JPH02193281A - Character recognizing device - Google Patents

Abstract

PURPOSE:To shorten the character recognition time by determining in advance the number of candidate characters which are brought to detailed calculation in accordance with a degree of 'complicated' and 'simple' of a character and storing it in a calculation candidate charac ter number correspondence table, and determining one candidate character number with regard to a candidate character group which is brought to rough classification. CONSTITUTION:In accordance with a difference of 'simplicity' and 'complicatedness' of a character, the number of candidate characters at the time of executing a feature quantity comparing calculation, which corresponds to each character is stored in advance in a candidate character number correspondence table 7. In this state, the candidate character selected by a rough classification discriminating means 3 is informed to a table retrieving means 8, and the table retrieving means 8 selects the candidate character from among the informed candidate characters by a prescribed reference. Subsequently, by referring to the candidate character number correspondence table 7, the number of candidate characters corresponding to this selected candidate character is retrieved, and a candidate character limiting means 9 determines one candidate character number, based on the number of retrieved candidate characters. In such a way, the number of candidate characters calculated in accordance with 'simplicity' of the character can be limited, and the recognition time of the character can be shortened as a whole.

Description

[Detailed Description of the Invention] <Industrial Field of Application> The present invention relates to a character recognition device, and more specifically, the present invention relates to a character recognition device, and more specifically, to recognize characters, symbols, etc. (hereinafter representatively referred to as “characters”) through an image input device or a communication medium such as a facsimile. ”), extract its feature amount, perform calculations based on the feature amount, select candidate characters of the − group, and select the character to be read from among the candidate characters. The present invention relates to a character recognition device that can precisely identify a character closest to a character and output a signal representing the character.

<Prior art> When performing character recognition using a conventional character recognition device, for example, as shown in FIG.
1) Input an image containing characters, and use the character cutting means (1) to input an image containing characters.
2) Cut out each unit character.

Then, the feature extraction means (13) extracts the feature amount based on the extracted character signal, and then the large classification identification means (13) extracts the feature amount based on the extracted character signal.
In step 4), narrow down multiple candidate characters using a simple method.

The detailed identification means (15) uses the recognition dictionary (1B) for each candidate character sent from the major classification identification means (14).
(The recognition dictionary (16) stores, in addition to the feature quantities, for example, the average value of the feature quantities, the state of distribution, and the rank in which each feature quantity affects recognition.) Obtain feature quantities and other information, compare them with the feature quantities obtained by the feature extracting means (18), calculate the degree of dissimilarity or similarity (hereinafter representatively referred to as "degree of dissimilarity"), and obtain the results. Output. The ranking determining means (17) rearranges the characters in descending order of dissimilarity, and outputs, for example, the character with the smallest dissimilarity as a recognition result.

In particular, as a specific example of this character recognition device,
An example of this is the character recognition device described in Japanese Patent Publication No.-28915.

In this character recognition device, the major classification identification means selects dictionaries with categories that have similar biases, and the detailed identification means selects dictionaries with categories that have similar biases.
For all characters in the dictionary of the selected candidate category, the side parts are masked and only the labor parts are identified in detail. This improves the recognition rate for groups of characters with similar biases.

<Problem to be Solved by the Invention> However, in the character recognition device shown in FIG. Therefore, setting a large number of candidate characters increases the amount of calculation and increases the recognition time. For example, in Kanji OCR that sets 100 candidate characters, the degree of difference is 10 to recognize a character.
You have to calculate it 0 times.

Reducing the number of candidate characters will shorten the calculation time, but in order to ensure a certain level of recognition performance, it is necessary to use a major classification identification method (
It is desirable to secure a sufficient number of candidate characters selected in step 14). Because, major classification identification means (14)
If a regular continuation character is omitted from the candidate characters selected in ,
This is because even if the detailed recognition means (15) performs detailed recognition calculations, it is not possible to obtain regular continuation characters.

For this reason, in the past, the number of candidate characters was set to be relatively large, resulting in a problem in that a large amount of time was wasted for detailed identification.

An object of the present invention is to provide a character recognition device that can shorten the overall character recognition time by employing a simple method.

Means for Solving the Problems> As shown in the figure, the character recognition device of the present invention for achieving the above objects first uses an image to obtain an image signal representing an object to be read including characters. a signal acquisition means (1); a feature extracting means (2) for extracting the feature of a character to be recognized in the image based on the image signal; A large classification identification means (3) to be selected, a recognition dictionary that stores feature amounts and other information necessary for recognizing each character, and a candidate character that stores the feature amounts of the character to be recognized in the recognition dictionary. a detailed identification means (4) that performs a feature comparison calculation by comparing the information with the information of the detailed identification means (4), and selects a character based on a certain standard from among the characters identified by the detailed identification means (4) and outputs a signal representing the character. Recognized character output means (5), a candidate character number correspondence table (7) that stores constants for specifying the number of candidate characters when performing feature value comparison calculations corresponding to each character, and a table search means (8) for selecting one or more candidate characters based on the criteria and searching for the number of candidate characters corresponding to the selected candidate characters; Candidate character limiting means (9) is provided for sending candidate character signals of the number of candidate characters to detailed identification means (4). The detailed identification means (4) performs feature quantity comparison calculation only for the candidate characters specified by the candidate character limiting means (9).

Effects> According to the character recognition device configured as described above, the feature amount extraction means (2) extracts the feature amount from the character signal included in the image signal captured by the image signal acquisition means (1). . The major classification identification means (3) selects a group of candidate characters by a simple method using this feature amount. The number of candidate characters is set to a sufficient number so that there are always consecutive characters.

By the way, if the character you are trying to recognize is complex, the candidate characters will also consist of complex characters similar to it, and if the character you are trying to recognize is simple, the candidate characters will also easily consist of similar simple characters. It can be inferred that Here, "complex" and "easy" refer to whether there is a high or low possibility of misreading other characters during recognition. Even if a character has a large number of strokes, it can be said to be a ``simple'' character if it has unique characteristics and there are almost no similar characters.

Therefore, when selecting candidate character groups using a simple method using the major classification identification means (3), even if fewer candidate characters are selected for "simple" sentences and characters, the correct reading characters will be ranked high. On the other hand, we focused on the fact that for "complex" characters, unless many candidate characters are selected, the correct reading characters will not always be ranked high.

In the present invention, the number of candidate characters corresponding to each character at the time of feature value comparison calculation is stored in the candidate character number correspondence table (7) according to the difference in characteristics such as "simpleness" and "complexity" of the characters. Then, based on the actually selected candidate characters, candidate characters to be subjected to detailed calculation are limited while referring to the candidate character number correspondence table (7).

That is, the candidate characters selected by the major classification identification means (3) are notified to the table search means (8), and the table search means (8) selects one or more candidate characters from among the notified candidate characters based on predetermined criteria. Select candidate characters. Then, referring to the candidate character number correspondence table (7), the number of candidate characters corresponding to the selected candidate character is searched. The candidate character limiting means (9) determines the number of candidate characters based on the number of searched candidate characters, and sends the limited candidate characters to the detailed identification means (4). The detailed identification means (4) performs comparative calculations on only the limited candidate characters while referring to the recognition dictionary (6). The recognized character output means (5) selects a specific character from among the characters identified by the detailed identification means (4) and outputs a signal representing the character.

<Example> Embodiments will be described in detail below with reference to the accompanying drawings showing embodiments.

FIG. 2 is a block diagram showing the configuration of the character recognition device of the present invention, in which (1a) shows an image camera such as a bidicon that captures the entire document, and obtains a formatted signal by binarizing its output signal. It represents an image input section consisting of a binarization circuit. The image signal is subdivided character by character by character cutting means (lb). The feature extraction means (2) extracts the feature amount of each character. For example, it is a histogram of directional vectors of character outlines or a distribution of blank area amounts. This feature quantity is input to the major classification identification means (3), which uses a simple classification function that takes a short calculation time, such as a city block function or a linear-order function, to identify all character types. Calculate the dissimilarity of Then, a plurality of candidate characters are selected.

The data of each candidate character is sent to the table search means (8), and the table search means (8) uses the candidate character number correspondence table (7) to search for the highest character among the candidate characters (the one with the least degree of difference). (a certain character) or the number of candidate characters for the top characters (
7) Search for the number of candidate characters. Candidate character limiting means (9
) determines the number n of one candidate character based on the number of candidate characters, and sends the top n candidate characters from the candidate character group to the detailed identification means (4). Candidate Character Number Correspondence Table (The contents of the force may be determined in advance when creating the Candidate Character Number Correspondence Table (7), but it is preferable to update the contents based on the results of actual use. For example, when using a character recognition device It is sufficient to add a learning function to the system so that it can be updated through learning.

Alternatively, the information may be updated artificially without using the learning method.

The detailed identification means (4) calculates the degree of dissimilarity for each of the n candidate characters. For example, the recognition dictionary (6
) and the feature extraction means (2).
) may also be used to find the distance to the feature amount found in ( ).

In this way, since the degree of dissimilarity is determined for each candidate character, the ranking determining means (51) rearranges the candidate characters in descending order of degree of dissimilarity and outputs the character code of the first rank.

Next, we will change the operation of the above character recognition device to "It's sunny today".
An example of recognizing the first character "hon" in a manuscript written as "hon" will be explained.

When the image input unit (1a) reads a manuscript that says "Today is sunny", the character extraction means (1b) reads "Book".
Image signals are converted into single characters such as ``1st'', ``ha'', ``sunny'', ``heaven'', ``na'', and ``su''. Then, the feature extraction means (2) extracts the feature amount X (Xl, X2.-, X
Extract n). The major classification identification means (3) uses the method described above to identify a plurality of (for example, about 50) candidate characters "hon" and "hon".
Select "Thursday", "Wednesday", "Sho"... Table search means (
In step 8), for example, the number of candidate characters is searched for the highest character "hon" among the candidate characters, and a result of 3 is obtained from the candidate character number correspondence table (7). The candidate character limiting means (9) limits the top three candidate characters "hon", "wood" and "water" and supplies them to the detailed identification means (4). The detailed identification means (4) performs distance calculations on only the three characters mentioned above. Recognition dictionary (6
) is the feature amount of the candidate character “hon” in A (A1.A
2. -An), the feature amount of "tree" is B (Bl, B2.-
... Bn), r water' feature amount C(CL, C2, -Cn
), then the distance
-A I -[(XI-Al)2 +(X2-A2)2
+-+ (Xn-An) 2 ko 1/2X -B
l −[(Xi−Bl)2 +(X2−82)2+−
+ (Xn-Bn)2] 112X -CI -[(
Xi-C1)2 + (X2-02)2+-+ (X
Find n-Cn) 2 ko l/2.

Note that this city-block function is sometimes used for the above-mentioned major classification identification, but for the major classification identification,
Instead of taking the sum of all terms (1 to n),
For example, the calculation procedure differs from the detailed classification calculation procedure in that the calculation is performed at high speed by "thinning out" three or five parts at a time.

A ranking determining means (51) rearranges the characters in descending order of distance and outputs the character code of the first ranking.

Note that even if the number of candidate characters is limited and detailed classification is performed in this way, the recognition performance will not deteriorate because input characters with a small number of candidate characters have regular continuation characters at the top.

FIG. 3 shows a graph showing the number of candidate characters determined for all characters of ``It's sunny today'' in the manner described above. For ``simple'' characters such as ``hon'' and ``日'', the number of candidate characters is small at 2 to 3, but for ``complex'' characters such as ``sunny'' and ``heaven'', the number of candidate characters is small. The number is 5-8.

Figure 4 shows the time required to recognize "It's sunny today" using the above method.

FIG. 4(a) shows the time required for recognition by a conventional character recognition device. Since the same calculation time t is required for each character, it takes X7 time for all characters. On the other hand, FIG. 4(b) shows the time required for recognition using the character recognition device of this embodiment, and the calculation time t1. t2°X7 is different. As a result, the overall recognition time becomes t=Σtn, and it can be seen that it can be shorter than the case of FIG. 4(a).

