JPS6118081A - Candidate character selection system for recognition of handwritten character - Google Patents

Abstract

PURPOSE:To improve man-machine performance by using a display where both input and output are unifired to display with emphasis with which the candidate characters can be selected within a prescribed range and displaying successively the candidate characters with the pen-down at a specific area. CONSTITUTION:The characters are written with on-line and displayed by a stylus pen 2 at a display where a coordinate input device 1 for input is unified with a liquid crystal display device 3 for output. The characters within a prescribed frame range covering the last five characters, etc. are emphasized with a reverse display, etc. as those characters with which the candidate characters can be selected. Under such conditions, it is decided that the stroke input due to the pen 2 is equal to a mode corresponding to a character stroke formed through a coordinate input part 4 of a processing part 12, a character frame calculating part 5, and a mode discriminating part 6 or a stroke for selection of candidate characters formed by an indication given from a lower right area of a frame of the emphasized display character. Then the candidate characters are displayed successively via a candidate character selection part 9, etc. in a candidate character selection stroke mode. Thus the recognized characters are corrected. Thus the man-machine performance is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an online handwritten character recognition device, and particularly to a candidate character selection method suitable for correcting misrecognized characters.

[Background of the invention]

As a candidate character selection method for correcting erroneously recognized characters of a handwritten character recognition device, for example, Japanese Patent Application Laid-Open No. 59-27382
As shown in the publication, a candidate character selection area for displaying and selecting candidate characters from 2nd to 0th place is provided so as to correspond to the handwritten character entry area (multiple characters),
Devices that select and display the correct characters are known.

This method has the advantage that the correct character can be selected immediately because all candidate characters are displayed from the second highest priority character to the highest priority character. On the other hand, there is only one handwritten character entry area.
It was limited to lines, and was far different from the multiple line input of conventional character input methods.

As another conventional method, as shown in Japanese Patent Application Laid-Open No. 58-203579, after all characters are recognized, all the characters are displayed, and in the next step, all erroneously recognized characters are A method is shown in which the user instructs ``/''L on the keyboard, and in the next step switches to the correction mode to display candidate characters corresponding to the previously specified misrecognized character to correct it.

This method has the advantage that consistency of thinking is maintained, accurate correction, and high-speed processing are possible because each step is processed at once. However, in online handwritten character recognition, you may have to revise the information while inputting it, so any misrecognized characters can actually hinder your thinking. Therefore, it is necessary to quickly correct misrecognized characters using the simplest possible method. In this respect, this method requires many steps such as specifying the misrecognized character, then specifying the correction mode, and then indicating the location of the misrecognized character to display candidate characters. As a candidate character selection method for handwritten character recognition, there is a problem in that it is not man-machine friendly.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a candidate character selection method for an online handwritten character recognition device that is easy to use on a machine and can deal with such problems.

[Summary of the invention]

In order to achieve the above object, the candidate character selection method according to the present invention takes into consideration the following points: ■ In online manual character recognition, multiple characters recognized before the character stroke currently being input are focused; When you use an input-integrated display device with an integrated output, it feels like traditional paper and pencil, drawing attention to your hand. ■In the method of recognizing character strokes by inputting them into a predetermined square (frame), the starting point of the first stroke that moves to the next square must first be located at the lower right corner of the eye due to the characteristics of the character. Focusing on two points, the range indicating that candidate characters can be selected is highlighted by displaying character attributes so that the operator can easily understand this. Furthermore, the present invention is characterized in that candidate characters are sequentially displayed by pressing down on a specific part of the square of the highlighted character, that is, the lower right corner of the eye, so that correction can be easily made. Therefore, the operator can input character strokes and modify characters one after another without being particularly conscious of mode identification.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, in which 1 is a coordinate input device, 2 is a stylus pen, and 3 is a display device, such as a liquid crystal (ELD) or electroluminescent (ELD) device.
) is used. The coordinate input device 1 and the display device 3 are integrated to provide a working environment similar to that of a conventional paper and pencil.

