JPS60217483A - Recognizer of character - Google Patents

Abstract

PURPOSE:To obtain a character recognizer capable of easy correction by storing plural recognition candidates, by displaying the 1st-ranked candidate, then by displaying other candidate when the 1st-ranked candidate is not sufficient as a recognition result and by displaying the second candidate with the aid of commands. CONSTITUTION:Writing strokes of character/mark patterns inputted on a tablet face 11a with the aid of a pen 12 are detected as a time series of position coordinate data, and character-recognized in a recognition processing part 13. Plural recognition candidates may be obtained sometimes, and stored in a recognition condidate memory in the highly-evaluated order. A display control part 14 displays the 1st-ranked recognition candidate on a display device 15, however, when there is little difference between the evaluation of the 1st-ranked candidate and that of the next and subsequent candidates, the display control part 14 performs a different display such as underlining them. When a mode swithing information input part 11b is commanded for inputting, an operation is shifted to a correction mode. If characters entered on the tablet face 11b are specified and inputted, other recognition candidate is read out from the recognition candidate memory, which is displayed on the display device 15 instead of an image pattern of the recognition candidate.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a character recognition device that can easily correct recognition results obtained by recognition processing of input character/symbol patterns.

[Technical background of the invention and its problems]

Recently, various devices have been developed that use character recognition devices to recognize characters and symbols input by hand, encode the recognition results, and input them into a computer. However, the characters that make up the Japanese language consist of kanji, hiragana, katakana, alphanumeric characters, etc., and number over several thousand characters. For this reason, it has been extremely difficult to accurately recognize and code the various characters and symbols mentioned above from handwritten character and symbol patterns.

Therefore, in recent years, various Japanese word processors have been put into practical use that convert kana character code strings input using a kana keyboard into kana-kanji characters to create Japanese sentences containing a mixture of kana and kanji characters. However, this type of Japanese word processor is still expensive and requires some experience in operating the kana keyboard, so there is a problem that not everyone can use it right away.

By the way, a coordinate input device 1 such as a tablet, as shown in FIG. - A device has been developed that performs character recognition processing by providing time-series coordinate data indicating the writing strokes of symbols to the recognition processing unit 3. This recognition processing unit 3 is configured with a CPU, etc., and detects the characteristics of the input character/symbol pattern from the time-series coordinate data of the writing strokes, performs character recognition, and outputs a character code indicated by the recognition result. It is something to do. Further, images of characters and symbols indicating the above-mentioned recognition results are displayed on the display section 4. However, in this type of device, since the character types that make up the Japanese language are diverse as described above, a menu part 1a is provided on the tablet surface of the coordinate input device 1 for specifying the input character type. , the character/symbol pattern can be input while specifying the character type to be input by hand. For this reason, there was a problem in that the input operation was extremely troublesome. Apart from such problems, character and symbol patterns input by hand have so-called quirks peculiar to scribes, and erroneous recognition is likely to occur due to pattern deformation due to these quirks. If this erroneous recognition occurs, you can selectively display other recognition candidates for the handwritten character/symbol pattern to search for the desired recognition result, or does not exist in the recognition candidates, enter the character/symbol pattern again by hand, or use the numeric keys 1b to enter the character/symbol pattern to be input.
It was necessary to input the symbol code directly.

In this way, each time one character/symbol pattern is input and the recognition result is confirmed, the handwritten input data of the next character/symbol pattern is imported and the recognition process is repeated. There is. Therefore, inputting characters and symbols was troublesome.

In addition, in a device that uses an optical character recognition device (OCR) to read characters and symbols handwritten on a form 5, etc., for example, as shown in Fig. 2, characters read and rejected by the 0CR6 are used as input patterns. displayed on the display device 7,
At the same time, a plurality of hypothetical character patterns of recognition candidates determined for the input character pattern are displayed on the display device 7. Then, a selection instruction for the plurality of recognition candidates is inputted from the keyboard device 8, and the recognition candidate is selected by the control section 9.

However, each time the above recognition reject occurs, a correction mode is set for that input character pattern, so
When the quality of the characters and symbols written on the form 5 is poor and many recognition rejects occur, there is a problem in that the processing efficiency is significantly reduced. Moreover, the operator is required to constantly wait and deal with the above-mentioned recognition rejects, which poses a problem of a heavy burden.

[Purpose of the invention]

The present invention was made in consideration of these circumstances, and its purpose is to recognize and process handwritten characters and symbols, and to easily and efficiently modify the recognition results to achieve desired results. An object of the present invention is to provide a highly practical character recognition device that can obtain recognition results and input the recognition results into a computer or the like.

[Summary of the invention]

The present invention is applicable to, for example, input character/symbol patterns given as time-series coordinate data indicating handwriting strokes inputted via a coordinate input device, or characters/symbol patterns inputted as a character pattern image via an optical character reading device. A symbol pattern is recognized and a plurality of recognition candidates are generated for each input character/symbol pattern, and these multiple recognition candidates are stored in correspondence with each input character/symbol pattern, and each of the input characters/symbols is When the first recognition candidate with the highest recognition evaluation value among the plurality of recognition candidates selected for each pattern is insufficient as a recognition output result, that state is stored in correspondence with each input character/symbol pattern. This is how it was done. Then, the image pattern of the first-rank recognition candidate category found for each of the input character/symbol patterns is displayed, and the display image pattern of the first-rank recognition candidate is in an insufficient state as the recognition output result. In some cases, this recognition candidate is displayed as a result of other recognition outputs in a manner different from the image patterns of sufficient recognition candidates, and when these displayed image patterns of recognition candidates are designated and input as their display position coordinates. , Display the display image pattern of another recognition candidate determined for the input character/symbol pattern from which the recognition candidate was obtained, in place of, for example, the display image pattern displayed at the display position seated image inputted with the above-mentioned instruction. This is how it was done.

〔Effect of the invention〕

Thus, according to the present invention, a plurality of recognition candidates determined for an input character/symbol pattern are stored in correspondence with each input character/symbol pattern, and the first-ranked recognition candidate is visualized. Display the image, indicate what should be corrected using the display position coordinates, and display the next recognition candidates in order, so you can easily and efficiently specify the desired recognition output result and make corrections to the recognition result. It becomes possible to do this. Therefore, while inputting character/symbol patterns continuously, it is possible to make corrections to each recognition candidate all at once, making it possible to improve input processing speed and recognition processing efficiency. . Moreover, since it is not necessary to specify the input character type as in the conventional case, the handling can be simplified, and other effects can be achieved.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 3 is a schematic block diagram of the embodiment apparatus, and FIG. 4 is a schematic block diagram of the recognition processing section in the embodiment apparatus. This embodiment device uses a tablet device 11 as a coordinate input device,
The writing strokes of character/symbol patterns input on the tablet surface 11a of the tablet device 11 using a writing implement 12 such as a pen are detected as a time series of position coordinate data, and this time series coordinate block is detected by a recognition processing unit. 13 for character recognition. However, this recognition processing section 13 includes, for example, a coordinate detection circuit 13a, a stroke number detection circuit 13b, a character detection circuit 13,
C1 standard pattern memory 13d1 collation processing circuit 13e1
and a recognition candidate memory 13f. The stroke number detection circuit 13b detects the number of strokes of an input character/symbol pattern by detecting each writing stroke from the position coordinate series detected by the coordinate detection circuit 13a, and the character detection circuit 13c detects the number of strokes of the input character/symbol pattern. The completion of input of one character/symbol is detected. Therefore, the matching processing circuit 13e selects the 4jii of the recognition target category of the corresponding number of strokes read out from the standard pattern memory 13d according to the number of detected strokes.
Recognition candidates for the input character/symbol pattern are determined by comparing the Q pattern with the characteristics of each writing stroke of the input character/symbol pattern detected by the coordinate detection circuit 13a. A plurality of recognition candidates may be selected for one input character/symbol pattern, and these recognition candidates are ranked in descending order of their evaluation values, for example, as shown in FIG. It is designed to be stored in the candidate memory 13f. In addition, in FIG. 5, the coordinate data (X.

Y) indicates the input position coordinates of the input character/symbol pattern, and the characters/symbols shown corresponding to the coordinate data indicate a standard pattern displayed as an image of a recognition candidate.

In this way, the display control unit 14 selects the first-rank recognition candidates for each input character/symbol pattern determined by the recognition processing unit 13, converts the standard pattern of the recognition candidates into an image, and outputs the image to the display device 15. are doing. The display of each of the recognition candidate standard patterns on the display device 15 is performed in accordance with the input position coordinates on the tablet surface of the input character/symbol pattern whose recognition candidates have been obtained based on the coordinate data. As a result, as shown in FIG. 3, each first-rank recognition candidate of the input character/symbol pattern is displayed corresponding to the input position.

Therefore, the display control unit 14 displays the above-mentioned evaluation values as images on the display device 15 because there is a small difference between the evaluation values of the recognition candidates in the first ranking and the evaluation values of the recognition candidates in the second ranking and below. When the first-ranked recognition candidate is insufficient to be used as a recognition output result, in order to display this status, the display output format is made different from other ones by, for example, adding an underline to the corresponding display character. There is. Incidentally, such status information is simultaneously stored in, for example, the recognition candidate memory 13f.

Such handwritten input of characters and symbols, their recognition processing, and display output of recognition candidates are performed by the recognition processing unit 12 from the mode switching information input unit 11b provided on a part of the tablet surface 11a of the tablet computer 1 to the device 11. The input mode is set according to the mode information input to the input mode. In this input mode poem, characters and symbols are continuously input regardless of the presence of misrecognition of the input characters and symbols, and recognition processing is performed sequentially for the input character and symbol patterns, and each of the above input Recognition candidates for characters and symbols are displayed. The recognition candidates found for each input character/symbol are stored in the recognition candidate memory 13f in the order according to their evaluation values.
As mentioned above, the information is stored in the .

By the way, after inputting a series of characters/symbols as described above, when an instruction is input to the mode switching information input section 11b, the operation mode of the apparatus is switched from the input mode to the correction mode. However, at this time, the display device 15
Look at the character/symbol image patterns of recognition candidates displayed in・Image patterns such as symbols are displayed together with image patterns of correctly recognized recognition candidate characters/symbols.The above-mentioned erroneous recognition may occur between similar characters due to deformation of the input pattern, or u) IL It may also occur between similar characters of different character types.

In other words, with this device, characters and symbols can be input continuously in the same way as when creating a normal sentence, without specifying the input character type as in the past. For this reason, for example, the kanji characters "ma" and "open" may be misrecognized, the number "1" and the alphabet "I" may be misrecognized, and even parentheses "(" may be misrecognized as the alphabet "C". or

Corrections to the character patterns of the recognition candidates displayed in this way can be made using the tablets 1 to 11 of the tablets 1 to 11.
This is done by inputting the corresponding characters written above a)]. The displayed character image of the recognition candidate is associated with the input character from which the recognition candidate was obtained using the coordinate data, and therefore, the recognition candidate to be corrected can be specified by inputting the input position coordinates of the corresponding character. It is now possible to do so. The display control unit 14 then reads other recognition candidates for the input character/symbol pattern specified by this position coordinate information from the recognition candidate memory 5f, and displays the image pattern of this recognition candidate on the display device 15. It is displayed instead of the image pattern of the recognition candidate currently displayed.

This display switching of recognition candidates is performed each time the input position coordinates of the input character/symbol are specified. This display switching control is performed, for example, by sequentially loading the storage order of recognition candidates in the recognition candidate memory 13f, as shown in FIG. When the desired recognition result among the plurality of recognition candidates is displayed by this display control, the modification of the input character/symbol is completed. This correction processing is performed on all the erroneously recognized recognized ml corrections, and the correction mode is canceled when all the corrections are completed. At this time, the character/symbol code of each recognition candidate stored in the first rank position of the recognition candidate memory 13f is output to a predetermined computer, and then the recognition candidate memory 13f is cleared and used for the next input. Make it standby. Note that this modification process may be performed all at once after a certain amount of character input, but it may also be performed sequentially.

In this way, multiple recognition candidates are determined for the input pattern,
According to this device, which selects a desired recognition result from among these recognition candidates and performs correction processing, the probability of obtaining a correct recognition result increases as the number of recognition candidates increases, as shown in Figure 7, which shows the recognition cumulative classification rate. Therefore, it becomes possible to input desired characters efficiently through the above correction processing.

Further, as described above, since the recognition candidates can be appropriately corrected while continuously inputting characters, input efficiency is high and handling is not troublesome. Moreover, since the character to be corrected can be easily specified as the written coordinate position of that character, effects such as extremely easy correction work can be achieved.

In particular, since this is different from the conventional method of inputting characters while specifying the input character type, it is possible to input characters with a natural feel.

