JPS6068486A - Character recognizer - Google Patents

Abstract

PURPOSE:To prevent the generation of misrecognition of characters without increasing the memory capacity by writing the final coordinates into the final memory area of a memory in case a character pattern having such a long stroke that exceeds the memory capacity is supplied. CONSTITUTION:A character recognizer is provided with an input part 1, ROM3 and 4, an arithmetic part 6, etc. The part 1 uses a touch switch to detect the input coordinates and to display it to a display part 2. At the same time, plural memory areas are provided to the RAM5 and the coorinates detected by the part 1 are stored successively to the memory areas according to the program of the ROM4. While an operation decoder 7, an address part 8, a timer part 12, etc. are provided to the ROM3. In case a character pattern having such a long stroke that exceeds the memory capacity is supplied, the final coordinates of the pattern are written to the final area of the RAM5. Thus a character pattern having a long stroke can be recognized accurately without increasing the capacity of the RAM5.

Description

[Detailed description of the invention] [Technical field of invention] The present invention relates to a character recognition device.

[Prior art]

In recent years, a number of character recognition devices have been developed that recognize characters, figures, etc. that are handwritten and input on a tablet using a pen-shaped jig or the like and convert them into input information. In this type of character recognition device, when a character/zoom is input, the coordinates of the character pattern are detected, and the detected coordinates are sequentially written into a predetermined memory, and then the coordinates are read according to the contents of this memory. It is now recognized as such. In this type of device, if a large capacity memory is used, the control circuit becomes complicated and the cost increases, so it is preferable to use a device whose memory capacity is suppressed to a certain extent.

[Problems with conventional technology]

However, when loading coordinates into the memory sequentially,
For example, even if characters have the same stroke,
When a character with a long stroke is input, the memory overflows during the writing process, and the coordinate memory cannot be accessed any further, resulting in erroneous recognition.

[Purpose of the invention]

This invention was made against the background of the above-mentioned circumstances, and its purpose is to effectively write the coordinates k MJ into memory without increasing the memory capacity when characters with long strokes are input. An object of the present invention is to provide a character recognition device which prevents the occurrence of erroneous recognition.

[Key points of the invention]

In order to achieve the above-mentioned object, this invention calculates coordinates from the first coordinate of an input character pattern to a predetermined number.
The gist is that when writing to memory sequentially, if a character pattern with long strokes that exceeds the memory capacity is input, its final coordinates are written to the final storage area of the memory. It is.

〔Example〕

Hereinafter, one embodiment of the present invention will be congratulated with full reference to the drawings. FIG. 1 is a circuit diagram of, for example, an electronic wristwatch equipped with a character recognition device. In this case, the input section 1 should consist of transparent touch electrodes arranged in, for example, a 4x4 matrix on a transparent surface glass fixed to the top surface of the watch case. The touch electrodes are arranged on the XY coordinate system, and are arranged on the XY coordinate system, for example, 16X,
16=256 coordinate positions are formed by electrical processing. Also, when a human body such as a finger touches the touch electrode,
At this time, a contact capacitance component generated in the touch electrode is detected, a coordinate position is input, and a character pattern is input by kneading. This type of technology can be achieved by utilizing a patent application already proposed by the applicant (Japanese Patent Application No. 55-35660, title of invention: touch switch device). Then, depending on the input character pattern, Messeno is displayed on the display section 2, for example, when the alarm time comes. Note that this display section 2 is made of a liquid crystal display device.
Of course, time data is also displayed.

The ROM (read only memory) 3 is
1, and outputs microinstructions AD, I)A, OP, NAf in parallel. Then, the microinstruction ADld, ROM4,
Each of the data is applied to a RAM (random access memory) 5 as address data.

Further, the microinstruction DA is applied to the RAM 5 or the arithmetic unit 6 as data. Furthermore, the microinstruction OPU is applied to the operation decoder 7, so that the operation decoder 7 outputs the control signals cs, , Cs2, x%Y, Z%R1S, in response to the I″1.vc.
Output G, etc. The microinstruction NA is in the address section 8.
In response to this, the address section 8 receives micro instructions AD, D necessary for the next processing from signals d, c, etc., which will be described later.
Address data for reading A, OP, and NA is output and supplied to the ROM3.

RO'M4 stores vector strings of standard character patterns (n to be described later) for characters such as numbers and alphabets, and vector strings of characters C and turns input from the input section 1 (to be described later). Character recognition is performed by comparing the control signal C with the control signal C.
8, the data is read out.

The RAM 5 has various registers to be described later, and is used for time measurement processing performed by the arithmetic unit 6 and character recognition processing J! It is used in various processes such as lj, and data reading and 4H loading are performed under control signals C81 and R/W.

