JPH0344355B2 - - Google Patents
Info
- Publication number
- JPH0344355B2 JPH0344355B2 JP57220676A JP22067682A JPH0344355B2 JP H0344355 B2 JPH0344355 B2 JP H0344355B2 JP 57220676 A JP57220676 A JP 57220676A JP 22067682 A JP22067682 A JP 22067682A JP H0344355 B2 JPH0344355 B2 JP H0344355B2
- Authority
- JP
- Japan
- Prior art keywords
- section
- transformation
- deformation
- character pattern
- character
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
- 238000011426 transformation method Methods 0.000 claims description 30
- 238000000034 method Methods 0.000 claims description 21
- 230000009466 transformation Effects 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000001131 transforming effect Effects 0.000 description 2
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Character Discrimination (AREA)
- Character Input (AREA)
Description
【発明の詳細な説明】
(A) 発明の技術分野
本発明は、筆者毎に入力文字パターンに対し最
適な変形を施すことによつて認識率を向上し得る
文字認識装置に関する。DETAILED DESCRIPTION OF THE INVENTION (A) Technical Field of the Invention The present invention relates to a character recognition device that can improve recognition rate by optimally transforming input character patterns for each writer.
(B) 技術の背景
文字認識装置においては、入力文字パターンの
特徴を抽出し予め準備した標準文字パターンの特
徴とを照合する、いわゆるパターンマツテング方
式をはじめ、各種の認識方式が用いられている
が、いずれの方式においても、一般に、認識率の
向上と認識速度の向上は互に他の妨げになつてお
り、両者をともに満足させなければならないとい
う困難な課題を抱えている。(B) Background of the technology Various recognition methods are used in character recognition devices, including the so-called pattern matching method, which extracts the features of an input character pattern and matches them with the features of a standard character pattern prepared in advance. However, in either method, improvements in recognition rate and recognition speed generally impede each other, and it is a difficult problem to satisfy both.
(C) 従来技術と問題点
文字認識装置、とくに手書文字認識装置におい
ては、認識率を向上するために、従来、個人毎の
標準文字パターンを格納する、いわゆる個人辞書
を用いる方法が知られている。しかし、このよう
な方法は、辞書の記憶容量の増大、あるいは辞書
作成のための筆者の負担が増大するという問題を
生じている。(C) Prior Art and Problems In order to improve the recognition rate of character recognition devices, especially handwritten character recognition devices, it has been known to use a so-called personal dictionary that stores standard character patterns for each individual. ing. However, such a method has the problem of increasing the storage capacity of the dictionary or increasing the burden on the author for creating the dictionary.
(D) 発明の目的
本発明の目的は、辞書の記憶容量の増大あるい
は筆者の負担を増加することなく、多数の筆者に
利用し且つ認識率を向上し得る文字認識装置を提
供することにある。(D) Purpose of the Invention The purpose of the present invention is to provide a character recognition device that can be used by a large number of writers and can improve the recognition rate without increasing the storage capacity of the dictionary or increasing the burden on the writers. .
(E) 発明の構成
本発明になる文字認識装置は、文字種毎の標準
文字パターンの特徴を格納する辞書部と、筆者毎
の入力文字パターンと該入力文字パターンに複数
種の所定の変形法による変形を施して得られる複
数の変形文字パターンとのうち前記標準文字パタ
ーンとの類似度が高い入力文字パターンもしくは
変形文字パターンに施した無変形を含む前記変形
法を最適変形法として登録する最適変形法記憶部
と、被認識入力文字パターンを前記最適変形法記
憶法に登録する最適変形法に従つて変形する変形
部と前記変形部によつて得られる変形入力文字パ
ターンの特徴を抽出する特徴抽出部と、前記特徴
抽出部によつて得られる特徴と前記辞書部に格納
する特徴とを照合することによつて認識結果を得
る照合部とを備えるものである。(E) Structure of the Invention The character recognition device of the present invention includes a dictionary section that stores the characteristics of standard character patterns for each character type, an input character pattern for each writer, and a plurality of predetermined transformation methods for the input character pattern. Optimal deformation in which the deformation method including no deformation performed on the input character pattern or deformed character pattern that has a high degree of similarity to the standard character pattern among the plurality of deformed character patterns obtained by applying the deformation is registered as the optimal deformation method. a transformation unit that registers the input character pattern to be recognized in the optimal transformation method memory method, a transformation unit that transforms the input character pattern according to the optimal transformation method, and a feature extraction that extracts features of the transformed input character pattern obtained by the transformation unit. and a matching section that obtains a recognition result by comparing the features obtained by the feature extracting section and the features stored in the dictionary section.
