JPS6161151B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a segmentation method for automatically segmenting characters in an online handwritten character recognition device.

When entering characters into an online handwriting recognition system, an input device typically consists of a tablet and a stylus pen. When an operator writes characters, he places a manuscript paper with a frame surrounding the area where the characters are to be written on the tablet and begins writing.In this case, the character data is sampled by sampling the character coordinates by moving the pen down. is started and continues until pen up.

When writing characters continuously, data for one character must be extracted from such a series of coordinate points and used as a recognition target. The above extraction is called segmentation. As a segmentation method, a method is known in which an operator notifies a recognition unit of a character end signal in a specific manner every time a character is written. Although such a method is simple and error-free, it is an operational burden and inefficient for the operator to input a control signal indicating the end of one character each time he finishes writing one character. In order to perform character segmentation automatically, it is conceivable to use the starting point or end point of a stroke as a representative point and perform character segmentation using these representative points, as shown in Figure 7. If the distance between character frames becomes small to a certain extent, stable segmentation cannot be performed.

The present invention aims to eliminate the above-mentioned drawbacks and to provide a segmentation method for an online handwritten character recognition device that can always perform stable segmentation. Therefore, the segmentation method in the online handwritten character recognition device of the present invention is based on the stroke information and time series information of characters that are chronologically written in the character frames on the manuscript paper placed on the input tablet. In an online character recognition device that performs character recognition, the representative point is the midpoint of the line segment connecting the start and end points of the stroke, the center of gravity of the stroke, or a point on the stroke that is 1/2 of the total distance of the stroke from the start point. Determine whether or not the representative point of the stroke exists in the same character frame for each piece of stroke information that is input in chronological order. At the time when the stroke information is obtained, the stroke information before the stroke information is determined to be one character recognition target, and one character is thereby extracted. Hereinafter, the present invention will be explained with reference to the drawings.

FIG. 1 is a perspective view of an example of an input device used in the present invention, FIG. 2 is a diagram showing manuscript paper, FIG. 3 is a diagram showing the system configuration of an online handwritten character recognition device, and FIG. 4 is a diagram showing a segment FIG. 6 is a diagram showing the function of the station section, and is a diagram for explaining the principle of one embodiment of the present invention.

In Figures 1 and 2, 1 is a tablet, 2
indicates manuscript paper, 3 indicates a stylus pen, R ₁ to R _o indicate character frames, and M indicates a mark.

When inputting characters, a manuscript paper 2 is placed on a tablet 1, and a mark M is touched with a stylus pen 3. As a result, coordinate values representing character frames _R1 to Rn are registered on the online handwritten character device side. Characters are written in the character frame Ri using the stylus pen 3. When a character is written with the stylus pen 3, a switch (not shown) inside the stylus pen 3 is closed, sampling is started, and the coordinate points of the stroke are sent to the online handwritten character recognition device. When one stroke is completed and the stylus pen 3 is raised, a switch inside the stylus pen 3 is opened and stroke end information as described below is written.

3, 4, and 5, 4 is an input device, 5 is a segmentation unit, 6 is a recognition unit, and 7
1 is a display, 8 is a noise removing means, 9 is a positioning means, 10 is a character frame generating means, 11 is a character buffer, and 12 is a character area determining means.

The input device 4 consists of a tablet 1, a manuscript paper 2, and a stylus pen 3, as shown in FIG. The segmentation section 5 is composed of, for example, a microcomputer, and performs the functions shown in FIG. 4. The recognition unit 6 is composed of, for example, a large-sized computer, and performs character recognition based on the character information of one character sent from the segmentation unit 5. The display 7 displays recognized characters.

FIG. 4 explains the segmentation unit 5. The positioning means 9 determines the position of the document paper 2 based on the coordinate values of the mark M. The character frame generating means 10 calculates the coordinate values of each character frame from the coordinate values of the manuscript paper 2. The character buffer 11 stores coordinate point series of strokes forming characters. Character area determining means 12
determines the set of strokes that make up one character,
This is to notify the recognition unit 6 of the coordinate point series of the stroke set.