The calculation unit 6 performs various calculations ft under the control t11 signal X.
'i! Is it fast to finish one volume during the process? -4kF Adult 11 - Give to display section 2, RAM 5. Furthermore, when a judge operation is performed, a signal d indicating the presence or absence of data as a result of the operation, and a signal 4 indicating the absence of a carry (in signal ek, respectively) are output and supplied to the address section 8.

The S R type flip-flop 9 is set or reset by the control signals i; , signal R is output at the end of input. Then, the set output of C number +3 is input to the AND gate 10 as control (j8). Then, to the other end of the AND gate 10, input the clock of a predetermined frequency output by the U5 oscillation circuit 1. In addition, the clock output from the AND-DART 10 is input to a timer unit 12 consisting of a counter and counted. When the next stroke is input, the OFF pasting color amount f!L@i'' is applied.
The timing and opF timing are respectively controlled by the signal () when the signal G is output.
J-J to the RAM 5 and the arithmetic unit 6 through the gate 13
λ is received. Note that the timer unit 12 is cleared by the control signal S.

Note that the input section 1 and the display section 2 are each controlled fBIj by a control signal 2 or YK.

An oscillation circuit 11 constantly oscillates a reference frequency signal of, for example, 32.7681h, and supplies it to the frequency dividing circuit 14. and,
The frequency dividing circuit 14 outputs a 16-bit signal and applies it to the address section 8. Timing processing is executed once every 1/16 seconds depending on the stiffness.

Next, the standard vector string stored in the ROM 4 and the vector string of letters 4 and y inputted from the input section 1 will be explained. Figure 2 shows a number with a stroke number of 1.
2" The situation when inputting from all input section l is the whole building,
As shown in FIG. 2(A)l', when the character pattern "2" is input, its coordinate position data is sequentially written into the RAM 5. In this case, a maximum of 32 pieces of coordinate position data can be stored in the RAM 5 in one stroke. Therefore, as shown in Figure 2 [B], the character pattern "2"
After inputting , when the finger leaves the touch electrode of input section 1, R
The stroke length is calculated from the coordinate position data written in AM5, and then divided into six equal parts. Then, Figure 2 (
As shown in c), each equally divided point is approximated by a straight line from the starting point side to the ending point side, and the vector (0 to 708 Fi
The vector of each part is determined according to IW+'i).
The vector sequence is output. The selected vector sequence is then compared with the standard vector sequence in rLOM4;
The most similar character pattern is output.

Figures 4, 5, 6, and 7 respectively show standard vector sequences for each character pattern with a stroke count of 1, 2, 3, and 4, and are described in Lt OM, 4 above. .

Next, the configuration of the RAM 5 will be explained with reference to Figure 8'. .α.In this case, the 1゛∆ Luginuta is set to one foot B and 9 songs to detect the character separator υ, and the OF I” special J of the timer 8t112 is set to the constant value of the TM register. If it is smaller than the time, it will be the end of one stroke, and on the other hand, if it is thicker, it will be the ``4'' ending (of the - character), that is, the character separation.Also, FA, F, , F,
The register is for storing flags. 2 is a stroke number rank, which counts the number of strokes of the - character. Moreover, Y is an address counter, and input Fj
Each time the coordinates from B1 change, 32 memory areas M
It counts the addresses for specifying the i-th 7th address from 1 to Mst. Furthermore, memory candy areas M, ~M.

, Nike% Up to 32 coordinate position data are sequentially stored in one stroke. Also, memory area “1, m@, mf
i, m, - letter ρl stroke, 2nd stroke,
3rd stroke 1st% Write down the vector sequence of 4th stroke 1
Therefore, in this embodiment, one character can be input with a maximum of four strokes.

Next, the operation of this embodiment will be explained at station FP4 with reference to FIGS. 9 to 11.

First, an overview of the overall operation will be explained with reference to general 7o- in FIG.
It starts running every 16 seconds. And first, step S
, the calculation unit 6 executes the time measurement process of T of the RAM 5.
A predetermined operation is performed on the previous data in the register to calculate J1 current time data. Then, this current 1 is sent to the r-time data U display unit 2 and displayed.

Next, timer processing in step S2 is executed.

This timer processing requires some processing to be performed for a certain period of time in the flow described later, and if a certain period of time is preset in the 1M cash register, the predetermined period of time is counted each time this process is executed.

Next, in step S, a message setting mode 11, ie, power/failure/disconnection processing, is executed. On the face of it, this judgment process U1 determines whether the mode switch (not shown) for message setting is on or not, and whether or not the message setting mode is disabled. Cut L,
r Y E S J T ;l ij, &i, Sufu
7, S, 1/C proceeds to the literary magazine d1 (processing routine), and if ``No'' is answered in step S4, tl- proceeds to the publication[)i process of step S4. In this judgment process, it is judged whether or not the alarm time preset in the L register has been reached. If YES, the process proceeds to step S, where the fjlJE message data is read out from the RAM 5 and displayed. When the message is displayed again for a certain period of time, Stella fSa K determines that this is the case, and in step S, the display of the message is deleted.・On the other hand, Stetsu fS4
, When it is determined that the collapse is also "No" in S6,
This general flow process ends, and other processes (not shown) begin.

Further, when the setting of the message setting mode is determined in step S and the process proceeds to step s8, it is determined in step sIl whether or not the flag FA in the RAM 5 is rOJ. Therefore, when the character recognition device is not running, it is normally set in rOJ, and therefore the first step S and VC proceed, data "1" is set in the flag FA, and character recognition processing is being executed. It is remembered that Then, according to the flow described later, the recognition process of the character/gutter whose coordinates are inputted from the touch electrode of input $1 is executed (step S+o), and the inputted message data is processed (step S+o). 7'S,,l, After the display is confirmed on the display unit 2 (step 5I2), the flag FA is cleared (step fS
+s) and the character 8@ processing execution state is canceled. Note that in step S, when the flag FA is equal to rOJ, K returns to the process currently being executed at F.

FIG. 10 is a flowchart showing the specific contents of the character recognition process in step 5loVc. That is, when entering the character recognition process Stella 7'', first, the initialization process of step SA is executed.

Namely, F, register and F in RAM5.

The flag "1n" is set in the register, the contents of the stroke number counter 2 are cleared, and the address counter YVc"1n is set.

In this case, the sF1 register is 0 at the beginning of the - character.
2. 1# is set at the end of the stroke, and the F register is 0'' at the beginning of one stroke, and 11 at the end of the stroke.
'' is what you need to know.

Then, in the next step S, after executing the process of reading the input h1 or 14 marker position data, the process proceeds to step 7''S, and it is determined whether or not a character pattern is input from the input section l, in other words, the touch It is determined whether the electrode is touched or not, and if there is no touch input yet, the process moves to Stella fS, and it is determined whether the content of the Fl register is 1#.Since it is 1# at first, Returning to step s2, the same process is repeated, and a standby state is entered until a touch input is received.

Then, when there is a touch input, the process proceeds to step s4,
In the above step, the same process as 8 points is executed. Since the first digit is 1#, after clearing the contents of the timer section 12, the timing operation of the above OFF time is started (step 5s) 0 and , After clearing the contents of register F in the next step S6, proceed to the next step S.
It is determined whether the contents of the F2 register are 'l#' or not.

At first, since it is l#, the process advances to step 7''s8, and the increment processing of the contents of the stroke number counter Z is executed. The contents of the stroke number counter 2 are "1" and the i-th stroke is the target.Next, the contents of the register in VC1F are cleared (Stella 7'S9), and the process proceeds to step S+o, where the coordinate change is The presence or absence is determined.

In this case, when the character pattern is written by moving the finger, the coordinates change accordingly, but if there is no significant change, the process returns to step S2. That is, when the touch electrode is merely touched and the finger is not moved, it enters a standby state in which the loop of Stella fS2, S8, S4, S7, and SIo is repeated. Therefore, when there is a coordinate change, Stella fS1
．． , and checks whether the contents of the address counter Y are less than "32", that is, in this embodiment, one stroke has a maximum of 32
This judgment process detects whether or not the final coordinates of one stroke have been read. At first, the value of the address counter Y is "1", so the process proceeds to Stella 7' S t s and the memory ftt area ivir to M8 . Of these,
According to the contents of the address counter Y, the first input coordinate position data is written into the addressed memory area MY, that is, the memory area M. Then,
In the next step 814, the contents of address counter Y are +
After executing the increment process of 1, Stella fS
, return to. Therefore, the value of address counter Y is "3".
As a result of the repeated execution of steps 81m and S14 until %2 is reached, the %
The 31 coordinate position data are sequentially written into memory areas M, M, 1 according to the contents of the address counter Y. So, address? Count JY value is “32”
Then, proceed from step S, 1 to step Sat,
After the 32nd coordinate position data is written into the final memory area M, 8, the process returns to step S. In this case, 1
During stroke input, once the value of the address counter Y reaches "32", step Si1 is repeatedly executed, and 32 or more coordinate position data are stored in the memory area M1. will be written to.

Therefore, when inputting a character with a long stroke, for example, the number "8" as shown in FIG. The VC is loaded sequentially, and the landmark position data from point 0 to the [phase] switch is loaded into the final memory area M, 2 km. Therefore, memory area M
Since the contents of 3 are rewritten one-dimensionally starting from point O, the coordinate position data of point 3 is finally obtained, that is, the final coordinate position data in one stroke. This result 1. Point) ill of Q and ■ is approximated by a straight line in terms of coordinates.

On the other hand, - the touch electrode of the input 17b1 is used to start writing the next stroke after the first stroke of a character has been written, or in the case of a character with a stroke count of 1, to start writing the first stroke of the next character to be input. When you release your finger, step S is detected as a lever, and step 81 s
VC advances, but in this case, the contents of the Fl register are 0.
'', the process proceeds to step S16, and it is determined whether the contents of the %F register are 60#.In this case as well, since it is 0'', the process proceeds to step S16. Here, the stroke length t is determined from the coordinate appended data stored in each memory area M1 to M8 t of RAM:5.
) is calculated. Therefore, the stroke length calculated in this way is divided into 6 equal parts in the next step S, 8.
After the equally divided points are extracted, the peg torque of each part is determined according to the vectors in the diagram P3, and the village torque sequence is calculated (step S1). And this vector sequence is RA
Among the memory areas m1 to m4 of M5, the memory area mzK is addressed by the contents of the stroke number counter 2.
is written (step 5z-o). Therefore, 1
The vector sequence of the stroke is written into the memory area m. In the next step S1, 1'' is set in the F register, and then the process proceeds to Stella 7'Stt, where the OFF time of the timer section 12 is set in the TM register in the RAM S. Here, if the above-mentioned OFF time is longer than the above-mentioned fixed time, it is determined that the - character input has finished, but if it is smaller, the process returns to step S2. and wait for the next stroke to be input.

When the second stroke is input, the coordinate reading process is executed in the same manner as described above, and then step S1.
~Lx is added. For each character with a stroke count of 2.3.4, the process in step S'2Q is repeated 2, 3, and 4 times, so RAM 5
The vector sequences of the second stroke, third stroke, and fourth stroke are written in the memory areas m2, ln3, and m4 within.

In this way, - character input is performed, step S8
．． So this is the test 1j and proceed to step S2'3.%
"1-" is set in the Fl register.Then, as mentioned above, the vector strings written into memory areas 1 to 17+4 are classified into VCs by number of strokes, and the vector strings by number of strokes are ]70 It is compared with the standard vector sequence in M4 (step S, 4), and as a result, the most similar character pattern is extracted (step S2
．． ). Detects the difference in direction between the vector q and each vector in the standard vector sequence, which are equated by inputting a character pattern, and extracts the smallest sum of them as the most similar character pattern. be done. The character pattern extracted as described above is displayed on the display section 2 (
Displayed on STETSU f82g).

In the above embodiment, the XY coordinate system is composed of 256 points arranged in a 16×16 matrix, but the scale of the coordinates is not limited to the above embodiment. Furthermore, in the above embodiment, the number of strokes of characters to be input is up to 4 strokes, but the number can be arbitrary, and the types of characters can also be katakana, hirasu, kanji, -p, etc. Good too. Furthermore, in the above embodiment, all 32 pieces of seat position data per stroke are read, but the number may be arbitrary.

〔Effect of the invention〕

As explained above, this invention prevents the user from making a stroke that exceeds the memory capacity while sequentially loading the coordinates from the first coordinate of the input character pattern to the 75r foot into the memory. When a character pattern is input, its final coordinates are ripped into the final storage area of memory, so even character patterns with long strokes can be recognized accurately without increasing memory capacity. can.

[Brief explanation of drawings]

The drawings show an example of the implementation of this invention; e) Figure 11 shows character recognition W4. Electronic time-opening C circuit configuration diagram looking down on the device, No. Figures 2 (A) to (C) show the number “2” as a character pattern.
Figure 3 is a diagram showing the process by which a vector sequence is expected when inputting . Figures 5, 6', and 7 show Nutrok general 1.2.3.4, respectively.
Indicates the Yanagimu vector sequence of the character pattern consisting of [T
hjl, Oi, 8 + Nu1, ICA M configuration diagram, Figures 9 and 10 are 7 o
-f ya-), ZI+111? <1 is the figure shown in the figure below. 1...Human Resources Department, 3.4...ROM,
5...[LAM, 6... Arithmetic unit, M1
~M6. ...Memory area. Figure 2 (A) CB) (C) Figure 3 Figure 6 Figure 7 Figure 8

Claims (1)

[Claims] a detection means for detecting the coordinates of the 9th turn of the character; a first storage means having a predetermined number of memory areas and storing the coordinates detected by the detection means; and a detection means for detecting the predetermined number or more of coordinates between the patterns. and means for recognizing a character pattern according to the coordinates stored in the first and second storage means. Character recognition device.