(F) 発明の実施例
以下、本発明の要旨を実施例によつて具体的に
説明する。(F) Examples of the invention Hereinafter, the gist of the present invention will be specifically explained using examples.
第１図は本発明一実施例のブロツク図を示し、
１は入力文字パターンを観測し２値化入力文字パ
ターンを得る観測部、２は観測部１において得ら
れる２値化入力文字パターンに対し正規化処理を
施し正規化入力文字パターンを得る正規化部、３
は正規化部２において得られる正規化入力文字パ
ターンに無変形を含む所定の変形法による変形を
施し変形入力文字パターンを得る変形部、４は変
形部３において得られる変形入力文字パターンの
特徴を抽出する特徴抽出部、５は特徴抽出部４に
おいて得られる特徴と後記辞書部に格納する特徴
とを照合し認識結果を得る照合部、６は照合部５
において得られる認識結果を格納するバツフア、
７は認識結果を表示する表示部、８は変形部３の
制御をおこなう制御部、９は文字種毎の標準文字
パターンの特徴を格納する辞書部、１０は後記最
適変形法記憶部へ筆者毎の最適変形法を登録する
ための操作をおこなう操作部、１１は後記複数種
の所定の変形法を記憶する変形法記憶部、１２は
筆者毎の所定の文字種毎の入力文字パターンと該
入力文字パターンに複数種の所定の変形法による
変形を施して得られる複数の変形文字パターンと
のうち標準文字パターンとの類似度が最も高い入
力文字パターンもしくは変形文字パターンに施し
た無変形を含む前記変形法を最適変形法として登
録する最適変形法記憶部、１３は最適変形法記憶
部１２に登録する前記最適変形法を判定する判定
部、１４は前記最適変形法を求めるために用いる
前記所定の文字種すなわち最適変形法判定用文字
種を記憶する文字種記憶部である。 FIG. 1 shows a block diagram of an embodiment of the present invention,
1 is an observation unit that observes an input character pattern and obtains a binary input character pattern; 2 is a normalization unit that performs normalization processing on the binary input character pattern obtained in the observation unit 1 and obtains a normalized input character pattern. ,3
4 is a transformation unit that transforms the normalized input character pattern obtained in the normalization unit 2 by a predetermined transformation method including no transformation to obtain a transformed input character pattern; A feature extraction unit 5 extracts features, a collation unit 5 collates the features obtained in the feature extraction unit 4 with features stored in the dictionary section to be described later, and obtains a recognition result, 6 a collation unit 5
A buffer that stores the recognition results obtained in
7 is a display section that displays recognition results; 8 is a control section that controls the transformation section 3; 9 is a dictionary section that stores the characteristics of standard character patterns for each character type; and 10 is an optimal transformation method storage section that will be described later. An operation unit performs operations for registering an optimal transformation method; 11 is a transformation method storage unit that stores a plurality of predetermined transformation methods described later; 12 is an input character pattern for each predetermined character type for each writer and the input character pattern; Among the plurality of transformed character patterns obtained by transforming the input character pattern using a plurality of predetermined transformation methods, the transformation method includes no transformation performed on the input character pattern or the transformed character pattern that has the highest degree of similarity to the standard character pattern. 13 is a determination unit that determines the optimal transformation method to be registered in the optimal transformation method storage unit 12; 14 is a determination unit that determines the predetermined character type used to determine the optimal transformation method; This is a character type storage unit that stores character types for determining the optimal transformation method.