Note that the coordinate input device 1 has a coordinate area TNPAI (, GA) for inputting character strokes and the like.

Blocks 4 to 11 represent functions processed by the processing device 12, such as a microcomputer. The outline of the process will be explained using the following drawings.

First, the output of the coordinate input device 1 is sampled at appropriate sampling intervals, and stroke coordinates are input (
Block 4). When the stroke coordinates are input, it is calculated where the starting point of the stroke coordinates falls on the character frame numbers shown in FIG. 2 (block 5). For the sake of explanation, it is assumed that there are 50 character frames in total, with 10 frames on the horizontal axis and 5 frames on the vertical axis.

Next, mode identification processing is performed to identify whether the input stroke is a character stroke or a stroke for selecting candidate characters (block 6). If the stroke is identified as a character and a character stroke for one character is extracted, character recognition processing (block 7) is performed. In this recognition, the first to eighth candidate characters are output, and these are stored in the candidate character buffer (block 8).

On the other hand, when the stroke is identified as a candidate character selection instruction, the next candidate character after the already displayed character is displayed. The next candidate character is displayed by storing it in the character code buffer (block 10) and displaying the character (block 11) K.

Note that the results of character recognition are similarly stored in the character code buffer and displayed.

When the above steps are processed one after another, a display format as shown in FIGS. 3 to 4 is obtained. For example, in FIG. 3, a character stroke is currently being input into character frame number 22 (area S). This has not yet been processed for character recognition. However, the characters displayed in the RO~R5 and DO~D4 areas have already been recognized and displayed, and in particular, candidate characters can be selected for DO~1) 4. is highlighted by reverse display (in the present invention, the selection range is five characters).

This group of characters in reverse display is not a confirmed character yet, but it means that you can select candidate characters. on the other hand,
Characters from R0 to R5 are already determined characters and cannot be selected as candidates.

Next, when the stroke of frame 22 in FIG. 3 ends and the stroke moves to frame 23, frame 22
Strokes that match the characters are recognized and displayed at the same time as shown in FIG. That is, the characters that can be selected as candidate characters are DO~D4''character selection method'' and are shifted (updated) by one character compared to FIG. On the other hand, ``6 Sentence'' becomes a confirmed character, making selection of candidate characters impossible and eliminating the reverse display.

Character code buffer (CODE-BO) at the time shown in Figure 4
The contents of F) are shown in Figure 5.

Here, it is shown in JIS 2 byte code.

Note that blank codes are inserted where no characters are displayed.

FIG. 6 shows the candidate character buffer (C
The contents of ANI)I-BUF) are in a state where 5 candidate characters of 5 characters are stored. When the character recognition is completed, the frame number (FRAMENOJ and the one for candidate character selection) is stored in the buffer pointed to by the pointer l (POINTEF(,I). Pointer 2 (POINTi2) is set to 2, and the codes of candidate characters are stored in the table below. In this example, the codes for '°style, 2, sei, and desk' are stored, respectively.

If the degree of pattern matching is significantly different in character recognition, a blank code (2120H) is stored. Here, the subscript H indicates hexadecimal notation.

Now, if the operator determines that the 6'' character in frame 20 number is a mistake, as shown in Figure 7, when the operator moves the pen down on a specific point, the display state in (a) changes to ``mountain''. ``Select'' will be displayed.

Further down the pen (C) → (d) → (e) → (a
) are displayed in sequence. Therefore, if pen-down is stopped when a correct candidate character is displayed, the code of that candidate character is stored in the character code buffer. In addition, as shown in FIG. 7, whether the candidate character selection instruction is specified or not is determined if the starting point coordinates of pen down are in the area 203 at the lower right corner of the character frame divided into four parts 17, and the stroke from this pen down to pen up. The judgment is made based on whether the projected length of the X and Y axes or the stroke length at the start and end points is smaller than a predetermined threshold. The reason for this is that, as mentioned above, there are no strokes with short stroke lengths or X- and Y-axis projected lengths at the lower right end, as described above. That is, the starting point of the stroke often appears in the area 200° 201.202 due to the nature of the layout.

