JPH03502841A - character recognition device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] character recognition device (Technical field) The present invention relates to an apparatus and method for identifying characters.

(Background technology) Trade between non-English and Western European countries has expanded dramatically, leading to For example, in the past, communication between English countries and Kanji countries has increased in importance. When a document written in English is received in a Kanji country, it must first be sent to the government's translation center. transferred to the center. There the document was translated, rewritten by hand into Kanji, and finally The document is delivered to its destination. Once a response to a translated document is prepared, the response is It is translated from kanji to English at the government's translation center and forwarded to the correspondent in the English country. death However, from English to Kanji and vice versa, calligraphy 1! Use translator to fix This was problematic in that it caused serious delays in the popular process. The ninth step to overcome this difficulty is to have a key that displays the ideograms of the kanji. A plyter device was developed.

This device is equipped with an operator who is proficient in kanji and capable of using a typewriter. It is possible to produce hard copies of documents written in kanji by employing can.

However, since Kanji contains over 51,000 different ideograms, Typewriter device VC has 0 functions that are problematic in that it requires a large number of keys. This type of device reduces the number of keys required by using keys. improvements have been introduced, but the problems described above still remain. Besides, type When using a writer device, learn how to fully use the keyboard device. This requires extensive training for operators, and the process is Another problem remains in that it is expensive and time consuming.

A handwritten ideogram for your friend who overcomes the problems encountered when using keyboard devices. An ideograph detection device has been developed that receives and identifies ideograms. Ideographic characters must be written on an input device and must be written characters. A predetermined basic line element that is a typical line element used by anyone writing in It needs to be formed from a primitive (word root). Ideographic characters are in the device After inputting the input ideograms, the device generates the input ideograms formed by the input ideograms. and compares the input primitive with the P3 content of the index table. index table In order to accommodate a wide variety of users' handwriting, each predetermined number of Remember the changed primitives. Each primitive stored in the table The primitives that form characters are usually determined by the device. It will be done. Tables are also used to form each letter in ideographic languages. It also remembers several sets of primitives. several pairs of preforms that form the input character If a primitive matches one of several sets of primitives in the index table, then Output code in combination with the primitives is generated and communicated to an output device. this is Allows a hardcopy image of the input ideogram to be formed. deer However, due to the large variety of each primitive stored in the table, the device The processing speed of 5J! The problem is that it makes timekeeping inappropriate. There is a problem.

Additionally, a predetermined number of basic line elements or primitives used in the device teeth,! As rtgL, 5 Rima friends will be lower than that, and Maku will be 2o or higher. sea urchins have been selected. of a subset to form any ideogram of a language. By using only five basic primitives, Tomoe ideographs appear unique. Although many different ideograms can be formed from the same set of primitives, There is a problem in that it is done. This allows the device to distinguish between various ideographs. This results in a decrease in performance.

To try to overcome this problem, we can add 20 or more primitives to the subset. Next, include bu. However, a set of primitives that represent one or more ideograms Although the number of ideographs has decreased, it has become clear that various ideograms are primitives of the same family. The same problem still exists in that rays are formed. However, a subset By increasing the number of primitives in the circle, the processing time of the device is further extended. Another problem exists in that

Furthermore, these devices are capable of detecting characters written in one language. However, another problem exists in that multilingual character detection is not possible. follow Therefore, improved character recognition equipment is needed.

It is therefore an object of the present invention to eliminate or alleviate the above-mentioned disadvantages.

(Disclosure of invention) According to the invention, nine characters are made of several primitives, i.e. a predetermined set. A character recognition device is provided which identifies the character variant and the primitive t which are the constituent elements of However, the device successively receives each of the primitives forming a character, and this receiving primitive input means for generating an input signal for each of the blocks; receiving all input signals and identifying each of said primitives received by said input means; A process that generates a character code to display the character by identifying the primitive. means, storing a character code and a combined output code for each of said set of characters; storage means, for this inputted character in order to identify the inputted character; comparison means for comparing the generated character code with each of the character codes in the storage means; and is continuous with the comparison means, and the previous entry is made by character identification by this comparison means. output means resulting in the reproduction of input characters; Contains.

The apparatus further includes: examining the input signal t generated for each of the primitives; When the character code is equal to the character code combined with the output code of multiple features, the combination with the character An identification means is also provided for operator error to identify the output cord tube. It is desirable that

In this device, the input character code is the same as the character code stored in the storage means. The probability that the character code generated for the input character is equal to comprising replacement means for selecting the highest character code stored in the storage means; is desirable. Also, the output means is the formation of the reproduction of the received ideogram or ideogram. printer, audio synthesizer or video player that allows the formation of audio playback. at-include at least one device selected from the group consisting of video display terminal devices. It is also desirable to

The character recognizer recognizes all ideograms, uppercase English characters, and Rocha characters. ii the written characters! It is desirable that you have the ability to understand.

The predetermined set of basic primitives contains 20 unique primitives. It is also desirable that the various combinations of those primitives On the other hand, the same series of letters form virtually all the letters of languages with different numbers. This will reduce the occurrence of diverse characters formed from primitives. follow The use of 20 individual primitives means that one or more international output codes Reduces the occurrence of input characters that display a character code equal to the character code combined with Ru. This, of course, increases the probability of detecting the correct ideogram.

(Brief explanation of the drawing) Embodiments of the invention will now be described - by way of example only and with reference to the accompanying drawings. Now.

FIG. 1 is a functional block diagram of a device for character identification; Figure 2 is a diagram showing ideograms; FIG. 3 is a diagram showing the basic primitives used in the device shown in FIG. 1; Figures 4a-4c represent the letters shown in Figure 3 and from Nine Pleasatigue to Figure 2. Figure 5 shows a more detailed basic block diagram of the apparatus shown in Figure 1. lock diagram, Figure 6 is a detailed basic block diagram of a portion of the apparatus shown in Figure jlE1. , FIG. 7 is a diagram showing a coding method used in the device shown in FIG. 1; Figures 8a and 8b are diagrams showing input basic strokes; 9th heat and 9bg are diagrams showing more ideograms, FIG. 10 is a diagram showing a probability matrix used in the device shown in FIG. 1; Figure 11 shows one English character, Figure 12 shows more English characters. This is a diagram.