すなわち、本実施例は、観測部１において得ら
れる２値化入力パターンに対し正規化部２におい
て正規化を施したあと、更に、変形部３において
変形を施すことによつて、照合部５において高い
認識率による認識がおこなわれるようにしたもの
であり、このために、変形部３において施すべき
最適な変形法を筆者毎に最適変形法記憶部１２に
予め記憶するようにしたものである。 That is, in this embodiment, after the normalization unit 2 normalizes the binarized input pattern obtained in the observation unit 1, the transformation unit 3 further transforms the binary input pattern, so that the matching unit 5 Recognition is performed with a high recognition rate, and for this purpose, the optimal deformation method to be applied in the deformation section 3 is stored in advance in the optimal deformation method storage section 12 for each author.
最適変形法記憶部１２の生成は次のようにして
おこなう。筆者は操作部１０に設けられている鍵
盤を操作し筆者名を制御部８に入力する。続いて
筆者は、文字種記憶部１４に記憶してある文字種
すなわち最適変形法判定用文字種を所定の最適変
形法判定用紙に記入し該文字認識装置に入力す
る。 The generation of the optimal deformation method storage unit 12 is performed as follows. The author operates a keyboard provided on the operation unit 10 and inputs the author's name into the control unit 8. Next, the author writes the character type stored in the character type storage unit 14, that is, the character type for determining the optimal transformation method, on a predetermined optimal transformation method determination form and inputs it into the character recognition device.
このようにして入力された最適変形法判定用文
字種は、１文字づつ、観測部１による２値化と正
規化部２による正規化をうけたあと、変形部３に
おいて、変形法記憶部１１に記憶する複数種の所
定の変形法にしたがつて変形がおこなわれる。 The character type for determining the optimal transformation method input in this way is binarized character by character by the observation unit 1 and normalized by the normalization unit 2, and then stored in the transformation method storage unit 11 in the transformation unit 3. Deformation is performed according to a plurality of stored predetermined deformation methods.
第２図は変形法の例を示し、(a)は変形を施さな
い正規化入力文字パターンであり、(b)は縦方向に
正の変形を施した正縦変形パターン、(c)は縦方向
に負の変形を施した負縦変形パターン、(d)は横方
向に正の変形を施した正横変形パターン、(e)は横
方向に負の変形を施した負横変形パターン、(f)は
斜めに正の変形を施した正斜め変形パターン、(g)
は斜めに負の変形を施した負斜め変形パターンで
ある。 Figure 2 shows an example of the transformation method, where (a) is a normalized input character pattern without transformation, (b) is a normal vertical transformation pattern with positive transformation in the vertical direction, and (c) is a vertical transformation pattern. (d) is a positive horizontal deformation pattern with negative deformation in the horizontal direction, (e) is a negative horizontal deformation pattern with negative deformation in the horizontal direction, ( f) is a positive diagonal deformation pattern with positive diagonal deformation; (g)
is a negative diagonal deformation pattern in which negative deformation is applied diagonally.
第２図は無変形のほか６種類のパターンを例示
したが、変形の量を変化させることにより、また
縦変形と横変形を組合わせること等によつて多様
多様な変形法による変形をおこなうことができ
る。 Figure 2 shows six types of patterns in addition to no deformation, but it is possible to perform deformation using a variety of deformation methods by changing the amount of deformation or by combining vertical deformation and horizontal deformation. I can do it.
前記入力した最適変形法判定用文字種のすべて
に対し、このようにして変形をおこない、無変形
パターンすなわち正規化入力文字パターンを含
め、すべての変形入力文字パターンについて、パ
ターンマツチングによる認識をおこない、最適変
形法判定用文字種毎に得られた認識結果をバツフ
ア６に格納する。 All of the input character types for determining the optimal transformation method are transformed in this way, and all transformed input character patterns, including unmodified patterns, that is, normalized input character patterns, are recognized by pattern matching, The recognition results obtained for each character type for determining the optimal transformation method are stored in the buffer 6.