FIG. 5 shows the structure of the character buffer 11. X _1-1 , Y _1-1 ...X _1-i , Y _1-i is the stroke S ₁
shows the coordinate point series, X _2-1 , Y _2-1 ... is the stroke
It shows the coordinate point series of stroke S ₂ written after S ₁ . Coordinate point series X _1-1 , Y _1-1 ...X _1-i , Y _1-
Stroke end information is inserted between _i and the coordinate point series X _2-1 , Y _2-1 . . . in order to distinguish between strokes.

FIG. 6 is a diagram for explaining the principle of segmentation in one embodiment of the present invention. In Figure 6, (X _S1 , Y _S1 ), (X _S2 , Y _S2 ) are the coordinate values of the upper left corner of the character frame, (X _E1 , Y _E1 ), (X _E
2 , Y _E2 ) are the coordinate values of the lower right corner of the character frame, P _S1 , P _S2 ... are the starting points of the stroke, respectively,
P _E1 , P _E2 ... are the end points of the stroke, M ₁ , M ₂ ,
M ₃ and M ₄ indicate the midpoints of the start and end points of the stroke, respectively.

Now, let us assume that a set C of character frames R set in advance is C={R|R ₁ , R ₂ . . . Rn}. However, the character frame R is defined by a point (X _S , Y _S ) and a point (X _E , Y _E ).

When the operator fills in characters sequentially,
The _midpoint M ₁ (X _{M1 ,} Y _M1 ) of the starting point P _S1 of the first stroke _of "A _" and P _E1 satisfies _{the following} : In this case, M ₁ 〓R ₁ ………………………. Similarly, the midpoint of the second stroke of “a”
It turns out that M ₂ is also M ₂ 〓R ₁ . However, when the first stroke of "i" is entered, it is known that the midpoint M ₃ is M ₃ 〓R ₂ , so the stroke information up to the first stroke of "i" is written for one character. is the recognition target data.

In the above example, segmentation was performed at the midpoint between the start and end points of the stroke, but instead of this, 1/2 of the total distance of the stroke is calculated, and the segmentation is performed at a position 1/2 of the total distance from the start point. It is also possible to use a point on the stroke as a representative point, and perform segmentation using such a representative point. Alternatively, the center of gravity of the stroke may be used as a representative point, and segmentation may be performed using such a representative point. To find the center of gravity of a stroke, find the maximum and minimum values of the X coordinate value and the maximum and minimum values of the Y coordinate value from the coordinate point series of the stroke, and then draw a rectangular frame defined by the maximum and minimum values above. All you have to do is find the center of

As is clear from the above description, according to the present invention, stable and reliable segmentation can be performed at all times compared to conventional segmentation methods.

[Brief explanation of the drawing]

FIG. 1 is a perspective view of an example of an input device used in the present invention, FIG. 2 is a diagram showing manuscript paper, FIG. 3 is a diagram showing the system configuration of an online handwritten character recognition device, and FIG. 4 is a diagram showing a segment FIG. 5 is a diagram showing the outline of the character buffer, FIG. 6 is a diagram for explaining the principle of one embodiment of the present invention, and FIG. 7 is a diagram showing the conventional segmentation section. FIG. DESCRIPTION OF SYMBOLS 1...Tablet, 2...Manuscript paper, 3...Stylus/pen, 4...Input device, 5...Segmentation section, 6...Recognition section, 7...Display, 8...Noise removal means, 9...Positioning means, 10...Character Frame generating means, 11...Character buffer, 12...Character area determining means.

Claims (1)

[Scope of Claims] 1. An online character recognition device that performs character recognition based on stroke information and time-series information of characters chronologically written in character frames on a manuscript sheet placed on an input tablet, Find the midpoint of the line segment connecting the start and end points of the stroke, the center of gravity of the stroke, or a point on the stroke that is 1/2 the total distance of the stroke from the start point as the representative point. It is determined whether or not the stroke information exists in the character frame for each piece of stroke information that is input in chronological order, and when stroke information whose representative point does not exist in the same character frame is obtained, the stroke information before the relevant stroke information is determined. A segmentation method in an online handwritten character recognition device, characterized in that stroke information is determined as one character recognition target, and one character is thereby extracted.