JPH0729002A - Handwritten graphic recognition device - Google Patents

Abstract

PURPOSE:To prevent the deterioration in a recognition rate owing to the fluctuation of a pen and the quantizing error of a tablet at the time of input. CONSTITUTION:Data which a writer enters is converted into data of a coordinate point string expressed by an electric signal on the display integrated tablet 1 and the data is transferred to a stroke processing part 3. When the termination of the entry of a stroke is judged in a stroke termination judgement part 2, data is processed for respective strokes whenever the stroke is inputted and the number of sides included in the stroke and a part which seems to be an arc are obtained in the stroke processing part 3. When a graphic termination judgement part 4 judges the termination of a graphic, a graphic recognition part 5 is started, and whether the inputted stroke is a closed graphic or an opened graphic. Characteristics formed on all the strokes are extracted after the graphic is inputted, and the closed graphic is judged. Then, the graphic is synthesized with the characteristics for the respective strokes and the graphic is recognized without depending on a stroke order an the number of the strokes.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten figure recognition apparatus, and more specifically, it recognizes a figure without using a figure dictionary.
The present invention relates to a handwritten figure recognition device that displays a figure closer to the input and limits the figure to be recognized according to the size of the input figure and is suitable for actual use. For example, it is suitable for a word processor, a personal computer, or the like that identifies a figure that is input from a tablet using a pen and displays the result.

[0002]

2. Description of the Related Art As a method of recognizing a graphic entered on a tablet, an input graphic stroke is converted from a shape feature into a set of line segment element sequence data by arcs or straight lines, and is previously converted into similar line segment element sequence data. The recognition was performed by matching the defined registered figure. Although there is a method that introduces ambiguity when exchanging with line segment element sequence data, it was necessary to register the figure. In addition, there is a method in which the distribution of the angle histogram in the writing direction is examined, line segments are determined based on this distribution status, and the number of sides or the like is identified. In either method, uniform processing was performed regardless of the size of the input figure and the length of the stroke. Further, as a method of selecting a candidate of the figure recognition result, there is a method of indicating the next candidate. That is, this is a method in which the next candidate appears when the defined area is touched with the pen. There is also a method of tapping (lightly touching the tablet with a pen) to display a list of candidates and selecting from them. Further, in the case of canceling the input figure, there is also a method of canceling the input figure by touching the area defined as cancel with a pen.

As a known document describing the prior art of the present invention, for example, a "pattern recognition method" has been proposed in Japanese Patent Laid-Open No. 63-82584. The one disclosed in this publication is suitable for a recognition device for handwritten character / figure with ambiguity. Even if there is ambiguity in the stroke shape, the shape of the input stroke is changed in order to enable correct matching with the registered figure. Separation into a part that is clearly an arc and a part that is a straight line and a part that is an arc, a straight line or an ambiguous part, and by changing the interpretation of whether this ambiguous part is an arc or a straight line, the arc or straight line of the input stroke A plurality of structural patterns with segment elements are created, the degree of difference is determined by comparing the structure of the structural pattern and the structural pattern of each dictionary figure, and checking the arc-like or straight-line likeness of the line segment element, and the difference is small. It is to choose one.

[0004]

As described above, in the conventional graphic recognition device, there are the following problems in making a low-priced device and mounting it in a personal device. That is, in the method of converting into a line segment element sequence, in order to perform accurate recognition, it is necessary to trace the coordinate sequence in detail for all the written strokes and determine the line segment and the arc element. Further, it takes a long processing time to determine the element sequence, and a personal device having a low processing capacity such as a CPU (central processing unit) cannot perform the processing in real time. There is a cost problem in increasing the processing capacity. In addition, the method using the distribution of the angle and the histogram cannot recognize correctly an absolutely small figure because of the quantization error caused by the resolution of the tablet. Further, increasing the resolution of the tablet has manufacturing problems and cost problems, and cannot be adopted for personal devices.

Further, when a person manually writes a figure on the tablet with a pen, the pen fluctuates. This fluctuation may adversely affect recognition when the figure is small. For example, a circle written in a very small size may be erroneously recognized as a polygon having a large number of sides such as a regular hexagon. In general, when a person writes a figure, if the size of the figure becomes smaller than a certain size, it is usually deformed and cannot be written correctly. On the other hand, in the case of a display-integrated type tablet, the figure may be deformed and displayed due to the low resolution on the display side, and the third party may not be able to judge. For example, when writing a regular hexagon with a maximum width of 5 mm for a 3 mm figure, the handwriting on the tablet is often seen as a circle by a third party. In such a case, in the above method, a hexagon or the like may be output as a recognition result, which may cause a mismatch with the written handwriting.