(Ninth Best Mode for Carrying Out the Invention ta> With reference to Figure 1, handwritten character recognition A device 1° is shown. The device 1o is connected to the data processing device 14 and the next input includes device 12; Input * The IH2 receives handwritten characters and converts them into a series of signals. , and then transmits it to the data processing device 14. Data processing device 14#i, this receiver is All signals are processed and entered! The next character input at t12 is detected. Output device 16 is also connected to the data processing device, and the handwritten text accepted or rejected by the input device 12 is connected to the data processing device. Receive an international ASCII output code that displays all characters. This is reproduction of handwritten characters allows to be generated.

The device 10 is operable in several modes, each mode being able to handle handwriting in a different language. Allows characters to be recognized and reproduced. The selection means 18 selects which It is provided so that you can choose how to operate the device jut by language. Therefore, processing The means 14 responds to the selection means 18W and records appropriate information for the six words F# individually. stored and accessible according to the mode selected by the selection means 18. It is divided into sections 14a, 14b---14yaK.

For the sake of brevity, the apparatus shown in FIG. Characters in other languages can be detected in a similar way by subjecting It must be understood that the processing means 14 for detecting ideograms I will explain this when it is conditioned.

Referring to FIG. 2, an ideographic IC is illustrated.

As can be seen in the figure, the ideogram IC is formed by several basic line elements, i.e. primitives. The primitives are named from Prl to Prs, respectively.

Primitives Prl to Pr3 are basic line elements that are sometimes used in writing in ideographic languages. Ru.

The writing order of the ideographic line element order is the king of logical efficiency, experience and human innate Based on habit. Several research results show that there are some basic rules when writing ideograms. There is a rule of thumb and it is as follows: Horizontal - Vertical Left tilt - right tilt First-in - last-close Each Kanji may use one or more of the above rules in forming the character. . An example of the basic line element order of an ideogram is shown in Table 1 below: Table-1 Reduce the number of primitives users need to write when forming ideograms This reduces the amount of data that must be processed by the five processing units 14. In order to Use Pr, to Pr. The 15 primitives Pr, to Pr□ are the table A variety of words that are members of a set of basic line elements, which are used to form letters. All of the ideograms in the word are in various combinations from the primitives Pra to Pro. Such a subset of primitives is selected because it can be made by Ru. The device is working to detect characters written in another language as described below. When the primitive Prp to Prt goes from the primitive Pra to Pro Used with Tsuka 5, apparatus 10 is better illustrated. The input device +112 is Online digitizer tablet 2ot-includes stylus 201.

The ideograms to be recognized are written on the tablet 20 with a stylus 20a. child From rL#C, ideogram IC=i-shaped base forming force and friend primitive Pra For each of Pro, a series of Cartesian coordinate data point signals PNo to PN N is generated. The capital letter °N' in the data point signal forms the letter ICt. refers to the order in which the primitives were input, while the subscript 1N” refers to the order in which the primitives are input along the primitives. indicates the number of extraction points F. The data point signal is then sent to data processing unit 14. communicated.

The storage device f122 is positioned within the data processing device 14 and is connected to the digitizer 20. Connected. Storage device 22 receives the raw Cartesian coordinate data point signal. , store that signal before processing. Preglosse? 24 is each input primitive receives a copy of PNN from Cartesian coordinate data point signal PNo for Perform pre-processing to remove redundant and false data, and perform thorough data processing. The Cartesian coordinate data signal is sent from the preprocessor 24 to the function extraction unit 26. The extraction unit extracts Cartesian coordinate data for each of the input primitives Pr. The data point signal is converted into a vector code and a series of Nuka2vc transforms.

The vector code and series of scalars generated by the feature extraction unit 26 are The input signal is given to the primitive detection unit 28, and the detection unit receives each input of the character ICI type bottom formation. The vector code generated from the primitive Pra to ProK is indexed. Compare with the existence of a table or dictionary. This means that the processing device 14 Whether the primitive is a component of the 15 primitives Pri to Pro enable detection. The input primitive Pr is the primitive detected A vector code combined with only one of the 15 primitives stored in section 28 primitive code 1 at time t which results in the formation of a vector code equal to the code O is generated from and sent to the storage device 50. This process For each primitive Pr f that forms the character ICi, a row for each vector code to be displayed. be called. Therefore, the set of primitive or character codes is the ideogram I Ct-C is generated for the input character. However, the input If the vector code generated for primitive Pr is 　MitipPra to a vector code combined with one or more of Pro, etc. If the input primitive is correct, the detection unit 2B Tests are performed on a series of scans combined with the vector code.

The generated character code is sent from the storage device sO to the character detection unit 32, and is stored in the second index. Compares with the contents of a table or dictionary. The detection unit 32 detects each of the ideograms of the language. Memorize the character code that is displayed. The character code to be memorized is It is made by combining the 15 primitives shown in Figure 3, and these characters are It is a necessary condition to input data into the tablet 20 in the order determined by the rules mentioned above. This is a matter of concern. The rules mentioned above are mostly used when writing in ideographic languages, so from a primitive typed in the wrong way. The formed character code is omitted from the index table.

The t character code generated for the input ideographic character IC is read by the character detection unit S2. If it is equal to the output character code, the combined output code or international A8CII output The power code is output to storage device 34. However, if there is more than one character code If the character codes representing the above ideograms are equal to 9, the character detection unit 32 stores the 2. Perform operations on the raw Cartesian coordinate data point signals stored in 2. Determine the correct ideogram IC=i displayed by the character code. This is an accurate international ASC II code can be output to storage device 54.

Equipped with 56 substitution/correction devices, the input character code is stored in the character detection unit 28. The zero substitution device 56 checks the input character code when it is not equal to the input character code. , it is assumed that the input character code will be displayed for the next input character code. The character detection unit 3 performs the above-mentioned processing by substituting all the most reliable character codes. Send it back to 2.

In the storage device 54: The international A19CII code stored in the ideogram C1 represents Output device or device 115 [has been added, which device outputs the ideograph phonetic and A video display terminal (VDT) 16a on which a visual reproduction can be formed.

The printer 16b and/or printer 16b also have a video synthesizer 16c and the like.

Referring to FIG. 6, processing device 14 is better illustrated. preprocessor 24 is explained to remove redundant and false Cartesian coordinate data point signals. It includes a comparator 24a and a storage device 24b which function in the manner described.

