JPS6162985A - Recognition order determining system - Google Patents

Abstract

PURPOSE:To improve the recognition rate of the titled system without requiring an enormous knowledge part by determining priority levels of individual character type categories in accordance with frequency and determining the recognition order of candidate character type categories on the basis of calculated degrees of resemblance and priority levels. CONSTITUTION:A binary character pattern outputted from a character observing part 1 is inputted to a pattern matching part 4 through a feature extracting part 2, and degrees of resemblance between a feature pattern and standard feature patterns are calculated. A priority level is determined for each character type category in a dictionary part 3 in accordance with frequency. Operating parts 4a-4c select candidate character type cetegories from category sets in the order of the degree of resemblance, and their character type codes and degrees of resemblance are stored in a candidate memory 5 together with priority levels of category sets. A knowledge part 6 determines the recognition order of candidate character type categories on the basis of relations between degrees of resemblance and priority levels and stores character type codes in a result memory 8 in accordance with the recognition order.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a character recognition system for recognizing handwritten Chinese characters, and more particularly to a recognition ranking determination method for candidate character type categories.

(b) Conventional technology Generally, in a character recognition system, the degree of similarity between a feature pattern extracted from an input character pattern and a standard feature pattern for each character type category registered in advance in a dictionary section is calculated,
Select n candidate character type categories with a high degree of similarity.

Then, the candidate character type category with the highest degree of similarity is output as the recognition result, and in order to correct misrecognition, the selected n candidate character type categories are ranked from 1st to 0th in order of the degree of similarity. Decide on the recognition ranking up to the first rank.

However, as mentioned above, using only similarity to determine the recognition ranking often results in false recognition, so after selecting multiple candidate character categories based on similarity, some post-processing is performed to determine the recognition ranking. Now I can think of a way to do that.

Conventionally, as post-processing, Japanese Patent Application Laid-Open No. 59-3208
As disclosed in Publication No. 2, there are methods that perform grammatical processing, and methods that determine whether the characters before and after the recognized character are kanji, katakana, or hiragana, as in Japanese Patent Application Laid-open No. 59-27381. Something was proposed.

0→ Problem to be solved by the invention In conventional technology, grammatical processing is performed as post-processing, so the knowledge section such as grammatical dictionaries becomes enormous, and furthermore, the processing content becomes extremely complex. There is a problem that
Furthermore, in the method of determining whether the preceding and succeeding characters are kanji or katakana, no improvement in recognition rate could be expected if the selected candidate character categories were only kanji or hiragana.

B) Means for solving the problem The present invention sets a priority level for each character type category according to its frequency, and selects candidate characters based on the similarity obtained by calculation and this priority level. This determines the recognition ranking of species categories.

(Function) According to the present invention, character type categories that have a large degree of similarity and are frequently used are ranked higher in the recognition order, and the knowledge section has a combination of similarity and priority. All you have to do is remember the relationship.

(F) Embodiment FIG. 1 is a block diagram of a character recognition system to which the present invention is applied, and (1) reads characters written on an input manuscript and outputs the reading result as a binary character pattern. Character observation unit, (2) is a feature extraction unit that extracts feature patterns from input character patterns, (3) is a dictionary unit that stores standard feature patterns for each character type category, and (4) is the extracted feature pattern and standard feature. This is a pattern matching unit that performs matching with patterns and calculates the degree of similarity between both patterns.

The character type categories in the dictionary section (3) are categorized according to their frequency, such as those with high frequency in category set 1, those with medium frequency in category set 2, and those with low frequency in category set 3. Priorities 1 to 3 are set for each category set 1 to 3 in order.

The pattern matching section (4) selects category sets 1~
3, each of which selects n candidate character type categories in descending order of similarity from each category set, and selects n candidate character type categories from each category set in descending order of similarity The code and the similarity as the calculation result are stored in the candidate memo (January 5). At this time, the calculation unit adds the priority of the corresponding category set to the character type code and similarity, and these three pieces of information are stored in the candidate memo as information of each candidate character type category. In this manner, a total of 3n candidate character type categories, n from each category set, are stored in the candidate memo January 5).