In the conventional method of selecting a candidate, the amount of movement of the pen is large, and it takes time to input a figure with the pen.
In addition, the stroke may be input to the tablet against the intention while the writer is writing the figure, but in this case also,
In the above method, misrecognition may occur. Thus, there is a problem of operability.

The present invention has been made in view of the above circumstances, and does not depend on the stroke order, and does not depend on the input figure, except for the reduction of the recognition rate due to the fluctuation of the pen which occurs at the time of input and the quantization error of the tablet. It is an object of the present invention to provide a handwritten figure recognition device with less difference.

[0008]

In order to achieve the above-mentioned object, the present invention creates (1) detection means for detecting the trajectory of a pen written on a tablet and stroke data recorded in time series. An online handwritten figure recognition device comprising a processing means for performing recognition processing using the stroke data and a display means for displaying a processing result by the processing means on a tablet, and data for each stroke when the stroke is input. Data processing means for carrying out the processing, recognition means for obtaining the number of sides included in the stroke and a portion recognized as an arc, extraction means for extracting features formed in all strokes after inputting a figure, and closed figure It has a judgment means for judging (a figure that forms a loop with all strokes and does not have an intersection), and integrates it with the characteristics of each stroke to create a stroke order or image. To recognize the figure without depending on,
Further, (2) as a feature of each stroke, a calculating unit that calculates a histogram of an angle component indicating a writing direction, an estimating unit that estimates the number of sides based on the distribution of the histogram, and a category after classifying The linearity of the section is determined with respect to the detection means for detecting the candidate positions of the vertices estimated in each case and the coordinate sequence between the vertices and the coordinate sequence between the adjacent sides input in the same stroke. By having a discriminating means, a figure having an arc such as a fan shape, and a discriminating means for discriminating a figure that does not have it, by performing detailed discrimination of the figure based on features such as side length and angle obtained from the coordinates of the vertices. Recognizing a figure without depending on the stroke order and the number of strokes, or (3) detecting means for detecting the trajectory of a pen written on a tablet, and creating stroke data recorded in time series, The stroke is input in an online handwritten figure recognition device including a processing means for performing recognition processing using trooke data, a display means for displaying a processing result by the processing means on a tablet, and a selection means for selecting a recognition result. Data processing means for processing data for each stroke each time, and a recognition means for determining the size and rotation angle of the graphic element included in the stroke, after the graphic input is completed, by the data of all strokes, To find the size and rotation angle of a graphic element,
(4) In any one of (1), (2), and (3) above,
Depending on the size of the input graphic, the category of the graphic output as the recognition result is restricted, and further, (5)
In any one of (1), (2), and (3), when it can be determined that the input stroke is shorter than other strokes, the stroke is excluded from the recognition target, or (6) display The detection means for detecting the trajectory of the pen written on the integrated tablet, the processing means for creating the stroke data recorded in time series and performing the recognition processing using the stroke data, and the processing result by the processing means In an online handwritten pattern recognition device comprising a display means for displaying on a tablet and a selection means for selecting a recognition result, when the recognition result is displayed on the screen, by tapping the display screen with a pen, Of the candidate for displaying the candidate and the next figure input, or (7) online handwritten figure recognition having a function of selecting the candidate. On the device, when a tablet on which the recognition result is displayed is tapped with a pen, processing such as candidate selection or input cancellation is performed depending on the relationship between the tapped position on the tablet and the position of the displayed graphic. In an online handwritten figure recognition device having selection means for selecting or (8) a function for selecting candidates, when tapped with a pen on the tablet on which the recognition result is displayed, the tapped position on the tablet And a stroke position forming a displayed graphic, and selecting means for selecting a stroke that is erroneously recognized, and selecting processing such as selection of a candidate for only that portion, or (9) Detecting means for detecting the trajectory of the pen written on the tablet and stroke data recorded in time series are created, and the stroke data is recorded. In an online handwritten figure recognition device comprising a processing means for performing recognition processing and a display means for displaying the processing result by the processing means on a tablet, the recognition result is output depending on the number of strokes input to the tablet. It is characterized by having a limiting means for limiting the categories of figures.

