JPH03160581A - Pattern recognition device - Google Patents

Abstract

PURPOSE:To stably segment a pattern by judging the termination of the input of one pattern by means of a position relation between an input window and an inputted stroke so as to segment the pattern. CONSTITUTION:An input part 2 composed of a tablet and an input pen has plural input windows inputting pattern information, and a character segment part 6 which judges whether the input of the pattern such as one character terminates or not by the position relation between the inputted stroke and the input window and segments the character is provided. Then, the termination of the input of one character is judged, and strokes until the previous one is set to be one character. Then, a recognition part 7 recognizes the character and a recognized result is displayed in a display part 9. Thus, the the pattern can stably be segmented in spite of the habit and errors of a user.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a pattern recognition device that recognizes patterns such as characters and figures. [Prior Art] In a conventional pattern recognition device, for example, an online character recognition device, the end of input is detected by inputting an instruction to a portion indicating the end of input, or the like.

However, this method hinders the user's thinking. Therefore, we created multiple input windows and cut out a pattern based on the termination condition of one pattern when the input moves from the input window currently being input to another input window, that is, when the input of another character or figure begins. There is a way to start recognition. Furthermore, in OCR, there is a method of cutting out and recognizing patterns by providing multiple input windows for cutting out patterns. [Problem to be solved by the invention] However, in the above conventional example, the user's habits, carelessness,
Or, if part of the pattern protrudes from the input window during input and is input into another input window due to reasons such as the input window is small or the pattern is complex, the pattern cannot be extracted correctly and recognition is impossible. There was a problem that the user would have to input the information again due to incorrect recognition. The present invention eliminates the above-mentioned drawbacks of the conventional technology, stably cuts out patterns regardless of user's habits or mistakes, and has a small input window. To provide a pattern recognition device that increases the recognition rate of pattern recognition regardless of the complexity of the pattern and improves operability.

[Means for Solving the Problem] In order to solve this problem, the pattern recognition device of the present invention is a pattern recognition device that recognizes patterns such as characters and figures, and includes a plurality of The apparatus includes an input means having an input window, and a pattern cutting means for determining the end of input of one pattern and cutting out the pattern based on the positional relationship between the input window and the input stroke. Here, when the stroke currently being input belongs to a plurality of input windows, the pattern cutting means extracts one pattern based on the positional relationship between the input window to which the previous stroke belonged and the stroke currently being input. Determine whether the input is finished. Moreover, the pattern recognition device recognizes patterns of characters, figures, etc., and includes an input means having multiple input windows for inputting pattern information, and when the input pattern belongs to multiple input windows. The apparatus includes pattern cutting means for determining the input window to which the input pattern belongs based on the positional relationship between the window and the input pattern and cutting out the pattern. [Operation] In this configuration, by determining whether the input of one pattern ends or belongs to the input window based on the positional relationship between the input window and the input stroke and cutting out the pattern, stable pattern cutting is achieved. Make it possible. [Embodiments] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this embodiment, characters will be explained as an example of a pattern.

FIG. 1 is a block diagram showing the configuration of the character recognition device of this embodiment. 1 shows the character recognition unit of this embodiment. 2
is an input unit consisting of a tablet and an input pen, and the up/down information of the pen tip is sent every 0.02 seconds. X
Coordinate. Detect the y-coordinate. 3 is a central processing unit (hereinafter referred to as CPU) for arithmetic processing that controls the entire device, and 4 is a processing program for CPU 3, that is, a control program. ROM 4 stores an error processing program and a program according to the flowchart shown in FIG. 5, which will be described later.
This RAM is used as a work area for various programs stored in the computer and as a temporary save area during error processing. A midpoint memory 5b that stores the midpoint found from the stored start and end points of the stroke, and a feature that stores feature points such as the start point, end point, and cusp extracted from the coordinates of the input point stored in the coordinate memory 5a. Memory 5 cs It has a display memory 5d for storing display data such as character recognition results, and a work memory 5e necessary for executing character recognition and the like.

Further, 6 is a character cutting unit that determines whether input of one character is completed based on the positional relationship between the input stroke and the input window and cuts out the character, 7 is a recognition unit that recognizes characters, etc., and 8 is a recognition unit. The recognition dictionary used in Section 7 contains character codes.
Total number of strokes. Feature point coordinates, etc. are described. Reference numeral 9 denotes a display section for displaying recognition results, etc., 10 a bus line used for memory addresses, data transfer, etc., and 11 an interface section for interfacing with an external output device, that is, a printer 12.

