JP3202852B2 - Handwritten character recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten character input device, and more particularly to a handwritten character recognition device having a learning function.

[0002]

2. Description of the Related Art In the conventional handwritten character recognition, when learning a handwritten character, a method of designating a character to be learned and then inputting handwritten character data to the character again is adopted. . In addition, as a method of this character specification, a method of inputting a character code from a keyboard,
There is a method of inputting a correct character by handwriting and designating a character to be registered (Japanese Patent Laid-Open No. 1-103787).

Hereinafter, a learning method and apparatus in the conventional online handwritten character recognition apparatus will be described with reference to the drawings. FIG. 8 is a configuration diagram of an example of a conventional handwritten character recognition device. In this figure, 81 is a character input unit, 82 is a character recognition unit, 83 is a dictionary unit for recognition, 84 is a recognition result output unit, 85 is a learning character designation unit, and 86 is a learning character designation unit. Character data input section,
Reference numeral 87 denotes a character registration unit.

FIG. 9 shows a case where a character is designated during learning.
FIG. 11 is an explanatory diagram showing a state in which a combination with handwriting data corresponding to the character is registered in a dictionary unit. In the figure, reference numeral 91 denotes a character input for designating a learning character, reference numeral 92 denotes a result of converting the input character into a character code, and reference numeral 93 denotes handwriting data of the character to be learned.
First, handwritten character recognition in a conventional online handwritten character recognition device will be described.

[0005] The character input section 81 allows the input person to input characters by handwriting. In general, a configuration of an electronic pen (a stylus pen, hereinafter also referred to as an “input pen” or simply a “pen”) and a tablet (an electronic input handwriting detection board, a digitizer) is used for input. This is because the Z (depth) direction is shifted inside the flat electric insulator and then X
The alternating current flows through fine metal wires densely provided at equal intervals in the direction and the Y direction, while the electronic pen has magnetism, and determines whether or not the electronic pen touches the flat plate, the contact coordinate position, etc. It is such that it is detected as disturbance of alternating current in a thin line. However, since these are well-known techniques described in, for example, "Microcomputer Handbook (Showa 60) II, Chapter 8, Chapter 3, 3 Pen Touch Input" published by Ohmsha, etc. principle,
Detailed description of the configuration and the like is omitted.

The handwritten character data input by the character input unit 81 is converted into a character recognition unit 8 in the form of handwriting data such as a sequence of coordinate points.
Passed to 2. The character recognizing unit 82 obtains candidate characters by referring to the received handwriting data to the dictionary unit 83, and obtains a character code for the candidate character as a recognition result. The obtained character code is passed to the recognition result output unit 84, and finally the character corresponding to the code is presented to the input person. Basically, the technology for obtaining character codes from handwriting data is based on electronic matching of input strokes with registered strokes, and dictionary matching of determined input strokes with characters. Becomes However, for these, see also “Tomimoto et al .: Online Handwritten Kanji / Hiragana Recognition, IEICE Technical Report,
PRL82-37, p. 49 (1982) ", and a detailed description of its principle, configuration, and the like is omitted. Further, the inputting person selects or confirms a character intended for input from the candidate characters, which is basically the same technology as the keyboard type Japanese word processor. For this reason,
Also in this regard, a detailed description of the principle and configuration will be omitted.

Next, learning of handwritten characters in the conventional online handwritten character recognition device will be described. The learning character specification unit 85 allows the input user to specify the character code of the character that the user wants to learn. As a method of designating a character code, there is a method of inputting a character by a code from a keyboard, a method of carefully inputting a character by handwriting in a square style, and obtaining a character code from the recognition result. For example, when the character of “mountain” is designated as a learning character as shown at 91 in FIG. 9, a character code of 3B33 may be directly input from the keyboard, or the character code of the learning character may be input using character recognition. 3B33 shown as 92 as the character code (here, the JI
S code). Next, the learning character data input unit 86 causes the input person to input handwritten character data corresponding to the specified character code using a pen and a tablet, and extracts handwriting data to be learned from the data. And
The resulting handwriting data is passed to the character registration unit 87.

[0008] The character registration unit 87, which has received the input learning character code and handwriting data, generates new data for character recognition based on the data and generates the dictionary unit 8 for recognition.
Register it in 3. For example, a broken character of “mountain” as shown by 93 in FIG. 9 is not recognized as a “mountain” as it is, and is newly generated as handwriting data corresponding to the code 3B33 of “mountain”. Dictionary unit 8
3 is registered for learning.

