JPH0443315B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to figure recognition, and particularly to a figure recognition preprocessing device suitable for online handwritten figure recognition.

[Background of the invention]

Conventional online handwritten figure recognition, like character recognition, had severe restrictions on stroke order, number of strokes, tilt, etc., and had the disadvantage of poor man-machine performance.

[Purpose of the invention]

An object of the present invention is to provide an online handwritten figure device and a recognition method with good operability.

[Summary of the invention]

A feature of the present invention is that a figure recognition device uses connection information indicating the presence or absence of a connection between the end point of each line segment and the end point of another line segment of a plurality of line segments obtained from strokes input sequentially by hand. A line segment connection detection means is provided which determines a line segment group in which all line segments are in a connected state to be a closed figure, and the line segment group determined to be a closed figure by the line segment connection detection means is cut out, and the cut out line segment group and By comparing shapes with dictionary shapes and performing shape recognition, it is possible to automatically cut out groups of line segments that make up closed shapes from continuously input handwritten strokes, improving operability. be.

Furthermore, a feature of the present invention is that strokes constituting a figure are input by hand, and when the input strokes constitute a closed figure, the closed figure is cut out, and the cut out closed figure and the dictionary figure are compared. By using the corresponding dictionary figure as the recognition result, it is possible to automatically cut out a group of line segments constituting a closed figure from the strokes that are continuously input by hand, thereby improving operability. It is in.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIGS. 1 to 4.

FIGS. 1a and 1b show block diagrams of an online handwritten figure recognition apparatus according to the present invention. The coordinate sampling device 1 is a device such as a tablet that samples the current coordinates at appropriate time intervals. When sample coordinates are input from here, a direction code is generated from the two sample points in direction code b detection 100. Note that the sample coordinates include information on the beginning and end of the stroke, and if the stroke ends,
Direction code detection 100 processes line segment connection detection 20
Pass to 0. Line segment connection detection 200 determines whether two strokes are connected, creates a connection list, and when all connection lists are in a connected state, passes processing to relative direction code generation 300. if,
If there are any unconnected points in the connection list, the process returns to direction code detection 100 to check the next sample point. The recognition unit 6 recognizes the input figure by matching it with a dictionary using the relative direction code obtained in this way, and passes the matched figure information to the figure code information generation 8. The function of each part will be explained in detail below.

The flow of the process of direction code detection 100 shown in FIG. 2 is shown. At 101, processing is started by a stroke start signal from the coordinate sampling device at the previous stage.
Here, we will use a doublet as an example of the coordinate sampling means, and hereafter the start of the stroke will be referred to as pen down, the end as pen up, and the sampled coordinates will be referred to as pen coordinates. At step 102, the coordinates of sample point 1 when the pen is down are taken in and stored at step 103 in a preset storage circuit. At 104, the next sample point 2 is fetched, and at 105, a direction code is calculated based on the previously stored sample point 1 and the newly fetched point 2. Here, the direction code is defined in advance as
This is the number given to each direction when 360 degrees is divided into L parts. Next, it is determined in step 106 whether sample point 2 has pen-up information, and if no, it is determined in step 107 whether the previous direction code and the current direction code are the same.
If the codes are the same, the process returns to 104 to perform the next coordinate sample. If the direction codes are different, then in step 108, sample point 2 is stored next to sample point 1 in the storage area (109). Here at 108,
Pen coordinate storage X2 means that the stroke is divided into multiple line segments due to changes in the direction code.
This is because the coordinates of the end point of the previous line segment are newly stored as the coordinates of the start point of the next line segment. After this, the process returns to step 104 to acquire the next sample point. 10
If a pen-up is detected in step 6, it is determined in step 110 whether the total number of strokes is smaller than the preset maximum number of strokes N, and if it is, 1
In step 11, the coordinates of the sample point 2 are stored, and the process jumps to the line segment connection detection routine 200. On the other hand, if the number of strokes becomes greater than N, the routine jumps to a routine 400 that outputs an unrecognized error message.

FIG. 3 shows the flow of processing 200 for detecting connection of line segments. In step 201, the numbers of unconnected line segments are output from the connection flag list on the storage device. Here, the connection flag list is the number of line segments x 2 x
It consists of 1 bit, and "1" is stored when both ends of each line segment are connected to each other line segment, and "0" is stored when the ends are not connected to any line segment. At step 202, the coordinates corresponding to the unconnected coordinate numbers output as described above are read. 20
3 reads the other coordinate numbers, and 204 reads these 2
Determine whether the points are the same point. If they are the same point, the next stored coordinate number is read out as the other coordinate. At step 205, the coordinates are taken in. 20
At 6, the exclusive OR of the two points is taken, at 207 the lower n bits set in advance are masked, and at 208 it is determined whether the result is 0 or not. Here, the lower n
The reason why the bits are masked is to allow for slight deviations in connections and intersections due to handwritten input. If the result in step 208 is 0, it is assumed that the two points are connected, and in step 209, "1" is written to the corresponding address in the connection flag list. At 210, the entire flag list is checked, and when all flags are "1", it is determined that the input figure is closed, and a relative direction code of 300 is jumped to the generating routine. If all the flags are not "1", the process jumps to 201 to find the next unconnected coordinate number. On the other hand, if it is determined in step 208 that the two points do not match, it is determined in step 212 whether the connection check for all sample points has been completed, and if it has not been completed, the process returns to step 211, and if completed, a new Return to step 100 to input a line segment.

FIG. 4 shows the process flow of relative direction code generation 300. In step 311, the first direction code 1 of the line segment stored in advance in the process 100 is read out.
The next direction code 2 is read in 302, and if direction code 2 is smaller than direction code 1 in 303, 30 is read.
After adding the code number L to the direction code 2 in 6, 30
4, subtract the direction code 1, and store the result as the new direction code 2 value. Also, in 305, it is determined whether the omnidirectional code processing is finished,
If not completed, the process returns to step 302 to read the next direction code. On the other hand, if the process has ended, the process proceeds to recognition processing.

FIG. 5 shows the memory arrangement of the storage device when implementing this method. Broken lines in the figure indicate the relationship between the direction code corresponding to each coordinate and the connection flag list.

FIGS. 6a and 6b show an example of a graphic input and how it is converted into the same relative direction code when the stroke order is the same, regardless of the direction code.

〔Effect of the invention〕

According to the present invention, the timing of cutting out a recognition target in figure recognition is automatically detected.
The user can concentrate on inputting information, and operability is improved.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the online handwritten figure recognition device of the present invention, Fig. 2 is a flowchart of direction code detection processing, Fig. 3 is a flowchart of line segment connection detection processing, and Fig. 4 is a relative direction code A flowchart of the generation process, FIG. 5 is a memory map of the storage device, and FIGS. 6a and 6b are diagrams of graphic input examples. 1... Coordinate sample device, 100... Direction code detection device, 200... Line segment connection detection device, 30
0...Relative direction code generator, 6...Graphic recognition device, 8...Graphic generator.

Claims (1)

[Scope of Claims] 1. An online handwritten figure recognition device that recognizes a figure made up of at least one stroke input by handwriting by comparing it with a dictionary figure, comprising: , line segment connection detection means for determining a group of line segments in which all line segments are connected to be a closed figure using connection information indicating whether or not the end point of each line segment is connected to the end point of another line segment. An online handwritten figure recognition device, comprising: cutting out a group of line segments determined to be closed figures by the line segment connection detection means, and performing figure recognition by comparing the cut out group of line segments with the dictionary figure. 2. An online handwritten figure recognition method in which a figure consisting of at least one stroke is input by hand, and the input handwritten figure is compared and recognized with a dictionary figure. An online handwritten figure recognition method, characterized in that when strokes constitute a closed figure, the closed figure is cut out, the cut out closed figure is compared with a dictionary figure, and the corresponding dictionary figure is taken as a recognition result. 3. The online handwritten figure recognition method according to claim 2, characterized in that when all line segments constituting the input stroke are connected, the figure is cut out as a closed figure.