JPH0120797B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a pattern recognition device that distinguishes the shapes of characters, numbers, symbols, etc. that are handwritten into a watch, calculator, etc. and performs a predetermined operation. Comparisons are made according to a predetermined permutation of standard glyphs, and the first standard glyph that matches is recognized as the desired handwritten character.

Conventionally, input means for wristwatches have often been push-button keyboards, but inputting characters such as alphanumeric characters requires a large number of key buttons, making it difficult to put this into practical use as a wristwatch. The character input device used in the present invention, for example, draws a character shape by touching and connecting a group of sensor terminals arranged on the top glass of a wristwatch with a standard character pattern registered in a memory circuit. The character shape is determined by comparison. However, handwritten character shapes have their own unique characteristics, and in order to always match a standard character shape pattern, it is necessary to prepare dozens of standard character shape patterns for a single character, as will be explained later. .

In this way, the capacity of the memory circuit increases,
Since the number of discrimination steps increases, there is a method of extracting only the characteristic parts of the character shape and making a judgment based only on those characteristic parts. In that case, there is no restriction on sides other than the characteristic parts, so it is safer against curly characters. However, when comparing all the shapes or all sides of a handwritten glyph and a standard glyph, a perfect match signal can be obtained, but when comparing with features, all sides match. There is no such thing, and therefore no output is obtained by matching. The glyph discrimination method of the present invention, when assuming one character, determines the standard glyph judgment scanning order based on whether or not there is another standard glyph or its characteristic part in the standard glyph of that character. (For example, among the sides of the character E, there is the character 〓, among the sides of the character 〓, there is the character 〓, and the character E is the character 〓
〓 is expressed as covering, and the judgment scanning order is E〓〓
) The side of the handwritten glyph and the side of the standard glyph or the characteristic part are compared from the beginning according to the judgment scanning order, and the first matching glyph is found. It is determined that it is a character.

According to the determination method of the present invention, even if the handwritten character shape does not completely match the character shape registered in the memory circuit, if it is the closest character shape, it will be selected, which has a great effect on curly characters. In addition, it has the advantage that the judgment circuit is simple and the judgment time is short. The present invention is most effective when performing pattern recognition using the characteristic parts of the character shape, but it is also applicable to a method in which all sides of the character shape are compared and determined.

A detailed explanation will be given below with reference to the drawings.

FIG. 1 is a top view of a wristwatch showing a part of the structure of the present invention. 1 is a case, 2 is a top glass, 3 is a gasket, and 4 is a band. This provides waterproofing and insulation.
In the gap, sensor terminals 11 to 1 are inserted as shown in the figure.
9 is placed. These sensor terminals 11
~19 is case 1 due to contact with fingertips, etc.
It is configured so that the sensor terminals are grounded according to the order of the letters drawn by the fingertips. Therefore, these sensor terminals 11 to 19
However, by detecting the order in which they are grounded, it is possible to determine the drawn character shape.

FIG. 2 is a partial sectional view of FIG. 1 showing the state in which the sensor terminal is attached, and shows the center sensor terminal 1.
5 has a hole in the center of the upper glass 2 and is taken out to the outside by a metal terminal 15A, and the surrounding sensor terminal 11 is connected to the case 1 by a gasket 3.
It is insulated from the top glass and taken out to the outside by a metal vapor-deposited wire 11A from the second side of the upper glass. The metal vapor-deposited wires 11A and 22 are connected to terminals 29 and 30 of the circuit board 27 by springs 25 and 26. As shown in FIG. 1, there are eight ambient sensor terminals 11, each of which is independently connected to the circuit board 27. As is clear from this figure, each sensor terminal is insulated from the case 1 which is grounded, and when a fingertip touches it, it is grounded to the case 1 through the human body, and the input order of the character shape is changed to the circuit board 27. It's like it's given to you.

FIG. 3 is a configuration diagram in which sensor point inputs are configured as sensor side inputs in order to determine the character shape detection that is a part of the present invention as side inputs.

11 to 19 are nine sensor point inputs, and if each point input position is connected with a straight line to the adjacent point input position, side inputs 41 to 6 shown in this figure are obtained.
8 is obtained. There are 28 side inputs 41 to 68,
By using these edge inputs, the 26 alphanumeric characters, numbers, general symbols, etc. can be easily expressed. An example is shown in FIG.

Figure 4 is a diagram of the character shapes E, G, B, and Q expressed by the method shown in Figure 3 using nine sensor terminals.
1 to 19 are in the ON state as the fingertip passes through them, and the shape of the letters cannot be determined just by the combination of ON and OFF of the sensor terminals. However, when identifying by edge inputs, in glyphs E and G, glyph G has edge input 62 shown in FIG. 3, and glyph B has edge input 5.
5, 57, and for the character Q, the side inputs are 62, 58
The shapes E, G, B, and Q are the third
The discrimination can be made by the side inputs 41 to 68 shown in the figure. Moreover, the glyph patterns that are registered in the memory circuit in order to distinguish each glyph shape only need to register the different parts of the side input for each glyph, and only detect whether or not that side input was received. The capacity of the memory circuit can be very small, and the number of determination steps and time required can be reduced.

FIG. 5 is a glyph diagram in which the pattern recognition part registered in the memory circuit in the glyph shown in FIG. 4 is shown with solid lines.

For character E, side inputs 68, 67, and 64 are registered, and for character G, side inputs 45, 67, and 6 are registered.
2. For character B, side inputs 67, 64, 56, 5
7. For the character Q, side inputs 45, 55, and 62 are registered. Each registered edge input has features unique to its glyph shape, which are determined by the interrelationship of all alphabets and numeric glyph shapes. For example, when drawing the letter G, the letter E
can be distinguished from the presence of the side input 62, the character B from the absence of the side input 57, and the character Q from the absence of the side input 55. . Therefore, as mentioned above, the glyph pattern to be registered in the memory circuit only needs to be input from a part of the glyph, that is, the characteristic part of the glyph that is not present in other glyphs, and handwritten input is limited to only that characteristic part. It would be good to compare it with. As a result, the effects are a reduction in memory capacity and a faster comparison time.

In this way, the method of registering only the characteristic parts of each character shape in the memory circuit and making judgments is not limited to the edge input method adopted in this example for convenience of explanation, but also the method explained in Figs. 3 and 4. A point input method may also be used.
However, in order to accurately distinguish character shapes, it is necessary to increase the number of sensor terminals or change the order of strokes drawn on the sensor terminals.
It is necessary to register it in the memory circuit.

FIG. 6 is a waveform of a character shape determination circuit that is a part of the present invention, which converts the sensor terminal input waveform into a side input using the character G in FIG. 5 as an example, and shows the radius for comparison with the memory circuit. In the diagram and block diagram,
101 is a waveform diagram when the sensor terminal is touched with a fingertip, 13F, 12F, and 11F are the waveforms of each sensor terminal, and 113 is the sensor terminal waveform 13F
triggered by the sensor terminal waveform 12F
Similarly, the flip-flop output waveform 112, which is turned off by the delay time 112D of
Flip-flop output waveform turned OFF, 111
is the flip-flop output waveform triggered by sensor terminal waveform 11F. Waveform 113
The trailing edge of 112 is indicated by diagonal lines 13, 12 and 12,
The parts indicated by 11 overlap with waveforms 112 and 111, respectively, and are the sensor terminal waveforms, 13F and 12F.
This shows that the sensor terminal waveforms 12F and 11F were input successively. In other words, the point input of the sensor terminal is converted to the side input, and the diagonal lines 13, 1
2 is the side input 51 indicated by the same number in Fig. 3, diagonal line 1
2 and 11 can now be handled as edge inputs 45.

The waveform 130 is a collection of these side input waveforms using an OR gate, and is input into the handwriting input register 140 from the beginning as shown in the figure. Of course, the side input 51 is, for example, a bit
0001, the side input 45 is bit 0101, as shown in the diagram.
It is stored in code converted form.

On the other hand, the memory circuit 141 includes, for example, the side input characteristic portions 45, 6 of the character "G" shown in FIG.
7 and 62 are converted into bit codes and stored, and at the same time as the completion of one character in the handwritten input, the character 45 of the character "G" exists in the handwriting input register 140, and if so, the characteristic Sections 67, 62, and so on are compared by the comparator circuit 150, and if all side input bit codes exist, at least the condition of the letter "G" is satisfied, and as shown in FIG. ``It is not a glyph EBQ'', the thing is determined.

FIG. 7 is a general example of cursive characters based on side input, which depicts all possibilities of handwriting for the character "P". The method for comparing standard glyphs is 1 in this figure.
~ As shown in 5, the sensor terminal 1 used
1 to 17 are different for each character shape, so by registering all of them in a memory circuit that registers standard character shapes, no matter what shape of the letter "P" is input by hand, any of the characters 1 to 5 in Figure 7 will be registered. The purpose of this is to make sure that they match. The standard character shape to be registered in the memory circuit can be configured by inputting the sides as explained in FIG. 4. For example, in FIG.
5, 51, 53, 68, and 4 in this figure 5.
4, 67, 45, 50, 56, and 68 are registered so that matching inputs can be detected.

However, when registering side inputs, the capacity of the memory circuit can be further reduced as described above by registering only the characteristic parts that are not found in other glyphs. This will be explained with reference to FIG.

Figure 8 is an example of a method in which only the characteristic parts of the side input of the character "P" in Figure 7 are represented by solid lines.
correspond to the character shapes in FIG. 7, respectively.

However, there are only two or three sides of the standard glyph registered in the memory circuit, sides 45 and 53 in FIG. 1, and sides 44, 50, and 56 in FIG.
Compared to the method shown in FIG. 7, the amount is one-half to one-third. Characteristic parts of such glyphs are determined by the mutual relationship with other glyphs used, and the edge input of the extracted feature parts is, for example, ``features that exist only in the letter ``A'' and ``B''. It is composed of ``sides with characteristics not found in the characters.'' If the character shape memory is constructed by "inputting the edges of characteristic parts" in this way, the dotted line portion shown in this figure can be drawn in any way, and the range of application to curly characters will be further expanded. This is shown in FIG.

FIG. 9 shows another way to draw the character 1 in FIG. Side 81
84 changes as shown in the figure, indicating that there are quirks created by the person. However, the edge inputs 45, 53
If there is, all the characters will be judged as "P", so
1 to 5 are all determined to have the character shape "P".

Similar changes can be applied to Figures 2 to 5 in Figure 8, so in this example, approximately 25 types of curly characters are countered for one character "P". Become.

FIG. 10 shows the characteristic parts of uppercase letters, numbers, symbols, etc., using solid lines. As explained in FIG. 8, these characteristic portions, ie, side inputs, are unique characteristics of each character shape, and are composed of combinations that are not found in other characters.

Also, some of the glyphs shown in this figure, similar to the glyph "P" in Fig. 8, have the same letter represented by various shapes, and are numbered like P ₁ R ₂ ... P ₅ . can be,
It is registered in the memory circuit so that it can be correctly recognized no matter what shape is drawn, in response to the drawing habits of people. As explained in FIG. 9, the several dotted lines of one character shown in this figure are not related to character shape discrimination no matter which path they take, and can be treated as a human habit of character shape. I can do it.

In this figure, when drawing a glyph, all the sides where the fingertip connects the sensor terminal are shown with dotted lines and solid lines.In particular, the dotted line is the side where the fingertip is input due to the stroke order of the glyph. Even when separated from the terminal board, that is, when connecting one sensor terminal to another sensor terminal in space, it is displayed as one side input.

For example, the character shape "〓" is drawn in one stroke, but the character shape "A ₁ " shows the state in which the fingertip is once removed from the input terminal board, as shown by sides 150 and 151. Also, as shown in side 170 of the character "Z", sides 171,
Even if you intend to draw 172, if your fingertip does not touch the sensor terminal 13, the sensor terminal order will be 1.
2, 13, the edge input becomes 170, edge 1
This is because 71 and 172 may not be input. The reason why a number of dotted lines are shown in one glyph in this figure is due to the reason mentioned above. It expresses the case.
Therefore, there are many different ways to draw a single character, and there are many cases in which the sides of characteristic parts of other characters are used as covers. For example, if you draw the glyph "C", the set of dotted lines and solid lines of the glyph "C" will form the glyphs "J ₁ " and "L".
It can be seen that the characteristic parts of "V ₃ " are covered. Similarly, if you draw the character ``P <sub>3</sub> '', you can see that it covers the features of the characters ``C'' and ``L''. In this way, when drawing one character, if there are many glyphs covered, select only the correct glyph from among them,
A determination circuit must determine whether the character shape is the desired character.