[0009]

The handwritten figure recognition apparatus of the present invention processes the data for each stroke every time a stroke is input.
Obtain the number of sides and arc-like parts included in the stroke, extract the features formed in all strokes after the figure input, determine the closed figure, integrate with the features of each stroke, stroke order, Recognize shapes without depending on the number of strokes. The amount of processing can be suppressed by using the simple process of calculating the distribution of the angle histogram and by performing the process for each stroke independently. The method using the distribution of the angle histogram absorbs the blurring of the pen. There is a merit that can be done. In addition, since the number of vertices is estimated in the vertex extraction method, it is possible to easily detect the position of the vertices without the need for strict tracking determination as compared with the case of decomposing into line segment element strings. By performing detailed identification again using the detected vertices, it is possible to solve the problem of erroneous recognition due to the low resolution of the tablet when the histogram is used again. With this method, it is possible to provide a graphic recognition system which can be mounted on a small personal device having a low processing capacity, has a high degree of freedom without being influenced by the stroke order, the number of strokes, the rotation, etc., and can suppress the processing amount. Further, the category of graphics to be output as a recognition result is limited by the size and the number of strokes of the entered graphics, and a recognition result close to the graphics actually displayed on the tablet can be obtained. From the viewpoint of operability, even when selecting candidates, by eliminating complicated operations and shortening the input time,
The efficiency of figure input with a pen is improved. In this way, a compact, high-performance and easy-to-use graphic recognition device can be realized.

[0010]

Embodiments will be described below with reference to the drawings. FIG. 1 is a block diagram for explaining an embodiment of a handwritten pattern recognition device according to the present invention, and is a block diagram of an online pattern input device for inputting a pattern by rotating a pen on a tablet. In the figure, 1 is a tablet, 2 is a stroke end determination unit, 3 is a stroke processing unit, 4 is a figure end determination unit, 5
Is a figure recognition unit, 6 is a display coordinate generation unit, and 7 is a display unit. Note that the numerical values in the following embodiments are examples, and appropriate settings can be made in consideration of the usage status of the figure recognition device.

Embodiment 1 A handwritten figure recognition device according to the present invention detects a trajectory of a pen written on a tablet, creates stroke data recorded in time series, performs a recognition process using the stroke data, and The present invention relates to an online handwritten figure recognition device that displays results on a tablet. Data written by the writer on the display-integrated tablet 1 is converted into data of a coordinate point sequence represented by an electric signal, and the data is transferred to the stroke processing unit 3. Next, when the stroke end determination unit 2 determines that the stroke has been entered, the stroke processing unit 3 performs stroke processing. That is, each time a stroke is input, data processing is performed for each stroke, and the number of sides included in the stroke and an arc-like portion are obtained. Next, when the figure end determination section 4 determines the figure end, the figure recognition section 5 is activated to determine whether the input stroke is an open figure or a closed figure. That is, after the figure input is completed, the features formed in all strokes are extracted, the closed figure (the figure that forms a loop in all strokes and has no intersection) is determined, and it is integrated with the features of each stroke, and the stroke order is determined. Recognize figures without depending on the number of strokes or strokes. The result of the recognition process is sent to the display coordinate generation unit 6, is calculated at the coordinates corresponding to the input position, and the result is sent to the display unit 7 using liquid crystal for display.

FIG. 2 is a flow chart for explaining the operation of the handwritten figure recognition apparatus according to the present invention. Less than,
Each step will be described in order. First, the stroke is input (Step 1), and the stroke processing unit removes the same point and the intermediate point in the coordinate point sequence from the data as preprocessing (Step 2). Next, it is judged whether or not the stroke is shorter than the threshold value (Step 3). If the stroke is shorter than the threshold value, the stroke is skipped (Step 4).
Return to 1. If the stroke is not shorter than the threshold, the histogram of the absolute direction is calculated and the stroke feature is extracted (Step 5). Next, in the figure end determination unit, it is determined whether or not the figure input is completed (Step 6). If the figure input is not completed, the procedure returns to Step 1. If the figure input is completed, the absolute direction histogram is calculated and the stroke feature is extracted (Step 7). Next, in the figure recognition unit, it is judged whether or not the input stroke is a closed figure (Step 8). If it is not a closed figure, an open figure recognition process is performed (Step 10) to calculate the feature amount of the figure. (Step 1
1). If the figure is a closed figure in Step 8, the closed figure recognition process is performed (Step 9), the feature amount of the figure is calculated (Step 11), and the figure is displayed on the tablet. (Step 1
2).

