JPS608991A - Registering system of on-line handwriting character recognizing dictionary - Google Patents

Abstract

PURPOSE:To generate a plural kinds of picture dictionaries against the same character category, and to raise the recognition rate of a handwriting character by making a user to designate a part where it seems that a picture variation occurs once at the time of a registration. CONSTITUTION:A handwriting character input part 11, a keyboard 12 for inputting a character code and a stroke connecting part, a display part 13, a character recognizing part 15, a character variation generating part 18, a dictionary registering part 19 and a dictionary 10 are connected to the control part 14 of an on-line handwriting character recognizing device. Also, the generating part 18 is constituted of a memory 20, a control part 21, a stroke detecting part 22, a stroke connecting part 23, a picture number calculating part 24, a writing order calculating part 25 and a synthesizing part 26. In this state, a part where it seems that a picture variation occurs once at the time of a registration is designated by the user from a keyboard 2, several kinds of picture dictionaries against the same character category are generated by the generating part 18 under the control of the control part 14, registered in the registering part 19, and a recognition rate of a handwriting character is raised.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention relates to personal recognition dictionary registration in an online handwritten character recognition device. The present invention relates to a Ma& dictionary registration method that allows features for recognizing a character to be registered in a dictionary.

[Prior art and problems]

For example, when a handwritten character is written on an input section such as a digitizer, the on-twin character recognition device immediately reads the handwritten character.

Recognition is based on the number of strokes and line order information of a character written with a stylus pen or the like by detecting the X-Y coordinates and the top and bottom of the pen (whether it touches or leaves the paper surface). For example, when the input character is 10 strokes.

Character recognition is performed by checking a 10-stroke dictionary in order of stroke order.

Therefore, if you write horizontally with the correct number of strokes and stroke order, you can get a high recognition rate, but if you write with 9 strokes or the order of strokes.

It cannot be recognized if the number of strokes or stroke order differs from the recognition dictionary prepared in advance. Therefore, although nine different improvement methods have been considered, the current situation is that no definitive solution has yet been found. However, it is known that stroke order is relatively constant within a particular individual. Therefore, it is being considered that a personal handwriting recognition device would register only the person's writing style in a dictionary and recognize it.

That is, as shown in FIG. 1, if a standard writing style (standard number of strokes, stroke order, etc.) is registered in advance in the dictionary V1, and it is not recognized.

The character is input from the handwritten character input section 2, and the character code is inputted using a character input section 3 similar to a keyboard. Then, via the control section 5, the dictionary registration section 6 additionally registers the characters of the individual that could not be recognized in the dictionary 1. Upon this registration.

For example, when the handwritten character ``Saku'' cannot be recognized, the operator displays ``Saku'' written in a more standard way than in the dictionary 1 on the display unit 4, displays the number of strokes and stroke order, and refers to this. input. For example, as shown in Figure 3 (a), input the character written as one stroke in succession from the handwritten character input section 2, and then input the character code for "Saku" from the character code input section 3. input. As a result, this new input character is registered in the dictionary portion where the number of strokes is one less than in the dictionary-like 1-letter case. As a result, when "saku" is inputted as an unknown input in the state shown in FIG. Since it can be displayed, this unknown input is recognized as a "work".

In this way, each time a handwritten character cannot be recognized, the handwriting style of that person is registered.

As the user uses it more and more, the dictionary for each person's unique writing style increases, so the recognition rate can be improved.

By the way, handwritten characters can be written in various ways even by the same person.1 For example, there are multiple ways to write the character "saku" as shown in FIGS. 6(a) to 6(c).

However, with the above method, when registering - times, only -m number of strokes (one letter combining the number of strokes and the order of strokes)) can be registered, so as shown in Figure 3 (a) to (c), if the person is the same If you want to write the character in several ways, you must register it each time.

Therefore, if it takes a long time to converge until all of your writing styles are registered and the recognition rate becomes high, you will have to perform the registration process many times.

[Purpose of the invention]

The purpose of the present invention is to solve such problems by -
To provide an online handwritten character recognition dictionary registration system that enables the creation of several types of stroke number dictionaries for the same character category by having the user specify a place where the number of strokes is likely to change during registration.

[Structure of the invention]

In order to achieve this purpose, the online handwritten character recognition dictionary registration method of the present invention includes a dictionary in which nine character patterns are registered together with the number of strokes and stroke order, and inputs unknown input by comparing it with the patterns registered in this dictionary. Stroke selection means to select certain strokes in online handwriting to recognize characters with force.

The character pattern is provided with a stroke composition means for connecting the strokes selected by the stroke selection means, and a stroke number calculation means for calculating a new number of strokes changed by the connection of these strokes, and connecting strokes that have been indicated as possible to connect. [Embodiment of the Invention] Before describing an embodiment of the present invention in detail, its outline will be explained with reference to FIG. 4.

In the present invention, as shown in Fig. 4 (a), the strokes of characters written in standard writing style are added with numbers ■, ■, etc. according to the order of strokes, and stored in the registered dictionary together with the number of strokes. I'll keep it. This number is.

For example, when registering individual strokes input from a tablet, this is displayed on the display and the stroke order is numbered using the keyboard in places where the number of strokes may fluctuate due to continuous strokes. Enter using . For example, when there is a possibility that strokes ■-■ and ■-■ are consecutive strokes,
Input each stroke as shown in Figure 4 (b). As a result, the stroke ■-■ shown in Figure 3 (a) is connected to the character pattern, and the stroke ■-■ shown in Figure 3 (b) is connected.
The character pattern in which ■ is connected, s shown in the same (c)) 1=
A character pattern in which -ri ■-■ and ■-■ is created, the number of strokes becomes the new number of strokes, and the order of strokes is also corrected is obtained and registered in the dictionary. This allows multiple character patterns to be registered in the dictionary in a single operation.

An embodiment of the present invention will be described with reference to the narrow opening in FIGS. 5 and 6.

12 is a keyboard for inputting character codes and stroke connection parts; 13 is a display section; and 14 is a control section.

15 is a character recognition unit, 16 is an identification unit, 17 is a feature extraction unit,
18 is a stroke number variation creation section, and 19 is a dictionary registration section.

20 is a memory, 21 is a control section, 22 is a stroke detection section, 23 is a stroke connection section, 24 is a stroke count calculation section, 25 is an end order calculation section, and 26 is a composition section.

The stroke number variation creation unit 18, which is the most characteristic feature of the present invention, will be explained.

The stroke number variation creation unit 18 creates a continuous pattern of instructed strokes and also creates a new number of strokes.

This is to calculate the order of new strokes, as shown in FIG.

Memory 20. Control unit 21. Stroke detection part 22゜)
+=−ri connection section 232 stroke number calculation section 24. Stroke order calculation unit 25
9 composing unit 26 and the like.

For example, from the characters in Figure 4 (A) to the characters in Figure 4 (B).

It is also set in this memory 20 that the strokes ■-■ and ■-■ are connected, and that / is ■-■ and ■-■. The control unit 21 then learns that there are two pairs of strike pairs to be connected.

When these data are input into memory 20.

First, the control section 21 outputs the stroke ■-■ from the memory 20 to the stroke detection section 22 in order to perform the connection process of the stroke ■-■. This stroke detection section 22 detects that the stroke connection section 22 is connected to the stroke ■-■.
31 stroke count calculation unit 24 and transmission to line order calculation unit 25.

As a result, the stroke four-way connecting section 23 reads the strokes ■ and ■ from the memory 20 and connects them so that the shortest parts of the strokes ■ and ■ are integrated.

As a result, a stroke similar to the stroke I in FIG. 3(a) is synthesized. This connected stroke is then sent to the memory 20. At this time, the stroke number calculation unit 24 calculates -1 from the stroke number 7 set in the memory 20 by connecting the strokes -■. When the connected new strokes are set in the memory 20, the control unit 21 transfers the strokes ■ to ■ of the character pattern set to the compositing unit 26 to the memory 20.
Then, a new stroke in which the above ■-■ is connected is read out and combined to create a character similar to that shown in FIG. 3(A). Then, the character ``■'' outputted from the stroke number calculation unit 24 is added to ``--■'', and ``■~■'' is added to ``--■'', and these are set in the memory 20 as the first stroke connection character.

Next, the control unit 21 executes the stroke ■-■ connection process in the same manner as the stroke ■-■ connection process, and sets it in the memory 20 as a second stroke connection character.

Then, the control section 21 synthesizes the characters connecting the strokes ■-■ and ■-■ in the same manner.

This is set in the memory 20 as the sixth stroke connection character.

When three types of stroke connecting characters are obtained in this way, this is reported to the dictionary registration section 19.

These three types of characters are thus registered in the dictionary 10.

The operation of the present invention will be explained with reference to the narrow opening in FIG.

(1) When a handwritten character is input from the handwritten character input section 11, the feature extraction section 17 of the single character recognition section 1.5 searches the dictionary 10 using the number of strokes as a key.

Characteristics are sequentially extracted based on the stroke order, and the identification unit 16 compares the characters, and outputs those with the highest degree of similarity to the display unit 13 to recognize nine handwritten characters.

(2) If there are characters that could not be recognized at this time,
The user inputs the standard Kanji code of the input character from the keyboard 12, which is a character code input section.

This code may be a JIS code, or the reading may be input in kana and converted into kanji. In this way dictionary 10
Instruct the display unit 13 to display the input character pattern that could not be recognized along with its stroke number.2 For example, in the case of "Saku" in Figure 4 (A), enter these as shown in (B) and register. Enter the key. As a result, the stroke connection process as described above is performed in the stroke number variation creation unit 18,
The dictionary registration unit 19 registers these stroke-connected character patterns in the dictionary 10. At this time, if multiple character patterns are already registered in the dictionary 10 with the same standard kanji code, all of them are displayed on the display section 16, so each one is displayed.

Similar stroke connection processing can be performed.

Of course, these newly registered character patterns are registered in the dictionary 10 after their features are created by the feature extractor 17.

The present invention can also be used in the following nine ways.

In other words, the stroke order and number of strokes are displayed not only for those that have already been registered, but also for the character pattern that has just been input for reference. Since the user can see the differences between the data and the original, it can be used as a training tool for writing, and can be used as a check when data is not input correctly due to instability such as the data sizer.

Of course, if there is nothing already registered for that category, or even if there is, the writing style (especially stroke order) is different, then register it using the method of the present invention. Instead, you can simply register what you wrote at that time as before.

〔Effect of the invention〕

According to the invention, characters with connected strokes.

At the same time, it can be easily registered in a dictionary, so the recognition rate of 9 handwritten characters can be greatly improved. In addition, by displaying input cover patterns and dictionary registered patterns in a way that shows the stroke order and number of strokes, the user can check whether the digitizer did not input the data correctly and could not be recognized, or whether the person's writing style was not in the dictionary and could not be read. Figure 1 shows a conventional online handwritten character recognition device.

FIGS. 2 and 3 are explanatory views of conventional problems.

FIG. 4 is a schematic explanatory diagram of the present invention, FIG. 5 is a configuration diagram of an embodiment of the present invention, and FIG. 6 is a detailed diagram of the stroke number variation creation section.

In the figure, 10 is a dictionary, and 11 is a handwritten character input section.

12 is a keyboard for inputting character codes and stroke connection parts; 16 is a display section; and 14 is a control section.

15 is a character recognition section, 16 is an identification section, 17 is a feature extraction section,
18 is a stroke number variation creation section, and 19 is a dictionary registration section.

20 is a memory, 21 is a control section, 22 is a stroke detection section, 23 is a stroke connection section, 24 is a stroke number calculation section, 25 is a flower order calculation section, and 26 is a composition section.

Patent applicant: Fujitsu Limited Representative Patent Attorney Akira Yamatani 533-

Claims (2)

[Claims] (1) An online handwritten character recognition device that is equipped with a dictionary in which character patterns are registered along with their stroke counts and stroke orders and that recognizes input characters by comparing unknown input with patterns registered in the dictionary. a display means for displaying an instruction number for indicating the stroke of a character along with a character pattern; A stroke p- that selects strokes that may be connected
The stroke selection means includes a stroke selection means, a stroke synthesis means for connecting strokes selected by the stroke selection means, and a stroke number calculation means for calculating a new number of strokes changed by connecting these strokes. An online handwritten character recognition dictionary registration method characterized by registering in a dictionary a character pattern in which strokes whose connectability is indicated are connected together with a new number of strokes. (2) The online handwritten character recognition dictionary registration system according to claim 1, wherein a keyboard is used as the stroke selection means, and a stroke order is used as the instruction number for instructing the stroke.