JPH05143738A - On-line graphic recognizing device - Google Patents

Abstract

(57) [Abstract] [Purpose] Online figure recognition that recognizes figures by extracting feature points by the ratio of the angle between two lines or stroke length and adding one rule in the order in which one stroke can be written. Provide the device. [Structure] A handwriting input acquisition means for acquiring time-series coordinate data of a line figure input by handwriting, a memory for temporarily storing the coordinate data, and the acquired coordinate data are converted into a required data amount for normalization. In an online figure recognition device equipped with a data pre-processing means, a dictionary 5 for recognizing data, and a figure recognizing means for comparing and recognizing an input figure and a dictionary pattern figure, three consecutive data points are connected by a straight line. A feature point extracting means is provided for extracting a feature point of one stroke and a feature point which is a connection point of a plurality of strokes drawn in a row only by the angle between the line segments created in (1).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an on-line graphic recognition device used in a document editing device or the like.

[0002]

2. Description of the Related Art As a conventional online graphic recognition device, there are various devices described below. As the first example, Japanese Patent Laid-Open No. 55-13
Some of them are disclosed in Japanese Patent No. 8172, entitled "Online Character Feature Extraction Device". This is to extract a time series of four types of characteristic amounts of stroke linearity, relative length ratio, direction angle from the start point to the end point, and further, the change in angle of stroke bending as the character characteristic amount. It is possible to detect even a slight difference in stroke and correctly identify similar characters.

Further, as a second conventional example thereof, Japanese Patent Laid-Open No.
There is one disclosed in the title of "Online handwritten figure recognition device" in Japanese Patent Publication 1-28180. This is to recognize the handwriting handwriting by dividing it into a plurality of line segments from the feature quantity based on the line connecting any two sample points of the handwriting figure, and to recognize the handwriting figure online automatically without depending on the figure input state. Therefore, the man-machine characteristic can be further improved as compared with the conventional case.

Further, as a third conventional example thereof, there is one disclosed in Japanese Patent Application Laid-Open No. 61-29980 under the title "Online handwritten figure recognition device". This is because the handwriting handwriting is divided into a plurality of line segments by the feature amount obtained from a straight line connecting arbitrary two points of the handwriting handwriting sample points, and the recognition is performed without depending on the handwriting of the figure input time. By doing so, it is possible to greatly relax the restrictions on the number of strokes, stroke order, rotation, and division when inputting a handwritten figure.

Further, as a fourth conventional example thereof, there is one disclosed in Japanese Patent Application Laid-Open No. 62-82486 under the title "Online handwritten figure recognition device".
This is not only the angle change amount of two adjacent line segments,
By taking into account the angle change of two or more line segments and comparing it with the angle change of the line segment sequence of the dictionary, it is possible to absorb the ambiguity of the handwriting and enable the recognition. Even if the figure is greatly deformed or is in a rotated state, it can always be recognized correctly.

[0006]

As can be seen from the above-mentioned first conventional example, there are four types of means for extracting character feature amounts: stroke straightness, relative length ratio, direction angle, and angle change. Although it is performed by detecting, there is a problem that such an extraction method takes time and becomes complicated.

Further, in the recognition method by pattern matching, the correspondence of the feature points takes time because it calculates the distances of all the points with respect to one input point in the case of a character and without limiting the stroke order. Therefore, there is an attempt to improve the processing time by reducing the number of distance calculations by predicting the stroke order variation to some extent, but the configuration of the device may be complicated. In addition, in many cases, the features are similar to each other, and it may not be possible to sufficiently recognize the distance calculation. In such a case, there is a method in which the order of the feature points is written by one stroke, but there are a plurality of writing methods depending on the figure, which is troublesome.

[0008]

According to a first aspect of the present invention, there is provided a handwriting input obtaining means for obtaining time-series coordinate data of a line figure input by handwriting, and a memory for temporarily storing the coordinate data. An online graphic provided with a data preprocessing means for converting the acquired coordinate data into a required data amount and normalizing it, a dictionary for recognizing the data, and a graphic recognizing means for comparing and recognizing the input graphic and the dictionary pattern graphic. In the recognition device, a feature point of one stroke and a feature point which is a connection point of a plurality of strokes drawn in a row are determined only by the angle between line segments formed by connecting three consecutive data points with a straight line. A feature point extracting means for extracting is provided.

According to a second aspect of the present invention, handwriting input acquisition means for acquiring time-series coordinate data of a line figure input by handwriting, a memory for temporarily storing the coordinate data, and the acquired coordinate data are stored in the memory. An online graphic recognition apparatus comprising a data preprocessing means for converting the data into a required data amount and normalizing it, a dictionary for recognizing data, and a graphic recognition means for comparing and recognizing an input graphic and a dictionary pattern graphic. By dividing the stroke by the ratio of the length of the stroke drawn in Fig. 1 and the length connecting the start point and the end point of the stroke, the characteristic point of one stroke and the connecting point of a plurality of strokes drawn in succession A feature point extraction means is provided for extracting the feature points identified as

According to a third aspect of the present invention, in the first and second aspects of the present invention, feature point information is obtained by the feature point of one stroke and the feature point which is a connecting point of a plurality of strokes drawn in a row. A feature that associates the feature points of the figure input by handwriting with the figure recognized by the dictionary by creating a list and adding the order of drawing with one stroke and one rule to the information of each feature point in this feature point information list Point matching means was provided.

[0011]

According to the invention described in claim 1, since the feature points are extracted only by the angle between the two line segments when the figure is approximated by the polygonal line, the feature points can be sufficiently extracted and the processing is performed. Can be reduced.

According to the second aspect of the present invention, it is determined whether or not the line is a straight line based on the ratio of the length of the stroke and the length of the straight line connecting the start point and the end point of the stroke. Further, by adding the other data point to each of the temporarily separated strokes and determining whether or not the stroke is a straight line, it becomes possible to extract the characteristic point with high accuracy.

According to the third aspect of the present invention, the correspondence between the feature points of the input graphic and the dictionary data to be matched is classified by the combined information, and then one rule is added in the order of one stroke writing. It is possible to obtain a result in which the matching result is remarkably different even for the similar figure of points.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the invention according to claim 1 is shown in FIGS.
It will be explained based on. First, FIG. 1 shows the overall configuration of an online figure recognition device. This apparatus includes a graphic input unit 1, a data acquisition unit 2 as a handwriting input acquisition unit for acquiring time-series coordinate data of a line graphic handwritten and input from the graphic input unit 1, and the coordinate data. A memory 3 for temporary storage, a data preprocessing unit 4 as a data preprocessing unit for converting the acquired coordinate data into a required data amount and normalizing it, a recognition dictionary unit 5 as a dictionary for recognizing data, and an input A graphic recognition unit 6 as a graphic recognition means for comparing and recognizing the graphic and the dictionary pattern graphic, and a display unit 7 for displaying the graphic on the screen are provided.

In the present embodiment, in the online figure recognition apparatus having the functions of the respective parts as described above, the characteristic point of one stroke, and the characteristic point of one stroke are determined only by the angle between the line segments formed by connecting three consecutive points of data with a straight line. A feature point extracting means for extracting a feature point which is a connecting point of a plurality of strokes drawn in a row is provided. A detailed description of this feature point extraction means will be given later.

In such a configuration, the graphic input unit 1
Includes a tablet 8 for inputting handwritten figures and a stylus pen 9. The figure drawn on the tablet 8 is sampled at regular intervals by the data acquisition unit 2 and stored in the memory 3 as coordinate points. The data preprocessing unit 4 reads the coordinate point sequence data of the input graphic stored in the memory 3, performs preprocessing such as data noise removal and smoothing, data sampling, and normalization processing, and the graphic recognition unit 6 performs the preprocessing. The recognition is performed by referring to the recognition dictionary unit 5, and the recognition result obtained by this is displayed on the display unit 7.