FIGS. 3A to 3D are diagrams showing an example of stroke processing, showing an example of characteristics of each stroke. FIG. 10A shows an example of the input stroke, and the black circles indicate the points after the column processing. FIG. 6B is an example of a direction code, FIG. 6C is an angle histogram (windowing sequence), and FIG. 7D is an angle histogram (after windowing).

The stroke processing unit 3 removes the same points and intermediate points in the coordinate point sequence from the data as preprocessing. The midpoint here means that when connecting the coordinates of the points before and after with a straight line, or if there is no point on this line,
It means the point where it first intersects when proceeding in parallel with the X axis or the Y axis. The black circles in FIG. 3A correspond to the remaining removed parts. After that, the direction and length from that point to the next point are obtained for all points except the end point of the coordinate sequence that continues according to the writing direction in the stroke. As the direction, as shown in FIG. 3B, a value obtained by quantizing the entire circumference of 360 degrees in 32 directions is used. This is shown in FIG. 3C for each direction.
As shown in, the angle histogram obtained by summing the lengths is obtained. Next, as shown in FIG. 3D, after windowing with a width of 3 is performed, the angle having the maximum value of the histogram and
Find the peak or angle with a local maximum. If the maximum value of the peak is much smaller than the maximum value of the histogram, this peak is ignored. At the same time, the total of the circumscribing frame and length of the stroke is totaled. The above processing is repeated for each stroke input.

When the figure end judging section 4 judges that the figure is ended, the figure recognizing section 5 is activated, and the input stroke is first judged as a closed figure or an open figure as follows. That is, when the input figure is one stroke, the distance between the starting point and the ending point is 3 mm or less, and when the coordinates indicating the stroke are connected by a straight line and there is no intersection in the middle, the figure is a closed figure. If the input figure or two or more strokes, the end points of each stroke are connected and all strokes are included, and if all the end points are 3 mm or less, it is determined as a closed figure.

When it is determined that the figure is a closed figure, it is clear that the figure is an N-sided polygon (N> 3) if the figure is a closed figure, so that the peak of the histogram detected by the stroke processing unit 3 is obtained. Estimate the number of edges by the number of
Detailed identification is performed for each side. First, the vertices of the estimated number of sides are extracted. The midpoint is adopted as the apex formed by the end points of a plurality of strokes. When there are vertices in the stroke, the number of sides estimated to be included in the target stroke is first divided so that the length is the same. The above dividing points are arranged in the stroke writing direction between the start point and the end point of the stroke, and the following processing is performed. First, the first point of the division points is changed to the point farthest from the straight line connecting the points before and after. The same processing is repeated for the second and subsequent points. The position of the estimated vertex is detected by repeating this process until all the division points are unchanged. Then, after the vertices are extracted, the straightness defined by the formula below, straightness = (straight line distance between vertices) / (road riding on stroke), and between the vertices of the angle histogram of the corresponding section It is determined whether or not the line is a straight line based on the variance from the straight line direction of. The threshold for making the determination is arbitrarily set according to the tablet performance and the like. As described above, it is possible to distinguish between a polygon and a sector.

When it is determined that the figure is an open figure, line segment approximation is performed for each stroke, and a recognition result including a straight line or a plurality of line segments and an arc is output as a recognition result. The result of the recognition process as described above is sent to the display coordinate generation unit 6, and the coordinates of the display device corresponding to the input position are calculated. The result is sent to the display unit 7 using liquid crystal and displayed. After that, the same processing is repeated until the online figure recognition is completed.

In this way, the open figure and the closed figure are judged, and if the figure is a closed figure, an N-gon (N
Since it is clear that> 3), the stroke straight line portion is detected, and the input figure is recognized based on its position and length. When extracting the size and rotation angle of an input figure, in the case of an N-sided polygon, the rotation angle obtained by stroke recognition is converted into a constant rotation direction (for example, clockwise) at the time of stroke entry, and each figure is converted. The angle of the side. In this case, the figure is recognized and shaped by using the features such as the absolute angle of the other side depending on the position of the start point and the end point, the length of the stroke with respect to the size of the figure, and the length of the side. If it is an open figure, line segments and curves are approximated for each section.

In the recognition of a closed figure, after classifying stroke writing methods by using an angle histogram, an estimated number of feature points such as vertices are detected, and the feature points are used to make a detailed determination. Further, when the size of the graphic becomes smaller than a certain size, the recognition target is narrowed down to prevent the misrecognition of the small-sized graphic. Also, in terms of operability, a part is provided to determine whether the next pen input after displaying the result on the tablet is the next input or candidate selection. Select the processing of. When selecting this processing, the relationship with the display position of the immediately preceding result can be used.

