JPS60160481A - Reader of character - Google Patents

Abstract

PURPOSE:To allow recognition at every character by an optimum recognizing means by outputting the recognized result by the 2nd recognizing means if the recognized result of a character by the 1st recognizing means is included in a read-out object character by the 2nd recognizing means. CONSTITUTION:The recognized result by the 1st recognizing means 5 and that by the 2nd recognizing means 6 with respect to an input character string 8 are sequentially transmitted to an editing means 7. If the recognized result by the 1st recognizing means 5 is not a recognition object character of the 2nd recognizing means 6, that is, KANJI (Chinese character), the recognized result is outputted as a read-out result through a control means 4. If a recognition object character is HIRAGANA (cursive form of, Japanese syllable), said recognized result by the 1st recognizing means 5 is replaced with that by the 2nd recognizing means 6 which is suitable to HIRAGANA, and it is outputted as a read-out result through the control means 4.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention relates to a character reading device for reading character data in which characters of a plurality of different character types are mixed, for example, Japanese characters consisting of kanji, hiragana, numbers, etc. This invention relates to a character reading device that recognizes and reads text.

[Prior Art] A conventional device of this type was constructed as shown in FIG. FIG. 1 is a block diagram of a conventional example of a character reading device. In the figure, (1) is the paper on which the characters are read, (2) is the scanning means that reads the characters by photoelectric conversion, (3> is the recognition means that recognizes the photoelectrically converted character pattern, and (4) is the output of the recognition result. This type of character reading device operates as follows: Characters written on paper (1) are read by scanning means (2) and recognized by one recognition means (3). The recognition result is output via the control means (4).In this type of conventional device, when reading only one type of character set such as numbers, the reading rate is sufficiently high. However, it is possible to obtain characters from multiple character sets with different structures, such as kanji that have a complex sentence structure and consist mainly of combinations of straight lines, or kanji that have a simple structure and mainly consist of curved lines (such as hiragana, which is written in 1a). When mixed character data is read by a single recognition means (3), the processing becomes busy and complicated, and if this type of device is used to handle it, the device becomes expensive.

Furthermore, other conventional devices proposed to overcome this drawback include a plurality of character sets each configured to be suitable for recognition of characters of a plurality of character sets. For each character, one optimal recognition means is selected from the plurality of recognition means described above based on the characteristics of the character pattern, and recognition is performed using the selected recognition means.

The above-mentioned device which selects and uses a recognition means according to such a character mark 1- has the advantage that a high reading rate can be obtained if the recognition means most suitable for recognizing each character mark 1- is selected. However, in the case of handwritten characters that vary widely depending on the habits of the person writing them, it is difficult to select the most suitable recognition method for the input characters with high accuracy, and the overall reading rate is low despite the complexity of the device configuration and processing. The drawback was that it did not rise very high.

[Summary of the Invention] The present invention has been made in order to eliminate the drawbacks of the conventional ones as described above. A second recognition means suitable for recognition is provided, and the recognition result of the first recognition means is corrected using the recognition result of the second recognition means. The purpose is to provide a character reading device.

[Embodiments of the Invention] The present invention will be described in detail below using an embodiment shown in FIG.

FIG. 2 is a block diagram of an embodiment of the character reading device of the present invention. In the drawings, the same reference numerals and the same symbols indicate the same or equivalent parts as in the conventional example. In addition, (5) is especially
The first recognition means targets recognition of all characters, which can accurately recognize characters from a certain character set 1~, but can also recognize characters from other character sets, (6) accurately recognizes only characters from one character set The second recognition means (7) is an editing means that edits the recognition results of the first recognition means (5) and the second recognition means (6).

Next, the operation shown in FIG. 2 will be explained. First, characters written on paper (1) are read by scanning means (2) by photoelectric conversion or the like. This photoelectrically converted /j letter pattern is
A first recognition means (5) that can particularly accurately recognize a certain character set, but also recognize characters from other character sets;
The recognition means (5) is processed in parallel by a second recognition means (6) which can accurately recognize only characters from two character sets.
) and (6) are sent to the editing means (7). In the editing means (7), the first recognition means (
The recognition result of step 5) is sent to the control means (4), especially when the recognition result of the first recognition means (5) is a character of the character set to be read by the second recognition means (6). The control means (4) sends the recognition result obtained by the second recognition means (6) to the control means (4), and the control means (4) outputs the editing result of the editing means (7) as a reading result.

The operation of the embodiment of the present invention will be explained in more detail below, taking reading of Kanji and Kana characters as an example.

Figure 3 is an example of input characters written on form (1),
An example in which the character string "Rainy day" (8) is entered is acceptable. In general, kanji are mainly composed of combinations of straight lines, and hiragana are mainly composed of curved lines. Therefore, the first recognition means (5) uses conventional character recognition techniques mainly suitable for recognizing Chinese characters, such as a method of examining straight line portions of character patterns. In addition, the second recognition means (
In step 6), only hiragana characterlets are recognized using a character recognition technique suitable for hiragana, which has many curved parts, such as a method of examining the outline of a character pattern.

Figure 4 shows the input characters "rain" (9), [noj (10),
[fall J (11), [ru J (12>), ``日 J (13)
is recognized by the first recognition means (5) above, and the recognition results are
(15), [fall J (16), "ro" (17), and r day J (18). Figure 5 shows the same character string (
8) is recognized by the second recognition means (6) above.
For (11), −C is rejected and [J (10), “
J (12) and J (13) are respectively
(19), "ruJ (20) r is an example of recognition as J (21). Here, the symbol [◆J (22) means rejection. 1st
The recognition results of the second recognition means (5) and the second recognition means (6) are sequentially sent to the editing means (7). In the editing means (7), the recognition result of the first recognition means (5) is
If there are no characters to be recognized by the recognition means (6), in other words, if they are kanji, they are recognized with high accuracy, so this recognition result is output as a reading result via the control means (4). , and the recognition result of the first recognition means (5) is the character to be recognized by the second recognition means (6), that is, four characters,
If the character to be recognized is Hiragana, the recognition result of the first recognition means (5) is replaced with the recognition result of the second recognition means (6) suitable for recognizing Hiragana, and the recognition result is passed through the control means (4). Outputs the reading results. For example, in Figure 3, the input characters ``rain'' (9), ``Next J'' (11), ``Sun J''
Against (13) J゛ruf! The recognition results obtained by the recognition method (5) in 1 are the kanji ``Ame J (
14), [fall J (16), and [day J (18)], the editing means (7) sends it as is to the control means (4). However, the entering no j character [ no J (10), ``ru'' (12
), the recognition result of the first recognition means (5) is [tsuJ (15), [roJ (17)] because it is difficult for the first recognition means to read hiragana accurately. Therefore, the second recognition method (6
) is replaced with J (19), rJ (20) and sent to the control means (4). Therefore, it will be modified as shown in FIG. Figure 6 shows one character string (23) sent to the control means (4), and the recognition result of the first recognition means (5) is partially corrected to ensure correct reading. The result ``Rainy day'' is the control means (
You can see that it is being sent to 4).

In the above embodiment, only the recognition result of the first recognition means is corrected by the recognition result of the second recognition means, and the present invention is not limited to this, but the present invention is not limited to this. The recognition result of ``a'' and the candidate character ``a'' may be edited using the recognition result of the second recognition means and the candidate character. Moreover, although the case where there is one second recognition means has been described, the present invention is not limited to this, and for example, a plurality of second recognition means may be provided for each different character cella 1. Furthermore, although the case where the character data is a kanji/kana combination has been described, the present invention can also be applied to reading character data containing a mixture of other alphanumeric characters, symbols, etc.

[Effects of the Invention] As described above, according to the present invention, when the characters recognized by the first recognition means in the editing means are included in the characters to be read by the second recognition means, outputs the recognition result using the second recognition means, and for other characters,
It has a configuration that outputs the recognition result of the first recognition means, and performs character reading in which the recognition result obtained by the first recognition means is corrected by the recognition result obtained by the second recognition means. Because the recognition method is not complicated, classification errors caused by selecting a recognition method for each character in advance do not occur, and since each character can be recognized with the most suitable recognition method, it is possible to improve overall performance. This has the effect of obtaining a high cardinal rate.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a conventional character reading device.
FIG. 2 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 3 is an explanatory diagram showing an example of a kanji-kana combination written on a sheet, and FIG. 4 is an explanatory diagram showing an example of a recognition result. , FIG. 5 is an explanatory diagram showing another example of the recognition result, and FIG. 6 is an explanatory diagram showing a string of characters sent to the control means. In the figure, (1) is the paper, (2) is the scanning means, (3) is the recognition means, (4) is the control means, - (5) is the first recognition means, (7) is the editing means, (6) ) is the second recognition means. In the figures, the same reference numerals and the same symbols indicate the same or corresponding parts. Agent: Ookimi, 2 people from Chigumi Figure 1

Claims (4)

[Claims] (1) A scanning means for optically scanning characters and photoelectrically converting them; and a 10) recognition means for recognizing photoelectrically converted character patterns for the entire set of characters to be read, in parallel with the first recognition means. one or more second recognition means configured to operate and recognize a character pattern for some characterlets of characters to be read; and an editing means for editing the recognition results of each of the recognition means. In a character reading device that recognizes and reads characters that are read by the editing means, when the characters recognized by the first recognition means are included in the characters to be read by the second recognition means; . A character reading device characterized by outputting the recognition result of the second recognition means, and outputting the recognition result of the first recognition means if the character is a different character. (2) The character reading device according to claim 1, wherein the first recognition means is capable of accurately recognizing Chinese characters. (3) The M2 recognition means can accurately recognize Hiragana.
Character reading device described in 1rJ. (4) The character reading device according to claim 1, wherein the second recognition means is arranged in a plurality of stages for different character sets.