JPS63136182A - Input device for handwritten character - Google Patents

Abstract

PURPOSE:To reduce input time and labor by outputting a character including a written pattern as a candidate and therefore writing only the relevant part for input even though a desired character is not stored accurately. CONSTITUTION:Not only a character that is written completely and correctly but a character that is partly written is also available for the character to be written as an input. For instance, a Chinese character 'DARE' is available as long as only its 'FURUTORI' (the right-hand radical) is written. A matching part 2 reads an input pattern out of an input part 1 and one or more standard patterns out of an isolated character standard pattern memory part 3 respectively to obtain a partial pattern on a standard pattern sesponding best to the input pattern as well as the distance of said partial pattern with sespact to each standard pattern. A candidate character deciding part 4 obtains a candidate group including said input pattern, i.e., the recognizing candidates like other Chinese characters having the same 'FURUTORI'. These candidate characters are sent to an output part 6 and also displayed on a display part 5 for selection.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a character input device that recognizes and inputs the handwriting of written characters.

(Prior Art) Handwritten character input devices are quickly being put into practical use as excellent input devices that do not require training. Their principle is that characters to be recognized are registered as standard patterns, the distances between those patterns and the input character pattern are calculated, and the standard pattern name that gives the smallest distance is used as the recognition result. A matching method is known.

An example of a device based on this principle is the CR-100 online handwritten character input device manufactured by NEC ■.

This device is described in [Online Handwritten Kanji Input Terminal J, NEC Techniques, Vol. 37° No. 5. P
14-17 in detail. Input using the pattern matching method in the conventional device will be briefly explained below.

The input characters are converted into time series data of feature quantities. As the feature amounts, up-down information of the pen, direction angles of line segments forming characters, XY coordinates, etc. are used. Here, the description will be made using the direction angle as the feature amount. The input pattern is A-(a,
11≦i≦■). The standard pattern is
All characters to be recognized are stored in the input device in advance as time-series patterns having the same direction and angle as the input pattern. When the number of standard patterns is K, the standard patterns can be expressed as B, = (b, 11≦j≦Jk). Here, Jk is the number of line segments in the standard pattern. The distance between input pattern A and standard pattern B5 is defined as the distance when the correspondence between all line segments of A and all line segments of B1 is optimized. That is, input pattern AO
If the distance between the first data and the third data of standard pattern B is d(i, j), then the distance between patterns D (A, Bk) between A and B is d (t, j). As a method for associating the 0 time axis, which is the cumulative value of d(i, j) when the time axes A and B are associated so that the accumulated value in the time axis direction is the minimum, for example, Communication Society Technical Research Report PRL) 13-29 (1'133 September 27
The DP matching method described in the paper entitled "Online Handwritten Character Recognition by Stank DP Matching" published on pages 1 to 8 of Japanese) can be used. With this method, the distance from the input pattern is determined for each of all standard patterns, and the category of the standard pattern that provides the smallest distance among them is determined as the recognition result.

(Problem to be Solved by the Invention) In the handwritten character input device as described above, the distance from the input pattern to the entire pattern from the start to the end of each character of the standard pattern is calculated and recognized. It was about getting results. Therefore, unless the characters are written almost accurately from beginning to end, correct recognition is impossible because the distance from the standard pattern becomes large. In other words, it was impossible to manually write characters that could only be partially remembered or characters that were not accurately memorized unless a function such as kana-kanji conversion was added. Furthermore, even if the user remembers the information accurately, it is a burden for the user to write it down from beginning to end.

The purpose of the present invention is to eliminate such drawbacks and to
To provide a handwritten character input device capable of correctly inputting even partially written characters. It's in the second place.

(Means for Solving the Problems) The handwritten character input device according to the present invention requires the following parts. In other words, the elements constituting the present invention include an input unit that reads the characteristic amount of the handwriting of a character to be written as a human cover pattern, and an isolated character pattern that stores an isolated character pattern representing the handwriting from the beginning to the end of one character as a standard pattern. a character standard pattern storage section; a matching section that calculates the distance between the human cover pattern read by the input section and the entire or part of each standard pattern in the isolated character standard pattern storage section; a candidate character determination unit that determines recognition character candidates based on the determined distance; a display unit that displays one or more candidate characters determined by the candidate character determination unit to prompt the user to select; These are each part of the output unit that outputs selected candidate characters.

(Function) The present invention includes a matching unit that matches an input pattern with every part of a standard pattern, a candidate character determining unit that narrows down recognition candidates based on the determined distance, a display unit that prompts a user to select a candidate character, and a display unit that prompts a user to select a candidate character. By providing an output section that outputs a selected character among the characters, it is possible to input a partially written character. In other words, even if the user does not memorize the characters accurately and writes only a certain part of the characters, the input pattern is recognized as a partial pattern of the standard pattern, and the user can choose from among the categories of the standard pattern as a candidate. select. In this method of recognizing partial input patterns, various partial patterns and distances between the input pattern and the standard pattern are calculated, and if there is a pattern whose distance is less than or equal to the level, one of the standard patterns is calculated. Assuming that there is a high possibility that it is a standard pattern, the category of that standard pattern is selected as a candidate. Similarly, matching is performed with standard patterns of all target character types, and it is determined whether each standard pattern is a candidate.

(Example) Examples of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a handwritten character input device of the present invention. Here, an online handwritten character input device will be described as an example of an input device.

First, the isolated character standard pattern storage section 3 stores in advance one or more standard patterns for each character to be recognized. As described in the prior art, the standard pattern is described as time-series data of feature quantities such as direction angle and XY coordinates. Here, the standard pattern is expressed as Bm=(b=li≦j≦Jk), where the number of standard patterns is K. The input pattern is time series data A=(a, 11≦i≦I
) are read sequentially. The characters written as input may be characters written correctly from beginning to end as in the case of conventional devices, or may be characters written by only a part of the target character. For example, even if you want to recognize the character "Who", you only need to write down "Good;". In the matching section 2, the input pattern A is input from the input section 1, and the standard patterns Bl, k"' are input from the isolated character standard pattern storage section 3. l...
, K is read, and the distance between the partial pattern on the standard pattern to which the input pattern most closely corresponds and the distance at that time is determined for each standard pattern. The way to do this is
It is possible to use a "start-end open DP matching method" in which line segments of the standard pattern corresponding to the writing start (hereinafter referred to as the start end) and the writing end (hereinafter referred to as the end) of the input pattern are not fixed.This method Regarding the patent application 1986,
It is described in detail in the specification of No.-031179, and can be used. In other words, this method does not associate each pattern with line segments on the condition that the input pattern and the standard pattern's start and end points correspond, but allows the input pattern to be associated as part of the standard pattern. It's a method. FIG. 2 is a diagram showing such correspondence.

In Figure 2, human cover turn A is standard pattern B.

It is mapped to the line segment in the latter half of . In the matching section 2, all standard patterns By, k-1, . . .

Make a correspondence with K and find the inter-pattern distance D at that time.
Find (A, B,). Inter-pattern distance D (A, B,
) found correspondence (Lj(i)), i −1,
..., I, the distance between the line segments is d (i, j(i
)) ■, then D(A, B,) = Σd (i, j(i)) can be defined. When pattern distances are calculated for all standard patterns, the candidate character determination unit 4 determines the category group of standard patterns that are considered to include human cover turns, that is,
Find candidate characters. To find candidate characters, for example, a method using a threshold value can be used. That is,
A standard pattern category that provides a distance smaller than a predetermined distance threshold Thl is set as a standard pattern candidate. If you write only “Good” above,
In standard patterns that include ``佳'' such as ``Who,''``Shii,''``Kei,'' and ``Wai,'' the oI probability is such that the line segment is associated with the ``Wabi J'' part, and the distance is a small value. is high, so by setting Thl to an appropriate value,
Recognition candidates such as ``Who'', ``Shii'', ``Kore'', and ``IJ'' can be found.The candidate character determination section 4 then sends these candidates to the output section 6 and displays them on the display section 5 for the user to see. The user's decision is read into the output unit 6 and the final result is output.

In addition to the algorithm for the matching section, "Start end open end 1) P matching section", IEICE technical research group material, p RL85-11 (June +98s), 1
It is also possible to use the "optimal stroke matching method" described on pages 7 to 24.Furthermore, in the embodiment of the present invention, online text input was explained, but the method for determining the similarity of partial patterns can also be used. OCR by changing
(Optical Character Reader)
It is clear that the present invention is also applicable to non-online input such as.

(Effects of the Invention) The online handwritten character input device according to the present invention described above can output characters including a handwritten pattern as candidates. Therefore, even if the target character is not accurately memorized, it is possible to obtain a recognition result using only the memorized portion. Therefore, it is possible to save time and effort such as looking up a dictionary, which was conventionally necessary, and it is possible to improve the input speed. In addition, when a certain part of the target character has a partial pattern that is very unique to that character, it is possible to write and input only that part, which has the advantage of reducing the input effort. .

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a diagram for explaining processing in a matching section in the present invention. 1.---Input section, 2-.--1-.-Matching section, 3.--Isolated character standard pattern storage section, 4
−・−・−・−Candidate character determination section, 5−−−−−・−・−
Display section, 6...--Output section.

Claims (1)

[Claims] an input section that reads the feature amount of the handwriting of a character to be written as an input pattern; an isolated character standard pattern storage section that stores an isolated character pattern representing the handwriting of one character from the start to the end of writing as a standard pattern; and the input section a matching section that calculates the distance between the input pattern read in and the entire or part of each standard pattern in the isolated character standard pattern storage section; a candidate character determining section for determining candidates; a display section for displaying one or more candidate characters determined by the candidate character determining section to prompt the user to select; and an output section for outputting the selected candidate characters to the user. A handwritten character input device comprising: