JPH0535920A - Word recognition device - Google Patents

Abstract

PURPOSE:To improve recognition speed by deciding a word using a word score determined from the similarity degree, the existence rate and the form value among continuous three characters a three-character word in the recognition of sound/character, etc. CONSTITUTION:This device is provided with a word image storage means 101 storing a word image and a CPU. Characters are segmented from the CPU by changing the segment position variously from the word image, the segmented characters are recognized, the similarity degree, the existence rate, and the form value between continuous arbitrary number of character, calculated regarding the recognized character candidates, are a word score is calculated based on the numeric value and an appropriate word is decided from the calculated word scope.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a word recognition device for recognizing speech / characters in word units.

[0002]

2. Description of the Related Art In a conventional voice / character recognition device,
When recognizing a word, the recognition result candidates are expanded from the character candidates obtained from the combination of cut-out candidates in the word candidate area to sequentially create a new character string, and a language dictionary search is performed for the character string. By doing so, the word is decided (JP-A-59-078400, JP-A-63-21).
No. 6188).

[0003]

However, in the above-mentioned determination method, (i) when the number of cut-out candidates increases due to the contact or separation of characters, a large amount of character strings are created by the expansion process and the processing speed increases. Degradation occurs. Also,
(Ii) It is difficult to select a word (proper noun, numerical value, etc.) that is not registered in the language dictionary as the correct answer.

In the present invention, instead of considering all combinations of character string candidates, characters are determined (or limited to upper candidates) from the beginning, and only characters that may lead to the already determined character are determined. By considering
The above problem (i) is solved. Further, the above problem (ii) is solved by evaluating the existence rate of only three consecutive characters forming the character string.

That is, according to the present invention, when a word is determined from a word candidate area in the recognition of a voice / character, etc., the cutout candidates of the word candidate area are sequentially scanned from the left end to the right end, and three consecutive characters are extracted. A word recognition device that determines a word while sequentially determining (or limiting to upper candidates) the optimum character at each position of a cutout candidate using the word score obtained from the similarity between three characters, the existence rate, and the shape value Is provided.

[0006]

FIG. 1 is a block diagram showing the configuration of the present invention. As shown in the figure, the present invention is
Word image storage means 101 for storing an image to be recognized as a word read by an image reader,
A cutting-out means 102 for cutting out an image to be recognized as a character from the image stored in the word image storage means 101 at various cutting-out positions, and a cutting-out means 102.
Character recognition means 103 for recognizing a plurality of types of images cut out by character recognition as character candidates, and similarity, existence ratio, shape value between any number of consecutive characters for the character candidates recognized by character recognition means 103 And a word determining means 10 for determining a valid word from the word score calculated by the calculating means 104.
It is a word recognition device comprising 5 and 5.

The cutout means 102, the character recognition means 103, the calculation means 104, and the word determination means 10 of the present invention.
As 5, it is convenient to use a microcomputer consisting of a CPU, ROM, RAM, and I / O port.
The word image storage means 101 is usually R
AM is used.

[0008]

According to the present invention, the word image storage means 101
From the word image stored in, the character is cut out by changing the cut-out position variously, and the character is taken as a character candidate, and consecutive n (however, n
Is a natural number) and the similarity, existence rate, and shape value between the characters are calculated, and a valid word is determined based on the word score obtained from the numerical values.

Therefore, the optimum character is sequentially determined from the cut-out character candidates, so that the processing speed can be shortened when the cut-out character candidates are increased due to contact or separation. In addition, the recognition rate of words that are not registered in the language dictionary is improved.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below based on the embodiments shown in the drawings. The present invention is not limited to this.

FIG. 2 is a block diagram showing the configuration of an embodiment of the word recognition device according to the present invention. The present invention will be described below by taking a printed alphanumeric character OCR (optical character reading device) as an example. In this figure, 1 is a word image buffer memory for storing an image (word rectangular area) to be recognized as a word read by an image reader,
2 is a character candidate buffer memory for storing character candidates, 3 is a word candidate buffer memory for storing word candidates, 4 is a word selection buffer memory for storing a selected word, and 5 is a working memory. The word image buffer memory 1, the character candidate buffer memory 2, the word candidate buffer memory 3, the word selection buffer memory 4, and the working memory 5 are each composed of RAM.