Embodiment 2 In the on- line figure recognition apparatus, the minimum and maximum values in the X-axis direction and the Y-axis direction are calculated for the coordinate point sequences of all strokes input from the tablet, and the X-axis direction and the Y-axis direction are obtained. Calculate the respective widths and use the larger width as the maximum width of the figure. If this maximum width is less than 4 mm, only "circles" or "straight lines" will be recognized.
If it is larger than mm and smaller than 7 mm, use the figure above.
Add "regular triangle" and "square". If it is more than 7 mm and less than 10 mm, add "Right triangle" and "Rectangle".

Embodiment 3 With respect to the stroke coordinate point sequence input from the tablet, after the stroke preprocessing, the distance between the start point and the end point is calculated. In the stroke determination unit, the lengths of the strokes are compared with each other before the recognition process is performed, and the stroke having the maximum length is obtained. If there is a stroke having a length of 1/10 or less based on the length of this stroke, it is determined that the writer has made an error and the stroke is erased from the stroke buffer. After performing the above processing, the recognition processing is executed.

Embodiment 4 FIG. 4 is a flow chart for explaining the online graphic recognition processing. The steps will be sequentially described below. On the input device, online graphic recognition is started by operating a keyboard or the like. When you enter the online figure recognition mode, first enter the pen input waiting state (step
1). When a figure is input, the user inputs the figure using a pen (step 2). After inputting, figure recognition processing is performed (step
3), display the first candidate of the recognition result (step
4). After the display, it waits for pen input (step 5). When any input is made with the pen, it is determined whether or not it is a tap (step 6). If it is a tap, it is interpreted that the next candidate is requested, the figure of the current result is erased, the next candidate is displayed (step 7), and the input waiting state is set (step 7).
5). If it is not a tap, it is determined that the input of the next graphic is started, and the graphic input processing is performed (step 2). After that, the same processing is repeated until the online figure recognition is completed.

Embodiment 5 FIG. 5 is another flowchart for explaining the online figure recognition processing. Hereinafter, each step will be described in order. When you enter online figure recognition mode,
Enter the state of waiting for pen input (step 1). When a figure is input, the user inputs the figure using a pen (step 2). After the input is completed, graphic recognition processing is performed (step 3), and the first candidate of the recognition result is displayed (step 4). After the display, it waits for pen input (step 5). When some input is made with the pen,
It is determined whether it is a tap (step 6). If it is determined that the tap is made, the figure position of the displayed recognition result is compared with the tapped position (step 7), and the tapped position is inside or near the figure (for example,
If the distance is within 3 mm in a straight line), the figure of the current recognition result is erased and the next candidate is displayed (step 8). If it is outside the figure, it is determined that the cancellation has been input, the figure of the recognition result is erased (step 9), and the state of waiting for input is obtained (step 1). After that, the same processing is repeated until the online figure recognition is completed.

Embodiment 6 An explanation will be given according to the flow chart shown in FIG. Note that st
The processing from ep1 to step6 is the same as in the fifth embodiment. step
When it is determined that the tap is made in step 6, the stroke position forming the figure of the displayed recognition result is compared with the tapped position (step 7), and the tapped position is inside or near the figure. (For example, within a straight line distance of 3 mm)
If there is, the graphic of the current recognition result is erased, and the next candidate in which the recognition result of the tapped stroke is changed is displayed (step 7). If it is outside the figure, it is judged that the cancellation has been input, and the figure of the recognition result is deleted (step
9) Then, the input is awaited (step 1). After that, the same processing is repeated until the online figure recognition is completed.

Embodiment 7 Recognition is carried out in the same manner as in Embodiment 1. In the figure recognition unit 5,
When it is determined that the figure is a closed figure, the polygon is recognized by regarding the number of input strokes as one stroke forming at least one side.

[0026]

As is apparent from the above description, the present invention has the following effects. That is, each time a stroke is input, the data for each stroke is processed, the number of sides included in the stroke, the arc-like portion, etc. are obtained, and after the figure input is completed, the features formed in all strokes are extracted. , Closed figure (a figure that forms a loop with all strokes and does not have intersection points) is determined, integrated with the features of each stroke, and the figure is recognized regardless of stroke order, stroke count, so input The fluctuation of the pen at the time
It is possible to provide an online figure recognition device which prevents a reduction in recognition rate due to a quantization error of a tablet, is not affected by the stroke order, and has a small difference from an input figure. In addition, a candidate can be easily selected and re-filling can be selected.