The knowledge section (6) stores the relationship between similarity and priority, and the clustering control processing section (7) refers to the contents of this knowledge section (6) and stores it in the candidate memo 51. The recognition rankings of the top n characters from among the 30 candidate character categories are determined, and the result memo 1 is recorded in the recognition ranking order of the character type codes.
Stored on month 8). For example, if we use the city block distance as the similarity, and assume that the smaller the distance, the thicker the similarity, the knowledge section (6) specifically contains the distance and distance as shown in Figure 2. The relationship of whether or not the recognition order can be swapped based on the priority is stored.

The manner in which the recognition ranking is determined will be explained below using a specific example.

First, the character type code is ML, the city block distance is Di,
The priority is p (i:=a, b, c・”−, P
=2' +2.3) candidate character type category (MLD
For example, as shown in FIG. 3(a), the top five candidate character type categories with the highest similarity are selected from each of category sets 1 to 3, and each city block The relationship between distance is Dv<DP<D,
Assume that <DQ <Db <DIl <I)w.

Then, as in the conventional method, based only on the size of the city block distance, the recognition order becomes as shown in FIG. 3 (b).

However, now the threshold value in knowledge part (6), ,D,.

The relationship between D and the calculated city block distance is DI<D
If VI Da(D2 < DQI Dw < Ds, the clustering control processing unit (7) sets the character type code Mv.

M, ,M, are ranked as B rank, and character type code M0゜Mb MR9Mw is ranked as C rank, and in order to change the recognition order based on priority within these ranks, as shown in Fig. 3 P→ As such, within each rank, character type codes with high priority will be placed at the top, and as a result,
A recognition ranking is determined based on similarity and priority. Then, ``Top 5 character type codes M, , M, , Mv,
Mb and MQ are stored in the result memo (January 8) in order,
The answer output control section (9) sends the first character type code M to a character display device such as a word processor or a personal computer as a recognition result, and the character type is displayed. At this time, if there is a misrecognition, the character type code of the next recognition order is sent,
Thereafter, the character type codes of the next rank are sequentially sent out until a correct recognition result is obtained.

Therefore, priority is given to candidate character type categories with high similarity and high frequency, and the recognition ranking of such characters becomes high and the recognition rate improves. The distance Dv is sufficiently small compared to other candidates, Dv<DI, and D.

<Dr, :ob <D2, character type code M
v is ranked in the pararank and the recognition order is not changed, and the character type code MP2M12MQ9Mb of rank B is changed in the order based on the priority, and the recognition order is as shown in FIG. 3). In this case as well, when compared to FIG. 3 (b), the recognition ranking of the candidate character type category with high frequency comes to the top.

By the way, in this embodiment, the character type categories are divided in advance into category sets according to the priorities, and the priority information is obtained. You may remember it.

(G) Effects of the Invention According to the present invention, frequently used character categories are ranked higher in the recognition ranking, so the recognition rate improves.
Furthermore, recognition rankings can be determined in a short processing time without requiring a huge knowledge section.

[Brief explanation of the drawing]

Figure 1 is a block diagram of a character recognition system to which the present invention is applied, Figure 2 is an explanatory diagram showing the contents of the knowledge section, and Figure 3 (A).
-) are explanatory diagrams showing specific examples of recognition ranking determination. Explanation of main figure numbers (1)...Character observation section, (2)...Feature extraction section,
(3)... Dictionary section, (4)... Pattern matching section, (5)... Candidate memory, (6)... Knowledge section, (7)... Clustering control processing section, (
8)...Result memory, (9)...Answer output control section. Applicant Sanyo Electric Co., Ltd. and one other agent Patent attorney Shizuo Sano @l r:? 1 (mouth') Mv, MP 1Ma, MQ, Mb, MP, MW; (-') Mv, + Ma, M+A-7
Shihura -1), Mp, Ma' 3U〉U -1-111

Claims (1)

[Claims] (1) A feature pattern extracted from the input character pattern,
In a character recognition system that selects multiple candidate character categories by calculating the degree of similarity with standard feature patterns for each character type category registered in advance in the dictionary section, each of the character type categories is given priority according to its frequency. , and the recognition ranking of the plurality of candidate character type categories is determined based on the similarity obtained by the calculation and the priority.