Reference numeral 6 is a CPU, and 7 is a word score calculation section comprising a processor. The word score calculation section 7 is a 3-character similarity calculation section 7a, a 3-character existence rate calculation section 7b, and a 3-character shape value calculation. It has a portion 7c. Reference numeral 8 is a control unit including a processor that controls the entire system.

The CPU 6 cuts out an image to be recognized as a character from the image stored in the word image buffer memory 1 at various cutting positions and stores it in the character candidate buffer memory 2.

Next, the cut-out plural kinds of images are recognized as character candidates (cut-out candidates), and for the recognized character candidates, the similarity between three consecutive characters, the existence rate, and the shape value are respectively calculated between three characters. The similarity calculation unit 7a, the inter-character existence ratio calculation unit 7b, and the inter-character shape value calculation unit 7c calculate,
Based on these calculated values, the word score calculation unit 7 calculates a word score and sets it as a word candidate. Then, a valid word is determined from the calculated word score.

Here, the similarity between three characters means that three consecutive characters are consecutive.
It is the average value of the similarity of two characters.

The existence ratio between three characters is the probability that a combination of three consecutive characters appears in an English sentence. For example, t
It is considered that the combination of he and abl has a higher three-character existence rate than the combination of zzz and # $%.

The three-letter shape value is a value indicating the validity of the relative shape of three consecutive characters. For example, if the heights of three characters are almost the same, the combination of ace is considered to have a higher inter-letter shape value than the combination of aCe.

The word score is the similarity between three characters, the existence rate,
It is a value that is obtained from the shape value and represents the validity as a word up to the n-th (where n is a natural number) character. n is 1 or 2
For the similarity between three characters, the existence rate, and the shape value in this case, the similarity, the existence rate, and the shape value between one character and two characters are used.
The word score is calculated by the following formula. S (1) = Wr.R + Wp.P + Wf.F S (n) = {S (n-1) + Wr.R + Wp.P + Wf.F} / 2 S (n) Word score up to the nth character (however, n
Is a natural number) R, P, F 3 Character similarity, existence ratio, shape value Wr, Wp, Wf 3 Character similarity, existence ratio, shape value weight

The cutout candidates are candidates for characters and character breaks in the word rectangular area. The right and left ends of the word rectangular area are also counted as one of the cutout candidates. All characters and places where there is a possibility of character breaks are obtained as cut-out candidates, so normally divide into more than the actual number of characters. For example,
The letter O may be cut out like (),
The letter m may be cut out like rn.

FIG. 3 is an explanatory view showing the stored contents of the character candidate buffer memory 2. A character candidate is a candidate for a character rectangular area created from a combination of the cutout candidates. The character candidate has information such as a cutout candidate at the left end of the character rectangular area (hereinafter, the leading line), a cutout candidate at the right end (hereinafter, the ending line), and a recognition result candidate obtained by character recognition. All character candidates in the word rectangular area are stored in the character candidate buffer memory 2. For example, when there are four cutout candidates from No. 1 to No. 4, character candidates of the following combinations of cutout candidates are stored in the character candidate buffer memory 2. 1-2 1-3 1-4-4 2-3 2-4 3-4

FIG. 4 is an explanatory view showing the stored contents of the word candidate buffer memory 3 and the word selection buffer memory 4.
In the word candidate buffer memory 3, the characters determined from the beginning in the word rectangular area are sequentially stored. The tail line of the last character of a word candidate is called the tail line of the word candidate. The number of word candidates is not always one, and all word candidates having a word score larger than a certain threshold value are stored in the word candidate buffer memory 3. For each word candidate, a character having the same start line as the end line is added to the end, a new word score is calculated, and the word score is moved to the word selection buffer memory 4. After the word candidates in the word candidate buffer memory 3 are evaluated once, the high-rank candidates having a high word score are returned from the word selection buffer memory 4 to the word candidate buffer memory 3 again.

