JPH0690741B2 - Handwritten character recognition method - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to a handwritten character recognition method for performing character recognition based on data relating to handwritten characters.

[Technical background of the invention and its problems]

With the recent development of information processing technology by computers, the amount of data handled there is becoming enormous, and therefore, how to handle these data inputs is becoming an important issue. As a solution to this, an online handwritten character recognition device using a tablet or the like for reading the X and Y coordinate values of the position pressed by the pen has come to be used. In these handwritten character recognition methods, handwritten data is cut out for each character, the characteristics of each stroke that constitutes this character are extracted, and the character recognition is performed by matching this with the contents of the recognition dictionary. There is. As a character cutout method, a writer pushes a specific position of a tablet with a pen to indicate a character break, and a method of cutting out a character or as shown in Japanese Patent Publication No. 52-15187.
There is a method of setting a character frame on a tablet and cutting out a character. However, these methods impose a burden on the writer to show the division of characters in the text, and thus have serious problems in practical use such as a delay in the writing time and an increase in the fatigue of the writer. Further, in the method of fixing the entry frame, there are various drawbacks such as the fact that a form for freely setting the entry position cannot be used, and the entry position cannot be known when the sheet is placed on the tablet for entry.

[Object of the Invention]

The purpose of the present invention is to essentially provide only the character pitch to the writer in order to reduce the burden on the writer, reduce the fatigue of the writer, shorten the writing time, and free the writing position. The purpose is to provide a handwritten character recognition method characterized by limiting the above and removing other restrictions.

[Outline of Invention]

The present invention is a handwritten character recognition method for performing character recognition based on time-series input of coordinates representing a handwriting of data regarding a character input by handwriting, in which first data regarding the character (for example, “Hen”) is input. After a predetermined time has passed,
Character recognition is started with the first data relating to the input character as one character, and the second data relating to the character relating to the input first data (for example, "creation") is set after the elapse of the predetermined time. When additional input is made, the first data and the second input additionally made
If the distance from the data is within a predetermined value, character recognition is performed again with the first data and the second data as one character.

〔The invention's effect〕

Thus, according to the present invention, the writer can write characters without being aware of the position of the character frame set on the tablet and the operation indicating the division of characters, which reduces the burden on the writer and the writing time. The present invention has a great effect in practical use because the shortening is realized and the position of the paper can be freely set even when the paper is placed on the tablet and characters are written on the paper.

Example of Invention

Embodiments of the invention will be described below with reference to the drawings.
FIG. 1 shows the configuration of the online recognition device. The writer writes characters on the tablet 2 with the pen shown in FIG. The X and Y coordinate values obtained from this are sent to the detection / cutout unit 3, where a stroke group for each character is extracted and transferred to the character recognition unit 4. Based on this, character recognition is performed at 4 and the result is output at 5.

The detection / cutout unit 3 in FIG. 1 cuts out as follows. It is assumed that there is no character separation from the first stroke to the Nth stroke and that the N + 1th stroke is input. First, the character size and position are calculated from the first to Nth strokes. Reference numeral 6 in FIG. 2 is a virtual character frame showing the size of the character when the character "O" is input. On the other hand, the size of a character to be input and the character pitch are determined in advance by the writer's designation, the hardware switch designation, etc., and the inter-character distance is calculated from these values. FIG. 3 shows an example of this calculation method. Let a be the size of the character, b be the pitch of the character string, and the inter-character distance c is given by c = b−a. Based on this c, the distance between the stroke 8 to be input next and "o" is calculated. In the example of FIG. 2, when the center of the virtual character frame and the center of the character frame 7 determined by the size of the designated character are overlapped, the distance d from the character frame 7 to the stroke 8 is set to "o". And stroke 8 distance. When this value is outside the range of the fixed threshold value, it is determined that the stroke 8 belongs to the next character. Further, even when the time from the final stroke 9 of "O" to the stroke 8 is over a certain value, it is judged that the stroke 8 belongs to the next character and the recognition of "O" is started. However, in this case, since the stroke 8 is input after the end of the recognition, the character separation is determined based on the above-described stroke distance, and when it is determined that the character separation is not recognized, the stroke to be continued is also included. Perform character recognition again.

FIG. 4 shows this example, and 10 is the stroke sequence written first, and it is assumed that the time over the threshold has elapsed until the next stroke is entered. At this time, 10 is regarded as one character and character recognition is performed, and a recognition result of "tree" is obtained.
However, after that, if it is determined that a further 11 stroke sequences are input, and it is determined that 10 and 11 are within one character frame, 10 and 11 are collectively considered as one character and character recognition is performed. , This time, the recognition result of "bayashi" is obtained. In this case, two types of recognition results are obtained, but the correct one is output as the recognition result based on the recognition result, the similarity obtained at the time of recognition, the positional relationship of the stroke group, and the like. In this example, “wood” is output as the result instead of “wood” and “wood”.

The following method is known as an example of the character recognition method performed by the character recognition unit 4. Character stroke of K point
Expressed by X and Y coordinate values, when one character is composed of N strokes, a 2NK-dimensional vector is created from these coordinate values, and the correlation calculation between this vector and the vector registered in the dictionary as a standard pattern is performed. This is a method for recognizing characters based on the similarity between the two.

The present invention is not limited to the above embodiment. For example, any character recognition method may be used, and the character spacing, character size, character pitch,
Alternatively, the method of presetting data corresponding to these is not particularly limited. Further, it is also possible to adopt a system that allows a certain amount of variation without strictly controlling the size and pitch of characters.
The calculation method of the distance between the virtual character frame and the stroke to be entered next is not particularly limited. In short, the present invention can be variously modified and implemented without departing from the scope of the invention.

[Brief description of drawings]

FIG. 1 is a diagram showing an on-line handwritten character recognition device, and FIGS. 2 to 4 are diagrams for explaining the recognition method of the present invention. 1 ... pen, 2 ... tablet 3 ... detection cutout part, 4 ... character recognition part 5 ... output line, 6 ... virtual character frame 7 ... definition character frame, 8,9 ... stroke 10,11 ... stroke group

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Katsunori Oi 1 Komukai Toshiba-cho, Sachi-ku, Kawasaki-shi, Kanagawa Toshiba Research Institute, Inc. (56) References JP-A-57-106986 (JP, A) JP-A-54-91148 (JP, A)

Claims (1)

[Claims] 1. A handwritten character recognition method for performing character recognition based on time series input of coordinates representing handwriting of data relating to characters input by handwriting.
If a time equal to or more than the threshold value elapses before the second data related to the character is input, the first data is regarded as one character, the character recognition is started, and the first data regarded as the one character is continued. , When the second data regarding the character is input and the distance between the first data and the subsequently input second data is within a predetermined value, the second data continuously input to the first data A handwritten character recognition method characterized in that character recognition is performed again by regarding data regarding characters including the data of 1 as one character.