JPH0564398B2 - - Google Patents

Description

[Detailed Description of the Invention] [Summary] An online handwritten character recognition device that recognizes characters input by handwriting includes a quantization processing means for calculating handwritten strokes of characters input by handwriting, and similarity information in association with basic strokes. The present invention includes at least a basic stroke similarity dictionary in which characters are registered, and the characters input by hand are recognized using the similarity.

[Industrial application field]

The present invention relates to an online handwritten character recognition device configured to recognize characters input by hand by searching for similarity information registered in advance in association with basic strokes. .

[Problems to be solved by conventional technology and invention]

Conventionally, devices that recognize handwritten characters online compare the handwritten strokes of the handwritten characters with the basic strokes to be compared and calculate the degree of similarity to recognize handwritten characters. . For this reason, if the handwritten strokes input by handwriting are slightly different from the handwriting strokes when the correct characters are handwritten, some of them may be slightly different due to the habits of the person who inputted them by handwriting. There was a problem that it could not be recognized correctly.

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention registers similarity degree information in advance in association with basic strokes, and when recognizing handwritten characters, uses the registered similarity information. By adopting a configuration that recognizes handwritten characters using similarity information, it is possible to recognize even quirky handwritten characters.

FIG. 1 shows the basic configuration of the present invention. In the figure, 1
2 is a handwritten character input means, 2 is a quantization processing means, and 3 is a handwritten character input means.
4 represents a character similarity calculation means, 4 represents a basic stroke similarity dictionary, and 5 represents an output character candidate selection means.

In FIG. 1, handwritten character input means 1 is for inputting handwritten characters, and is, for example, a tablet.

The quantization processing means 2 calculates handwritten strokes of input handwritten characters.

The character similarity calculation means 3 searches the basic stroke similarity dictionary 4, reads basic strokes with a predetermined degree of approximation, and calculates the degree of similarity.

The basic stroke approximation dictionary 4 is designed to register basic strokes that approximate a basic stroke, such as those with an approximation level of 0, those with an approximation level of 1, and those with an approximation level of 2, etc., as described later. The information is registered in this manner.

The output character candidate selection means 5 selects characters to be output based on the calculated degree of similarity.

[Effect]

Using the configuration shown in FIG. 1, handwritten character input means 1
When handwritten characters are input from , the quantization processing means 2
calculates the handwritten stroke (expressed by direction and length) of the handwritten character, and notifies it to the character similarity calculation means 3. The character similarity calculation means 3 that received this notification uses the basic stroke similarity dictionary 4.
is searched and the degree of approximation is calculated, and the character corresponding to the calculated result is searched. The output character candidate selection means 5, which has been notified of the searched characters, recognizes, for example, a character with a small similarity value as a handwritten character and outputs it.

As explained above, even if the handwritten stroke of a part of a handwritten character does not match the correct basic stroke, the user's habits are investigated in advance and the basic stroke corresponding to this habit is registered in advance as similarity information. By doing so, it becomes possible to accurately recognize handwritten characters.

〔Example〕

FIG. 2 shows the configuration of one embodiment of the present invention. In the figure,
1-1 is a tablet, 3-1 is a character similarity calculation means, 6 is a character coordinate reading means, 7 is a normalization processing means, 8 is a basic stroke extraction means, 9 is a basic stroke dictionary, and 10 is a character dictionary. . Note that numerals 2, 4, and 5 in the figure are equivalent to those shown in FIG.

In FIG. 2, a tablet 1-1 is used to input handwritten characters using a pen.

The character coordinate reading means 6 reads coordinate values (x, y,
z).

The normalization processing means 7 normalizes the input handwritten characters to a predetermined size by reducing or enlarging them.

The basic stroke extraction means 8 searches the basic stroke dictionary 9 for handwritten strokes of characters input by hand, and calculates the degree of similarity.

The basic stroke dictionary 9 is a dictionary in which correct basic strokes of characters are registered.

As described above, the character similarity calculation means 3-1 reads the basic stroke corresponding to the handwritten stroke from the basic stroke similarity dictionary 4,
In addition to calculating the degree of approximation, the character dictionary 10 is searched for and read out characters corresponding to the combination of basic strokes adopted for the calculated degree of approximation.

The character dictionary 10 is a register of characters corresponding to combinations of basic strokes.

The output character candidate selection means 5 selects a predetermined character as a handwritten character from the recognized character candidates notified and outputs the selected character.

As explained above, according to this embodiment, handwritten characters are recognized by comparing the basic strokes that constitute a correct character using the basic stroke extraction means 8, and the recognition is performed using the basic strokes. If the handwritten characters are not recognized, the recognition rate of handwritten characters is improved by adopting a configuration in which the handwritten characters are recognized by comparing basic strokes corresponding to human habits and the like.