Next, the operation of the present invention will be described with reference to the word determination flowchart of FIG. In FIG. 5, the start state is a state in which the word rectangular area, the cutout candidate, and the character candidate have already been obtained. Of the cutout candidates of the word rectangular area, the cutout candidate that is the target of the processing at a certain point of time is called a line of interest. The line of interest is sequentially scanned from the left end to the right end of the word rectangular area, and when the scanning is completed, the one remaining in the word candidate buffer memory 3 is the correct answer candidate.

First, the word candidate buffer memory 3 is cleared, and one word candidate consisting of only blanks is created and stored in the word candidate buffer memory 3 with the cutout candidate at the left end of the word rectangular area as the end line. Further, the cutout candidate at the left end of the word rectangular area is set as the attention line (step 11). Then, the word selection buffer memory 4 is cleared (step 12). Next, one word candidate having the same end line as the attention line is selected from the word candidate buffer memory 3. That is, the word candidate is extracted and deleted from the word candidate buffer memory 3 (step 13).

Then, one character candidate having the same leading line as the target line is selected from the character candidate buffer memory 2 (step 14). Further, one recognition result candidate obtained by character recognition is selected from the currently selected character candidates (step 15).

Next, the word selected at the end of the currently selected word candidate and the character obtained above are stored in the word selection buffer memory 4 as a new word candidate. At this time, the word score of the word candidate is obtained from the last three characters by using the similarity between three characters, the existence rate, and the shape value (step
16). Then, if there is another recognition result candidate obtained by character recognition among the currently selected character candidates, the process returns to step 15 (step 17). If there is another character candidate in the character candidate buffer memory 2 that has the same leading line as the line of interest, the process returns to step 14 (step 18). Furthermore, if there is another word candidate in the word candidate buffer memory 3 that has the same end line as the line of interest, step 13
Return to (step 19).

Then, from the word selection buffer memory 4, a word candidate with the highest word score (or a high-rank candidate) is selected.
Is returned to the word candidate buffer memory 3. At this time, among the end lines of the word candidates in the word candidate buffer memory 3, the leftmost one in the word rectangular area is set as a new line of interest (step 20). If the new line of interest is not at the right end of the word rectangular area, there are still characters to follow, and the process returns to step 12 (step 21). Then, the word candidate having the highest word score is determined as the correct answer from the word candidate buffer memory 3 (step 22).

In this way, the character candidates cut out from the image of the word are sequentially scanned from the left end to the right end, and the cut out characters are extracted from the similarity between three characters, the existence rate, and the shape value of three consecutive characters. By determining the word while sequentially determining the optimal character at each position of the candidate,
Even characters that are difficult to determine due to separation or the like can be accurately and quickly recognized.

[0028]

According to the present invention, when a character candidate is cut out from a word image, the cutting position is variously changed and the word is determined while the optimum character is sequentially determined. The processing speed is reduced when the number of character candidates to be cut out increases due to reasons such as separation. In addition, the recognition rate of words that are not registered in the language dictionary is improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of the present invention.

FIG. 2 is a block diagram showing the configuration of an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing stored contents of a character candidate buffer memory.

FIG. 4 is an explanatory diagram showing stored contents of a word candidate buffer memory and a word selection buffer memory.

FIG. 5 is a flowchart showing the operation of the embodiment.

[Explanation of symbols]

 1 word image buffer memory 2 character candidate buffer memory 3 word candidate buffer memory 4 word selection buffer memory 5 working memory 6 CPU 6a 3 character similarity calculation unit 6b 3 character existence rate calculation unit 6c 3 character shape value calculation unit 7 Word score calculation unit 8 Control unit

Claims (1)

Claims: 1. A word image storage means for storing an image to be recognized as a word read by an image reader, and an image to be recognized as a character from the image stored in the word image storage means. The cutting means for cutting the cutting position in various ways, the character recognition means for recognizing a plurality of types of images cut out by the cutting means as character candidates, and the character candidates recognized by the character recognition means for consecutive arbitrary Calculation means for calculating the similarity, existence rate, and shape value between a certain number of characters, and a word determination means for calculating a word score based on the numerical values; and a word determination means for determining a valid word from the word scores calculated by the calculation means. A word recognition device comprising: