JPH049358B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION <Technical Field> The present invention relates to an apparatus for recognizing handwritten characters written within a predetermined character frame. In particular, the present invention relates to a device that accurately determines the completion of inputting one character written in the character frame without intervening other key operations, and recognizes the character.

<Prior art> Conventionally, in so-called online handwritten character recognition devices that recognize handwritten characters in real time, character frames are provided on a coordinate input panel such as a tablet to separate characters from each other (for cutting out characters). The characters handwritten within the character frame were recognized on a character frame basis. This type of handwritten character recognition device is known as described in Nikkei Electronics (published December 5, 1983).

In these conventional devices, coordinate points with a predetermined density are formed on the tablet, a predetermined character frame is set, and each stroke (start point, end point, point)
The feature points are extracted and used as input data, and each time the input of this one character is completed, a separate function key is operated to recognize that one character has been input, or a stroke is applied to another character frame. When the starting point was input, there was something that was recognized as one character input completed.

The former method of recognizing the completion of inputting one character requires operating a function key for each character input, which is very inconvenient for input operation, and the latter method requires inputting a character into a certain character frame. When consecutive strokes entered into adjacent character frames, there was a risk that this would be judged as the completion of one character input, which could lead to erroneous character recognition.

<Purpose> The present invention provides a handwritten character recognition device that recognizes the completion of inputting one character using the latter method described above.
The purpose of this is to make it possible to reliably and accurately determine the completion of inputting one character.

<Example> FIG. 1 is a diagram showing the external configuration of a handwritten character recognition device of the present invention, in which 1 is a tablet device, 2 is an input pen having a writing device such as a ballpoint pen at the tip, and 3 is a character recognition unit. It is.

This tablet device 1 is, for example, a generally well-known electromagnetic induction type coordinate input device, which generates coordinate points at a predetermined density (for example, 10 dots/mm) in advance, and an input pen. By bringing the tip of 2 close to the tablet surface, the nearby coordinate points xi and yi can be obtained.

In addition, a stroke switch is provided at the tip of the input pen 2 and is turned ON/OFF by pressing or releasing the input pen 2.
The OFF signal is output together with the above coordinate data.

On the writing surface 10 of this tablet, character writing frames 11 are formed for each predetermined coordinate group, and each character writing frame 11 has coordinate points in a predetermined range. Therefore, it is possible to determine which character frame it belongs to.
In addition, when inputting characters on the entry surface 10,
An entry form having a writing frame for writing input characters on the paper with the pen 2 is arranged.

Further, the character recognition unit 3 incorporates various control circuits to be described later, and is also provided with a connection cable for supplying recognized characters to another output device, such as a CRT display device.

This tablet device 1 uses the above coordinate data,
Stroke data can be obtained for each stroke of each character by turning the stroke switch signal ON/OFF. That is, when inputting characters, the switch is turned on at the start point of each stroke and turned off at the end point of the stroke, and the position of the entered stroke is obtained based on the coordinate data at that time. Therefore, stroke data corresponding to the number of strokes of the input character is obtained.

The operator enters the desired number of character strokes within one character frame above, and when one character input is completed,
Start writing characters into a new character frame. At this point (because the stroke starting point in the new character frame is obtained), completion of the previous character entry is determined.

However, depending on the writing method, the stroke of one character may end up entering an adjacent character frame (which is essentially an erroneous operation). In this case, character entry is not completed, and if it is determined that it is completed in this state, it may cause erroneous recognition.

Further, if the character entry is completed and left as is for a long time, the character entry is not completed, which is a problem.

Therefore, when a stroke in a character frame consecutively enters an adjacent frame, it is necessary to detect this, ignore coordinate data outside the frame, and control the character input so that it is not considered complete. This control will be explained below.

FIG. 2 is a circuit block diagram for the above-mentioned control built into the above-mentioned character recognition unit 3.

20 is a CPU that recognizes the corresponding character based on the stroke data input from the tablet and performs sequence control of the process up to output.
It is controlled according to a system program stored in the ROM 21 in advance. This ROM 21 includes a preprocessing program, a stroke feature point extraction processing program, a distance calculation program, a recognition processing program,
Five major system programs of output processing programs are stored.

The preprocessing program 31 performs a smoothing process to remove step-like or spike-like trajectory distortions in the locus of stroke data that is inputted sequentially in time series from a tablet, and interpolation between coordinate points input and imported from the tablet. interpolation processing to be performed, sampling processing to capture the necessary amount of each coordinate data input from the tablet, normalization processing to reduce/enlarge and normalize the position and size of the input stroke, and the present invention. Features such as input completion processing for input characters are provided.

The stroke feature point extraction processing program 32 calculates the start point, end point, midpoint, etc. of the stroke from the stroke data obtained through the above processing through arithmetic processing, and extracts the stroke feature points. It is a processing program.

The distance calculation program 33 is a program for comparing the standard pattern 23 described later with the above feature points, calculating the distance between them, and calculating distance data.

The recognition processing program 34 is a program that selects the most suitable character based on the distance data and uses it as a recognized character.

The output processing program 35 is a program that sends at least one character (TIS code) of a plurality of candidate characters to an output device, such as a CRT display device, through the above-mentioned processes.

Each of the above programs is processed in a time-sharing manner,
Corresponding characters are recognized from sequentially input stroke data.

22 is a data memory composed of RAM, and is provided with the following various buffers (Buf). That is, 41 is a coordinate data buffer that temporarily stores coordinate data input from a tablet and performs preprocessing on the data. 43 is based on the stroke data obtained by the above-mentioned preprocessing,
It is a buffer that stores characteristic points such as the start point, end point, and midpoint of the stroke. 44 is a result buffer that is compared with a standard pattern based on the data of the feature point buffer, stores the JIS code of the corresponding recognized character in the lowest order as a result of distance calculation, and is further stored together with the distance data. .
In addition to the above-mentioned buffers, an area 42 consisting of various buffers for each of the above-mentioned processes is provided.

23 is a memory for storing standard patterns;
Along with the JIS code of each character, standard feature points such as the start point, middle point, and end point of that character are stored for each stroke.

24 is an interface device to which a tablet 25 (tablet device 1 in FIG. 1) and an output device 26 (for example, a CRT device) are connected.

Next, the operation of the circuit configuration shown in FIG. 2 will be explained according to the operation flow shown in FIG. 3.

First, when the handwritten character input mode is set using a function key or the like, the system operates as shown in FIG. That is, a process of importing coordinate data from the tablet is performed, and coordinate data (xi, yi) of the position where the input pen 2 is close to the writing surface of the tablet is supplied via the interface 24 and stored in the coordinate data buffer 41. . Also, at this time, it is determined whether the stroke switch at the tip of the input pen is ON, and if it is OFF, the stored data is not used and is awaited until it is ON. Then, when the stroke switch is turned on, that is, when a stroke start point is entered, the coordinate data is taken in and converted into the input frame and the coordinate values within that frame. (Steps 100, 101, and 102 in FIG. 3) The operator continuously writes strokes corresponding to the input characters from this starting point. That is, the coordinate point next to the above-mentioned starting point is taken in, and the frame position and coordinate values within the frame are determined in the same manner as above (step 103,
104, 105). Further, smoothing processing, sampling processing, and interpolation processing are performed based on the above-mentioned coordinate values between the two points (step 108). Repeat this action and
Stroke data is input.

When one stroke input is completed, by removing the input pen 2 from the writing surface of the tablet,
The stroke switch is turned OFF. This state is determined at step 104, and the stroke is determined to have been completed, and the process moves to step 110. After that, a timer set to several seconds turns on (step 110) and counts the number of strokes (equivalent to the number of strokes).
Furthermore, the start point, middle point, and end point are calculated from the above stroke data, extracted as feature point data, and stored in the feature point buffer 43 (step
111).

Thereafter, the feature point data is compared with the standard pattern feature point data corresponding to the number of strokes in the standard pattern memory 23, and the distance is calculated. The smallest distance data is determined to be the recognized character (step 112).

Thereafter, recognition processing is performed based on this distance data (step 113), and characters corresponding to a predetermined distance range are sequentially transmitted and stored in the result buffer 43 (step 129).

This recognition process takes a long time, and it is not guaranteed that all characters have been filled in (all strokes).In this process, there is a process to import coordinate data and confirm whether the stroke switch is ON/OFF. Parallel processing is performed.

During this process, turn on the stroke switch.
When it is confirmed, the process moves to step 140 where it is converted into a frame position and frame coordinate values, and it is determined whether the stroke is within the same frame as the previously input stroke. If it is within the same frame, the recognition process is stopped. The input process for the next stroke is executed again, and the timer is turned off (steps 140, 141, and 142).

If the next stroke start point is in another frame, the character input end flag is turned on as if character input in the previous frame has been completed, the above recognition process continues again, and the recognized characters are output to the result buffer. be done. Note that at this time, although not shown, if a new stroke is input, it is processed in parallel with the previous input process.

When the recognition process is completed, the recognized character (the character with the smallest distance data) is sent to the output device via the interface 24, and the above-described stroke input process is executed again (steps 130, 131).

Here, the character input completion process will be further explained.

As is clear from the above explanation, when writing characters in one frame, if the start point of the next stroke is entered in a different frame from the previous character entry frame, the step
141 completes character input.

Also, if strokes from one character entry frame consecutively enter adjacent frames, step 106 is performed.
When it is determined that it is outside the frame, the process moves to step 107, the over flag is turned on, and the process returns to step 103.

That is, coordinate data that is not in the same frame is not used and is ignored until it returns to the same frame or the stroke switch is turned off.

When consecutive strokes return to the same frame again,
The step starts because the overbrag is ON.
Through the interpolation process in step 108, interpolation is performed to connect the coordinate points taken out from the frame and the returned coordinate points, resulting in a series of stroke inputs.

In this way, even if the stroke enters adjacent frames consecutively, the input is not completed, so that one stroke is not separated and can be recognized accurately.

Further, even if a stroke is completed in one frame and the starting point is not input in a different frame for inputting the next character, input of one character can be completed if the stroke is left unused for a predetermined time or more. In other words, when one stroke is completed (stroke switch
OFF), the timer is turned on (step 110), and when the stroke is started again within the same frame, the timer is turned on.
It will be turned off.

Further, recognition processing is performed after the stroke is completed, and if there is no input after the recognition is completed, it is determined in step 115 that the timer is turned off, but it is determined whether the stroke is to be started (115, 116, 117).

That is, if the stroke input is left unattended for a predetermined period of time (for example, several seconds), the timer is turned off, thereby completing the character input and outputting the recognized character.

If the input is left for a predetermined period of time in this manner, the timer completes the character input without starting a new stroke, and unnecessary operations are not required.

<Effects> As described above, in the case of the handwritten character recognition device of the present invention, when recognizing the completion of inputting a single character, the handwritten character recognition device of the present invention changes from a state input in a certain character frame to a character other than the input character frame. Detects the existence of a stroke start point in a frame and concludes character input, and also detects when a stroke enters an adjacent character frame consecutively and inputs the coordinate data outside the frame without determining that character input is complete. Since the control is performed so as to ignore the characters, the completion of one character input can be determined reliably and accurately, and the occurrence of erroneous recognition of characters, which is the case with conventional devices, can be avoided as much as possible.

[Brief explanation of drawings]

Fig. 1 is a perspective view showing the appearance of the device of the present invention;
The figure is a block diagram showing the control circuit configuration of the device, and FIG. 3 is a flowchart showing the operation of the configuration of FIG. 2. 1: Tablet device, 2: Input pen, 3: Character recognition unit, 10: Tablet writing surface, 11:
Character entry frame, 20: CPU, 21: ROM, 22:
RAM, 23: Memory for storing standard patterns;
24: Interface, 25: Tablet device, 26: Output device.

Claims (1)

[Claims] 1. Generating stroke data by detecting coordinate values from a trajectory drawn within a predetermined frame of a tablet device having a plurality of predetermined character entry frames;
In a handwritten character recognition device that recognizes characters by calculating feature points from the stroke data and comparing them with a standard pattern, a first step detects a stroke start point in a frame different from the frame after writing a character in the predetermined frame. means and
A second method for detecting a state in which a stroke enters an adjacent frame consecutively from the predetermined frame in which characters are written.
In response to the detection by the first means, it is determined that one character input is completed, and in response to the detection by the second means, the character input is not determined to be completed and coordinate data outside the predetermined frame is invalidated. A handwritten character recognition device characterized in that the handwritten character recognition device is controlled to