Next, the flow of recognition processing will be described based on the flow of FIG. First, in (a), the tablet 8
Draw a figure with the stylus pen 9 and input. The graphic data input by this is taken in as coordinate points in time series. At this time, stroke (stylus pen 9
The up and down information of the stylus pen 9 when drawing a continuous line from the time when the tablet is lowered onto the tablet 8 to the time when the tablet is raised is also captured. In (b), the coordinate data of the figure input by handwriting in (a) is removed from a singular point due to noise or camera shake, a point with the same coordinate when the stylus pen 9 is stopped at one point, and the like. In (c), data is sampled in order to limit the data to necessary points.
In (d), normalization processing is performed. The normalization method is
For example, it can be performed by a normalization method or the like in which the total stroke length is constant for any figure. (E)
Then, the characteristic points are extracted, and the strokes drawn by drawing the broken line or the curved line and the straight line in (f) are separated from the characteristic point information. In (g), rearrangement processing of feature points is performed in order to associate the feature points for recognition. In (h), the strokes are combined to remove unnecessary feature points. In (i), classification is performed according to each feature point to classify recognition target graphics. In (j), (i)
The information of the recognition target figure classified by is read from the dictionary and the input figure is recognized. In (k), the recognition result is output.

Next, a method of extracting feature points using the feature point extraction means will be described. It is assumed that the input graphic is represented by time-series coordinate points as shown in FIG. 3 and is drawn in the arrow direction as the first stroke, the second stroke, and the third stroke. As a result, the input figure is
As shown in (a) to (c), straight lines (horizontal lines, vertical lines, diagonal lines), (d), curves as shown in (e),
Drawing with connecting lines (combining straight lines and curves) as shown in (f) to (h), strokes such as circles and ellipses as shown in (i) and (j), and strokes made by combining these. Be done. The combination is a combination of straight lines and straight lines or curved lines and straight lines. These connected portions are separated as "feature points", and the input figure is represented only by the straight lines and curved lines. In this case, in the stroke, the starting point, the ending point, and the intermediate point become the characteristic points, and in the connecting line, the separated starting point or the ending point becomes the characteristic point.

Therefore, a method of extracting the characteristic points will be described with reference to FIGS. 5 to 7 show a method of extracting feature points by the angle of a polygonal line connecting three consecutive points of input data. First, in FIG.
(A) is a stroke figure which connects continuous input data with a straight line and is represented by a broken line. (B) is a stroke figure after preprocessing the input data and sampling the input data. (C) is an enlarged view of the inside of the broken line in (b). In this case, consecutive points Pi-
The angle θi of the polygonal line formed by connecting 1, Pi and Pi + 1 is P
The line connecting i-1 to Pi is the vector x, and Pi to Pi + 1
If the line connecting the two is a vector y, then θi = cos ~ ¹ (x · y / | x | · | y |) can be obtained by the equation (1).

Then, the angle θ is obtained for all points of the stroke, and when the angle θ is smaller than a predetermined threshold value α, that point is set as a feature point. In this case, the start point and the end point of the stroke are also feature points. In (d), strokes are sequentially separated by straight lines between characteristic points. A stroke in which the angle θ obtained by this method is all larger than the threshold value α is not a combined stroke (straight line, curved line) or a combined stroke feature point (combined portion) cannot be obtained.
As shown in FIG. 6, when the input figure is a combination stroke of a straight line and a curved line, the stroke connection portion may not be obtained because the angle is relatively large. Therefore, when the angle θi of the broken line and the angle θi-1 of the previous broken line shown in FIG. 6 are both smaller than the threshold value β, the point at the angle θi-1 is set as the feature point.
However, the angle threshold β has a relationship of β> α. By obtaining the points that satisfy these two conditions, it is possible to obtain the feature that is the separated portion of the stroke. In the case of the circle shown in FIG. 7, all the angles of the polygonal lines are in the above-mentioned conditions (θ ≦ α, θ
i-1, θi ≤ β) is not satisfied, and the angles that do not satisfy such a condition because the angle is comparatively large are classified and the recognition processing is performed separately for the graphics whose start point and end point are near each other.