判定部１３は、バツフア６に格納した認識結果
を文字種記憶部１４に記憶する文字種と照合し、
前記認識結果の正否を決定するとともに、該決定
結果によつて筆者毎の最適変形法を判定する。 The determination unit 13 compares the recognition result stored in the buffer 6 with the character type stored in the character type storage unit 14,
It is determined whether the recognition result is correct or not, and the optimal transformation method for each author is determined based on the determination result.
以上のようにして最適変形法辞書部１２の生成
すなわち登録をおこなつたあと、該当筆者による
入力文字の認識をおこなう。すなわち被認識入力
文字は観測部１による２値化と正規化部２による
正規化をうけたあと、変形部３において、最適変
形法記憶部１２に記憶する筆者毎の最適変形法に
従つて変形が施されたのち、パターンマツチング
法による認識をおこなう。 After generating or registering the optimal transformation method dictionary section 12 as described above, the characters input by the corresponding writer are recognized. That is, the input character to be recognized is binarized by the observation unit 1 and normalized by the normalization unit 2, and then transformed by the transformation unit 3 according to the optimal transformation method for each writer stored in the optimal transformation method storage unit 12. After that, recognition is performed using a pattern matching method.
以上のようにして、変形部３において変形を施
すことによつて、筆者の書法上の癖が除かれるの
で認識率を向上することができる。 By performing the transformation in the transformation unit 3 in the manner described above, the writer's writing habits can be removed, so that the recognition rate can be improved.
なお、上記実施例においては、正規化部２によ
る正規化を施した後に変形部３による変形を施し
ているが、この順序を入換えても本発明の目的を
達することができる。 Note that in the above embodiment, the normalization unit 2 performs normalization and then the deformation unit 3 performs deformation, but even if this order is reversed, the object of the present invention can be achieved.
また、上記実施例においては、予め最適変形法
記憶部１２を生成し、原則として、変形部３にお
いて被認識入力文字パターンに対し変形を施して
いるが、通常は、変形部３において変形を施さず
に認識をおこない、認識不能あるいは誤認識が生
じたときにおこなう修正作業を通じて最適変形法
を学習し、これを最適変形法記憶部１２に登録さ
せるようにすることもできる。このような方法に
よれば、最適変形法記憶部１２の生成に要せられ
る筆者の負担を軽減できる等の効果がある。 Further, in the above embodiment, the optimal transformation method storage unit 12 is generated in advance, and the transformation unit 3 transforms the input character pattern to be recognized in principle. It is also possible to perform recognition without performing recognition, learn the optimal deformation method through correction work performed when unrecognized or erroneous recognition occurs, and register this in the optimal deformation method storage unit 12. Such a method has the effect of reducing the burden on the author required to generate the optimal deformation method storage unit 12.
その他、変形部３・特徴抽出部４および照合部
５における処理をパイプライン制御することによ
つて認識時間の短縮を図ることができる。 In addition, the recognition time can be shortened by pipeline controlling the processing in the transformation section 3, feature extraction section 4, and matching section 5.
(G) 発明の効果
以上説明したように、本発明によれば、辞書の
記憶容量の増大あるいは筆者に与える負担を増加
することなく、多数の筆者の利用が可能であり且
つ認識率を向上し得る文字認識装置を得ることが
できる。(G) Effects of the Invention As explained above, according to the present invention, the dictionary can be used by a large number of writers and the recognition rate can be improved without increasing the storage capacity of the dictionary or increasing the burden on the writers. You can get a character recognition device.
第１図は本発明一実施例のブロツク図、第２図
は変形部にて行なう変形法を示す図である。図に
おいて、３は変形部、４は特徴抽出部、５は照合
部、９は辞書部、１２は最適変形法記憶部であ
る。
FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a diagram showing a deformation method performed by a deformation section. In the figure, 3 is a transformation section, 4 is a feature extraction section, 5 is a collation section, 9 is a dictionary section, and 12 is an optimal transformation method storage section.