FIG. 2 is an external perspective view showing the configuration of the character recognition device of this embodiment. 20 indicates an input vent that constitutes the input section 2.
21 is a transparent tablet with a liquid crystal display installed at the bottom. Reference numeral 22 indicates a menu area used for giving instructions for correction, editing, etc., and 23 indicates a character input area used for inputting characters. The above menu area 22 and character input amount 23 are displayed on a liquid crystal display. FIG. 3 is a diagram showing part of the character information described in the recognition dictionary 8 of this embodiment. As shown in Figure 3, the recognition dictionary 8 contains the number of strokes of a character. Character codes, feature point coordinates, etc. are stored, and characters input from the input section 2 are cut out by the cutout section 6 and compared with patterns to perform character recognition. FIG. 4 is an enlarged view of a part of the character input area 23 in an external perspective view of the character recognition device of this embodiment.

4l is the input window that is currently being input, 42 is the adjacent input window, and 43, 44, and 45 are the first characters of each input character.
．． Second. The third stroke 46, 47.48 is the starting point of the second stroke 44. Midpoint. The end point 49 is the center of gravity of the first stroke 43 and the second stroke 44. In this figure, the second stroke 44 belongs to two input windows 4l and 42. In the conventional method, input was made to another input window 42 by the second stroke 44;
It was determined that one character had been input, and character recognition was performed, resulting in recognition. According to this embodiment, first, the middle point 47 is found from the starting point 46 and the ending point 48 of the second stroke 44, and since the middle point 47 belongs to the input window 41 where input is currently being made, the second stroke 44 is It is determined that the input is not another character, that is, the input of one character has not been completed. This makes it possible to avoid cutting out incorrect characters due to inadvertent text that extends beyond the character frame. FIG. 5 is a flowchart showing the operating procedure of the CPU 3 of this embodiment. First, when an operator inputs a coordinate point to the tablet 21 using the input bench 20, it is determined in step Sl whether the input is from the menu area 22 or the character input area 23. If the judgment result is input in the menu area 22,
Proceeding to step S8, the characters displayed on the display section 9 are corrected, edited, etc. On the other hand, if the determination result in step S1 is that the input is in the character input area 23, coordinate data is stored in the coordinate memory 5a as strokes forming a character in step S2. When one stroke has been inputted, in step S3, the character cutting section 6 determines whether the input stroke is in the same input window as the previous stroke. In the case of the same input window, the input of one character is not completed, so the process returns to step S2 and the input of the next stroke is repeated. If the input windows are not the same, it is determined in step S4 whether the input strokes are input to two or more different input windows. If the determination result is input to two or more different input windows, in step S5, a midpoint is determined from the start and end points of the stroke stored in the coordinate memory 5a, and the midpoint is stored in the midpoint memory 5b. It is determined whether the midpoint stored in the midpoint memory 5b is within the same input window as the previous stroke. If this judgment result is within the same input window as the previous stroke, it is judged that the input of one character has not been completed, and step S
Return to step 2 and repeat inputting the next stroke. On the other hand, if the determination result in step S4 is that all of the input strokes are in a different input window from the previous stroke, and the determination result in step S5 is that the midpoints of the previous stroke and the input stroke are in different input windows, then one character It is determined that the input has been completed, and in step S6, the recognition section 7 performs character recognition using the previous stroke as one character, and in step S9, the recognition result is displayed on the display section 9, and the process ends.

Note that the flowchart in FIG. 5 shows the general flow and does not show the process of outputting characters from the printer l2. As described above, according to this embodiment, even if the user inputs the input outside the input window due to carelessness, it is possible to correctly cut out the characters, thereby improving the user's ease of use and recognizing the characters. It is possible to realize a character recognition device with improved performance. Although the character cutting section 6 of this embodiment determines the end of character input by finding the midpoint from the start and end points of the stroke, it is also possible to determine the end of character input by finding the center of gravity of the stroke instead of the midpoint. good.

In this case, the structure of the character recognition device of this embodiment may be changed by replacing the center point memory 5b in FIG. 1 with the center of gravity memory 5b' as in the structure of the RAM 5 shown in FIG. Also, step S5 in the flowchart in FIG. 5 is replaced with step S5 in FIG.
' Just replace it with '. The other operations are the same as in Figure 5. The above determination based on the center of gravity has the same effect as the previous embodiment. In addition, it is possible to stably cut out and recognize characters, even for characters that include curves and whose strokes tend to extend beyond the input window, for example.

Furthermore, instead of the midpoint or center of gravity, information on points that are considered to indicate the characteristics of the character or stroke, such as finding the point corresponding to the stroke length H from the starting point of the stroke, is input using an input means such as a keyboard. Judgment may be made by processing according to this input. In this case, the structure of the character recognition device of this embodiment or the middle point memory 5b in FIG. 1 is changed to the structure of the RAM 5 shown in FIG.
It is only necessary to replace step S5 in the flowchart of FIG. 5 with step 85" in FIG. 9, and the other operations are the same as in FIG. 5. The above determination based on point A from the starting point also takes into consideration the fact that the user's character input tends to start within the input window, and has the same effect as the above-mentioned embodiment. Not only can you obtain this effect, but you can also stably extract and recognize objects that protrude significantly from the input window. [Other Embodiments] Although the input unit in the above-mentioned embodiment was composed of a tablet, next we will explain the case of input using OCR composed of a scanner. The structure of the character recognition device of this embodiment is the same as that shown in FIG. 1 except that the RAM 5 shown in FIG. 6 is used. In this embodiment, first, consecutive black pixels forming a character captured by a scanner are extracted as one block.
In this case, in the input example shown in FIG. 4, it can be seen that the block consisting of the first stroke 43 and the second stroke 44 belongs to two different input windows 41 and 42.

Therefore, to determine which input window this block belongs to,
The center of gravity 49 is determined and stored in the center of gravity memory 5b'.
Let the input window to which belongs be the input window of the block. In the example shown in FIG. 4, since the center of gravity 49 belongs to the input window 41, the block of the black pixel portion is determined to belong to the input window 4l, and is treated as one character by the character cutting unit 6 along with the blocks after the third stroke 45. Characters are extracted. FIG. 10 shows the CP in this embodiment. It is a flowchart which shows the operation procedure of U3.

First, in step S21, image information input by an operator using a scanner is stored in the coordinate point memory 5a. In step S22, consecutive black pixels are extracted from the information stored in the coordinate point memory 5a, and each is defined as a block. In step S23, the character segmentation unit 6 determines whether there are blocks belonging to a plurality of input windows. If there are blocks that belong to multiple input windows, step S
In step 24, the center of gravity is determined, and it is determined to which input window the block belongs based on the position of the center of gravity. Steps S23 and S24 until there are no more blocks belonging to multiple input windows.
is repeated, and when it is determined that all blocks belong to the input window, the process proceeds to step S25 to cut out characters,
In step S26, the recognition unit 7 performs recognition, and in step S27, the recognition result is displayed on the display unit 9, and the process ends. As described above, even in this embodiment in which input is performed using a scanner, even if characters protrude from the input window due to the user's carelessness, as in the previous embodiment, characters can be correctly extracted. As a result, it is possible to realize a character recognition device that is easier to use for users and has a higher recognition rate. The embodiment described above is an example embodying the present invention, and the point at which belonging to the input window is determined is the midpoint. Center of gravity. In addition to the point 1A, various other points can be considered depending on the type of characters or patterns of other figures. As explained above, according to this embodiment, even if the input characters/figures belong to multiple input windows, the user's ease of use is improved by providing a character cutting section that correctly cuts out the characters. , which has the effect of increasing the recognition rate. [Effects of the Invention] According to the present invention, patterns can be cut out stably regardless of the user's habits or mistakes, and the input window is small. It is possible to provide a pattern recognition device that increases the recognition rate of pattern recognition regardless of the complexity of the pattern and improves operability.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the character recognition device of this embodiment, FIG. 2 is an external perspective view of the character recognition device of this embodiment, and FIG. 3 is a block diagram showing character information written in the recognition dictionary 8. 4 is a diagram showing a part of the character input area 23, FIG. 5 is a flowchart showing the operating procedure of this embodiment, and FIG. 6 is a diagram showing the configuration of the RAM in the case of character extraction based on the center of gravity. , Figure 7 shows the steps for character extraction using the center of gravity, Figure 8 shows the configuration of the RAM 5 when character extraction is performed using the H point, and Figure 9 shows the steps for character extraction using the H point. Figure 10 is a flowchart showing the operating procedure of another embodiment. In the figure, 1... character recognition device, 2... input section, 3...
・CPU, 4...ROM, 5"-RAM, 5a...
Coordinate memory, 5b--Midpoint memory, 5b'...Gravity center memory, 5b''...% point memory, 5c...Characteristic memory, 5d...Display memory, 5e...Work memory, 6... Character cutting section, 7... Recognition section, 8...
recognition dictionary, 9...display section, 10...pass line,
11... Interface section.

Claims (3)

[Claims] (1) A pattern recognition device that recognizes patterns of characters, figures, etc., which includes an input means having a plurality of input windows for inputting pattern information, and based on the positional relationship between the input windows and input strokes, 1. A pattern recognition device comprising: pattern cutting means for determining the end of input of one pattern and cutting out a pattern. (2) When the stroke currently being input belongs to a plurality of input windows, the pattern cutting means extracts one pattern based on the positional relationship between the input window to which the previous stroke belonged and the stroke currently being input. 2. The pattern recognition device according to claim 1, further comprising determining whether or not input has been completed. (3) A pattern recognition device that recognizes patterns such as characters and figures, which includes an input means having multiple input windows for inputting pattern information, and when the input pattern belongs to multiple input windows, 1. A pattern recognition device comprising: a pattern cutting means for determining the input window to which the input pattern belongs based on the positional relationship between the input window and the input pattern and cutting out the pattern.