In addition, in the above example, an example was shown in which the number of candidate characters is determined based only on the highest character "hon" among the candidate characters, but the number of candidate characters is determined only based on the highest character "hon" among the candidate characters. The number of characters may be determined, and the number of one candidate character may be determined based on the number of candidate characters found. For example, if the number of candidate characters for the top three characters ``book'', ``wood'', and ``water'' is 3 and 4.5, respectively, then the average value of these candidate character numbers, the maximum value, and the minimum value are taken to form one. The number of candidate characters may also be determined.

This is because each candidate character included in the candidate character group realized as a result of the major classification is considered to have approximately the same number of candidate characters in the candidate character number correspondence table (7). This is because the results do not differ much even if the number of candidate characters is determined for the top few characters.

Further, in the above embodiment, the detailed identification means (4) simply calculates the distance, but it is of course possible to use a more complicated calculation formula in order to increase the recognition rate.

Even in this case, the number of candidate characters to be calculated can be limited depending on the "simpleness" of the candidate characters, so the present invention is very effective in shortening calculation time.

Further, in the above embodiment, the image camera was used to input the document image, but the invention is not limited to this.
The upper part of the image may be input through a communication line such as a facsimile.

Various design changes can be made without changing the gist of the invention.

<Effects of the Invention> As described above, according to the character recognition device of the present invention, the number of candidate characters to be calculated in detail is determined in advance according to the degree of "complexity" or "simpleness" of the character, and a calculation candidate character number correspondence table is created. The number of candidate characters for one candidate character group was determined for each broadly classified candidate character group.

Therefore, in the past, detailed calculations were performed for all candidate characters, which took a long calculation time, but in the present invention,
- Character recognition time can be shortened because comparison calculations can be made for only limited candidate characters per human character. In particular, since the time required to recognize "simple" characters is significantly reduced, the average recognition time per -character can be reduced when recognizing general documents containing a mixture of "complex" and "simple" characters. Therefore, if the same recognition time is allowed, the precision of character recognition can be increased, and the recognition performance of the character recognition device can be improved.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of the character recognition device of the present invention, Fig. 2 is a block diagram showing an embodiment of the character recognition device, and Fig. 3 is a candidate determined corresponding to each input character. A graph of the number of characters, FIG. 4 is a comparison table of character recognition time with a conventional example, and FIG. 5 is a block diagram showing the configuration of a conventional character recognition device. (1)...Image signal acquisition means, (2)...Feature quantity extraction means, ■...Major classification identification means, (4)...Detailed identification means, ■...Recognized character output means, (6)...Recognition dictionary, (7)...Candidate character number correspondence table, (8)...
Table search means, (9)... Candidate character limiting means Patent applicant: Sumitomo Electric Industries, Ltd. Agent

Claims (1)

[Claims] 1. Image signal representing the object to be read including characters an image signal acquisition means for acquiring the image signal; Trying to recognize images based on image signals Feature extraction to extract the features of characters Calculations are performed based on the means and the above features. Broad classification identification means for selecting candidate characters for numbers and the features necessary for recognizing each character. The recognition dictionary that stores the information Use recognition words to identify the features of the character you are trying to recognize. Compare the candidate character information stored in the detailed identification means for performing feature value comparison calculation; From among the characters identified by detailed identification means Select characters based on certain criteria and select the characters Recognized character output means that outputs a signal representing In a character recognition device having Number of candidate characters when calculating feature comparison Stores constants specifying each character. The correspondence table for the number of candidate characters and the above candidates Select one or more characters from among the characters according to the specified criteria. Select a number of candidate characters and select this selection Search for the number of candidate characters corresponding to the candidate character Table search method and number of searched candidate characters Determine the number of candidate characters based on Detailed knowledge of candidate characters corresponding to the number of candidate characters Candidate character limiting means to send to another means and the above detailed identification means are candidates. To the candidate character specified from the character limiting means The feature is that it performs feature comparison calculations for character recognition device.