Claims (1)
る辞書部と、筆者毎の所定の文字種毎の入力文字
パターンと該入力文字パターンに複雑種の予じめ
決められた変形法を施して得られる複数の変形文
字パターンとのうち前記標準文字パターンとの類
似度が高い入力文字パターンもしくは変形文字パ
ターンに施した無変形を含む前記変形法を最適変
形法として登録する最適変形法記憶部と、被認識
入力文字パターンを前記最適変形法記憶部に登録
する最適変形法に従つて変形する変形部と、前記
変形部によつて得られる変形入力文字パターンの
特徴を押出する特徴抽出部と、前記特徴抽出部に
よつて得られる特徴と前記辞書部に格納されてい
る特徴とを照合することによつて認識結果を得る
照合部とを備えることを特徴とする文字認識装
置。1. A dictionary section that stores the characteristics of standard character patterns for each character type, an input character pattern for each predetermined character type for each author, and a plurality of input character patterns obtained by applying a predetermined transformation method for complex types to the input character pattern. an optimal deformation method storage unit that registers, as an optimal deformation method, the deformation method including no deformation applied to the input character pattern or deformed character pattern that has a high degree of similarity to the standard character pattern among the deformed character patterns; a transformation section that transforms the input character pattern according to an optimal transformation method that registers the input character pattern in the optimal transformation method storage section; a feature extraction section that extrudes features of the transformed input character pattern obtained by the transformation section; and the feature extraction section. 1. A character recognition device comprising: a matching section that obtains a recognition result by comparing features obtained by the dictionary section with features stored in the dictionary section.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP57220676A JPS59111582A (en) | 1982-12-16 | 1982-12-16 | Character recognizing device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP57220676A JPS59111582A (en) | 1982-12-16 | 1982-12-16 | Character recognizing device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPS59111582A JPS59111582A (en) | 1984-06-27 |
| JPH0344355B2 true JPH0344355B2 (en) | 1991-07-05 |
Family
ID=16754714
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP57220676A Granted JPS59111582A (en) | 1982-12-16 | 1982-12-16 | Character recognizing device |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JPS59111582A (en) |
-
1982
- 1982-12-16 JP JP57220676A patent/JPS59111582A/en active Granted
Also Published As
| Publication number | Publication date |
|---|---|
| JPS59111582A (en) | 1984-06-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wakahara et al. | On-line handwriting recognition | |
| TWI435276B (en) | A method and apparatus for recognition of handwritten symbols | |
| JP3761937B2 (en) | Pattern recognition method and apparatus, and computer control apparatus | |
| EP0476033A4 (en) | Objet recognition system | |
| US5659633A (en) | Character recognition method utilizing compass directions and torsion points as features | |
| JPS6175981A (en) | Recognizer of handwritten character | |
| JPH0344355B2 (en) | ||
| Huerst et al. | Interactive error repair for an online handwriting interface | |
| Rahman et al. | Notice of violation of IEEE publication principles: Modified syntactic method to recognize Bengali handwritten characters | |
| JPH09319828A (en) | On-line character recognition device | |
| JPS60116082A (en) | Individual discrimination method | |
| JPH0442316A (en) | Electronic computer | |
| CN108921155A (en) | A kind of hand script Chinese input equipment Uighur words Slant Rectify method | |
| JPS5842904B2 (en) | Handwritten kana/kanji character recognition device | |
| JP3198218B2 (en) | Online handwriting recognition method | |
| JPH01316889A (en) | Handwritten character recognizing system | |
| JPH0567241A (en) | Online handwritten character recognition device | |
| JP3365538B2 (en) | Online character recognition method and apparatus | |
| Rakate et al. | Integrated approach to handwritten character recognition using ANN and it's implementation on ARM | |
| JPH0253833B2 (en) | ||
| JPH0634259B2 (en) | Character recognition device | |
| JP2665488B2 (en) | Personal dictionary registration method | |
| JPH01125684A (en) | Character recognizing device | |
| JPH05189621A (en) | On-line hand-written character recognizing device | |
| JPS6293776A (en) | Information recognizing device |