The function extraction unit 26 extracts adjacent Cartesian coordinate data that completely forms each primitive Pr. A second comparator and a second comparator function to generate a vector for the point signal. and an index table or dictionary. Storage device 26c receives the vector; Next, the vector is sent to the fifth combiner 126d. Comparator 26d hahek Remove redundant information and create a set of unit vectors for each primitive. form a tor or vector cord and a series of squash. Sukashi is a member of each Primi The length of each unit vector in the vector code generated for the The vector code and set of scats generated for each primitive represent It is sent to the storage device 26e and stored before being sent to the primitive detection unit 28. .

Primitive detection s28 includes a second index table or dictionary connected to the second index table or dictionary 28b. Equipped with all 4 comparators.

Table 28b contains predetermined vectors from each primitive Pra to ProK. Memorizes a list of torque codes and primitive chords. Vector code is 1 Represents one or more of the five primitives Pra to Pro.

The primitive detection unit 28 stores the space generated for each vector code. It also includes a storage device 28c and a test section 28d. The test section 28d is The vector code in combination with the squash represents one or more of the 15 primitives. Performs an operation on a series of vectors to determine if the vector code is equal to Now, this allows the exact primitives to be determined. input program A vector code for each of the primitives Pr is located in the dictionary 28b. , and 0 is given to the storage device from the combined primitive code a.

A series of primitive codes generated for the input ideogram IC. The character code is determined by the fifth comparator S2a and the third index table or dictionary 52b. lji- is sent to the character detection unit 52 containing the character. The dictionary 52b stores each of the ideograms of the language. memorizes the list of character codes forming the character code and the combined international output code. Ru. The comparator 52a and the dictionary 52b output the input ideogram IC1-display. is equal to the character code representing one or more of the ideograms. It works to detect whether or not. The character detection section 52 has only one character code. raw Cartesian coordinates, whether equal to the character code representing an ideograph or higher A discriminator 52c is also included to test for data point signals. this is Makes it possible for all accurate ideograms to be detected. Identification of the correct ideogram is complete and the combined international ASCII code are sent to the storage device 34 and then to the output device 34. Sent to 16.

As mentioned above, when the character code is not equal to the character code found in the dictionary 32b An alternative or correction @56 is used. The substitute part 36 is the input ideogram I CK: Determine the character code that has the highest probability of becoming the corresponding character code. Probability matrix that works collectively to determine 6b and a storage device S6c. This is digitizer tablet 2 Increases the probability of detecting ideograms input as 0.

The ideogram IC should be input to the device 10 via the digitizer tablet 20. Sometimes a stylus 20a is placed on top of the tablet 20 to form an ideographic IC. Each of the primitives Pr is drawn separately. As explained above, ideograms Glyph C11 - Primitives used to form 15 primitives 1-1) K from Pra to Pro must be substantially equal. however, Such a restriction is due to the fact that each of the 15 primitives can be written in ideographic terms. Since it is a basic stroke that everyone uses, it poses many problems. Never. Moreover, primitives Pr, 11 to Pro are input to device 1G. In order to reduce the number of input characters that generate the same character code when This was selected to simplify the processing in step 14. Inputting primitive Pr After completing the force, the stylus 2011 moves the tablet 20 for a predetermined length of time. removed from. This means that the data processor 14 has the primitive Pr completely This is the result of a timeout signal being generated that allows the input to be 90. after that , the next primitive forming the character is entered and a timeout signal is generated.

This process continues until each primitive forming a character has been entered into the device 10. continue.

In order to form the primitive Prf, the stylus 20a and the entire tablet 20t- Movement across produces a series of Cartesian coordinate data point signals. day The data processor 14 processes the Cartesian coordinate data points generated for each primitive. The extracted locus signal is extracted at a sampling rate of x1oo samples per second. The standard data signal is stored in the storage device 22. The extracted data for each primitive data means that the data processing unit 14 has input a complete primitive. are continuously stored in separate registers until a time-out signal is received. tabulet T) While the next primitive Prs is formed at 20, the extracted Cartesian locus The target data point signal indicates that the next timeout signal is detected by processing unit 14. are stored separately in various registers within storage device 22 until the

This process is completed until each primitive forming the ideograph completes its input, and the primitive The Cartesian coordinate data signals generated for each are stored separately in the storage device 22 and then Continue until Notify the data processing device 14 that the input of the Zengoi character IC has been completed. To command, you must press the End of Character (EOC) key located on the tablet. No. This is entered into the tablet 20 with nine more data entered before This prevents disturbing the data combined with the ideographs.

Since digitizer tablet 20 is used, stylus 2011 and digitizer Due to the bonding spots on the surface 20 of the Iza tablet, temporary and irregular noises may occur during the extraction process. A sound is produced. Moreover, due to the uneven movement of the operator's hands, small amplitude pulse sounds occurs, which is the extracted Cartesian coordinate data point signal and the desired Cartesian This will lead to discrepancies between the coordinate data point signals. In addition, the processing device 1 For a sampling rate of 4, the digitizer-tablet surface 20i The slow movement of the lath 20a then requires a large amount of storage space and equipment 10(1) Introducing a large number of extra data point signals which increases processing time. become. Therefore, as mentioned above, preprocessing is performed to reduce redundant and false data. A processor 24 is used.

To accomplish this function, a copy of the extracted Cartesian coordinate data point signal is is applied to the comparator 24a. Stylus 20a and digitizer tab In order to reduce the noise caused by inadvertent coupling errors of the let 20, The extracted Cartesian coordinate data point signals are analyzed separately. If extracted The Cartesian coordinate data point signal crosses the digitizer-tablet 20 boundary. If you find that you have a set of coordinates that extend over Point signals are deleted. Second, to reduce the amount of redundant data, In order to increase the processing speed of the device 10, the first two Cartesian coordinate data points are The event signals are compared by a comparator 24m. If two Cartesian coordinate data points If the distance between the int signals is 4 less than the predetermined threshold value, the The two extracted data point signals are removed and the first and third extracted data point signals are The distance between Cartesian coordinate data point signals is determined. This process consists of two Continue until the distance between data point signals is greater than the threshold value . The first data point signal is sent to the storage device 24b and the first data point signal is sent to the storage device 24b. The point signal is compared to the next preceding data point signal.

Furthermore, if the distance between two Cartesian coordinate data point signals is If it is greater than the set threshold value, the second Cartesian coordinate data point The data point signal is compared to the third data point signal. If the 2nd and #I5 day If the point signal pressure separation is greater than the second threshold value, the second data The tap point signal may be caused by an inadvertent miscoupling of the stylus 20a and taplet 20. will be deemed to have occurred because of this and will be deleted.

However, if the distance between the second data point signal and the fifth data point signal is less than the second threshold value, the first data point signal is inadvertently will be deemed to have occurred and will be deleted. This process forms the input characters Extracted Cartesian coordinate data points for each of the input nine primitives The reduction in data tt performed on the signal, resulting in the need for processing S. It becomes... For example, if the ideogram IC shown in FIG. 1-The two ports from the forming primitive Pr1 are separate (input on tablet 2o) It will be done. The data processing device 14 uses the tablet 20 for the first primitive Prl. Extract the Cartesian coordinate data generated by , the extracted Cartesian coordinate data point signal P11 to Pl, 'i memory The information is stored in the device 22. Similarly, the processing device 14 processes the following two primitives Pry and the Cartesian coordinate data point signal P2. generated for Pr3. From P 2. and P51 to P3. t- respectively extracted and extracted to the storage device 22 Stores Cartesian coordinate data point signals.

With this Vcvc, the Cartesian coordinate data point signal is sent separately to the preprocessor 24. The signals are then sent to each other, where the signals are stored in a comparator 24a. 1st #C1 1st Extracted Cartesian coordinate data point signal P1m for primitive Prl is compared with the outer boundary Cartesian coordinates of digitizer tablet 2o. if The extracted friend data point signal is considered to be outside the boundaries of the tablet 20. If so, it will be deleted. Second, the remaining data point signals P1. From P l, is compared with the previous data point signal P11. For example, if the data point P1. If the distance between and Pl is smaller than a predetermined value, then the data Point signal P1. is deleted, and data point signal P1m is replaced by data point signal P1m. It is compared with No. P11. If the distance between the data point signals PIS and Pl is If the data point signal Ptm is greater than the threshold value, the data point signal Ptm is 4b, and the above-described processing is performed on the data point signal P1. and P14 investigation resume. In this process, input primitive Pr1t-representing coordinate data is Each data point extracted for the first primitive Pr1 until the reduction is completed. performed on the event signal. In addition to this processing, other input friend pools For the extracted Cartesian coordinate data point signal for each of the As a result, the storage device 24b is manually stored for each of the fixed primitives. Store the reduced Cartesian coordinate data point signal.

Extracted false and extra Cartesian coordinates for each input primitive When a data point signal is removed, the resulting next data point signal is stored. Sent from the device 24b to the function extraction unit 26.In the function extraction unit 26, each input The Cartesian coordinate data point signal for the primitive enters the tablet 20. Vector code and rounding simplifies the process of detecting input primitives. and a series of squash. However, for the input primitives The Cartesian coordinate data is pre-processed before forming the vector code and grid. Investigate to find out whether the coordinates are reduced to a pair of coordinates by the sensor 241C. It will be done. This is when the primitive Pre is input to the tablet 20. happen. If this primitive is detected, the primitive code C is The above processing is not required (output to the storage device. Function extraction unit 26Fi, When forming vector codes and sketches to determine other primitives, Using the modified Freeman coding system FC shown in FIG. do. The Freeman coding system is a series of Cartesian coordinate data points The signal (Pa, Pt, −”・−・Pie　Pint　) is a series of unit vectors (Each vector has a combined length) Allows to be K-transformed. however, This signal FiPo is (Xo, Yo) K’tl, <, Pi (Xi, Yi) K etc. Unit vectors are adjacent nine Cartesian coordinate data point signals. The class drawn between the numbers Pi and pi+1 is 8 in the Freeman code PC. By comparing with 19 of the Freeman unit vectors FV1 to FV, It is formed by

However, the shape of the nine primitives entered on the digitizer tablet 20 Due to the angle introduced into the shape, a pair of Cartesian coordinate data point signals Pi and Pint, which is not coincident with the Freeman unit vector FVN. In order to allow the different lines to be assigned to exact Freeman unit vectors, Tolerance is required. Adapt these drafting variations of input primitives To do this, the Freeman coding system FC1 uses the Freeman unit vector Use a 20° tolerance for each of the PVNs, so that the border line is A pair of Cartesian coordinate data point signals Pi and Pin contained in one of Any line formed between t and the appropriate Freeman unit combined with that boundary It becomes possible to allocate the data to the position vector FVN.

Between each adjacent Cartesian coordinate data point signal for each of the primitives To generate the Freeman unit vector FVN for each line formed, The preprocessed Cartesian coordinate data point signal is sent to the conbalameter 26a. Ru. In the conbalameter 26a, the adjacent nine Tekart coordinate data point signals are measured. and a line is formed between them. Carelessness of the stylus 20a by the operator A method that reduces errors introduced into the extracted Cartesian coordinates due to , the length of the line formed between each adjacent data point signal is compared to the threshold value be done. If the length is less than the predetermined threshold length, then The two data point signals are considered to be the result of a false hand movement by the operator. and therefore deleted. This process is performed by an operator with only a small amount of non-horizontal parts. The horizontal line drawn by the data creates the desired horizontal line. to ensure that it is removed.

After removing the inadvertent data point signal, the line represents the remaining adjacent data points is formed between the signals and compared with the modified Freeman code FC. If the line is Tolerance boundary 1! If it is included in one of As to A8, the free combination thereof Mann unit vectors PV1 to FV are sent to the storage device 26c. If two de The line formed between the cult coordinate data point signals is the Freeman code PCK. If the invalid boundary mX* to X8 contains one neutralization, then the second Cartesian coordinate data point signal is replaced by the next preceding Cartesian coordinate data point signal and a new line is formed between them. Similarly, the new line will be the border that the line is valid for. A! Free again to find out whether A, 01 is located from Compared with Mancode PC. If the resulting line is contained within the effective boundary group N In this case, the unit vector FVN of 9-wn combined with the boundary JAN is stored in the storage Sent to C. However, if a valid Freeman unit vector is not found If the second data point signal of the pair is The same process is repeated. If the line is within the effective boundary AN is the remaining Cartesian coordinate data point generated for the input primitive. If the coordinates are still not found after substituting each of the The invalid Freeman vector is represented by the unit vector U′, and the invalid Freeman vector is It is sent to the station 26c.