Next, with reference to FIGS. 3 and 4, the operation when recognizing characters using similarity information will be described in detail. FIG. 3 shows an example of a dictionary of basic stroke approximations, and FIG. 4 shows an example of identification of basic stroke approximations.

The stroke numbers shown in the upper part of FIG. 3 represent the addresses of basic strokes.

The pattern shown at the top of the figure represents a basic stroke pattern.

The approximate stroke numbers shown in the upper part of the figure are composed of approximation degrees 0, approximation degrees 1, and approximation degrees 2, and each represent the order of the approximation degrees with respect to the pattern indicated using the stroke numbers. The degrees of approximation 0, 1, and 2 were determined by researching the habits of people when writing by hand, and the degree of approximation 0 is the most frequent, and the degree of approximation 1 is the next most frequent. Degree 2 is the next most common.

Figure 4 shows an example of identifying the degree of similarity of basic strokes.
and their basic strokes (stroke numbers 1, 10, 1, and 6 in Figure 3), and Figure 4 C, D, Ho, and F show handwritten characters with peculiarities and their approximate basic strokes (stroke numbers 1, 1, and 6). 1,
9, 1, 6, and similarly quirky handwritten characters and stroke numbers 1, 1, 1, 6), respectively.

FIG. 4A shows the correct character ``TO'', and when this correct character ``TO'' is input from the tablet 1-1, the quantization processing means 2 calculates a handwritten stroke as shown in FIG. 4B. This handwritten stroke (stroke numbers 1, 10, 1,
6) is exactly the same as the basic stroke combination (stroke numbers 1, 10, 1, 6), so the character "door" can be easily recognized.

4C is a character with a peculiar character, and the part indicated by a in the figure is diagonal. When this character with a peculiar character is inputted from the tablet 1-1, the quantization processing means 2 A handwritten stroke as shown in FIG. 4D is calculated. This handwritten stroke (stroke numbers 1, 9, 1, 6 in Figure 3) is different from the correct combination of basic strokes (stroke numbers 1, 10, 1, 6) for the character "door".
It will not be recognized. Therefore, the character similarity calculation means 3 searches the basic stroke similarity dictionary 4 shown in FIG. 3, searches the basic stroke similarity dictionary 4 based on the basic strokes shown in FIG. Since the stroke number 9 is similar to the stroke number 10, the basic stroke combinations (stroke numbers 1, 10, 1, 6) are also extracted as matching targets and recognized as the character "door". be able to do so.

Similarly, the character ``e'' in Figure 4 has a peculiar character, and a part (the vertical part) of the character indicated by ``b'' in the figure is missing. For this character, as shown in Figure 4, the combination of basic strokes (stroke number 1 has a similarity of 2 to stroke number 10) , 6) are also extracted as matching targets, and the character ``door'' can be recognized.

As explained above, it is possible to improve the recognition rate of handwritten characters by investigating quirks in parts of characters as similarity information and registering them in advance with respect to basic strokes. It becomes possible.

〔Effect of the invention〕

As explained above, according to the present invention, similarity information is registered in advance in association with basic strokes, and when recognizing handwritten characters, the similarity is calculated using the registered similarity information. Since the system adopts a configuration that recognizes handwritten characters, even handwritten characters with peculiarities can be recognized, and the recognition rate can be improved.

[Brief explanation of the drawing]

FIG. 1 shows the basic configuration of the present invention, FIG. 2 shows the configuration of one embodiment of the invention, FIG. 3 shows an example of a basic stroke approximation dictionary, and FIG. 4 shows an example of basic stroke approximation identification. In the figure, 1 is a handwritten character input means, 2 is a quantization processing means, 3 is a character similarity calculation means, 4 is a basic stroke similarity dictionary, and 5 is an output character candidate selection means.

Claims (1)

[Claims] 1. An online handwritten character recognition device that recognizes characters input by hand, including a character dictionary that registers characters corresponding to combinations of basic strokes, and quantization for calculating handwritten strokes of characters input by hand. a processing means; a basic stroke similarity dictionary that registers similarity information consisting of a plurality of approximate strokes ranked according to the order of similarity, which indicates a stroke that approximates the basic stroke in association with the basic stroke; Character similarity calculation in which a stroke similarity dictionary is searched to extract a stroke that approximates a handwritten stroke, the degree of similarity is calculated, and the character dictionary is searched for a character that corresponds to the combination of the calculated basic strokes. and an output character candidate selection means for selecting a predetermined character from recognized character candidates notified based on the result obtained by the character similarity calculation means and outputting the selected character, the character being input by hand. An online handwritten character recognition device characterized in that it is configured to recognize.