Next, the flow of processing for extracting the characteristic points is shown in FIG.
It will be explained based on. First, in (a), as shown in FIG. 5 (b) described above, each θi of the broken line angles obtained by sequentially connecting three consecutive points is obtained. In (b), when the obtained angle θi is smaller than the threshold value α, the point is extracted as a feature point as shown in (c), and then in (d),
Judge whether all data has been processed and
If "s", the process is terminated, and if "No", the process returns to (a). Further, in (b), when the obtained angle θi is larger than the angle threshold α, the process proceeds to (e). In this (e), if both the previous angle θi−1 and the angle θi are smaller than the angle threshold β, the previous point is set as the feature point as shown in (c). Thus, if the condition of (b) is satisfied, the separation between straight lines can be performed, and if the condition of (e) is satisfied, the separation between a straight line and a curve can be performed, and (b), (e If both conditions are not satisfied, it can be regarded as a circle or an ellipse.

As described above, a figure or the like is often drawn by combining strokes unlike a character. By separating strokes, a figure is represented by a minimum unit stroke, and the separated points are used as feature points. The feature point extraction method can be extracted by the angle of the line segment approximated by the broken line. Therefore, by extracting features using only the angles of two line segments that are approximated by polygonal lines in this way, it is possible to reduce the processing time compared to the conventional extraction method, and in the case of circles, the angle threshold to be set is set. Depending on the value of, the extraction can be performed by not extracting the feature point in the middle of the stroke of the drawn circle.

Next, an embodiment of the invention described in claim 2 will be described with reference to FIGS. In the present embodiment, the method of the feature point extracting means for extracting the feature points is changed in the above-described online figure recognition apparatus of FIG. That is, the feature point extraction means here divides the stroke by the ratio of the length of the stroke drawn continuously and the length connecting the start point and the end point of the stroke, thereby
A feature point of a stroke and a feature point which is a connecting point of a plurality of strokes drawn in a row are extracted.

Therefore, a specific method for obtaining the characteristic point by the ratio of the length of the straight line connecting the starting point and the ending point of the stroke to the length of the stroke will now be described. FIG. 9 shows a method for determining whether one stroke is a curve. A stroke distance L is a value obtained by summing all the distances of the straight lines between the points when the input continuous points of the strokes are connected by straight lines and approximated by a broken line. Start point (point where pen is down) and end point (point where pen is up) of one stroke
Let Ls be the distance connecting the two with a straight line. When the ratio of these two distances is larger than the threshold value γ (L / Ls ≧ γ), the stroke is determined to be a curve or a connecting line. When it is smaller than the threshold value γ (L / Ls <γ), it is determined to be a straight line (hereinafter, this process is referred to as a curve determination process).

A method of extracting a joint from a stroke determined to be a curve or a joint line will be described with reference to FIG.
In (a), when it is judged as a curved line or a connecting line as a result of judging whether it is a straight line by the above-mentioned method, (b)
In, a position that is in the middle of the total stroke distance is separated as a tentative stroke with the intermediate point. In (c), the separated separation strokes 10 and 11 are separated.
The curve determination process is performed again. When the result of performing the curve determination process again is a separation stroke 10 that is not a straight line, in (d), the curve determination is performed by adding several points (points within the broken line) of the other separation stroke 11 to the separation stroke 10. I do. Such separation processing is performed until the stroke length reaches a predetermined value λ. However, the specified length λ
Is a length that allows the determination of straight lines and curved lines. In (e), when the separated results are straight lines and curved lines, (d)
At (f), it is determined which stroke is to be used as the feature data added from the other stroke, and the stroke separation is terminated at (f).