Therefore, a set of Freeman unit vectors for each of the input primitives PvN or U' is formed from PVi and stored separately in the storage device 26c　K . The series of unit vectors are then sent separately to comparator 26d. The data 26d compares each unit vector FVi+uy with the previous unit vector FVi. and if they are equal, the skasi count for that unit vector increases. In addition, the unit vector PV i + t is deleted. This process This is performed on the unit vectors generated for each of the primitives Pr. child The operation is a reduced one for each input primitive nine forming a character. The result is to form a series of unit vectors or vector codes, with each vector The code has a series of combinations of scales, and the scales are simple in the vector code. represents the length of each position vector.

For example, if the ideogram IC shown in FIGS. 1 and 4 is input into the device 10, In this case, the comparator 26m is first combined with the 11th C and the first primitive Prl. Examine the Cartesian coordinate data point signal, each instantaneous contact data point P11 to Pl, respectively @L1 凰 to Ll, t- form 41 strings L11 to Ll4 is then compared with the Freeman code FC and the combined Freeman vector FVi to FVM is assigned to the line. Therefore, as shown in Figure 1a4 , formed by Cartesian coordinate data points) Pl, to Pl, and 1lL 1t to Ll4 t-The generated primitive Prl is IBLll to Ll, Since the boundary is within (the length of each II is greater than or equal to the threshold value) None), 7 li-w y he/) le PV3, FV3, FV3, FVm't divided by "1".

A set of vectors, with each of the vectors generated for primitive Pr1K is sent to storage device 26c and stored therein. The process described above then Cartesian coordinate data point in combination with the primitives Pr2 and Prs) KI$! 3 and the resulting vector is also sent to the storage device t26C. . Following this, each primitive Pr has a comparably freeman vector. It is sent to the controller 24d. Then, the resulting adjacent free map for each primitive vectors are compared. If the adjacent vectors are identical, then the vector One of them will be deleted and the Sukasi count will increase. Results from comparator 26dK is then sent to storage device 26e.

For example, if the primitive Prl shown in Figure 4a is a series of Freeman vectors i ;’v, , FV, , pv, , FV, when processed to form the A comparator 2411 converts a series of vectors into a vector code F having 14 spaces. V, VC reduction.

For example, if a primitive is input and FVI - FVB, TVs -FV4, FV4, FV4, FVB -FVB -FVs K If a series of freeman vectors is generated, a series of unit vectors is vector code FVs, FV4. FV, , FV, K reduced S, A series of squash equal to 3.2.1 will be generated.

Vector codes and combinations for each primitive forming the input character The series of search results are sent to the primitive detection unit 28 from the memory 26e.

The vector code is given to the comparator 281, and a series of vector codes are given to the storage device 28. It is stored in c. The vector code received by comparator 28a is It is compared with the vector code stored in the dictionary 28b of the primitive. Dictionary 28 is It is divided into 16 primitive code terms, of which the first 15 terms is uniquely combined with one of the 15 primitives Pra to Pr6, and the primitive Stores vector codes uniquely combined with primitives. Clause 16 is one or more contains ambiguous vector code that can represent primitives.

Section 16 also contains unique test information for each ambiguous vector code. , which allows the exact input limit to be determined.

If the vector code for the input primitive is 5 (combined with the found vector code, if equal) The generated primitive codes 1 to 0 are sent to the storage device 30. This process For each vector code generated for each primitive forming the entire character will be carried out.

In this way, character codes are generated from a series of primitive codes, The character code represents the ideogram input into the digitizer tablet 20.

However, the resulting vector code for one of the primitives is and when it is equal to the vector code stored in the 16th term, The test information combined with the ambiguous vector code is given to the test section 28d. . Test part 28 Receives dFi test information 9 Which vector code is being examined? Examine that information to determine which Thereafter, the test section 28d was examined. Receive a sequence of vector codes combined with vector codes and perform operations on them. However, its operation is determined by proprietary test information. The results of the test are entered @2 Select the correct primitive food tv to represent the given primitive. Sent back to 8b. Although the vector code is ambiguous, Also, since the values of each successive square are very different from each other, Provide appropriate information to identify bad vector codes.

For example, the primitive PrB shown in the 8th diagram is input with 20 PCs (Tablet). If It becomes. However, the primitive shown in this vector code #1jllb diagram It is also used to represent the vector Prbt, so its vector is the number one in the dictionary 28b. It will be found in Section 16. The vector code for the two primitives is Although identical, the series of squashes associated with them are very different.

The series of squares combined with primitive Pr2 is 3.1°5, while the A series of skis combined with Mitib Prb is 1.5. becomes i. Therefore, one The exact prefix can be determined by comparing the relative lengths between the first two squashes in the run. If the primitive code can be determined, it is compared with the contents of the dictionary 28b. If the vector code is not equal to the vector code positioned within it, then The vector code is also an unconfirmed primitive code assigned to the storage device 50. Assigned a deU.

Therefore, the output of the primitive detection section 2B is a series of primitive codes or input Contains the character code representing the ideographic character ICQ.

The character code stored in the storage device 30 is given to the character code recognition section S2, and the character code is It is received by the parator 321. Comparator S2x can be coded by one person It is compared with the contents of the character dictionary ii 32b generated for the character.

As mentioned above, the dictionary 52b can be used with the compatible international ASCII output code. The character code for each ideogram is stored in a language. International ASCII output code The code is used internationally to represent ideograms. several ideograms in the same order It is formed by the input nine primitives, so the input friend primitive Some ideograms are the same character code, although the relative positions between them are very different. It has a code. When an ambiguous character code is received, the device 10 generates an appropriate representation. Rounding that allows you to detect characters, character dictionary 32bFi each ambiguous character code It also includes test information uniquely combined with the code.

When the character code is received from the storage device 30, it is passed through the conbalameter 321. The data is compared with the contents of the dictionary 52b.

If the received round character code is a word uniquely combined with only one ideogram The character code found in document S2b (if equal, the international A8 combined with it) The CII output code is output from the dictionary 32b and stored in the storage device 54. death However, the character code that occurs for the input ideogram may be one or more ideograms. When it is equal to an ambiguous character code that is combined with a character, Upon receiving the test information in which unique test information is given to the character identifier 32c, The discriminator 32c searches the storage device 22 for unprocessed Cartesian coordinate data, and test information to determine the International A,90II output code representing the nine ideograms. perform operations on that data as determined by. Complete the test operation raw Cartesian coordinate data is entered relative to each of the friend primitives. Since the information related to the location is stored, the raw Cartesian coordinate data points are , formed for the ideogram and used as opposed to the 几 series of sqashi.

