determining whether or not the threshold value exceeds the threshold value. If there is a partial pattern, all characters having that partial pattern are removed from the candidate characters. Next, the next partial pattern dictionary 8 constituting the remaining candidate characters is searched, and the threshold processing is repeated on the partial pattern dictionary 8. Here, the partial pattern matching unit 5 corresponds to a matching unit that matches the partial pattern of the input character with the partial pattern of the candidate character and a candidate character selecting unit that further selects the candidate character selected by the selecting unit.

With respect to the candidate characters remaining in the process of the partial pattern matching unit 5, the partial pattern position matching unit 6 checks the positional relationship between the partial patterns of the input characters and the partial pattern of the candidate characters, A candidate character whose positional relationship is the same as the positional relationship candidate of the partial pattern of the input character is selected, and the recognized character candidate is finally determined in combination with the penalty obtained by the processing of the partial pattern matching unit 5. . That is, the candidate character remaining in the processing of the partial pattern position matching unit 6 with the smallest penalty is output from the recognition result output unit 7 as the recognition result.

Next, Table 2 shows the characters {member}, {song},
Here are examples of partial patterns that define consumption, immersion, refinement, and powdering and numbers of those partial pattern dictionaries. letter〃
The members {song}, {song}, {soak}, {sin}, {powder} are represented by combining the partial patterns in the partial pattern dictionary numbers n1 to n13.

[0028]

[Table 2]

Table 3 shows the stored contents of the character dictionary 9 for the characters {member}. Characters and {members} are composed of partial patterns, mouths, eyes, and eight corresponding partial patterns, as is clear from Table 2. The number of partial patterns is {3}, and the number of segments in square writing is {12}. However, the minimum number of segments is $ 10 to allow for the continuation character case.
{, The maximum number of segments is assumed to be {24}, the major classification code of the first segment is {II}, the major classification code of the last segment is {I}, and the number of the partial pattern dictionary constituting the character is n1 , N2 and n3.

[0030]

[Table 3]

Table 4 shows an example of partial pattern dictionary numbers for the characters {member}, {song}, {consequence}, {soak}, [genki], and {powder}. It can be easily understood that the characters can be composed.

[0032]

[Table 4]

The segment code sequence stored in the partial pattern dictionary 8 includes a segment code sequence in a stroke order, a segment code sequence which always exists in the character structure, and a segment which may appear as a curved portion formed by continuation characters or scribbles. And a code sequence. The inter-segment relative position information is position information between two indispensable segments (positional relationship of end points of segments, intersection, T-type connection, etc.). Dictionary number n5 for the partial pattern illustrated in Table 2
As for the stroke order, the stroke order is as shown in FIG.
There are three types as shown in (b) and (c). Table 5 shows the number of stroke orders, stroke order data and positional relationship information stored in the dictionary of the partial patterns.

[0034]

[Table 5]

In FIGS. 3 (a), (b) and (c), the symbols (a), (b) and (c) indicate segments and the order of writing, and the broken lines increase when writing is continued. A possible segment. In the data of stroke order 1 in Table 5, {213} on the left side of the right column indicates that the stroke order is (a),
(B) and (c), and the direction codes are also the direction codes of the segments (a), (b) and (c).

In the data of stroke order 2 in Table 5,
{ 1 0 32} on the left side of the right column and { 1 0 0 2 0 3} on the left side of the right column in the data of the stroke order 3 are also direction codes. As indicated by the broken line in FIG. 3, {0} is the direction code 0.
This means that there is a possibility that the number of segments will increase, which is also shown in Table 5 with an underline. However, the segment indicated by a broken line that may increase in the stroke order 1 in FIG.
° and therefore have been removed by segmentation.

FIG. 4 shows a part of these relationships. In the example shown in FIG. 4, at the number n5 of the partial pattern dictionary,
(A), (b) and the direction code of the segment of the shape shown in (c) is 〃1 0 2 0 3〃, including broken line portion.
On the other hand, when the input character has no broken line portion, the direction code is {123}, which corresponds to the input character.

The leftmost direction code in the right column of Table 5 indicates the code of the segment constituting the partial pattern, and the position information following the leftmost direction code in the right column is 1
The second digit is the segment number, the second digit is the start or end point information of the segment, the third alphabet is the positional relationship information, the fourth digit is the segment number, and the fifth digit is the segment Is the start or end point information. The alphabet representing the third positional relationship is 〃
L, R, H, B}, etc., {left, right, top,
It means the following. For example, position information $ 21L11
The meaning of 〃 indicates that the starting point of the segment (b) is on the left side of the starting point of the segment (a).

FIG. 5 is a flowchart showing in detail the processing of the partial pattern collating unit 5 in the handwritten character recognition apparatus according to one embodiment of the present invention.

Hereinafter, the operation of the flowchart will be described with reference to FIGS. 1 and 2 and Tables 1 to 4 assuming that the input character is {member} (12 segments in square writing).

The input character {member} is read by the character input unit 1, preprocessed by the preprocessing unit 2, and segmented by the segmentation processing unit 3 based on FIG. Major classification unit 4 based on the segmentation result
Are narrowed down to the candidate characters shown in Table 4. The contents of the penalty buffer storing the penalties of these candidate characters are cleared.

Next, the dictionary numbers of the partial patterns in the same order of the candidate characters are read into the buffer (step 51). In the first case, the partial pattern dictionary numbers {n1, n1, n4, n4, n10,
n10} is read (step 51). Step 5
Duplicate ones of the partial patterns read in 2 are removed, and a partial pattern table to be collated is created (step 52). In this example, the first partial pattern to be collated is {n1, n4, n10}.

The dictionary data of the partial patterns is read in order from the partial pattern table to be collated (step 53).
In the case of this example, the dictionary data of {n1} is read for the first time. Subsequently, the direction and position of the segment of the partial pattern dictionary are compared with the direction and position of the segment of the input character, and a penalty is calculated (step 5).
4). Next, it is checked whether or not the comparison with all the segments of the partial pattern dictionary has been completed (step 55). If it is determined in step 55 that the process has not been completed, the process of step 54 is repeated.

If it is determined in step 55 that the process has been completed, the number of segments of the input characters that have been subjected to the collation processing is stored for each candidate character having a partial pattern to be collated (step 56). Then, it is checked whether or not there is an unmatched partial pattern in the partial pattern table to be matched (step 57). If it is determined in step 57 that there is an unmatched partial pattern, steps 53 to 53
The process of step 56 is repeatedly executed. If it is determined in step 57 that there is no unmatched partial pattern, the process of re-selecting candidate characters is performed based on the penalty (step 58).

In this example, since the input character is {member}, the penalty of the candidate character having the partial pattern {mouth} is the smallest, and the penalty of the candidate character having the partial pattern {US} is the least. Considered large. In the penalty threshold processing, candidate characters having the partial pattern {US} are removed from the candidates. The candidate characters left at this point are members,
Singing, disappearing, immersing.

It is checked whether or not there is a partial pattern that has not been collated with the remaining candidate characters (step 5).
9). If it is determined in step 59 that there is a partial pattern that has not been collated, steps 51 to 5
The processing up to 8 is repeated. In this example, the partial pattern dictionary number read from the second candidate character is {n2, n2, n5, n7} in the second column of Table 4. In this case, the overlapping pattern is {n2}, and partial patterns to be collated by removing one {n2} are {n2, n5, n7}
It becomes 〃. If it is determined in step 59 that there is no unmatched partial pattern, the process of the partial pattern position matching unit 6 is executed.

Table 6 is an example of a penalty table used when collating the dictionary segment of the partial pattern (the dictionary side segment is also referred to as a dictionary segment with respect to the input character segment) and the direction of the input character segment.

[0048]

[Table 6]

The code of the dictionary segment i currently being collated is dci, and the code of the input character segment j is icj
And The accumulated penalty up to the present of the candidate character having the collating partial pattern is defined as Pm. m represents the number of the candidate character. In code dci, {it always exists},
There are two types that can appear: Code dc
The matching method differs depending on the type of i. Next, a description will be given.

The case where (a) the code dci is {always present} and (a1) dci = icj will be described. (A) When the positions of the segment i and the segment j are the same, the segments i and j are made to correspond to each other, and Pm is not changed.
(B) When the positions of the segment i and the segment j are not the same, dci = icj + 1, and the position of the segment i is the same as the position of the segment (j + 1), i and j + 1 are made to correspond and icj is skipped. However, 1 is added to Pm as a penalty for skipping icj. (C) The positions of the segment i and the segment j are not the same, and dci + 1 = ic
j, and when the position of the segment i + 1 is the same as that of the segment j, the i + 1 is made to correspond to the j and the ici is skipped. However, 1 is added to Pm as a penalty for skipping ici. (D) In addition to the above, the segment i and the segment j are associated with each other, and when the positions do not match, 1 is added to Pm as a penalty.

Next, the case where (a) dci is {always present} and (a2) dci @ icj will be described. (A) When dci = icj + 1 and the position of segment i is the same as that of segment j + 1, i and j + 1 are made to correspond and icj is skipped. However, 1 is added to Pm as a penalty for skipping icj. (B) dci + 1
= Icj, and when the position of segment i + 1 is the same as that of segment j, i + 1 is made to correspond to j, and ici is skipped. However, 1 is added to Pm as a penalty for skipping ici. (C) In addition to the above, dci and icj are associated with each other, and a penalty is added to Pm based on Table 6.

Next, in the case (b) where the code dci is {possible to appear}, (b1) dci = icj, dci + 1
= Icj + 1 and segment i + 1 and segment j
When the position of +1 is the same, dci and icj are associated. (B2) In addition to the above, it is determined that the number of segments corresponding to dci in the input characters has not increased, and
Do not change m.

Next, an example of collating a partial pattern dictionary with input characters will be described with reference to FIG. 4 in relation to penalties.

For example, the partial pattern n of the input character {OFF}
The case where 5 is written in square style and the case where it is written continuously will be described. The stroke order is assumed to be the same as that of FIG. The segment direction code when n5 is written in square style is $ 123
}, And the segment direction code when written continuously is {10203}.

First, the collation between the partial pattern n5 written in the standard style and the partial pattern n5 of the partial pattern dictionary 8 will be described. Since the number of stroke orders of the partial pattern n5 of the partial pattern dictionary 8 is three, the partial pattern n5 of the input character is collated in order with the stroke order 1, the stroke order 2, and the stroke order 3 in the partial pattern dictionary. Since it does not match 2, it is considered that the penalty of the collation result is not 0.

Here, the collation with the stroke order 3 will be described in detail. Since the segment direction code of the partial pattern dictionary stroke order 3 of the partial pattern n5 is { 1 0 2 0 3 }, d
c1 = 1, dc2 = 0 , dc3 = 2, dc4 = 0 , dc
5 = 3. On the other hand, since the segment direction code of the input partial pattern n5 is {123}, ic1 = 1, i
c2 = 2 and ic3 = 3.

Assuming that dc1 = 1 in the partial pattern dictionary is a segment that always exists, dc1 = ic1, and that both segment positions are the same, dictionary segment 1 is made to correspond to input segment 1 as shown by the arrow in FIG. Do not attach.

Dc2 = 0 in the partial pattern dictionary is a segment that may appear, and dc2 ≠ ic2, and dc2 is skipped on the assumption that no segment corresponding to dc2 appears in the input character. As above, dc3
And ic2, skip dc4 = 0 , and dc5 and i
and c3. As a result, the penalty is zero.

Next, the collation of the subsequently written partial pattern n5 with the partial pattern n5 of the partial pattern dictionary will be described. Also in this case, the partial pattern n5 of the partial pattern dictionary
The description of the comparison between the stroke order 1 and the stroke order 2 will be omitted, and only the comparison with the stroke order 3 will be described. The segment direction code of the input character partial pattern n5 is $ 1020
3〃, ic1 = 1, ic2 = 0, ic3 =
2, ic4 = 0 and ic5 = 3.

Assuming that dc1 = 1 in the partial pattern dictionary is a segment that always exists, dc1 = ic1, and that both segment positions are the same, dictionary segment 1 is made to correspond to input segment 1 as shown by the arrow in FIG. Do not penalize.

Dc2 = 0 in the partial pattern dictionary is a segment that may appear, and dc2 = ic2,
Since dc3 = ic3 and the segment 3 of the partial pattern dictionary n5 and the position of the input segment 3 of n5 are also the same, ic2 is made to correspond to dc2 as a segment that appears as a continuation character. No penalty will be charged. Similarly, dc3 and ic3, dc4 and ic4, dc
5 and ic5 correspond to each other. As a result, the penalty becomes zero.

The above-described processing is carried out on the basis of a partial pattern forming an input character, a segment which always exists in the character configuration stored in the partial pattern dictionary 8 and a segment which may appear as a continuous character or a scribble curve. This is equivalent to collating a partial pattern with a certain partial pattern and determining a difference between a segment that always exists in the character configuration and a corresponding partial pattern in a partial pattern forming an input character.

As described above, the step 58 is executed to narrow down the candidate characters based on the penalty.
If it is determined in step 59 that there is no partial pattern remaining in the remaining candidate characters, the partial pattern position matching unit 6 checks the partial pattern position, and the final recognized character of the input character integrated with the penalty value is determined. Is determined.

[0064]

As described above, according to the handwritten character recognition apparatus according to the present invention, in the handwritten character recognition processing, the candidate characters are narrowed down by the large classification information, and the candidate characters are compared with the partial patterns. As a result, character recognition is performed more narrowly, so that the effect that character recognition is performed at high speed is obtained.

Further, according to the handwritten character recognition device of the present invention, a segment which forms a partial pattern by a segment and which may appear as a curved portion of a continuation character or a scribble in addition to a segment which is always present in the structure of a character. Since the character recognition is performed by collating with the partial pattern of the input character, it is possible to obtain the effect of being able to cope with continuous characters and the like.

[Brief description of the drawings]

FIG. 1 is a block diagram functionally showing a configuration of a handwritten character recognition device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a segment direction and a segment code in the handwritten character recognition device according to one embodiment of the present invention.

FIG. 3 is an explanatory diagram of a stroke order example of a partial pattern in the handwritten character recognition device according to the embodiment of the present invention;

FIG. 4 is an explanatory diagram of collation between a dictionary segment of a partial pattern and a segment of an input character in the handwritten character recognition device according to the embodiment of the present invention;

FIG. 5 is a flowchart for explaining the operation of the handwritten character recognition device according to the embodiment of the present invention;

[Explanation of symbols]

 Reference Signs List 1 character input unit 2 preprocessing unit 3 segmentation processing unit 4 major classification unit 5 partial pattern matching unit 6 partial pattern position matching unit 7 recognition result output unit 8 partial pattern dictionary 9 character dictionary

──────────────────────────────────────────────────続 き Continued on the front page (72) Naoto Ouchi, Inventor 401-62-24 Nishi-Nippori, Arakawa-ku, Tokyo 401 Di Aheim Nishi-Nippori 401 (72) Inventor Hu Jin-ling 1-14-6 Dogenzaka, Shibuya-ku, Tokyo (56) References JP-A-62-2229384 (JP, A) JP-A-9-198466 (JP, A) JP-A-2-266485 (JP, A) JP-A-62-271087 (JP) , A) Masanori Toda, Naoto Ouchi, Rei Hukin, Yoshihide Magago, Tadahiro Kubota, Online Character Recognition by Segment Analysis, IPSJ 55th Annual Meeting (2nd half of 1997) Proceedings (2), Japan, IPSJ , September 24, 1997, 55th (2), p. 2-194 to 195 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06K 9/62 JICST file (JOIS)

Claims (5)

(57) [Claims] 1. A storage means for storing, for a plurality of characters, a partial pattern forming a character having the same large classification information as the large classification information for performing a large classification of characters, and input characters read in a stroke order. Selection means for selecting a character having the same large classification information as the large classification information from the characters stored in the storage means as a candidate character; forming a partial pattern forming an input character and forming a candidate character stored in the storage means A matching means for sequentially and only once matching a corresponding partial pattern in a partial pattern to be compared, and a difference between a partial pattern to be matched and a partial pattern forming an input character for each matching, and the difference is set to a predetermined threshold value. Candidate character selecting means for removing candidate characters having a partial pattern to be compared exceeding from the candidate characters selected by the selecting means. Recognizing the input character on the basis of the remaining candidate characters Te handwritten character recognition apparatus according to claim. 2. The handwritten character recognition apparatus according to claim 1, wherein the large classification information includes information on the number of segments forming the character and information on a type of a predetermined segment direction. 3. A handwritten character recognition apparatus according to claim 1, wherein the partial pattern is composed of a segment, a radical pattern forming a character, a bystander pattern, a pattern forming part of a radical, and a part of bystander. A handwritten character recognition device characterized by including a pattern forming a character. 4. A handwritten character recognizing device according to claim 1, wherein the collated partial patterns collated by the collating means are stored in the storage means in the same order as the partial patterns in the stroke order forming the input characters. A handwritten character recognition device, which is a character partial pattern. 5. The handwritten character recognition apparatus according to claim 1, wherein the partial pattern stored in the storage means is composed of segments, and may appear as a continuous character or a scribble curve in addition to the segments which are always present in the character configuration. Pattern, which includes a segment that can be input, and a partial pattern consisting of a segment that always exists in the character configuration stored in the storage means and a segment that may appear as a continuation character or a scribble curve portion A handwritten character recognition apparatus for determining a difference between a segment necessarily present in a character configuration and a corresponding partial pattern in a partial pattern forming an input character.