As mentioned above, the characters displayed in reverse as shown in Figure 4 can be easily modified to the correct characters because their candidate character codes are stored in the candidate character buffer as shown in Figure 6. .

Hereinafter, details of one embodiment of the present invention will be explained using FIGS. 8 to 15.

FIG. 8 shows the main processing flow of the embodiment, which is processed in steps A to M. The processing of each step will be explained. First, when the program is started, initial processing (step A) such as table clearing, table initial setting, and initial calculation for frame number calculation is performed.

Next, in steps B and C, the stroke coordinates are input and the stroke is displayed, and in step D, a character frame number is calculated based on the starting point of the stroke (details will be described later).When the character frame calculation is completed, it is determined whether or not to select a candidate character. Mode recognition processing (step E; details will be described later) is carried out for identification, and if it is not the candidate character selection mode, that is, M
When OD=O, the input stroke is determined to be a character stroke, and one character stroke extraction determination process (step G) is performed. Until the stroke extraction for one character is completed, the process jumps to step B, executes the same process, and stores the strokes in a stroke buffer (not shown).

If stroke extraction for one character is completed in step G, that is, when the stroke start point moves to another frame number, character g recognition processing (step H).

For example, this character recognition process is
The method disclosed in Japanese Patent Application Laid-Open No. 58-123177 is used.

In character recognition, a pattern matching process is performed between the feature points of the input stroke and the feature points of the dictionary, so the character with the highest degree of similarity is output. At the same time, candidate characters are created for a plurality of characters, in this case a maximum of five characters, in descending order of similarity.

Therefore, a group of candidate characters obtained each time one character is recognized are stored in the candidate character buffer one after another as shown in FIG. 6 (details of the steps will be described later).

Next, when the stroke for one character is recognized, the displayed stroke is erased in step C (step K). moreover,
The first code of the recognition result is stored in the character code buffer (CONLY-BUF) as shown in FIG. 5 (step L). At the same time, the character is displayed in reverse at the position corresponding to the frame in which this first character code is input. If the candidate character buffer in FIG. 6 is full of five characters, the reverse display of the frame number of the oldest candidate character group is stopped. Therefore, the number of characters displayed in reverse is always five characters as shown in FIGS. 3 and 4.

On the other hand, in step F, when mode MOD is 1,
That is, in the candidate character selection mode, the process jumps to step I and a candidate character buffer as shown in FIG. 7 is created (details will be described later). Candidate character selection is performed character by character, the selected character code is stored in a character code buffer, and character display is processed in the same way.

The overall flow of the main processing has been explained above, and below, steps D and E will show the main points of the present invention.

J, 1. M will be explained in more detail.

Figure 9 shows the detailed flow of character frame calculation.
Find XNO and Y-axis frame number FYNO. Here, IN'r indicates that the decimal part is rounded down and only the integer part is taken, and Δ1(, ΔV is the frame width of one character on the Y axis and the Y axis, as shown in FIG. 2. DIO).

Next, from the X and YM direction frame numbers, frame number F
Calculate RAMENO and exit. (Step D20
).

FIG. 10 shows a detailed flow of mode identification, in which first the stroke length t is calculated. This stroke length can be easily determined by adding all the distances from pen down to pen up (step E10).

Next, this stroke length t is compared with a predetermined threshold value tSH, and if t, ) Lr■, this stroke is determined to be a character stroke, and the mode MOD is set to O (steps E20 and E70). On the other hand, L − (Lr
When R, the starting point of the stroke (X
Calculate which area in the character frame the character @ l Y, ) falls in, and if the starting point is at the lower right corner of the four-part area in the character frame (ΔFXNO=1.ΔF'YN
O=1 >, this stroke is judged as an instruction to select a candidate character, and mode Mo1) is set to 1 (
Step 110 to Step E60).

FIG. 12 is a detailed flowchart of the process step J for storing candidate characters in the buffer. As shown in the candidate character buffer table (CANI) E-BUP) in FIG. In order to stop displaying the frame number in reverse characters, temporarily change it to 01.1)-PR, AM.
I retreated to the ENO table. Figure 3.

In the example of Figure 4, this frame number is 17 at the position where it is lost to 1 sentence.On the other hand, the frame number of the newly recognized character is 22, where 'expression' is located from Figure 4.
This is stored in the buffer shown in FIG. 6 at the position where the frame number is inserted. In addition, pointer 2 (P
OTNTERl) is set to 2 (step JIO
).

Next, we will create a code for the five candidate characters obtained through character recognition processing.
)"CANDI-BUF oCANDIO~CANI
) I42) By the way, store it. If there are less than five candidate characters, a blank code is set in the empty buffer (step T20).

Finally, pointer 1 (POINTERl) is incremented to store the next candidate character in the buffer.

However, when pointer 1 is 5 or more, it is set to 0. In Figure 6, at the start penalty point in Figure 12, POrN'l'ER,
1=4-r A certain note, when processing step J30ft, P
OINTER1=0.

FIG. 13 is a detailed flowchart of candidate character selection processing step (2). This routine is started when there is a mode identification of a candidate character selection instruction, and first, steps 110 to 14 are performed.
0, a frame number search is performed. That is, a match is searched for by changing the pointer from =o to 5 to indicate which frame of characters is selected among the advance display characters in the candidate character selection range. If an unstored frame is specified, the value becomes 5 in step I40, and the process returns without doing anything.

On the other hand, if the frame number can be searched as determined in step I20, pointer 2 (POfNT12) is incremented to output the next candidate character code, as shown in FIG. As shown in FIG. 14, this pointer 2 is PR,
Since it is an offset from AMgNO, for example, P
When OIN'l'ER2=3, candidate character code KCO
For DE, a "select" code (42728) is obtained (steps I50-60). Therefore, each time this routine is activated, candidate characters can be retrieved one after another.

When a candidate character is extracted in the process shown in FIG. 13, or when a new character is output in the character recognition process, the character display shown in FIG. 15 is processed.

First, the reverse display of the character corresponding to the frame number (OLD-=RRAMgNOK stored) saved in the storage process in the candidate character buffer shown in FIG. 12 is stopped.

In the example shown in FIG. 4, this is the sentence "" displayed in frame number 17.

On the other hand, the characters corresponding to the new frame number are displayed in reverse. In the example in Figure 4, frame number 2
It becomes the character of ``expression'' in 2. In addition, when selecting a candidate character, displaying and canceling reverse is only repeated by resetting.

One embodiment of the present invention has been described above, and next, the effects of this embodiment will be described.

The first effect is that it uses an integrated input display that integrates input and display, so characters can be entered in the same way as traditional paper and pencil, and the handwriting is neat and clean. The second effect is that multiple rows and columns of squares are generated on the display, so characters can be entered in the same way as manuscript paper, and it is easy to correct incorrectly recognized characters. , Candidate characters can be easily selected because multiple characters are displayed using highlighted character display.

As a third effect, since the mode designation for candidate character selection is automatically identified, character input and candidate character selection instruction input can be performed without being conscious.

As a fourth effect, when a candidate character selection instruction is repeatedly executed, the display returns to the first character display, so even if the instruction is made by mistake, it is possible to easily return to the display.

As a fifth effect, since candidate characters are displayed in order of priority, incorrectly recognized characters can generally be corrected with one or two operations.

Next, another embodiment of the present invention will be described.

In one embodiment of the present invention, mode identification of candidate character selection is performed by pen-down (short stroke) at a specific location within the character input frame, but as shown in FIG. 16, candidate character selection is performed separately from the character input frame. An area may be provided so that candidate characters can be selected by specifying this area.

As another example, in the present invention, as shown in FIG. 7, when candidate character selection is instructed for the same frame, characters with a high degree of matching are displayed one after another, but if the number of instructions is exceeded, There is no immediate way to go back. Therefore, by determining that when a candidate character selection instruction is issued once within a certain period, the direction is forward, and when there are two instructions, it is determined that the direction is backward, it becomes possible to easily go back. On the other hand, in the example shown in FIG. 16, this can be easily realized by dividing the candidate character selection area of the corresponding character into two parts, one in the forward direction and the other in the reverse direction.

Furthermore, as another embodiment of the present invention, as shown in FIG. 1, a device in which input and display are integrated is used. The display may also be implemented separately, eg OR'l'. However, in this case, the method is not the same as using paper and pencil, and there is a drawback that the coordinate input device and the display must be looked at alternately. However, since the selectable range of candidate characters is highlighted, correction is relatively easy.

Further, although the attribute display in one embodiment of the present invention is highlighted by reverse display, the same effect can be produced even if the attribute is displayed in color, by brightness modulation, or by rope display.

As described above, in other embodiments as well, since the selectable range of candidate characters is highlighted, it is easy to modify.

〔Effect of the invention〕

As described above, according to the present invention, characters can be input in the same manner as conventional manuscript paper, and even for recognized characters, the range from which candidate characters can be selected is highlighted, making it easy to input characters. Misrecognized characters can now be corrected, greatly improving man-machine performance.

Further, in the present invention, since instructions for selecting candidate characters are automatically identified based on stroke information and frame information, character correction can be performed with fewer operations.

7 to 7 are diagrams for explaining the operation contents and table configuration of an embodiment of the present invention, and FIGS. 8 to 15 are diagrams for explaining the detailed processing flow and its operation of an embodiment of the present invention, 16th
The figure is a diagram explaining another embodiment.

1... Coordinate input device, 2... Stylus pen, 3...
. . Display device, 4. Coordinate input processing, 5. Character frame calculation processing, 6. Mode failure processing, 7. Character recognition processing, 8. Candidate character storage processing, 9.・Candidate character selection process, 10...Storing process to character code buffer, 11...0 ＼ ~ Gin Sun II ・Nagogo 70 80th Gi 9 Tomanto no Haku 11 Ki (χS・'!5 ) Cage 12 of Cage 130 IS

Claims (1)

[Claims] 1. Candidate characters for correcting misrecognized characters by an online handwritten character recognition device that samples handwritten characters, converts them into stroke information, and recognizes the characters by matching the features of this stroke information with dictionary patterns. In the selection method, storing candidate characters for each of a plurality of previously recognized characters from the last character recognized by the online handwritten character recognition device in a candidate character buffer, and highlighting the plurality of characters with an attribute display; Candidate character selection method for online handwritten character recognition characterized in that candidate characters are sequentially retrieved from the candidate character buffer and corrected by inputting a candidate character selection instruction stroke for the highlighted character group. . 2. In claim 1, the adjustment display is:
A candidate character selection method for online handwritten character recognition characterized by notifying in reverse display, color display, brightness display, or rope display. 3. In claim 1, the stroke input of the candidate character selection instruction further subdivides a character frame provided for character stroke extraction and assigns it to one of the subdivided frames. Candidate character selection method for online handwritten character recognition, characterized by: 4. In claim 3, the character frame is X,
The character is divided into 2 parts each on the Y axis and divided into 4 parts, and when the stroke length of the stroke input in the lower right area of the 4 parts or the projected length of the X and Y axes is smaller than a predetermined threshold, the candidate character selection mode is set. Candidate character selection method for online handwritten character recognition, characterized by: 5. The online handwritten character according to claim 1, wherein the candidate characters are stored in a buffer in order of matching degree, and if the matching degree is less than a predetermined value, a blank character is inserted. Candidate character selection method for recognition. 6. The candidate character selection method for online handwritten character recognition according to claim 1, characterized in that when the candidate character selection instruction is sequentially repeated, the system returns to the first candidate character selection in a ring manner.