Once the correct international ASCII output code is determined, it is also stored in storage 34. Sent.

For example, if the ideograms shown in Figure 1 are input into the device, @! ba″ A character code equal to is generated and compared with the contents of dictionary 32b. However, the first The ideograms IC2 and IC3 shown in figures 9m and 9b, respectively, also represent the same sentence. It is represented by the character code @abs, so the character code is ambiguous. It is regarded. Unique test information L1 storage device combined with character code "aba" along with the raw Cartesian coordinate data from the location 22 to the differentiator 32c. become. In this example, the test information causes the discriminator S2c to identify the second primitive P rz passes through the first primitive Pr11. The result is to check the position of the second primitive Pr with respect to the primitive Prl.

If the result of this test is negative, the ideogram ICFi, its function is It is recognized that the ideogram IC2ic is not equal because it does not exist in the character IC2. To distinguish between the corresponding 0 ideograms IC and IC3, the third primitive P r3 is compared with the first primitive Prl forming the input ideogram IC, and both The relative dimensions between are examined. The result of this test shows that primitive Pr4 is The discriminator 52e uses the correct international ASCII output code. allows you to select a mode. The dictionary 32bF'i is generated by the discriminator 32Cvc. After receiving the fixed result, the correct international ASCII output code is sent to the storage device 64. .

After the International Person 8CII output code is determined and stored in storage device 54, it is , printer 16a, VDT terminal device 16b, audio synthesizer I It is given to an output device such as Sc in order to create an image of the input ideogram.

However, if one or more of the primitives is an unidentified primitive A character code is formed from a series of nine primitive codes assigned a primitive code U. or if the character code is one of the character codes found in the character dictionary 52b. If neither is equal, the character code is given to the replacement/correction unit 56. Place The conversion/correction unit 36 is equipped with a probability matrix) +7x56at-, which has 16 horizontal lines. , in the form of an array of registers 361' with 15 vertical lines.

As shown in Figure 10, each row of the matrix represents an unidentified primitive code. combination with one of the 16 possible primitive codes a through O, including do U. and each column 1j15 of the matrix has a possible primitive code 1 to It is combined with one of O. Each of the registers 36a' is a row primitive. Contains a number representing the probability that a primitive code will be mistaken for a vertical primitive code. I'm reading.

Therefore, there is a high probability that primitive code will be found as such, so The friend probability values stored in registers along the diagonal from left to right of Rix sba are all All have a value of 1. Two dissimilar primitives can be mistaken for each other Since it is very unlikely that two dissimilar primitives The probability value stored in the combined register is zero. For example, the first column of matrix 56m, which is associated with primitive Pry, is As you can see, these primitives are formed in very different ways, so The probability that the primitive Pry is actually mistaken for the primitive Prc is (LO. A primitive that has some similarity to a primitive has the number of similarities between them. The exact IE['kl14 in the range cLl to LIL9 is assigned by.

The character code contains a small number (and one un-aN primitive code Ut) The 9 o'clock received by comparator sbb is the horizontal You can check the probability of a column. When the highest probability value of dying in the row is discovered, the unconfirmed A column of primitive codes is used to replace the primitive code U.

The resulting character code is sent back to comparator 321, which outputs the resulting character code. A character string used to identify whether a code is equal to a character code found in the dictionary. It is compared with the contents of Book 32b. If the resulting character code is a character code in the dictionary If it is equal to , the international module 8CII output code is a dictionary! Searched from 2b, it The data is sent to the storage device 34 where it is stored.

If the resulting input character code is equal to the ambiguous character code, then the exact country In order to determine the 9CII output code, use the same method as mentioned above to store the storage device. A test is performed on the Cartesian coordinate data stored in 22.

However, if the resulting character code is the character code found in dictionary S2b If they are not equal, or if they were entered from the beginning and the t character code is found in the dictionary 32b. If the character code does not match, a second substitution is performed. Example 1 above When a character code occurs, the character code is sent to comparator 36b to form a character code. It is possible to identify the number of primitive codes that make up the code. Following this, the same Each character code in the character dictionary 52b formed from numerical primitive codes is It is sent to the comparator 36b and compared with the unconfirmed character code. During this comparison, Each of the character codes forms a primitive code and an unidentified character code. The number of differences between the primitive codes is examined. If character code and unconfirmed sentence If the difference a found between the character codes is greater than the threshold value, then the character The code will be discarded.

However, every character code has a difference in number less than the threshold value. If you notice that the character code is combined with the International 8CII output The code is stored in storage device S6c. International output code stored in storage device 56cK The order of the codes is such that the first international A8CII output code in storage is the unidentified character code. The code is determined to be combined with the character code most similar to the code. Storage device 56C K stored nine international output codes are then retrieved from storage device 56c, and further 1c is sent to the VD'll terminal device, thereby indicating that the amount is equal to the input ideogram. Each possible ideogram is displayed to the user. Therefore, User Company, for appropriate editing Software t-Select an ideogram that matches the ideogram input to the stringing device 10 be able to. If the replacement unit 36 does not create the desired ideogram, the dictionary To find the exact international A8CII output code from 52b (, editing program The nine ideogram signals stored in the storage device 54WC can be used to store input ideograms. It is coupled to a printer 162 to enable reproduction of characters to occur. the Above, one user can see the characters that have already been input The character signal 1i can then be sent to the VDT screen 16b. The equipment 5110 is Beforehand, the user can change the ideographic signals stored in the storage device 34. It is also possible to work with known editing programs.

The device 10 may operate in other modes so that the device functions to recognize characters in different languages. When conditioned on one, the same set of primitives shown in Figure 5 is Used to form glyphs. The primitives shown in Figure 5 are Since all of the characters in can be formed from these primitives, ideograms and It should be obvious that it is particularly useful: for forming uppercase and uppercase English letters. death However, other primitives can be formed so that all letters in all languages can be formed. It must be recognized that more than 20 pts may be added; Unless it is possible for primitives to form virtually every character in any language Because something like this should be possible, something like this should be possible. It's not in Hametsuka.

As mentioned earlier, the dictionary in processing unit 1i14 contains each uppercase letter in other languages. Generous primitive codes, character codes and ASCII output codes for It is divided into sections including: Uppercase letters are the same for everyone familiar with the language. Characteristically, they are written in the same way and order, so such characters are stored in the location.

