JPH096897A - Handwritten input recognition method - Google Patents

Abstract

PURPOSE: To hold the number of sampling data nearly constant and improve character recognition, etc., without being affected by the speed of handwritten input by making a sampling time short when the speed of the handwritten input is fast and long when slow. CONSTITUTION: At the start of the handwritten input, a CPU 1 inputs coordinate data at initial sampling time intervals and decides whether or not the number of coordinates from the start of the handwritten input up to now reaches three. When the number reaches three, the CPU 1 finds the initial movement quantity ya of the handwritten input and finds the rate P to a reference movement quantity (y), and then refers to a sampling time table 11 according to the rate P to determines the sampling time for inputting the coordinates of the 4th and succeeding points.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwriting input recognition method in which a coordinate string detected from a handwriting input pattern is taken in as sampling data and a handwritten character is recognized based on this sampling data.

[0002]

2. Description of the Related Art In a conventional handwritten character recognition device,
When the pen is touched down on the touch panel and one stroke is input, the coordinate data is sequentially detected at fixed time intervals following the input,
Each detected coordinate data is taken in as sampling data for handwriting input. FIG. 6 is a diagram for explaining the sampling operation at the time of single-stroke input, and there are individual differences in the writing speed of characters, and shows the sampling operation when writing extremely fast and the sampling operation when writing slowly. . In this case, since the time for sampling the coordinate points is a fixed time T fixedly fixed in advance, the number of coordinates decreases when the input speed of one stroke is fast, and increases when the input speed of the stroke is slow.

[0003]

When the handwriting input speed is high, the sampling data becomes small and the probability that the required feature points cannot be detected becomes high. Therefore, the character recognition is performed based on the sampling data. If this is done, there is a problem that the probability of false recognition increases. Therefore, if the input speed of one stroke is slowed down, the number of coordinates increases, so that the character recognition rate should be increased.
However, since the memory for actually storing the sampling data is naturally limited, conventionally, the sampling data is limited to a certain number, and the sampling is discarded thereafter. That is, as shown in FIG. 6, when the number of data that can be sampled is set to “20”, the coordinates from that point until the pen-up is cut off, and as a result, the data cannot be obtained until the end of one stroke. If character recognition is performed based on various sampling data, the probability of erroneous recognition increases. An object of the present invention is to reduce the sampling time when the handwriting input speed is fast, and conversely, to lengthen the sampling time when the input speed is slow, so that the sampling time is not affected by the handwriting input speed. The goal is to keep the number of data almost constant and to improve character recognition.

[0004]

Means for Solving the Problem and Its Action The means and sled action of the present invention are as follows. Coordinate data of the handwriting input pattern is sequentially detected at every predetermined sampling time, each detected coordinate data is read as sampling data of the handwriting input, and in the character recognition device for recognizing the characters of the handwriting input based on the sampling data, At the start of handwriting input, the coordinate data is detected at every predetermined time as the initial sampling time, and the number of coordinate data from the start of handwriting input to the present is counted, and whether this count value has reached the predetermined number. When it is determined that the number has reached a predetermined number, the movement amount of the handwriting input is detected as the initial movement amount, and the sampling time corresponding to the speed of the handwriting input is determined based on the initial movement amount. The coordinate data is determined and subsequent coordinate data is detected for each determined sampling time. Here, "character" is a kana character,
Including kanji, numbers, alphabets, symbols, figures, etc. When determining the sampling time according to the speed of the handwriting input, the detected initial movement amount and a reference movement amount that is determined in advance corresponding to the standard speed of the handwriting input are set. You may make it compare and determine the sampling time based on this comparison result. Further, when determining the sampling time according to the handwriting input, the speed of handwriting input is obtained based on the detected initial movement amount and the time corresponding to the movement amount, and the sampling time is calculated based on the input speed. May be determined. Therefore, by shortening the sampling time when the speed of handwriting input is fast, and conversely, by increasing the sampling time when the input speed is slow, the sampling data is almost unaffected without being affected by the speed of handwriting input. It can be kept constant and the improvement of character recognition and the like can be realized.

[0005]

【Example】

(First Embodiment) A first embodiment will be described below with reference to FIGS. FIG. 1 is a block diagram showing a character recognition device with a handwriting input device. The CPU 1 is a central processing unit that controls the overall operation of the character recognition device according to various programs stored in the ROM 2 or the like. The handwriting input device 3 is a handwriting input device composed of a touch screen. A transparent touch panel 5 is laminated on the liquid crystal display unit 4, and characters (figures, symbols, etc. are touched while touching the upper surface of the touch panel 5 with a pen tip. (Including) is handwritten, the coordinate detecting unit 6 detects the coordinate points (XY coordinate points) following the input trajectory and gives them to the CPU 1.
Here, when the CPU 1 writes the coordinate sequence data obtained by the coordinate detection unit 6 in the video RAM (VRAM) 7,
The display control unit 8 accesses the VRAM 7 and causes the liquid crystal display unit 4 to display and output the image data in the VRAM 7. As a result, the handwritten input pattern drawn on the touch panel 5 is displayed and output from the liquid crystal display unit 4 according to the input locus.

The sampling data memory 9 temporarily stores the coordinate data detected by the coordinate detecting unit 6 as sampling data for recognizing a handwritten input character. The CPU 1 detects the coordinates of the handwriting input detected by the coordinate detecting unit 6. The coordinate data extracted from the column at every predetermined time is read as sampling data and stored in the sampling data memory 9. Here, in the figure, Y1, Y
2 and Y3 are coordinate data extracted every predetermined time which is predetermined as the initial sampling time at the start of handwriting input, and the CPU 1 counts the number of coordinate data from the start of handwriting input to the present, and the counted value Has reached a predetermined number (“3” in this embodiment). This allows coordinate data Y of three points from the start of handwriting input.
When 1, Y2 and Y3 are sampled, the CPU 1 obtains an initial movement amount at the start of handwriting input based on these coordinate data and sets it in the work memory 10.

The work memory 10 has a reference movement amount (fixed value) y which is predetermined corresponding to a standard speed during handwriting input, an initial movement amount ya during actual handwriting input, and a reference movement amount. The ratio P of the initial movement amount ya to y is stored, and the CPU 1 refers to the sampling time table 11 based on this ratio P to determine the sampling time. The sampling time determined by this is a time interval for sampling the coordinate data of the fourth point and thereafter, and the CPU 1 reads the coordinate data of the fourth point and thereafter at every sampling time and stores it in the sampling data memory 9. .

The sampling time table 11 stores the sampling time corresponding to the ratio of the initial movement amount to the reference movement amount. As shown in FIG. 2, when the ratio P = 1, it is associated with the initial sampling time T0. ing. Further, when the ratio P = 2, 4, 8 ..., The sampling time T is 1 / 2T0, 1 / 4T0, 1 / 8T0.
Is shorter than the initial sampling time T0,
On the contrary, when the ratio P = 1/2, 1/3, 1/4, ..., The sampling time T becomes longer than the initial sampling time T0 as in 2T0, 3T0, 4T0. Here, when referring to the sampling time table 11 based on the ratio P in the work memory 10, the CPU 1 searches for P in the sampling time table 11 that is closest to this value and finds the corresponding sampling time T as the sampling time table 11. Read from.

The character recognition unit 12 recognizes a handwritten input character and converts it into a character code, and the CPU 1 takes in the sampling data for one character stored in the sampling data memory 9 and gives it to the character recognition unit 12. Here, the character recognition unit 12 refers to the character recognition dictionary memory 13 based on the sampling data for one character, performs character recognition, and converts the character code. The character recognition dictionary memory 13 stores codes corresponding to patterns of various characters, symbols, figures, etc., and the character recognition unit 12 searches for a pattern close to the sampling data and reads the corresponding character code. . The character code converted by this is taken in by the CPU 1 and stored in the document memory 14.

Next, the operation of this embodiment will be described with reference to the flow chart shown in FIG. FIG. 3 is a flowchart showing the sampling operation at the time of one-stroke input,
The execution is started when the pen touches the touch panel 5 with the pen. First, the CPU 1 starts sampling at a predetermined initial sampling time T0 (step S1). Then, it is determined whether or not the coordinate data of the three points is sampled from the pen-down, and the sampling operation is repeated every initial sampling time T0 until the coordinates of the three points are fetched, and the first point obtained by this is Y
The first and second points are set to Y2 and the third point is set to Y3 and stored in the sampling data memory 9 (step S2). Next, the pen movement amount with respect to the initial sampling time T0 is obtained as the initial movement amount (step S3). That is, the coordinate Y1 of the first point is subtracted from the coordinate Y3 of the third point to obtain the movement amount (Y3-Y1) from the first point to the third point, and this movement amount is divided by "2". , A movement amount ya corresponding to one sampling interval corresponding to the initial sampling time T0 is obtained, and the movement amount ya is set as the initial movement amount to the work memory 1
Set to 0.

Next, the CPU 1 reads out the reference movement amount y stored in advance in the work memory 10 and the initial movement amount ya calculated in step S3, and calculates y / ya based on these values. Then, the ratio P of the initial movement amount ya to the reference movement amount y is calculated to obtain the work memory 1
It is set to 0 (step S4). Here, the reference movement amount y is a value determined in advance corresponding to the standard speed of handwriting input as described above, and the initial sampling time T0
This is the amount of movement per sampling interval when handwriting is input at a standard speed during the period.

When the value of P is obtained in this way, the CPU
1 searches the sampling time table 11 based on this value and selects the corresponding sampling time T (step S5). In this case, if the value of P is "2" or a value close thereto, 1 / 2T0 is read as the corresponding sampling time T from the sampling time table 11, and the coordinate data of the fourth point and thereafter is read at each sampling time T. Sampling is performed (step S6). Such sampling operation is repeated until pen-up (end of one-stroke input) is detected (step S7). And 1
When the handwriting input for the characters is completed, the CPU 1 takes in the sampling data for one character stored in the sampling data memory 9 and passes it to the character recognition unit 12.
Then, the character recognition unit 12 performs character recognition based on this sampling data and converts it into a character code.

As described above, in the present embodiment, when the handwriting input speed is fast, the sampling time is shortened, and conversely, when the handwriting input speed is slow, the sampling time is lengthened to obtain the handwriting input speed. It is possible to improve the character recognition and the like by keeping the number of sampling data substantially constant without being affected by the above. That is, FIG. 4 is a time chart showing the sampling operation for one stroke, (A) shows a case where the input is faster than the standard input speed, and (B) shows a case where the input is slower. Here, the operation is performed by the initial sampling time T0 from the pen-down until the coordinates of three points are fetched, but the sampling is performed by the sampling time T determined by referring to the contents of the sampling time table 11 after four points. Be seen. Therefore, as shown in (A), when the input speed is fast, the initial movement amount ya becomes larger than the reference movement amount y, so T0 / 2 is selected as the sampling time T according to the ratio, and When the input speed is slow as shown in B), the initial movement amount ya
Is smaller than the reference movement amount y, so 2T0 is selected as the sampling time T according to the ratio.

(Second Embodiment) Hereinafter, a second embodiment will be described with reference to FIG.
An embodiment will be described. In the first embodiment, the time T for sampling the coordinate data of the fourth and subsequent points is determined based on the ratio of the initial movement amount ya to the reference movement amount y. However, in the second embodiment, the initial movement amount ya is determined. The speed of handwriting input is calculated from the time and the time corresponding to this movement amount, and the sampling time after the fourth point is determined based on this input speed.

FIG. 5 shows the contents of the sampling time table 11 used in the second embodiment, and the pen movement speed P
Sampling time T1, corresponding to 1, P2, P3 ...
T2, T3 ... Are set in the sampling time table 11. Here, since the sampling operation in the second embodiment is basically similar to the flowchart shown in FIG. 3, the illustration is omitted, but in the process of step S4, in the second embodiment, the initial movement amount ya is set to the initial sampling time. Divide by T0 to obtain the input speed and work memory 10
, And the sampling time table 11 shown in FIG. 5 is searched based on this input speed to determine the corresponding sampling time. The second embodiment thus constructed has the same effect as that of the first embodiment.

In each of the above embodiments, the three-point coordinates are sampled from the pen down at the initial sampling time T0, but the coordinate points may be increased to about 5 to 6 points and averaged. Instead of the movement amount per sampling interval, the movement amount from the pen down to the predetermined coordinate number may be used as the initial movement amount. In addition, the handwriting input device is not limited to a pen touch on the touch screen, but a transparent resistance film or electrostatic film is placed on the touch panel to detect the electrical change caused by the touch of the human body and input coordinates. Alternatively, it may be a tablet or a light pen.

[0017]

According to the present invention, when the handwriting input speed is fast, the sampling time is shortened, and conversely, when the handwriting input speed is slow, the sampling time is lengthened to improve the handwriting input speed. Unaffected,
It is possible to keep the number of sampled data substantially constant and improve the character recognition and the like.

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing a character recognition device with a handwriting input device.

FIG. 2 is a sampling time table 1 according to the first embodiment.
The figure which showed the content of 1.

FIG. 3 is a flowchart showing a sampling operation according to the first embodiment.

FIG. 4 is a time chart showing sampling operations when handwriting input is fast and slow.

FIG. 5 is a sampling time table 1 according to the second embodiment.
The figure which showed the content of 1.

FIG. 6 is a time chart showing a sampling operation when handwriting input is fast and when it is slow.

[Explanation of symbols]

 1 CPU 2 ROM 3 Handwriting input device 4 Liquid crystal display unit 5 Touch panel 6 Coordinate detection unit 9 Sampling data memory 10 Work memory 11 Sampling time table 12 Character recognition unit 13 Character recognition dictionary memory

Claims (3)

[Claims] 1. The coordinate data of a handwriting input pattern is sequentially detected at every predetermined sampling time, each detected coordinate data is read as sampling data of the handwriting input, and the character of the handwriting input is recognized based on the sampling data. In the character recognition device, when the handwriting input is started, the coordinate data is detected every predetermined time which is predetermined as the initial sampling time, and
The number of coordinate data from the start of handwriting input to the present is counted, it is determined whether or not this count value has reached a predetermined number, and when it is determined that it has reached the predetermined number, the amount of movement of handwriting input Is detected as the initial movement amount, the sampling time according to the speed of handwriting input is determined based on this initial movement amount, and the coordinate data after that is detected at each determined sampling time. And handwriting input recognition method. 2. The detected initial movement amount when determining the sampling time according to the speed of the handwriting input,
The sampling time is determined based on a result of comparison by comparing with a predetermined reference movement amount corresponding to a standard speed of handwriting input. Handwriting input recognition method. 3. When determining a sampling time according to the handwriting input, a speed of handwriting input is obtained based on the detected initial movement amount and time corresponding to the movement amount,
The handwriting input recognition method according to claim 1, wherein the sampling time is determined based on the input speed.