_The present invention relates to the method of this determination circuit.
First, the determination order of each character shape is determined so that it is selected earlier than "V ₃ ". In other words, the glyphs to be covered are arranged after the glyphs to be covered.
In the case of the character shape "P ₃ ", the character shape "C" as mentioned above.
From "L", the arrangement is ahead, and the character "L" is
Since it covers the glyph "J ₁ ", make the following array,
Since the glyph "J ₁ " and the glyph "V ₃ " do not cover each other, either one can be arranged first.
As a result, if the character shape is only the above five characters, "P ₃ "
The arrangement order is "C", "L", "J ₁ ", and "V ₃ ". Here, if the character shape "C" is drawn, the determination circuit of the characteristic part in FIG. 6 should recognize that the character shapes "C", "L", "J ₁ ", and "V ₃ " match. Actually, the character shape "C" is selected based on the above determination order. If you draw the letter “L”, you can also draw the letter “L” and “J ₁ ”.
match, but the character shape "L" is selected based on the determination order. In this way, one correct character is recognized.

In Figure 10, the characteristics of each character are determined taking into consideration the above ranking, and not only the characteristics of each character, but also the characteristics of each character when covering other characters. The edges of the feature are configured so that they are not covered again by the characters. For example, even if the glyph "C" covers the glyph "L", it will not be possible to create a judgment ranking unless the edges of the feature are selected so that the glyph "L" does not cover the glyph "C". . Also, the character "C" covers the character "L", the character "L" covers the character "J ₁ ", and the character "J ₁ " covers the character "C" or "P ₃ ". too,
Judgment rankings cannot be created. In other words, the edges of the feature must be constructed so that they always form an open loop, rather than a closed loop where the glyphs they cover loop with each other.

The fewer the sides of the feature, the smaller the capacity of the memory circuit, and the more dotted line parts of the character shape in Figure 10 increase, so the response to cursive characters becomes better, but the risk of a closed loop increases accordingly. . The reason why the numbers 1 to 0 in this diagram are limited to half the size of the alphabet letters is to minimize the overlap between the features of the numbers and letters.
This is a device that allows many character shapes to be distinguished using a small number of feature edges. As a result, the character shape shown in FIG. 10 satisfies the above-mentioned open-loop condition and has the greatest number of countermeasures against curly characters with the least number of feature sides. In this diagram, the character shape that covers the most characters is "B ₂ ", which covers the following character shapes.
“A ₃ , C, D ₃ , E, F, H, I, J ₁ , K ₂ , K ₃ ,
L, M, O, P ₁ , P ₂ , P ₃ , P ₄ , P ₅ , S, T, U,
V ₂ , 1 ₁ , 3 ₁ , 3 ₂ , 4 ₁ , 4 ₂ , 7 ₁ , -, /" or more, 30 or more character shapes, not covered by other character shapes, and placed first in the judgment ranking do.

On the other hand, the most commonly covered glyph is "1 ₁ "
and is covered by the following glyphs: “A ₃ , B ₂ ,
D ₁ , H, M, N, O, P ₂ , Q, R ₁ , R ₃ , U,
V ₃ , W ₁ , W ₂ , W ₃ , W ₄ , 2 ₁ , 2 ₄ , 3 ₁ , 4 ₁ , 4 ₂ ,
7 ₁ , 8 ₁ , 8 ₃ , 9 ₁ , 9 ₂ 0'' and above are the 28 glyphs that do not cover any other glyphs and are placed at the end of the judgment ranking. In the same way for other glyphs, select the glyph to cover,
When the glyphs to be covered are examined from the dotted lines and solid lines of the glyphs and arranged in the same manner as the glyphs "P ₃ ", "C", "L", "J ₁ ", and "V ₃ " are arranged, the glyphs shown in Figure 10 are obtained. The ranking will be as follows. “B ₂ , B ₁ , Q,
G, W ₄ , X, W ₁ , W ₂ , Y, Z, 2 ₄ , E, S,
O, A ₃ , W ₃ , R ₁ , R ₃ , N, M, R ₂ , D ₂ , 8 ₁ ,
9 ₁ , 6 ₁ , D ₁ , P ₂ , P ₅ , D ₃ , A ₂ , D ₄ , P ₁ , K ₁ ,
P ₄ , 9 ₂ , A ₁ , 8 ₃ , 8 ₂ , 6 ₃ , U, H, K ₃ , F, P ₃ ,
6 ₂ , 2 ₁ , 0, _{2 3} , 3 ₂ , 4 ₂ , 2 ₂ , 7 ₂ , 5 ₁ , 5 ₂ , 3 ₁ ,
7 ₁ , 4 ₁ , C, V ₃ , K ₂ , L, J ₁ , V ₁ , T, J ₂ , V ₂ ,
−, 1 ₂ , /, I, 1 ₁ ”, resulting in a permutation of 71 characters.

As soon as one of these 71 glyphs is handwritten, it is scanned in the order of "B ₂ , B ₁ , Q, G...", and the first matching glyph is determined as the desired handwritten character. Ru. This circuit system is shown in FIG.

FIG. 11 shows an embodiment of the pattern recognition determination system of the present invention. The judgment circuit in this figure is the sixth
It employs the method of ``comparing the sides of glyphs'' as explained in the figure, and uses the glyph "C" as an example of a handwritten character. The memory circuit 140 has “B <sub>2</sub> ”
Characteristic parts 67, 64, 56 on the sides of "B <sub>1</sub> ", "Q"...
57, etc. are stored, and are arranged in the order of "character shapes that cover other character shapes" mentioned above. If you draw the letter C, for example, sides 51, 45, 44, 67, 6
4,61 is stored in the handwritten register 141 in the form converted to the bit code explained in FIG. One input 201 of the comparison circuit 150 is the input of the side of this handwriting register 141, and the other input 202 is the input of the side of the first glyph "B <sub>2</sub> " of the memory circuit 140 from the characteristic parts 67, 64... is input. If the same bit code as side 67 of ``B ₂ '' is found in the handwriting register, then the bit on side 64 is looked for, and if that is present, then the bit on side 56 is looked for, and if not found, ₂ ” and the letter “C”
is negated, then the glyph "B ₁ " and the glyph "C"
The comparison begins. Similarly, side 48 of the glyph “B ₁ ”
is in the handwriting register 141, the output of the comparator circuit 150 is checked, and if not, the match between the character "B ₁ " and the character "C" is denied, and the comparison with the character "Q" begins next. .

In this way, the shape of the memory circuit 140 is
The handwritten input of the letter “C” is compared one after another,
It is impossible for the characters to match up to the arrangement order of the character "C". The reason is that the configuration of the handwritten input sides of the character "C" listed here is "B ₂ ", "B ₁ ", "Q"...
This is because the sides of the characteristic part of "4 ₁ " are in the "uncovered arrangement" as described above, and it is not until the sides 45 and 67 of the characteristic part of the next character shape "C".
These two sides 45 and 67 are the handwritten register 14
This is because it is designed to exist within the side numbers 51 to 61 inputted in No. 1.

Therefore, first, the character "C" in the memory circuit 140 that satisfies all sides of the characteristic part is the desired character in the handwriting input register 141. The glyphs "V ₃ ", "L", and "J ₁ " that are arranged after the glyph "C" are also matched because they are "covered" by the glyph "C", but because they are arranged after the glyph "C", they match. , only the letter shape "C" is selected. Although the character shape "C" has been taken as an example here, matching outputs can be obtained by following the same process for other character shapes.

In FIG. 11, the arrangement order of glyphs is determined based on the characteristic parts of the sides, but not only the characteristic parts of the sides, but all the sides may be compared. For example, in the glyphs "E" and "F", it is clear that the glyph "E" covers "F", so "E" has a higher priority than "F", and the glyphs "〓" and "〓" Then, since "〓" covers "〓", "C"
is placed before "L".

Also, it is possible to target the stroke order of the character shapes when the fingertip passes through the sensor terminal, and it is also possible to increase the number of sensor terminals and use the point input of the sensor terminal itself as a comparison target. be.

As described above, the present invention includes an input terminal board in which a plurality of sensor input terminals are arranged for pattern input, and a handwritten input character shape drawn by passing through the sensor terminals as an input signal of the position of the sensor terminal. A handwriting input register to be stored; a memory circuit that stores a standard character shape drawn by passing through the sensor terminal as a signal of the position of the sensor terminal; and a character shape is determined by sequentially comparing the contents of the handwriting input register and the memory circuit. a comparison circuit for making a determination, and the scanning order of each standard character shape in the memory circuit is arranged in accordance with the order in which the components of the standard character shape are included in the components of the handwritten input character shape. The arrangement order is such that when all of the standard glyph information stored in the memory circuit is included in the constituent elements of one target handwritten input glyph, the included standard glyphs are When all the information of the standard character shape of the handwritten input character shape stored in the memory circuit is included in the constituent elements of the other handwritten input character shape, the handwritten input character form It is characterized by arranging the standard glyphs of other handwritten input glyphs in the memory circuit in a higher order than the standard glyphs of . In this way, the comparison circuit sequentially compares the handwritten character shapes input in the handwriting input register with the standard character shapes in the memory circuit arranged according to a certain rule, starting from the topmost standard character shape. Then, the handwritten character shape reaches a rank that satisfies all of the standard character shapes, and the standard character shape at that rank becomes the desired handwritten character shape.

The present invention is not limited to the uppercase letters of the alphabet used in this embodiment, but can also be applied to lowercase letters of the alphabet, foreign languages other than English letters, kana, kana, etc. using similar means.
It can be applied to pattern recognition of hiragana, kanji, etc. Furthermore, the terminal board for handwritten input can be installed not only in the wristwatch used in this embodiment, but also in clocks, pocket watches, calculators, etc., and it can be said that the range of application of the present invention is extremely wide.

[Brief explanation of drawings]

FIG. 1 is a top view of a wristwatch showing a part of the configuration of the present invention, FIG. 2 is a partial sectional view of FIG. 1, and FIG. The configuration diagram, Figure 4, is a glyph diagram with the configuration shown in Figure 3.
Fig. 5 is a glyph diagram showing the characteristic parts of the glyph in Fig. 4;
FIG. 6 is a waveform diagram and block diagram of a character shape determination circuit for determining the character shape shown in FIG. 5, and FIG. 7 is a diagram of the character shape "P"
A glyph diagram based on the configuration of Figure 3 showing the changes in , Figure 8.
The figure is a glyph diagram showing the characteristic parts of the glyph in Fig. 7,
The figure is a glyph diagram showing the change in glyph shape in Fig. 8, 1.
Figure 0 shows the changes in character shapes in Figure 9 applied to other character shapes, and shows the characteristic parts of uppercase letters, numbers, and symbols of Alphabet, as well as handwritten input. Figure 11 shows the pattern recognition of the present invention. FIG. 2 is an explanatory diagram showing an example of a determination system. 11-19...sensor terminal, 41-68...
Side input, 140...Memory circuit, 141...Handwriting register, 150...Comparison circuit.

Claims (1)

[Claims] 1. A plurality of sensor terminals disposed on an input terminal board, a handwriting input register that stores handwritten input glyphs of numbers and characters drawn by tracing the sensor terminals as input signals at the positions of the sensor terminals, and the sensor terminals. A pattern recognition device comprising a memory circuit that stores standard glyph shapes in advance as a position signal, and a comparison circuit that discriminates glyph shapes of numbers and characters by sequentially comparing the contents of the handwriting input register and the memory circuit, The order of arrangement in the standard glyph consisting of the constituent elements constituted by points or edges of the memory circuit is such that when the constituent elements of one standard glyph include the constituent elements of the other standard glyph, the one standard glyph is ranked as the other standard glyph. A pattern recognition device characterized by arranging characters in a higher order than character shapes.