Next, the flow of processing for extracting the characteristic points is shown in FIG.
It will be described based on 1. In (a), the curvature of one stroke is calculated. In (b), curvature (L / Ls)
If it is smaller than the curvature γ defined by, it is determined to be a straight line. When the curvature is larger than γ, the stroke length is calculated and the intermediate point is obtained in (c). In (d), the stroke is temporarily separated into two strokes at the intermediate point. The curvatures of the temporarily separated strokes are calculated again in (e).
In (f), curve judgment (L / Ls ≧ γ) is performed,
If “No”, it is determined to be a straight line, and the procedure proceeds to (i) to determine whether there is the remaining data (stroke).
If not, the separation ends. At this time, if there is remaining data, the process returns to (c). Then, in (f) again,
When the curve determination is “Yes”, the stroke length is determined in (g), and it is determined whether the stroke length Li is a length (Li ≧ λ) at which a straight line and a curve can be determined. When the stroke length Li is smaller than λ,
It is determined to be a curve as shown in (h), and if the stroke length Li is larger than λ in (g), the process returns to (c). The points separated by such a method are set as characteristic points.

As described above, in the method of separating the strokes drawn by being connected, the connecting point may be ambiguous, and the length of the stroke and the length of the straight line connecting the start point and the end point of the drawn stroke. The separation point can be accurately extracted by performing the temporary separation at a ratio with the above and repeating such processing.

Next, an embodiment of the invention described in claim 3 will be described with reference to FIGS. In this embodiment, a method for recognizing a graphic by the feature point associating means using the feature points extracted by the feature point extracting means in the above-described two embodiments will be described. The feature point associating means is
A feature point information list is created from the feature points of one stroke and the feature points that are the connection points of a plurality of strokes drawn in a row, and the information of each feature point in this feature point information list is drawn with one stroke. By adding the order and one rule, the handwritten input figure and the figure recognized by the dictionary are made to correspond to each other, and the figure is recognized. Hereinafter, the feature point associating means will be sequentially described with a specific example.

As shown in FIG. 12, the stroke of the figure from which the characteristic points are extracted is determined as a straight line or a curved line, and the straight line is represented by two points (bold circles) of a start point and an end point (hereinafter referred to as end points).
The curve is represented by three points (thick circles), an end point and an intermediate point. The midpoint of the curve is the input data point that is closer to the point at the intermediate distance position of the stroke length. FIG. 13 shows an example of a figure in which the input figure is represented by end points and intermediate points. In this case, if the end points and the intermediate points, which are the characteristic points, are separated from one stroke, the connection relation information with the already separated strokes is already obtained. For example, when a graphic is not a target, the connecting relation can be easily obtained by setting the end point within a distance defined from a certain end point as the end point of the connecting stroke. FIG. 14 shows a state of the feature point information list 12 in which the feature point information is created as the number of end points, the intermediate points, and the connection relation of each end point. This feature point information list 1
2, it is possible to determine the connecting order of the end points. That is, for example, many symbolic figures such as flowcharts are closed figures, and the end points of each stroke are combined with other strokes (the strokes here are strokes represented by the minimum unit of a straight line and a curved line). Because there are many
By determining the order of the feature points, it is possible to correspond the feature points when matching the patterns at the time of recognition.
FIG. 15 illustrates the rearrangement of the feature points based on the direction and the connection point information, which makes it possible to easily correspond the points of the pattern matching for recognizing the input graphic. In FIG. 15, the flowchart symbol ""
This is an example in the case of expressing "matching", and shows a figure in which triangles are combined with their upper and lower vertices aligned. Now, assume that the input figure is input as shown in (a). Indicates the writing order, and the arrow indicates the direction of the written stroke.The feature points of this input figure are obtained by the feature point extraction method described above, and the feature point information list is obtained.
It is possible to change the stroke direction (order) based on the combined information of each feature point.

Next, a method of changing the stroke direction will be described. Now, let us say that the characteristic point of the figure is the starting point S and the ending point E of each stroke. Only the characteristic points are extracted from the input figure data, and the order is as follows.

(1S, 1E, 2S, 2E, 3S, 3E,
4S, 4E, 5S, 5E, 6S, 6E) Note that each point end is a coordinate value.

In this case, when the first point of the rearrangement data can be written by one stroke, the point at which one stroke can be written is set as the first point. By thus setting the point where one-stroke writing is possible as the rearrangement start point, the difference in the degree of similarity of the matching result at the time of recognition becomes remarkable. Then, when FIGS. 16A and 16B are drawn, the points obtained by the above-described feature point extraction method are the same feature points as in FIGS. 16A and 16B. In this way, by associating the data in the order of one stroke and by associating the points in the order of the numbers shown in FIG. 16, the value of the recognition result by the distance calculation between the patterns at the time of recognition becomes significantly different. In addition, in addition to this, when there are a plurality of points that are the starting points of one-stroke writing, or when the leaves are such that one-stroke writing cannot be performed, a certain "rule" is set.

As an example thereof, the characteristic point of the graphic is at the upper left, and the point (start point or end point) of the pair of strokes having that point is the starting point of the rearrangement as the point of the downward stroke. The order of the feature points of the figure in FIG. 15A is (b)
, A, b → c in the order of ,,,,,
Are rearranged in the order of d → e → f, and (3S, 3E, 2
E, 2S, 1E, 1S, 4S, 4E, 5S, 5E, 6
E, 6S).

Then, according to the feature point information list, 3E,
Since 2E, 4S, and 6S are connected, (3S,
3E, 2S, 1E, 4S, 4E, 5E, 6S).

If the strokes can be combined in one stroke, the strokes are combined to form the figure shown in FIG. 15C, and the order of the characteristic points is (3S, 6E, 5S, 1).
E).

The feature points thus obtained are classified by the number of feature points and the number of intermediate points shown in the flow of FIG.
By performing pattern matching between the feature points of the input figure and the feature points of the dictionary pattern, the figure can be recognized.

As described above, by adding one rule to the order in which one stroke can be written, it is possible to limit the number of strokes to one and to deal with figures that cannot be stroked. It is possible to obtain a result in which the matching result is remarkably different even in the similar figure of the feature points.

[0038]

According to the invention described in claim 1, handwriting input acquisition means for acquiring time-series coordinate data of a line figure input by handwriting, a memory for temporarily storing the coordinate data, and the acquired coordinates. In an online figure recognition device provided with a data preprocessing means for converting data into a required data amount and normalizing it, a dictionary for recognizing data, and a figure recognizing means for comparing and recognizing an input figure and a dictionary pattern figure, A feature point that extracts a feature point of one stroke and a feature point that is a connection point of a plurality of strokes drawn in a row only by the angle between line segments created by connecting three consecutive points of data with a straight line Since the extraction means is provided, it becomes possible to extract the feature points only by the angle between the two line segments when the figure is approximated by the polygonal line, whereby the feature points can be sufficiently extracted. It reduces the processing time in which it is possible to work efficiently.

According to a second aspect of the present invention, handwriting input acquisition means for acquiring time-series coordinate data of a line figure input by handwriting, a memory for temporarily storing the coordinate data, and the acquired coordinate data are stored in the memory. An online graphic recognition apparatus comprising a data preprocessing means for converting the data into a required data amount and normalizing it, a dictionary for recognizing data, and a graphic recognition means for comparing and recognizing an input graphic and a dictionary pattern graphic. By dividing the stroke according to the ratio of the length of the stroke drawn to the start point and the end point of the stroke, the characteristic point of one stroke and the connecting point of a plurality of strokes drawn in succession can be obtained. Since the feature point extracting means for extracting the extracted feature points is provided, the stroke length and the
Determine whether it is a straight line by the ratio of the length of the straight line connecting the start point and the end point of the stroke, if it is not a straight line, perform tentative separation, add the other data point to each tentatively separated stroke, You can judge whether it is a straight line,
This makes it possible to extract characteristic points with high accuracy, reduce the processing time, and improve work efficiency.

According to a third aspect of the present invention, in the first and second aspects of the present invention, feature point information is obtained by the feature point of one stroke and the feature point which is a connecting point of a plurality of strokes drawn in a row. A feature that associates the feature points of the figure input by handwriting with the figure recognized by the dictionary by creating a list and adding the order of drawing with one stroke and one rule to the information of each feature point in this feature point information list Since the point associating means is provided, the correspondence between the feature points of the input figure and the dictionary data to be matched is classified by the combined information, and then one rule is added in the order of one stroke to add the feature points. It is possible to obtain a result in which the matching result is significantly different even for similar figures.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an online graphic recognition apparatus which is an embodiment of the invention described in claim 1.

FIG. 2 is a flowchart showing a flow of recognition processing.

FIG. 3 is an analysis explanatory diagram showing a state of an input figure represented by time-series coordinate points.

FIG. 4 is a waveform chart showing various strokes classified.

FIG. 5 is an analysis explanatory diagram showing how strokes are separated at the time of feature point extraction.

FIG. 6 is an analysis explanatory diagram showing a state in which the input figure is a combined stroke of a straight line and a curved line.

FIG. 7 is an analysis explanatory diagram showing a situation where an input figure is a circle.

FIG. 8 is a flowchart showing a flow of feature point extraction.

FIG. 9 is an analysis explanatory diagram for determining whether or not one stroke is a curve which is an embodiment of the invention described in claim 2;

FIG. 10 is an analysis explanatory diagram showing a feature point extraction method based on curvature.

FIG. 11 is a flowchart showing a flow of feature point extraction.

FIG. 12 is an analysis explanatory diagram showing a state of point approximation.

FIG. 13 is an analysis explanatory diagram showing a state of an approximate figure.

FIG. 14 is a classification configuration diagram showing a configuration of a feature point information list.

FIG. 15 is an analysis explanatory diagram showing a state of rearranging feature points based on direction and connection point information.

FIG. 16 is an analysis explanatory diagram showing a procedure of one-stroke writing.

FIG. 17 is a flowchart showing the flow of processing of feature point classification and pattern matching.

[Explanation of symbols]

 2 Handwriting input acquisition means 3 Memory 4 Data preprocessing means 5 Dictionary 6 Figure recognition means

Claims (3)

[Claims] 1. A handwriting input acquisition unit for acquiring time-series coordinate data of a line figure input by handwriting, a memory for temporarily storing the coordinate data, and converting the acquired coordinate data into a required data amount. In the online figure recognition device provided with the data pre-processing means for normalizing the data, the dictionary for recognizing the data, and the figure recognizing means for comparing and recognizing the input figure and the dictionary pattern figure, three consecutive points of the data are straight lines. The feature point extracting means is provided for extracting a feature point of one stroke and a feature point which is a connection point of a plurality of strokes drawn in a row only by the angle between the line segments formed by connecting with each other. An online graphic recognition device. 2. A handwriting input acquisition means for acquiring time-series coordinate data of a handwritten input line figure, a memory for temporarily storing the coordinate data, and converting the acquired coordinate data into a required data amount. In the online figure recognition device provided with the data pre-processing means for normalizing, the dictionary for recognizing the data, and the figure recognizing means for comparing and recognizing the input figure and the dictionary pattern figure, By dividing the stroke according to the ratio of the length and the length connecting the start point and the end point of the stroke, the feature point of one stroke and the feature point which is a connecting point of a plurality of strokes drawn in a row are determined. An on-line figure recognition device characterized in that a feature point extracting means for extracting is provided. 3. A feature point information list is created from the feature points of one stroke and the feature points that are the connection points of a plurality of strokes drawn in a row, and the information of each feature point of this feature point information list is created. A feature point associating unit that associates a feature point between a figure input by handwriting and a figure recognized by a dictionary by adding an order of drawing with one stroke and one rule to the character string is provided. 2. The online pattern recognition device described in 2.