JPH06187504A - On-line handwritten character recognizer - Google Patents

Abstract

PURPOSE:TO ensure the vecorization with suppression of movements occurring at the positions near a dividing boundary and to improve the recognizing precision of handwritten characters by increasing the size of a slant area compared with the vertical and horizontal areas of each divided area. CONSTITUTION:A coordinate input device 31 inputs the designated characters, and these input data are sent to a RAM 36 and stored there. The stored coordinate data are sent to a vector processing part 43 and approximated to a segment having an angle. This obtained segment is sent to a dictionary comparison part 44 and compared with a dictionary of designated characters to receive the attribute of either one of eight directional vectors. Then the angular information and the information on the eight directional vectors are sent to a dividing area deciding part 45 so that the divided areas are corrected. In this correction, a divided area is obliquely increased so that its angle is included in the slant direction when the slant direction vector originally has an angle larger or smaller than that of the divided area. Thus the number of recognizing combination can be decreased fr identification of characters.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an on-line handwritten character recognition device for improving recognition rate by minimizing blurring of eight-direction vectorization due to difference in handwriting in an on-line handwritten character recognition device. is there.

[0002]

2. Description of the Related Art Generally, a character recognition device has a structure in which character data input by an input device is preprocessed and features are extracted and then identified. This pre-processing normalizes the character data of the characters handwritten online and removes noise contained in the normalized character data. In the feature extraction, a stroke (line segment) is detected from the information on the writing point movement in the process of writing a character, and a representative point (feature point) of this stroke is extracted. In addition, the character is identified by comparing the feature points extracted for the input character with the standard feature points for the standard characters prepared in advance, and as the distance between the corresponding feature points, each stroke is the minimum sum standard. Recognize the character as the corresponding character.

In particular, when vector quantization of each stroke is performed in feature extraction, areas are divided into vertical, horizontal, and diagonal directions, and vectorization is performed based on the areas.

As described above, in the online character recognition, since the time information of writing a character can be sequentially used, there are many advantages as compared with other character recognition, and high accuracy can be achieved by a relatively simple method. it can. For example, the stroke of a character can be easily cut out by detecting whether the writing point is raised or lowered, and the stroke order can be used as extremely effective information.

[0005]

Conventionally, in this type of vectorization, as shown in FIG. 6 (A), vectorization was performed in equal areas in eight directions. It occurred greatly and had a great influence on the recognition rate. In particular, as shown in FIG. 6B, in the vicinity of the boundary line between the vertical line and the diagonal line, and the horizontal line and the diagonal line, the vector that should originally be the diagonal line is classified into the vertical line and the horizontal line due to individual differences in character input. There has been a problem that a lot of blurring occurs, which becomes a factor of reducing the recognition accuracy.

The present invention has been made in order to solve the above-mentioned problems, and an on-line handwritten character recognition apparatus having a high recognition accuracy by eliminating unstable line segments near an oblique boundary line of a direction classification region due to vector quantization. The purpose is to provide.

[0007]

In order to solve this problem, according to the present invention, a vertical / horizontal area of each of the vertical / horizontal / diagonal divided areas which becomes a reference of vector quantization processing. On the other hand, by taking a large diagonal area, vectorization is performed while suppressing blurring near the division boundary line.

[0008]

With this configuration, the process of quantifying the coordinate data, which is one process of the on-line handwritten character recognition device, into a vector of 8 directions (16 directions),
By utilizing the property that the horizontal direction is stable and making the diagonal region larger than the vertical and horizontal regions for vectorization, vectorization in the vicinity of the dividing line compared to the conventional quantification of eight-direction equal regions. Since blurring is absorbed and unstable line segments are eliminated, the number of combinations in recognition is reduced, and recognition accuracy is improved.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a system configuration diagram of an online handwritten character recognition apparatus according to an embodiment of the present invention.

In FIG. 1, reference numeral 31 is a coordinate input device for inputting coordinate data, and sends the input coordinate data to a data storage area 38 of a random access memory (RAM) 36. A central processing unit 32 controls various devices according to a program stored in a program storage area 34 of a read only memory (ROM) 33. A ROM 33 has a program storage area 34 in which programs are stored and a character recognition dictionary area 35 in which a recognition dictionary is stored. 36 is a RAM,
It has a data storage area 38 for storing the input coordinate data and a vector division area storage area 37 for storing the vector division area. A display device 39 obtains data from the program stored in the program storage area 34 and displays an input frame or the like. The pen input computer system configured as described above is shown as a block for each functional element, as shown in FIG.

FIG. 2 is a functional block diagram of an online handwritten character recognition apparatus according to an embodiment of the present invention.

In FIG. 1, the online handwritten character recognition apparatus according to this embodiment has a coordinate input device 31 for inputting the coordinates of each point of the online handwritten character as coordinate data, and a RAM 36 for storing the coordinate data. A central processing unit 32 that performs vector quantization on the stored coordinate data and compares it with standard data as a dictionary, corrects the divided area based on the comparison result, and judges an online handwritten character, and the central processing unit. The ROM 33 is provided with a character recognition dictionary area 35 for the arithmetic operation program of 32 and the standard data, and a display device 39 for displaying online handwritten characters judged by the central processing unit 32.

The central processing unit 32 controls the arithmetic operation of the entire device based on a program stored in the program storage area 34 of the ROM 33, and the coordinate data output from the RAM 36 for each line segment of a character. A vector processing unit 43 for approximating the approximated line segment, a dictionary comparison unit 44 for comparing each approximated line segment with the data in the character recognition dictionary region 35 of the ROM 33, and an 8-divisional region based on the comparison result. And a divided area determination unit 45 that corrects the divided area and determines a divided area.

Next, the operation of the apparatus of this embodiment based on the above-mentioned structure will be described with reference to FIG. First, in the coordinate input device 31, a designated character is input. The input data is sent to and stored in the RAM 36. The stored coordinate data is sent to the vector processing unit 43 and approximated to an angled line segment. The obtained line segment is sent to the dictionary comparison unit 44 and compared with the dictionary of the designated character to give any attribute of the 8-direction vector.
Next, the angle information and the information of the eight-direction vector are sent to the divided area determination unit 45 to correct the divided area. This correction is to include the angle in the diagonal region by expanding the diagonal direction of the divided region as shown in FIG. 3A when the vector in the diagonal direction originally has an angle equal to or greater than or equal to the divided region. As a result, the number of combinations in recognition is reduced as shown in FIG. By repeating the above operation for a plurality of characters, it is possible to obtain a highly accurate 8-direction divided area.

FIG. 4 is a perspective view showing a concrete structure of an embodiment of the present invention. In FIG. 4, 11 is an input pen used for input. Reference numeral 12 is a coordinate input device. Reference numeral 13 is a display device using a liquid crystal or the like, in which an input frame and input coordinates are echoed back and displayed. 14 is a main body part.

FIG. 5 is a diagram showing a usage state of the on-line handwritten character recognition device in one embodiment of the present invention.

In FIG. 5, reference numeral 21 is an input pen used for input. Reference numeral 22 is a coordinate input area forming a coordinate input device, and the user inputs data in this frame.
A display device 23 displays a normal computer screen or the like.

[0019]

As described above, according to the present invention, since vectorization can be performed with highly accurate 8-direction divided areas, unstable line segments near the dividing line are eliminated, and as a result, the number of vector combinations in recognition is reduced, and recognition accuracy is improved. The effect that it can improve is produced.

[Brief description of drawings]

FIG. 1 is a system configuration diagram of an online handwritten character recognition device according to an embodiment of the present invention.

FIG. 2 is a functional block diagram of the device shown in FIG.

FIG. 3 is an explanatory diagram of each operation of the 8-segment area and the recognition combination in the apparatus illustrated in FIG.

FIG. 4 is a specific configuration perspective view of the device shown in FIG.

5 is a diagram showing a usage state of the apparatus shown in FIG.

FIG. 6 is an explanatory diagram of each operation of an 8-segment area and a recognition combination in a conventional online handwritten character recognition device.

[Explanation of symbols]

 11, 21 Input pen 12, 22, 31 Coordinate input device 32 Central processing unit 33 ROM 34 Program storage area 35 Character recognition dictionary area 36 RAM 37 Vector division area storage area 38 Data storage area 13, 23, 39 Display device 40 control unit 43 vector processing unit 44 dictionary comparison unit 45 divided area determination unit

Claims (1)

[Claims] 1. A direction in which handwriting information in the process of writing a character is input from a coordinate input device such as a tablet, and each input handwriting information is classified by areas divided into a plurality of vertical, horizontal, and diagonal directions. An on-line handwritten character recognition device for performing vector quantization processing to identify an on-line handwritten character based on each of the vector-quantized handwriting information, wherein the divided vertical / horizontal / diagonal regions are diagonally divided. An online handwritten character recognition device characterized in that the area of is larger than the vertical and horizontal areas.