[0009]

However, in the above-described conventional handwritten character recognition device, character learning registration must be performed in a procedure different from ordinary character input, which is not only complicated but also complicated. In some cases, a difference may occur between data input for learning and data input in a state of handwriting in a hurry, and the learning effect may be impaired. In addition, the method of specifying the character code to be learned has been troublesome.

SUMMARY OF THE INVENTION In view of the above problems, the present invention makes it possible to learn handwriting data input during the input of ordinary handwritten characters as they are, and it is not necessary to input a new character code. It has been made to provide a recognition device.

[0011]

In order to achieve the above object, according to the first aspect of the present invention, a dictionary unit in which characters and handwriting data are registered in association with each other and a handwriting data based on the input handwriting data are used. A character recognizing unit that performs character recognition, a candidate character displaying unit that displays candidate characters obtained by the character recognizing unit in a predetermined procedure, a receiving unit that receives a gesture input for the candidate character from an input user, and a candidate character displaying unit. the character selection unit for selecting a marked character of gesture received by the receiving section, shows trajectories is a learning instruction of the gesture that has been accepted by the accepting unit to Jesuchi from the displayed candidate characters
And determining learning determination unit or catcher or not, in the Rubun shape recognition apparatus wherein a selected character in the character selection section in association with the handwriting data having a learning unit to be registered in the dictionary part,
The learning determination unit indicates that the trajectory of the gesture indicates a learning instruction.
If it is determined that the gesture is not
Select a candidate character by the writing character selection unit and issue a learning instruction.
If it is determined that the gesture is
With the selection of the candidate character by the selection part,
Registering the handwriting data in the dictionary section in association with
I am going to learn .

[0012]

[0013]

[0014]

According to the above construction, in the invention of claim 1,
The receiving unit receives a gesture input from an input person.
Character selection section selects the characters provided to the gesture that has been received from among the candidate characters displayed in the candidate character display unit. Learning determination unit, of the accepted gesture trajectories
Is a gesture indicating a learning instruction . The study
The learning determination unit determines that the trajectory of the gesture indicates a learning instruction.
If it is determined that the character is not a gesture, the character
Selects candidate characters by the selection unit and indicates a learning instruction
If it is determined that the gesture, the character selection unit
A candidate character handwriting data entered with the selected character with selection by correspondence registered in the dictionary part
learn.

[0015]

[0016]

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a handwritten character recognition device according to the present invention will be described based on embodiments. (First Embodiment) FIG. 1 is a configuration diagram of a first embodiment of an online handwritten character recognition device according to the present invention. In this figure, 11 is a character input unit, 12 is a character recognition unit, 13 is a candidate character display unit, 14 is a character selection unit, 15 is a learning determination unit, and 16 is a recognition result output. 17 is a character registration unit, and 18 is a dictionary unit for recognition. In addition to the above, a floppy disk storage unit, various operation units, and the like are provided.

FIG. 2 is a diagram for explaining how a screen is displayed in the present embodiment. In this figure, 21 indicates a group of characters input by handwriting, 22 indicates an input frame for inputting handwritten characters, 23 indicates a group of displayed candidate characters, and 24 indicates whether or not to learn. Button. Because it is directly related to the gist of the present invention,
Here, the principle and configuration of the button will be described. In this embodiment, thin metal wires are buried at coarse intervals near the surface of the button, and an alternating current is passed through the thin metal wires. On the other hand, a magnet is embedded in the tip of the pen. For this reason, when the handwriting input user touches the pen with the button, the magnetic force disrupts the alternating current of the thin metal wire inside the button, and by detecting this, it is possible to input that the touch button has been operated, and thus, it is possible to input a predetermined instruction It has been.

FIG. 3 shows a change in the screen state when the "confirm" button for selecting or designating the character intended by the input person from the characters displayed as candidates from the state of FIG. 2 is pressed. FIG. In the figure, reference numeral 31 denotes a change in the display of the "OK" button, and reference numeral 32 denotes the state of the input text. FIG. 4 is a diagram showing handwriting data and character codes to be learned. In the figure, reference numeral 41 denotes handwriting data, and reference numeral 42 denotes a character code of a character to be learned.

Next, the learning operation of this embodiment will be described with reference to FIGS. First, an input user inputs a character by hand from the character input unit 11. The character input unit 11
Upon receiving the input handwritten character data, information is extracted in the form of a coordinate point sequence, stroke data, segment data, and the like, and the extraction result is passed to the character recognition unit 12 as input handwriting data. The character recognition unit 12 recognizes the input character by comparing the received handwriting data with dictionary data in the recognition dictionary unit 18. Here, character recognition means, for example, obtaining a character code corresponding to the character. The character recognizing unit 12 extracts a plurality of candidates in descending order of the possibility in consideration of the degree of matching and the frequency of use of the characters in the dictionary unit according to the recognition result, and transfers them to the candidate character display unit 13. The candidate character display unit 13
The received candidate character is displayed as shown at 23 in FIG. At the same time, the character selection unit 14 determines that one of the candidate characters is intended, and sets an instruction for learning. The first character that can be designated is the first candidate character with the highest possibility. It should be noted that characters that can be normally designated can be distinguished from other candidate characters by being highlighted. The input person designates one of the displayed candidate characters that the user intends to input an instruction.
The character selection unit 14 changes the candidate character designated by the input person to a state in which the candidate character can be designated. In FIG. 2, the input user specifies the character of “heaven” as the fourth candidate, and the character selection unit 14 is capable of instructing that the character is the character intended for input and instructing learning. This is indicated by reverse display. Next, the input person is instructed whether to learn the candidate character.

In this embodiment, two buttons of "confirm" and "learn" are prepared as instructions by the input person. In the case of learning the “confirm” button, and in the case of learning, each of the “learning” buttons is lightly pressed with a pen to identify the presence or absence of learning. If the input user presses the "OK" button in the state indicated by 31 in FIG. 3, the character "heaven" is simply determined as the character intended for input, and as shown in FIG. Enters the word "heaven". At this time, of course, the learning of the characters is not performed. Similarly, when the input user instructs the “learning” button, the learning determination unit 15
Informs the character registration unit 17 that the input handwriting data should be learned as the character "heaven". As shown in FIG. 4, the character registration unit 17 stores the handwriting data 41 extracted by the character input unit 11, the character “heaven” selected by the character selection unit 14, and its character code 42 ( Here, new data is generated in association with JIS code 4537) and registered in the dictionary unit 18 for recognition. In the subsequent handwritten character input, if a character similar to the handwriting registered here is input, this is registered in the recognition dictionary unit 18 and is immediately recognized as the character "heaven". Will be done. At the same time, characters may be input to the sentence in the same manner as “OK”. (Second Embodiment) Next, another embodiment of the handwritten character recognition apparatus according to the present invention will be described with reference to the drawings. In the present embodiment, a button for instructing whether or not to learn is selected,
This is displayed adjacent to a character for which a learning instruction can be given. Specifically, the only pair of buttons for instructing the input person whether or not to learn is not provided at a dedicated fixed position,
When the candidate character display unit 13 displays a candidate character, the candidate character is displayed adjacent to a character that can be selected and instructed to be learned by the current character selection unit 14. This is shown at 51 and 52 in FIG. (However, in FIG. 5, the two operations and the two types of buttons are widely described so that the following operation can be easily understood.) Hereinafter, the specific operation and operation of this embodiment will be described with reference to FIGS. Will be described. For example, in 51 of FIG. 5, when the input user designates the character “ma” as a character for inputting an instruction, the “confirmation” and “learn” buttons for instructing whether or not to learn are designated by “ Is displayed adjacent to the lower portion of the portion where "ma" is displayed. 52 shows a case where the input person has selected the character "heaven". Like the case of 51, two buttons are displayed adjacent to the lower part of the portion where "heaven" is displayed as a candidate. In this case, "heaven" is selected for the two buttons immediately below the "ma",
Needless to say, it is rejected when a learning instruction is made. (Third Embodiment) In this embodiment, an instruction as to whether or not to learn is made not by pressing a button but by using a gesture, and this is shown in FIG. Here, the gesture means that various commands are determined in advance according to the shape of the trajectory written with the pen, and the device side analyzes the input of the trajectory with the pen for the predetermined command by the handwriting input person. Performs an operation corresponding thereto. Note that the basic principle of the gesture is to recognize the character from the handwriting data and to perform the operation by recognizing an operation command instead of displaying the character on the display unit, and the configuration is not different from the character recognition. Also, this is "handwriting input O
S. Creates New Market for Personal Computers: Nikkei Electronics No. 525, pp. 115-133 (199
1.4.15) ”and the like, and the recognition of various trajectories and the correspondence between the trajectories and the commands are electromagnetic signal collation, which is not different from character recognition. . Therefore, a detailed description thereof will be omitted. Now,
In the present embodiment, a downward convex gesture as shown at 61 in FIG. 6 is not learned for a character indicated by a convex portion of the gesture (bending point portion), and the input person inputs the character. Is intended to be determined, and an upwardly convex gesture as indicated by 62 indicates that learning is to be performed. For example, in 61 of FIG. 6, since the character is convex downward at the position of the character “ma”, the input person selects the character “ma” from the characters displayed as candidates as the character intended to be input, It also indicates that the user has instructed not to learn. When the input user inputs this gesture, the character selection unit 14 determines that the character “ma” is a character intended for input by the input user, and the learning determination unit 15 does not learn this character. judge. Also, 62 in FIG.
In this example, the input person uses the convex gesture
Is designated as a character intended to be input, and indicates that an instruction to learn that the input handwriting data is the character "heaven" is given. Thereby, the character selection unit 14 determines that the character “heaven” is a character intended to be input and displays the character at a predetermined position in the text being input, and the learning determination unit 15 determines the input handwriting data as a character. It is determined to be learned as “heaven”. (Fourth Embodiment) In the present embodiment, the candidate character display unit 13 has redundancy, that is, it can be not only a candidate character obtained as a recognition result by the character recognition unit 12 in a normal procedure but also a character to be recognized to some extent. If it is determined that it is replaced to be displayed in addition to the candidate characters,
This is shown in FIG. This is done by outputting all characters having some stroke characteristics and the like included in the handwriting data, even if there are some differences.

The operation of this embodiment will be described below with reference to FIG. The character recognition unit 12 compares the input handwritten data with data corresponding to characters in the recognition dictionary unit 18, and passes a recognition candidate character group to the candidate character display unit 13. At this time, the recognition candidate character group received from the character recognition unit 12 includes characters that are considered to be somewhat similar to the input characters. Then, the candidate character display unit 1
3 displays a list of all these characters received from the character recognition unit 12. It should be noted that the display device of this embodiment is a portable device, and therefore has a small display portion, so that only 27 display devices can be displayed at the same time. For this reason, a display movement button is provided so as to be able to display the character group of the front candidate and the rear candidate as shown at 71.

In the present embodiment, since the input user can select an intended character from characters having a certain possibility, the character recognizing unit 12 inputs an extreme broken character or an abbreviated character, so that the character recognizing unit 12 recognizes a candidate character in a normal procedure. Characters that are not output as characters will also be presented. For this reason, it is possible to select and learn a character intended for input from these wide recognition candidate characters. As a result, even books and curricula can be learned. Further, it can be used for use of a person who has a disability in the arm or the like, or for studying a character that also serves as a play for a younger elementary school child or an infant. Characters other than the characters that are considered as candidate characters in the normal procedure may be displayed in various orders such as the order of the number of strokes, the order of the Japanese syllabary, the order of the character codes, and the like. May be set so that a character to be selected can be easily found.

As described above, the present invention has been described based on the embodiments. However, the present invention is not limited to the above-described embodiments, but includes the following. (1) In the first embodiment and the like, the button operation by the pen is detected not by disturbance of alternating current by magnetism but by distortion or change in capacitance. (2) In the second embodiment, the buttons of “confirm” and “learn” are one set, and the candidate characters targeted are moved and displayed above these buttons. (3) In the first and second embodiments, the handwriting input of the next character is also used as the "confirmation" instruction of the candidate character displayed as being capable of inputting the instruction at present. (4) In the first and second embodiments, “confirmed”
The “learn” button is provided in front of the displayed candidate character. This is a CRT that displays candidate characters as used in automatic cash dispensers at banks.
It can be easily realized by providing a transparent input panel on the front surface of the device. (5) In the first and second embodiments, the positions of the buttons for instructing "confirmation" and "learning" are located above the candidate characters due to manufacturing and arrangement with other buttons and the like. Or, even if it is below, it is a position separated by 1 or 2 cm. (6) In the third embodiment, a gesture is a concept including not only handwriting data but also an operation called tapping of tapping a character lightly with a pen tip. Thus, for example,
If tapping is once, select for inputting the instruction of the character displayed in the next order. If it is twice within one second, confirm or select that the character was intended to be input.
If it is a round, it is called learning. (7) In the fourth embodiment, the presence or absence of the function of this embodiment is determined by a separate instruction from the handwriting input user. Alternatively, furthermore, the recognition range is a step type (multi-stage type). (8) In each embodiment, the convenience of operations such as the cancellation of an erroneously given instruction can be input by a pen or by pressing a button on the top on the side opposite to the input side of the pen is provided. (9) When learning handwriting data for a character that is not registered in the dictionary in advance or a character that is not recognized in combination with the character, the input user inputs the character intended to be input using a character code. Rather than having the redundancy as in the fourth embodiment, a function is provided by writing characters intended for input correctly and in large size separately. In addition, a function has been added to make the entry frame for one character large accordingly. This allows
When no candidate character intended for learning is displayed, the handwritten input user does not need to refer to the character code each time. Further, it is possible to learn characters and the like that are not registered in JIS or the like. In this case, it is more convenient to display the characters handwritten in the standard style even in the text currently being created. (10) The characters targeted by the present invention are not limited to kanji and kana, but may be alphabets, Arabic numerals, Korean characters, Arabic characters, Russian characters, Greek characters, German characters, and the like. It also includes gestures, of course. The character is one (singular)
Of course, includes a plurality of characters (Japanese is generally described as a "character" in the present specification, since the distinction between two or more is not strict). For this reason,
For example, a word input in a cursive alphabet or a idiom composed of a plurality of kanji may be made learnable. In these cases, the input of the correct kanji and the like intended by the input person can be performed by using a phrase such as “patented invention” if the function performed in square writing or printed form is added as described above. Needless to say, learning can be performed with handwriting “/” or the like, which is convenient for preparing various documents such as a specification of a patent application.

In addition, when many types of characters can be input by handwriting, a function is added to enable to specify in advance which type of character is to be input in order to speed up the extraction of recognition candidate characters. This will be more effective when kanji is excluded from the input. (11) For convenience of manufacture or the like, one component (requirement) of the present invention may be a plurality of components, conversely, a plurality of components may be physically integrated, or an appropriate combination thereof. Further, a part is configured as software.

[0026]

As described above, according to the first aspect of the present invention, the selection of a character displayed on the candidate character display section and the registration in the dictionary section are performed in a single operation by using a gesture. This makes it possible to learn handwritten characters without performing cumbersome character registration operations.

[0027]

[0028]

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a first embodiment of a handwritten character recognition device of the present invention.

FIG. 2 is a diagram showing a display state of a screen in the embodiment.

FIG. 3 is an explanatory diagram showing a change in a screen state when an enter button is pressed from the state shown in FIG. 2 in the embodiment.

FIG. 4 is a diagram showing a state of handwritten data and character codes to be learned in the embodiment.

FIG. 5 is an explanatory diagram of a second embodiment of the handwritten character recognition device of the present invention.

FIG. 6 is an explanatory diagram of a third embodiment of the handwritten character recognition device of the present invention.

FIG. 7 is an explanatory diagram of a fourth embodiment of the handwritten character recognition device of the present invention.

FIG. 8 is a configuration diagram of an example of a conventional handwritten character recognition device.

FIG. 9 is a conceptual diagram showing a state in which a learning character is designated and registered in a recognition dictionary in combination with handwriting data in the conventional example.

[Explanation of symbols]

11 Character Input Unit 12 Character Recognition Unit 13 Candidate Character Display Unit 14 Character Selection Unit 15 Learning Judgment Unit 16 Recognition Result Output Unit 17 Character Registration Unit 18 Recognition Dictionary Unit 21 Sentence Displayed on Input 22) Input frame for inputting handwritten characters 23 Displayed candidate character group 24 Button for instructing input person whether or not to learn etc.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06K 9/03-9/68 G06F 3/03 G06F 17/22

Claims (1)

(57) [Claims] 1. A dictionary section in which characters and handwriting data are registered in association with each other, a character recognition section for performing character recognition based on input handwriting data, and a candidate character obtained by the character recognition section being specified. A candidate character display unit to be displayed in the procedure described in the above, a reception unit that receives an input of a gesture for the candidate character from the input user, and a gesture received by the reception unit from the candidate characters displayed on the candidate character display unit. said character selection unit for selecting a character, the a trajectories of the gesture that has been received by the receiving unit determines a learning determination unit whether indicate to the gesture learning instruction, a character selected by the character selector in Rubun character recognition apparatus having a learning unit in association with the handwriting data registered in the dictionary part, the learning determination unit trajectory of the gesture indicates learning instruction
If it is determined that the gesture is not
Select a candidate character by the writing character selection unit and issue a learning instruction.
If it is determined that the gesture is
With the selection of the candidate character by the selection part,
Registering the handwriting data in the dictionary section in association with
A character recognition device characterized by learning.