Various parts of the processing unit also recognize a wide variety of characters that generate the same character code. It also includes testing information that will allow you to test your skills.

Words that use the primitive Prp to PrtK@ similar nine elements when forming characters. For words, one of the primitives Prp to Pr< is the tablet 20 Same as above using Freeman coding except when input to The method performs primitive detection and primitive code determination. Therefore , when the primitives are manually input to Tablet 20, Freeman coding is also used. Before using the primitive bay aK (namely 'f) 1800° 270°, 56 0”), the The function extraction unit 26 examines the tangent of the class.

If the primitive has substantially 270 or s6o''' curvature [t- If it is found, the input Pr9 (combined with Tig Pr3 Mayu FiPrt) The primitive code $ or t determined immediately without further processing. determined. If the curvature of the primitive is found to be 'f180@ If the primitive start and end coordinate data signals are in the tangent direction ( In other words, right′! iA or counterclockwise direction). This is a primitive Allows Prr to be identified from Prp without further processing. do. In addition, if you check the tangent and the input primitive is If it cannot be determined that the gradient is qualitatively constant, use the processed coordinate data. Freeman coding the signal to determine the exact primitive code. Use and process.

For example, referring to Figure 11, if the device is conditioned to recognize English letters, If the character 1MI is entered in tuplet 20, then The primitives Prba Prg and Prc* Prb are used. These The primitives are connected to the feature extractor 26 and the primitives in the same way as the companions described earlier. processed by the live detection unit.

Therefore, a character code equal to 'bgcd' will result. This code is Since it is only combined with the letter 'M' of the word, the combined person 5CIX output code The code will be output.

For example, if the English letters 'D' and 'P' appear on the tablet as shown in Figure 12, )20, the character code generated for each character is Since the primitives forming the character are Prb and Prq, 'bqI' is obtained.

Therefore, if one of these characters is entered, it will be stored in the character dictionary and the friend test The information is used in a manner similar to that described previously, and the length of the primitive Prb and the length between the start and end points of primitive Prq is checked. this is, Even if the character codes generated for the two characters are the same, be identified.

For other languages such as German, French, etc., the device may override the means 18 and use a suitable module. Although the code must be conditioned, the method for detecting all handwritten characters is the same. Ru.

This means that although the human 8CII output code is different, the characters forming the 4 words are the same. In some languages such as German, French and English this may not even be necessary. Ru. A substitution matrix is not necessary since a small number of characters are used for non-ideographic languages. Although not available, it can also be used for each of the other languages.

Also, if the device 10 is conditioned to detect uppercase letters in a language, If desired, the device will provide the ASCII equivalent of the lowercase letter for the uppercase letter detected. ? Companion software for output is also provided. Lowercase letters are similar to uppercase letters in a similar way. Although it can be detected, lowercase letters may vary from person to person and may be written in a good form. This makes the detection process even more difficult, and the possibility of a large number of characters being written. Requires more storage space to allow character detection in a certain way Ru.

This device has a 20Mb bar)'de 4X function to store information for the dictionary. Used in the IBM PCXT personal computer equipped with Ru. In order to perform the identification process described above, the combiator has an input Cartesian A suitable software that allows the coordinate data point signals to be processed in the manner described above. Be prepared with your own software. The dictionary 32b contains a large amount of data, approximately s u　o　.

character codes for various ideographs and their associated international output codes. Since the code is memorized, the character code generated for the input ideogram is and memorized character codes to increase the detection speed. A well-known B-tree algorithm is used. B-tree algorithm has high processing speed However, index files are required, and memory is required. Condition books increase.

The device 10 is a small integrated unit that can be connected to a conventional personal comviator. It can also be fabricated on a circuit board, and the integrated circuit board can store a wide variety of dictionary contents. The ROM component of the storage friend and the appropriate structure to perform the data processing functions. Equipped with software gold-biased microprocessor.

Therefore, the device receives the same input IJ (formed from TIG) input in the same order. This provides the advantage of being able to distinguish between characters that are present. This is the exact table The operator must stop the data entry operation to select the desired character. Reduce if this occurs. Additionally, substitution means are provided if an input character is found in the device 10. If not, the device will select another character code that is most similar to the nine-character code entered. As mentioned above, it is possible to select de! j! further reduce the occurrence of Ru. Additionally, the device can be generated using knotware or Since all hardware components can be manufactured individually, the device is versatile and and can be used in a wide variety of environments.

The device of the invention also provides a method VC for processing input line elements in the device. , tablet the line elements except for a few characters that produce ambiguous character hoods. It provides the further advantage of being able to write virtually anywhere on the surface of the sheet.

Also, processing operations applied prior to determining the primitives that form a character , there is a small exception, gold line, where the input character is unrelated to the length of the input primitive. It becomes possible to do what is determined to be appropriate.

In addition, with a simple approach and process, handwritten text in virtually any language can be improved. The characters can be recognized quickly, which allows the device to be used in real-time applications. becomes.

The device of the invention is such that suitable information relating to the character K1m to be detected is located within the device. It is possible to detect every character that is entered and stored in nine dictionaries. It will be obvious to those skilled in the art that many modifications may be made.

FIG, 9a　　　　　　　　　　　FIG, 9b International Search Report 1mema seagull-＾----轡’nPCT/GB8B10110゜International search report PCT/GB 88101100SA 25916

Claims (23)