[Brief description of drawings]

FIG. 1 is a configuration diagram for explaining an embodiment of a handwritten figure recognition device according to the present invention.

FIG. 2 is a flowchart for explaining an operation of handwriting recognition according to the present invention.

FIG. 3 is a diagram showing an example of features for each stroke in the present invention.

FIG. 4 is a flowchart for explaining an online figure recognition process according to the present invention.

FIG. 5 is another flowchart for explaining the online figure recognition processing according to the present invention.

[Explanation of symbols]

1 ... Tablet, 2 ... Stroke end determination unit, 3 ... Stroke processing unit, 4 ... Graphic end determination unit, 5 ... Figure recognition unit,
6 ... Display coordinate generation unit, 7 ... Display unit.

Claims (9)

[Claims] 1. A detection means for detecting a trajectory of a pen written on a tablet, a processing means for creating stroke data recorded in time series, and a recognition processing using the stroke data, and the processing means. In an online handwritten figure recognition device comprising a display means for displaying a processing result on a tablet, a data processing means for processing data for each stroke every time the stroke is input, the number of sides included in the stroke, and The feature for each stroke includes a recognition means for obtaining a portion recognized as an arc, an extraction means for extracting a feature formed in all strokes after the graphic input is completed, and a determination means for determining a closed graphic. A handwritten figure recognition device that integrates and recognizes figures without depending on the stroke order or the number of strokes. 2. A feature of each stroke, a calculation means for calculating a histogram of an angle component indicating a writing direction, an estimation means for estimating the number of sides based on the distribution of the histogram, and after classifying the categories, each category The detection means for detecting the candidate positions of the vertices estimated in 1. and the discrimination for determining the linearity of the section for the coordinate sequence between the vertices and the coordinate sequence between the adjacent sides input in the same stroke. By having means, a figure having an arc such as a fan shape, and a determining means for determining a figure that does not have, by performing detailed determination of the figure by features such as side length and angle obtained from the coordinates of the vertices, The handwritten figure recognition device according to claim 1, wherein the figure is recognized without depending on the stroke order or the number of strokes. 3. A detection means for detecting a trajectory of a pen written on a tablet, a processing means for creating stroke data recorded in time series, and a recognition processing using the stroke data, and the processing means. In the online handwritten figure recognition device consisting of display means for displaying the processing result on the tablet and selecting means for selecting the recognition result, data processing means for processing data for each stroke every time the stroke is input, A handwritten figure recognition having a recognition means for obtaining the size and rotation angle of the graphic element included in the stroke, and for obtaining the size and rotation angle of the graphic element from the data of all strokes after the figure input is completed. apparatus. 4. The handwritten figure recognition device according to claim 1, wherein the category of the figure output as a recognition result is restricted depending on the size of the inputted figure. 5. The handwritten figure recognition device according to claim 1, 2 or 3, wherein when the input stroke can be judged to be shorter than other strokes, the stroke is excluded from recognition targets. 6. A detection means for detecting a trajectory of a pen written on a display-integrated tablet, a processing means for creating stroke data recorded in time series, and a recognition processing using the stroke data, In an online handwritten figure recognition device comprising a display means for displaying the processing result by the processing means on the tablet and a selection means for selecting the recognition result, when the recognition result is displayed on the screen, the display screen is displayed with a pen. A handwritten figure recognition device having a determination means for making a determination between a request to display the next candidate and a next figure input by tapping. 7. An online handwritten figure recognition device having a function of selecting a candidate, when the tablet on which the recognition result is displayed is tapped with a pen, the position tapped on the tablet and the position of the displayed figure. A handwritten figure recognition device having selection means for selecting processing such as selection of candidates and cancellation of input according to the relationship with. 8. An online handwritten figure recognition device having a function of selecting a candidate, when a tablet on which a recognition result is displayed is tapped with a pen, a tapped position on the tablet and a displayed figure are formed. A handwritten figure recognition device having selection means for selecting a stroke that has been erroneously recognized depending on the relationship with the stroke position to be selected, and selecting processing such as selection of candidates for only that portion. 9. A detection means for detecting a trajectory of a pen written on a tablet, a processing means for creating stroke data recorded in time series, and a recognition processing using the stroke data, and the processing means. An online handwritten figure recognition device comprising a display means for displaying a processing result on a tablet, further comprising a limiting means for limiting a category of a figure to be output as a recognition result depending on the number of strokes input to the tablet. Handwritten figure recognition device.