By the way, the above-mentioned embodiment is a character recognition device that recognizes characters/symbol patterns in real time by inputting them in writing via the TABLEF 1 to device 11, but OCR is used to recognize handwritten characters written on a form or the like. It can also be implemented as a device. FIG. 8 shows a schematic configuration of the apparatus according to the embodiment. Reference numeral 21 denotes an OCR that optically reads character/symbol patterns written on a form as an image pattern. This OC
The character/symbol pattern input via R21 is input to the recognition processing section 22, where the character is recognized. Recognition processing unit 2
2 uses a so-called inspection process to cut out a single character pattern from an input pattern and performs character recognition between it and a standard pattern prepared in advance. It is stored in the recognition candidate memory 23 correspondingly. The plurality of recognition candidates found for each input character/symbol pattern are stored in the same manner as in the embodiment described above. And the first
The image pattern of the recognition candidate for the rank is displayed on the display device 24 under the same display control. The coordinate input device 25 is
For example, by moving a cursor on the display screen of the display device 24, a recognition candidate character to be corrected is specified. It should be noted that the apparatus using this OCR 21 has a faster processing speed than the above-mentioned real-time input apparatus, and generally, if a highly reliable recognition result is not obtained, character rejection occurs. Normally, in conventional devices, a reject symbol is attached to the above character reject, but in this device, as shown in FIG. 9, a reject flag RF is simply set in the recognition candidate memory 23, and the first Recognized ranking candidates are displayed. The display pattern of the characters/symbols to which this reject flag is attached is displayed in a different manner from other characters to indicate that they are candidates for correction.

By specifying the character to be corrected using the cursor on such a display screen and displaying the next recognition candidates in order, it is possible to easily select the correct recognition result from among the plurality of recognition candidates, as in the above-mentioned embodiment. It becomes possible to choose,
It becomes possible to improve the processing efficiency.

As described above, according to the present device, a plurality of recognition candidates obtained by recognition processing of input characters/symbols are stored in correspondence with the input characters/symbols, and the first-ranked recognition candidate is displayed. Then, the characters/symbols to be corrected are simply specified as position coordinates, the plurality of recognition candidates are displayed in order, and the desired recognition candidate is selected, making it extremely easy and efficient to correct the recognition candidates. It becomes possible to do this.

It should be noted that the present invention is not limited to each of the embodiments described above. For example, a character type selection function may also be used to select recognition candidates. Further, as a means for specifying characters to be corrected, it is possible to use a coordinate input device using a transparent touch sensor attached to the display screen of the D2F display. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of drawings]

1 and 2 are diagrams each showing an outline of a conventional device, FIG. 3 is a schematic configuration diagram of an embodiment of the device of the present invention, FIG. 4 is a configuration diagram of a recognition processing section in an actual example device, and FIG. 6 is a diagram showing the display control of recognition candidates, FIG. 7 is a diagram showing the cumulative classification rate with respect to the number of recognition candidates, and FIG. 8 is a diagram showing the configuration of the recognition candidate memory, FIG. A schematic configuration diagram, FIG. 9, is a diagram showing an example of writing of a recognition candidate memory register flag. 11...Tablet device, 12...Writing instrument, 13.
... Recognition processing unit, 13f... Recognition candidate memory, 14 River display control section, 15... Display device, 21... 0CR1
22... Recognition processing unit, 23... Recognition candidate memo 1c2
4...Display device, 25...Coordinate input device. Applicant's Representative Patent Attorney Takehiko Suzue Figure 1 Figure 2 Figure 3 Figure 4 Figure 6 Figure 7 Figure 8

Claims (1)

[Scope of Claims] (1) means for recognizing input character/symbol patterns to generate a plurality of recognition candidates for each input character/symbol pattern; When the recognition output result is insufficient when the first recognition candidate having the highest recognition evaluation value among the plurality of recognition candidates selected for each input character/symbol pattern is insufficient as a recognition output result. , means for storing each of the recognition candidates and information on the input position coordinates of the input character/symbol pattern in correspondence with each of the input character/symbol patterns; means for displaying an image pattern of a first-ranked recognition candidate category; and a means for displaying an image pattern of a first-ranked recognition candidate category;
means for controlling the display of a display image pattern of a symbol pattern recognition candidate category to be different from other image patterns; means for inputting an instruction as position coordinates of an image pattern to be a correction pair of the displayed recognition candidate category;
A character recognition device characterized by comprising means for reading out the stored result at the closest position specified by the means and displaying display image patterns of other recognition candidate categories. (The character recognition device according to claim 1, wherein the two-character/symbol pattern is given as time-series coordinate data indicating a handwritten stroke inputted via a coordinate input device. (3) Character/symbol pattern The character recognition device according to claim 1, wherein the symbol pattern is input as a character pattern image via an optical character reading device. (4) Recognition candidates selected for each input character/symbol pattern. The state in which the recognition candidate in the first rank is insufficient as a recognition output candidate is a state in which the difference in recognition evaluation value between the recognition candidate in the first rank and the recognition candidate in the second rank or below is small. The character recognition device according to item 1. (5) The means for displaying a display image pattern of another recognition candidate category is configured to display a display image pattern of the other recognition candidate category instead of the display image pattern displayed at the display position coordinates inputted. 2. The character recognition device according to claim 1, wherein the character recognition device outputs a display image pattern.