[Claims] 1. In character recognition devices that identify characters formed from several primitives , the characters and primitives are members of a predetermined set, and the device: successively receives each of the primitives forming said character, and input means for generating an input signal for each of the input signals; receiving said input signal and each of the primitives received by the input means; A process that generates a character code representing the character by identifying the primitive. means and storing a character code and a combined output code for each of said set of characters; and a memory means for the input character to identify said input character. a comparison means for comparing the character code generated by each character code with each of the character codes in the storage means; , is connected to a comparison means and produces a reproduction of the input characters based on the identification of this comparison means. A character recognition device characterized in that it includes an output means for displaying. 2. Examine the input signal generated for each primitive, and check if there are multiple character codes. When the output code of the character is equal to the character code combined with the output code of the character, the output code of the character is further comprising identification means for performing an operation to identify the power code. The character recognition device according to claim 1. 3. Primitives can form any character in multiple languages. and the storage means stores character codes and character codes for each of the characters in all of the languages. 3. The character recognition device according to claim 2, wherein the character recognition device stores an output code and an output code. 4. The storage means further stores character test information, and the test information is output to one or more outputs. It is given to each character code combined with the code, and the means of identification is Receives test information and input signal and detects output code corresponding to the character. be equipped with a processor that operates according to the character test information in order to The character recognition device according to claim 3, characterized in that: 5. The predetermined set of primitives contains 20 distinct primitives, and Various combinations of the 20 primitives form every letter in a language. 3. Device according to claim 2, characterized in that it enables. 6. The processing means emits a primitive code for each of the identified primitives. and the character code is equal to any of the character codes of the storage means. a replacement means that receives a character code when the character code is set the primitive code to the input character code generated for the received character. Corresponding primitives of character codes of the storage means having the same number of primitive codes comprising comparator means for comparing with the code; When compared with the generated character code, the difference is less than a predetermined number. a memory means for storing output codes combined with each of the character codes of the storage means having the storage means; Mori and 4. The apparatus of claim 3, further comprising: 7. The substitution means has a definite room matrix, and this matrix has at least one girl. The replacement means receives a character code containing two unidentified primitive codes. generates an alternative primitive code similar in quantity to the unconfirmed primitive code. forming a character code equal to the character code stored in the storage means, and In order to represent the character with a character similar in quantity, the unidentified primitive code is substituted. 7. The apparatus according to claim 6, characterized in that the replacement primitive code is substituted. 8. The input means inputs Cartesian coordinate data for each of the primitives forming the character. an on-line digitizer tablet that provides data, and the processing means is Examine the Cartesian coordinate data for each primitive and derive a set of unit vectors. 2. The apparatus of claim 1, further comprising encoding means for forming. 9. Modified Freeman encoding means includes multiple Freeman unit vectors 9. Device according to claim 8, characterized in that it is an encoder. 10. The processing means Receive a set of unit vectors for each of the primitives, and to form a vector code for each and a series of scalars combined A feature extractor that removes redundant unit vectors and represents each primitive in a set. The vector code and the combined primitive code are the vector code generated for the character; comparator means for comparing the code with a vector code stored in the holding means; When the vector code is equal to the vector code stored in the holding means, the A primitive code is output, and the vector code is stored in a storage means and a vector is stored. When it is not equal to the torque code, it outputs an unconfirmed primitive code, 10. The device according to claim 9, characterized in that it comprises: 11. The holding means further comprises primitive test information, the information being one or more Uniquely combined with the vector code representing the above primitive, the processing means Receives a series of scalars combined with primitive test information and vector code and the vector code represents one or more primitive codes. The exact primitive code combined with said vector code when equal to Claim 10 further comprising a test section that performs an operation to detect the The device described. 12. The output means is a printer, audio synthesizer, and video display end. 12. The device according to claim 11, wherein the device is selected from a group comprising terminal devices. 13. further comprising pre-processing means for receiving the Cartesian coordinate data, the means for receiving the Cartesian coordinate data; and the distance between adjacent second coordinates, and this distance is determined by a predetermined thread. Extra data is removed by removing the second coordinate if it is smaller than the shjord value. 9. Apparatus according to claim 8, characterized in that it reduces the amount of data. 14. In a method for identifying characters made from several primitives, characters and primitives are members of a predetermined set, and the method successively receive each of the primitives that constitute the generate an input signal and identify each of the input primitives forming said character. examine said input signal in order to Forming a character code by identifying the primitives that form said character generate primitive code for each of said primitives for Record the character code and combined output code for each of the characters in the set. Remember, when said character code is equal to the character code combined with only one output code Compare the character code with the stored character code to detect the output code and detect the output code combined with the input character. When the code is equal to the character code combined with one or more output codes, the above statement Examine the primitive code generated for the character and perform operations on it. and further generating an image of the character by deleting the combined output code; A character identification method characterized by comprising each step. 15. Character test information is provided and the information is combined with one or more character signals. Uniquely combined with a character code, the method receives said character test information and input signal, and outputs an output code corresponding to said character; further steps of performing operations according to said character test information to detect 15. The method of claim 14, comprising: 16. Forming the set of primitives from 20 distinct primitives Furthermore, various combinations of these 20 primitives can be used to create synonyms in multiple languages. 16. A method according to claim 15, characterized in that it is possible to form any character. 17. The character code is not equal to any of the stored character codes. code and enter each of the primitive codes forming said character code. the same number of primitive codes as the character codes generated for the character Compare the stored character code with the corresponding primitive code having The stored character code is a predetermined number when compared with the character code. the output combined with each of the stored character codes when having less difference; memorize the code, 17. The method of claim 16, further comprising the steps of: 18. Input character code containing at least one unconfirmed primitive code Receive the Set the unconfirmed primitive code to a character code equal to the stored character code. to form a primitive that appears most similar to the unidentified primitive code. Replace with primitive code, 18. The method of claim 17, further comprising the steps of: 19. Generates Cartesian coordinate data for each primitive that represents an ideogram. Equipped with a digitizer tablet that the decal to form a series of unit vectors for each of the primitives. Claim 1, further comprising the steps of encoding the coordinate data. 8. The method described in 8. 20. The encoding is performed using a modified Freeman encoder containing multiple Freeman vectors. 20. A method according to claim 19, characterized in that it is carried out by: 21. Examine the set of unit vectors for each of the primitives and to form a combined series of scalar vector codes for each of the , remove the extra unit vector, Vector codes and combinations representing each of a predetermined set of primitives memorize the primitive code that was Compare the vector code for each primitive with the stored vector code, When the vector code is equal to the stored vector code, the primitive code is generate a vector code and make the vector code equal to any of the stored vector codes. If not, each step that generates unverified primitive code can be further included. 21. The method according to claim 20, characterized in that: 22. Prints combined with vector codes representing one or more primitive codes Prepare primitive test information, and set the vector code to one or more primitives. When combined with a vector code, it is equal to one vector code representing the code. combined with the vector code to find the exact primitive code. Each Utep performs an operation according to test information about a set of scalars. 22. The method of claim 21, further comprising: 23. Measure the distance between first and adjacent second Cartesian coordinate data points , Compare the distance with a predetermined threshold value and If it is smaller than the threshold value, the second adjacent Cartesian coordinate data point remove the 23. The method of claim 22, further comprising the steps of: