JP3560361B2 - Character recognition device and character recognition method - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a character recognition device for recognizing characters input online, and more particularly to a character recognition device and a character recognition method for recognizing handwritten characters in a device for setting reservation information for recording reservation of a VTR by inputting handwritten characters. About.
[0002]
[Prior art]
FIG. 7 is a block circuit diagram showing an example of a conventional character recognition device. In the figure, 1 is an input medium, 2 is coordinate value detecting means, 3 is pen information extracting means, 4 is coordinate value accumulating means, 7 is recognizing means, 14 is display means, 15 is area identifying means, 16 is time measuring means, 17 is an input end detecting means.
[0003]
Next, the operation will be described. The input medium 1 is a tablet that detects a coordinate position of a writing operation using a rod-shaped object (hereinafter, referred to as a “pen”) capable of performing a writing operation with the same feeling as a pen, a touch panel, or a flat plate shape called a digitizer. Is a coordinate input device. Here, the means for performing the writing operation is not limited to a rod-shaped member capable of performing the writing operation with the same feeling as a pen, but may be a finger or the like. The principle of detecting the coordinate position of such an input medium includes, for example, a resistance pressure-sensitive method, a capacitance method, an electromagnetic induction method, an electromagnetic transfer function method, and the like. Pens and tablets differ in structure, input accuracy, power consumption, price, etc., depending on the principle of detecting this writing movement. Differences such as wireless or wired, transparent or opaque, pressure detection or contact detection (proximity detection), etc. However, a coordinate input device based on any principle may be used.
[0004]
The coordinate value detecting means 2 sequentially detects the coordinate values of the input medium 1 that have been touched by writing with a pen (they may be pressed or non-contacted by the principle of coordinate detection, but are simply referred to as “contact” for simplicity). . The coordinate value may be detected at a constant cycle or may be detected each time an event occurs. In the case of a fixed period, the period may be set to a fixed period, but may be configured to be changeable according to the use form of the input medium. In the case where the detection is performed every time an event occurs, a configuration may be adopted in which the coordinates of a point at which the input by the pen changes, for example, a point of a curved writing locus are detected. For example, when the input medium 1 can detect coordinates with an accuracy of 1024 points in the horizontal direction and 1024 points in the vertical direction, the horizontal coordinates are set to X (X = 0 to 1023), and the vertical coordinates are set to Y (Y = 0 to 0). 1023), the coordinate value (X, Y) is output. The coordinate value detecting means 2 supplies the coordinate values to the coordinate value accumulating means 4 and at the same time to the area identifying means 15.
[0005]
The area identifying means 15 identifies whether the coordinate values sequentially supplied from the coordinate value detecting means 2 belong to any of several selected areas or entry frame areas formed on the surface of the input medium 1, and identifies the area. The identification signal shown is supplied to the input end detecting means 17.
[0006]
The pen information detecting means 3 detects whether the pen is in contact with the input medium (pen down) or not (pen up) and supplies the same to the timing means 16. The timer 16 measures the elapsed time since the pen-up, and supplies the elapsed time to the input end detector 17. The measurement is stopped by the pen down, and the measured elapsed time is reset.
[0007]
The input end detecting means 17 supplies an input end signal to the coordinate value accumulating means 4 when the identification signal output from the area identifying means 15 changes, that is, when it comes into contact with an area different from the area immediately before. . Alternatively, when the elapsed time output from the timer 16 exceeds a predetermined time, an input end signal is supplied to the coordinate value storage 4.
[0008]
The coordinate value accumulating means 4 sequentially acquires and accumulates coordinate values from a coordinate value series sequentially supplied from the coordinate value detecting means 2. The coordinate values stored when the input end signal is supplied from the input end detection unit 17 are supplied to the recognition unit 7 as a coordinate value sequence for one character. Here, the coordinate value series for one character may be a coordinate value series included in the area range of one entry box, but the coordinate values when the pen-up state changes to the pen-down state are included in the entry box area. If so, even if the subsequent input coordinate values are out of the range of the entry frame, it may be determined that the sequence is a series of one-character coordinate values.
[0009]
The recognition means 7 supplies the character recognized from the coordinate value series data for one character to the display means 14. The display means 14 displays the recognized character in an area corresponding to the entry frame in the form of an input handwriting locus. The recognizing means 7 is configured, for example, as shown in the block circuit diagram of FIG. In FIG. 8, reference numeral 21 denotes a feature extraction unit, 22 denotes a character dictionary storage unit, 23 denotes a comparison unit, and 24 denotes an identification unit.
[0010]
The feature extracting unit 21 extracts a feature amount of the coordinate value group and supplies the feature amount to the comparing unit 23. The character dictionary storage means 22 is a storage circuit corresponding to a dictionary in which characteristic amounts of each character are stored in advance, and is sequentially searched from the comparison means 23. The feature amount of each character is extracted and stored in advance in the same procedure as the feature extracting operation of the feature extracting means 21. The comparing unit 23 compares the input characteristic amount of the character with the characteristic amount of each character stored in the character dictionary storage unit 22 in order, and compares the characteristic amount input from the characteristic extracting unit 21 with the character dictionary. The distance from each character stored in the storage unit 22 to the characteristic amount is determined in order and supplied to the identification unit 24. The identification unit 24 determines that the character in the character dictionary storage unit 22 having the shortest distance between the feature amounts is the recognition character, and supplies the character to the display unit 14.
[0011]
The feature extraction is performed, for example, as follows. Even if the written character is a character having a plurality of strokes, interpolation is performed between the writing locus of the n-th stroke and the writing locus of the (n + 1) -th stroke so as to have a continuous writing locus as if written with one stroke. That is, even when pen-up and pen-down are performed in the middle of a coordinate value series for one character, the coordinate value series is handled as one coordinate value series without being separated there. Next, the sample points are reset so that the number of sample points forming the writing trajectory is constant for all input characters. The resampling of the sample points is performed by taking the sum of the distances between the sample points to obtain the length of the entire writing locus, and dividing the writing locus length by a predetermined number of sample points to obtain the average distance between the sample points. Then, from the starting point of the writing trajectory, sample points for each average distance are sequentially calculated and obtained, and the sample points are resampled as new sample points to obtain a coordinate value series. Next, considering the input plane as a complex plane, a vector between each sample point is divided by an average distance to obtain only a phase component. A transform coefficient obtained by performing a discrete Fourier transform on the phase component is used as a feature of the written character. In some cases, all of the conversion coefficients are used as a feature amount in the recognition operation. However, the calculation time may be reduced by excluding higher-order components as long as the recognition rate is not reduced.
[0012]
As the feature extraction method, not only a frequency domain feature extraction method but also many other methods such as a structural analysis method using a basic stroke and a method using pattern matching are used.
[0013]
Next, an example in which a conventional character recognition device is applied to time setting will be described. As a specific example of the time setting, there is a VTR recording reservation. The information of month, day, hour, minute and second related to date and time is represented by a maximum of two digits. Therefore, the input surface of this application example is configured to provide two entry frames in units of one character and to prompt input of one character each in the two entry frames. In this input form, a one-digit number may be entered in two frames, or a two-digit number may be entered in two frames. FIG. 9 shows an example of entry in the case where a single digit number is entered in these two frames. In the case where a one-digit number “2” is entered in the two entry boxes shown in FIG. 9A, an example is shown in the entry box on the right side as shown in FIG. 9B, and the entry box on the left side as shown in FIG. There is an example to fill in. In this example, only one of the entry boxes is filled, and the other is left unfilled. In addition, there is an entry example in which “0” is filled in a blank space on the left side as shown in (d). In the entry example in which “0” is filled in the blank space on the right side as in (e), “20” is apparently an error. Although any of the entry patterns (b), (c), and (d) can be recognized as "2", the user will be wondering how to enter when making the entry.
[0014]
FIG. 10 shows an entry example in which a state in which a two-digit number is already written is corrected to a one-digit number in a case where the correction is made by overwriting without depending on a method of erasing the already-written characters and re-writing. . In the case where a two-digit number “11” is already entered in two entry boxes shown in FIG. 10A and a one-digit number “2” is entered, the entry is made in the right entry box as shown in FIG. Since “1” in the left entry box remains and becomes “12”, it is necessary to fill “0” in the left entry box as shown in (d). Also, when filling in the left entry box, "1" in the right entry box remains as shown in (c) and becomes "21", so "0" is entered in the right entry box as in (e). When it is filled in, it becomes "20", which is obviously wrong. In this case, only the entry example (d) is a correct correction procedure, and trial and error is required until the user finds a correct correction method.
[0015]
As another example of the conventional character recognition device, there is Japanese Patent Publication No. Hei 6-10829 as a method capable of recognizing a plurality of characters without writing characters one by one in a predetermined entry box. In this method, the recognition process is performed for all combinations of natural strokes as characters, and the recognition distances (differences between the input at the time of recognition and the characteristics of the dictionary) are obtained. The sum of the recognition distances of each character is calculated, and the combination of characters that minimizes the sum of the recognition distances is used as the recognition result of the plurality of characters. Stroke combinations are recognized for each pen-up by sequentially retrieving strokes from a one-stroke writing locus ending at that time to a writing locus whose stroke number satisfies the maximum number of strokes of the character to be recognized.
[0016]
Next, an example in which another example of the conventional character recognition device is applied to time setting will be described. Another example of this conventional character recognition device is configured to be able to handle a character string of any number of characters, but is an application example of time setting. explain.
[0017]
For example, when a two-digit number is input by a combination of two-stroke numbers such as “45” (each of which is generally written in two strokes), the number of strokes of the recognition target character is set to a maximum of two strokes. When set to, since it is a 4-stroke input,
With the first pen up
Recognition of the writing trajectory of stroke 1 (recognition result 1),
In the second pen-up
Recognition of the writing locus of stroke 2 (recognition result 2),
Recognition of a combination of writing trajectories of strokes 1 and 2 (recognition result 12),
In the third pen-up
Recognition of the writing locus of stroke 3 (recognition result 3),
Recognition of a combination of writing trajectories of strokes 2 and 3 (recognition result 23),
In the fourth pen-up
Recognition of the writing locus of stroke 4 (recognition result 4),
Recognition of a combination of writing trajectories of strokes 3 and 4 (recognition result 34),
Are recognized seven times, and a recognition result and a recognition distance are obtained.
[0018]
Combine this with all four strokes and without duplication,
Recognition result 1 + recognition result 2 + recognition result 3 + recognition result 4
Recognition result 1 + Recognition result 23 + Recognition result 4
Recognition result 1 + Recognition result 2 + Recognition result 34
Recognition result 12 + Recognition result 3 + Recognition result 4
Recognition result 12 + Recognition result 34
The combination that minimizes the sum of the recognition distances of the five combinations is determined as the optimal combination, and the final recognition result is obtained.
[0019]
In another example of the above conventional character recognition device, the number of strokes of the character to be recognized is set to a maximum of two strokes. However, although a numeral can be represented by a maximum of two strokes, it is described by two or more strokes. You also need to recognize numbers. It is difficult for a user to use a character recognition device in which the number of handwritten strokes is specified to the user and the other is not accepted.
[0020]
For example, if three strokes are described here, normal recognition is impossible because only two strokes are combined in an actual recognition operation. Therefore, when the recognition target character is set to, for example, a maximum of three strokes,
With the first pen up
Recognition of the writing trajectory of stroke 1 (recognition result 1),
In the second pen-up
Recognition of the writing locus of stroke 2 (recognition result 2),
Recognition of a combination of writing trajectories of strokes 1 and 2 (recognition result 12),
In the third pen-up
Recognition of the writing locus of stroke 3 (recognition result 3),
Recognition of a combination of writing trajectories of strokes 2 and 3 (recognition result 23),
Recognition of a combination of writing trajectories of strokes 1, 2 and 3 (recognition result 123),
In the fourth pen-up
Recognition of the writing locus of stroke 4 (recognition result 4),
Recognition of a combination of writing trajectories of strokes 3 and 4 (recognition result 34),
Recognition of a combination of writing trajectories of strokes 2, 3 and 4 (recognition result 234),
And recognition are performed nine times, and a recognition result and a recognition distance are obtained.
[0021]
Combine this with all four strokes and without duplication,
Recognition result 1 + recognition result 2 + recognition result 3 + recognition result 4
Recognition result 1 + Recognition result 23 + Recognition result 4
Recognition result 1 + Recognition result 234
Recognition result 1 + Recognition result 2 + Recognition result 34
Recognition result 12 + Recognition result 3 + Recognition result 4
Recognition result 12 + Recognition result 34
Recognition result 123 + Recognition result 4
The combination that minimizes the sum of the recognition distances of these seven combinations is determined as the optimal combination, and the final recognition result is obtained. As described above, when the number of strokes of the recognition target character to be set is increased, the number of times of recognition and the number of times of calculating the total sum of the recognition distances increase accordingly.
[0022]
[Problems to be solved by the invention]
The conventional character recognition device is configured to enter only one character in a predetermined entry frame, and the number of characters to be entered may not correspond to the number of entry frames, so that the user does not know the entry method. There was a problem that gave an impression and made trial and error. Also, for example, in the human consciousness such as month, day, hour, minute, and second, one piece of information is treated as two pieces of information that are decomposed into numerical units for interface with the system. It was unnatural.
[0023]
Other conventional character recognition devices are configured to recognize any number of character strings. Even if only two characters are recognized, the processing procedure may be realized by hardware or software. However, there is a problem that it is complicated and large-scale, and the recognition speed is slow. Also, if the number of strokes is larger than the set number of strokes, normal recognition is not performed, and if the number of strokes is set to a relatively large number, the recognition time becomes longer.
[0024]
The present invention has been made in order to solve the above problems, and has an excellent human interface in accordance with the human thinking process, which enables a user to be confused in a filling method or to perform natural input without trial and error. The object is to obtain a character recognition device. It is another object of the present invention to provide a character recognition device that can be realized on a smaller scale with hardware or software at a recognition speed that does not make the user feel uncomfortable with the input.
[0025]
[Means for Solving the Problems]
A character recognition device according to the present invention includes a plane input medium including coordinate detection means for sequentially detecting coordinate values at which a writing means is positioned according to a writing operation, and coordinates written within a predetermined time with respect to the area. Accumulating means for sequentially accumulating the value series; contact state notifying means for notifying a contact state between the plane input medium and the writing means; and a pen-up state (the writing means is separated from the plane input medium) from the contact notifying means. Every time is notified All of the accumulated Recognition means for recognizing a writing trajectory written on the plane input medium based on a coordinate value series, and accuracy for adopting a recognition result having the highest accuracy among the recognition results obtained by the recognition means obtained for each pen-up With comparison means, the recognition result adopted is the last pen-up At Accumulated by All of When the recognition result is obtained from the coordinate value series, the recognition result is output as one character, and the adopted recognition result is Obtained from all coordinate value series accumulated up to the point of pen-up in the middle In this case, the recognition result is output as the first character of the two characters, and the remaining coordinate value series not corresponding to the adopted recognition result is recognized and output as the second character of the two characters. It is a feature.
[0026]
The character recognition method according to the present invention includes the steps of sequentially detecting coordinate values at which writing means on a plane input medium is located in accordance with a writing operation, and generating a coordinate value series written within a predetermined time for the area. Accumulating in order, notifying a contact state between the plane input medium and the writing means, and each time a pen-up (the writing means is separated from the plane input medium) state is notified All of the accumulated A step of recognizing a writing trajectory written on the plane input medium based on the coordinate value series, a step of employing the most accurate recognition result among the recognition results obtained for each pen-up, Is the last pen up At Accumulated by All of Outputting the recognition result as one character when the recognition result is obtained from a coordinate value series; Obtained from all coordinate value series accumulated up to the point of pen-up in the middle Outputting the recognition result as the first character of the two characters, and recognizing and outputting the remaining coordinate value series not corresponding to the adopted recognition result as the second character of the two characters; It is characterized by including.
[0031]
【Example】
Embodiment 1 FIG.
FIG. 1 is a block circuit diagram showing a character recognition device according to one embodiment of the present invention. In the figure, the same reference numerals as those in FIG. 5 is coordinate value extracting means, 6 is first switching means, 8 is second switching means, 9 is accuracy comparing means, 10 is storage means, 11 is third switching means, 12 is fourth switching means, 13 is a digit display means, 14 is a digit display means, 18 is a counting means, and 19 is a digit number judging means.
[0032]
Next, a case where the operation is applied to time setting will be described as an example. The information of month, day, hour, minute and second related to date and time is represented by a maximum of two digits. The operation of the common part with FIG. 7 is the same as the operation of the conventional example shown in FIG. 7, and the description of the operation is omitted here. The coordinate value detecting means 2 sequentially detects the coordinate values of the input medium 1 contacted by the writing operation with the pen, and supplies the coordinate values to the coordinate value accumulating means 4 and the area identifying means 15 at the same time.
[0033]
The pen information detecting means 3 detects whether the pen is in contact with the input medium (pen down) or not (pen up) and supplies it to the coordinate value accumulating means 4 and at the same time supplies it to the timing means 16. .
[0034]
The area identifying means 15 identifies whether the coordinate values sequentially supplied from the coordinate value detecting means 2 belong to any of several selected areas or entry frame areas formed on the surface of the input medium 1, and determines the area. Is supplied to the input end detecting means 17. The timer means 16 measures the elapsed time since the pen-up, and supplies the elapsed time to the input end detecting means 17. The measurement is stopped by the pen down, and the measured elapsed time is reset.
[0035]
The input end detection means 17 outputs an input end signal when the identification signal output from the area identification means 15 changes, that is, when it comes into contact with an area different from the area just touched immediately before. Alternatively, when the elapsed time output from the timer 16 exceeds a predetermined time, an input end signal is output. The input end signal output from the input end detecting unit 17 is supplied to the storage unit 10, the counting unit 18, and the digit number judging unit 19.
[0036]
The coordinate value accumulating means 4 sequentially acquires and accumulates coordinate values from the coordinate value series sequentially supplied from the coordinate value detecting means 2. When the pen information supplied from the pen information extracting means 3 changes from pen down to pen up, the stored coordinate values are supplied to the recognizing means 7 via the first switching means 6 as a coordinate value series to be recognized.
[0037]
The recognition means 7 is configured, for example, as shown in the block circuit diagram of FIG. In FIG. 8, the comparison unit 23 compares the input characteristic amount of the character with the characteristic amount of each character stored in the character dictionary storage unit 22 in order, and the characteristic amount input from the characteristic extraction unit 21. Then, the distance between the character and the characteristic amount of each character stored in the character dictionary storage unit 22 is obtained in order and supplied to the identification unit 24. The identification unit 24 determines the character in the character dictionary storage unit 22 having the shortest distance between the feature amounts as the recognition character. It can be said that the accuracy of the recognition result is high if the distance between the feature amounts when the recognition character is judged by the identification means 24 is short, and the accuracy of the recognition result is low if the distance is long. The recognition means 7 supplies the character C1 and the accuracy E1 recognized from the coordinate value series data to be recognized to the accuracy comparison means 9 via the second switching means 8.
[0038]
Each time the character C1 and the accuracy E1 are input from the second switching circuit 8, the accuracy comparing means 9 notifies the counting means 18 of the input. The counting means 18 counts the number of inputs in order. Since the number of inputs depends on the number of pen-ups based on the operation of the coordinate value accumulating means 4, it corresponds to the number of strokes of the writing locus for the entry frame. The counting means 18 is reset by an input end signal supplied from the input end detection means 17.
[0039]
The storage means 10 stores the character C0, the certainty E0, and the number of strokes S0 which have the highest certainty in the comparison by the certainty comparing means 9. The accuracy comparing means 9 compares the accuracy E1 supplied from the second switching circuit 8 with the accuracy E0 stored in the storage means 10. If the certainty E1 is higher than the certainty E0, the certainty E1 and the corresponding character C1 and the number of strokes S1 are written as the certainty C0, the character E0 and the number of strokes S0 in the storage means 10 to update the contents. If the accuracy E1 is lower than the accuracy E0, no update is performed. The storage unit 10 is reset by an input end signal supplied from the input end detection unit 17. When the character C1 and the certainty E1 are input to the certainty comparing means 9 in a state where the storage means 10 is reset, the character C1, the certainty E1 and the stroke S1 are respectively used as the first stroke, the certainty C0, the character E0 and the number of strokes. The data is written to the storage means 10 as S0. The storage unit 10 supplies the character C0 having the highest accuracy to the third switching circuit 11 in the comparison by the accuracy comparison unit 9.
[0040]
The number-of-digits judging means 19 is supplied with the stroke number S1 of the current writing locus from the counting means 18 and, at the same time, is supplied with the stroke number S0 at the time of the highest accuracy from the storage means 10. The digit number determining means 19 compares the stroke numbers S1 and S0 when the input end signal is supplied from the input end detecting means 17. When the stroke numbers S1 and S0 are equal, it means that the accuracy of recognizing the entire writing trajectory obtained at the end of the input is the highest. That is, the input writing locus is a one-digit number. When the number of strokes S0 is smaller than S1, it means that the writing trajectory with which the most accurate recognition can be performed is obtained as a trajectory that is stopped up to the stroke during input. That is, the input writing locus is a two-digit number. The number-of-digits determining means 19 determines whether the number is one or two, and stores the coordinate value accumulating means 4, the first switching means 6, the second switching means 8, the third switching means 11, and the fourth A one-digit control signal or a two-digit control signal is supplied to the switching means 12.
[0041]
When the two-digit control signal is supplied from the digit number judging means 19, the coordinate value accumulating means 4 supplies the stored coordinate values to the first switching means 6 as a coordinate value series to be recognized, and simultaneously outputs the coordinate values. It is supplied to the extraction means 5. The coordinate value extracting means 5 supplies the remaining coordinate value series excluding the coordinate value series corresponding to the stroke number S0 supplied from the storage means 10 to the first switching means 6.
[0042]
In the first switching means 6, the coordinate value series supplied from the coordinate value accumulating means 4 is added to one input terminal, and the remaining coordinate value series supplied from the coordinate value extracting means 5 is added to the other input terminal. When the two-digit control signal is applied to the control terminal, the coordinate value series supplied from the coordinate value extracting means 5 is used. When the two-digit control signal is not applied, the coordinate value supplied from the coordinate value accumulating means 4 is used. Switching control is performed so as to select a sequence, and the sequence is supplied to the recognition unit 7.
[0043]
The recognition means 7 supplies the character C1 and the accuracy E1 recognized from the coordinate value series data for one character to the second switching means 8. When the two-digit control signal is applied to the control terminal, the second switching means 8 performs switching control so that the character C1 is supplied from one output terminal to the fourth switching means 12. When the two-digit control signal is not applied to the control terminal, switching control is performed so that the character C1 and the accuracy E1 are supplied to the accuracy comparing means 9.
[0044]
The third switching means 11 performs switching control so that the character C0 is supplied from one output terminal to the tens display means 13 when a two-digit control signal is applied to the control terminal. When a one-digit control signal is applied to the control terminal, switching control is performed so that the character C0 is supplied to the fourth switching means 12. When neither the one-digit control signal nor the two-digit control signal is applied to the control terminal, none of the output terminals is selected. The tens place display means 13 displays the character C0 at a position on the left side of the entry frame.
[0045]
In the fourth switching means 12, the character C0 supplied from the third switching means 11 is added to one input terminal, and the character C1 supplied from the second switching means 8 is added to the other input terminal. When the two-digit control signal is applied to the control terminal, the character C1 supplied from the second switching means 8 is used. When the one-digit control signal is applied, the character C1 supplied from the third switching means 11 is used. Switching control is performed so as to select C0, and the result is supplied to the ones display means 14. The first place display means 14 displays the character C0 or C1 at a position on the right side of the entry frame.
[0046]
In the above embodiment, the recognition result of the input digit of a maximum of two digits is finally displayed only by the ones display means 14 and the tens display means 13, but control is performed using the recognition result. A timer device or the like may be configured.
[0047]
Next, the flow of processing of the character recognition device of FIG. 1 will be described with reference to the flowchart of FIG. When the pen is up (the pen is not in contact with the tablet) (102), the apparatus waits while detecting the pen state until the pen is down. When the pen is down (the pen touches the tablet) (102), the coordinate value is detected (103) and the coordinate value is stored (104). If the pen is still down (105), the next coordinate value is detected (103), the coordinate value is stored (104), and a loop is repeated to sequentially accumulate the coordinate values.
[0048]
When the pen is in the up state (105), as the end of one stroke, the characteristics of the writing trajectory of the stroke are extracted (106), the characters of the writing trajectory are recognized from the extracted features (107), and the recognition is performed. The result and accuracy are stored (108). Since there is no certainty in comparing the first stroke (109), the recognition result stored in 108 is stored as the provisional recognition result, and the certainty is stored as the highest certainty (110). When pen-down is performed again in the same entry frame within a predetermined time, and it can be determined that the writing operation is continued (111), the process returns to step 103.
[0049]
Similar to the above steps, the coordinate values are sequentially accumulated in accordance with the handwriting, and the recognition result and accuracy of the second stroke (not the handwriting locus of only the second stroke but the handwriting locus up to the second stroke) are stored (108). When the stored accuracy is higher than the previously stored maximum accuracy (109), the recognition result newly stored at 108 is stored as the provisional recognition result, and the accuracy is stored as the maximum accuracy (110). When the stored accuracy is lower than the previously stored maximum accuracy (109), the provisional recognition result does not change the maximum accuracy. If it can be determined that the pen down is performed again in the same entry frame within a predetermined time and the writing operation is continued (111), the process returns to step 103 again and the same processing (the expression of the second stroke is the nth stroke) And generalize).
[0050]
For example, the pen moves to a different entry frame, points to a separately provided selection area, or an input operation is not performed even if a predetermined time has elapsed since the pen was raised, and it can be determined that the writing operation has ended. At time (111), the process proceeds to the next process. When the provisional recognition result having the highest accuracy is obtained from the writing locus up to the final stroke (112), the provisional recognition result is output as a one-digit number (113).
[0051]
When the provisional recognition result having the highest accuracy is obtained from the writing locus up to the stroke in the middle (112), the provisional recognition result is output as the tenth digit of the two-digit number (115). Coordinate values constituting the remaining writing trajectory excluding the writing trajectory up to the stroke for which the provisional recognition result was obtained are extracted (116), the features of the remaining writing trajectory are extracted (117), and the writing trajectory is extracted from the extracted features. Is recognized (118), and the recognition result is output as the ones digit of the two-digit number (119).
[0052]
Next, the operation of the character recognition apparatus of FIG. 1 will be described with reference to FIGS. FIG. 3 shows an example of entry in the case where, for example, "2" (generally handwritten in one stroke) is input as one digit in one digit in the entry frame. In this case, the input is completed in one stroke, and the accuracy is determined. Is recognized as a single digit number "2" without needing to be compared.
[0053]
FIG. 4 shows an entry example in which, for example, "4" (generally written in two strokes) is input as a two-digit number with one digit in the entry frame. In this case, the input is completed in two strokes. When the stroke is decomposed for each stroke and the stroke of stroke 1 is recognized, the character is determined to be a character, but its accuracy is considerably low. (If the accuracy is too low, it may be rejected without assigning any character.) When the stroke of stroke 2 is recognized, it is probably recognized as "4" with high accuracy. When the accuracy of the recognition result of the writing trajectory of the stroke 1 and the stroke 2 is compared with the accuracy of the recognition result of the writing trajectory of the stroke 2, the recognition result of the stroke 2 is employed. Since the writing trajectory of the stroke 2 is all of the writing trajectories in this input, a one-digit number “4” is obtained as a final recognition result.
[0054]
FIG. 5 shows an example of entry in the case where, for example, "23" (each of which is generally handwritten in one stroke) is input as a two-digit number by combining a single stroke in the entry frame. The input ends with the stroke. When the stroke is decomposed for each stroke and the writing trajectory of the stroke 1 is recognized, it is probably recognized as “2” with high accuracy. When the writing trajectory of the stroke 2 is recognized, it is determined as a character, but its accuracy is considerably low. (If the accuracy is too low, it may be rejected without assigning any character.) Comparing the accuracy of the recognition result of the writing trajectory of stroke 1 and stroke 2, the accuracy of the recognition result of the writing trajectory of stroke 1 is better. Since it is high, the recognition result of stroke 1 is adopted. Since the writing trajectory of the stroke 1 is a part of the writing trajectory in this input, the recognition result “2” of the stroke 1 is in the tens place of the two-digit number. Only the remaining writing trajectory is extracted and recognized, and "3" is obtained as the first place of the two-digit number. As a result, a two-digit number “23” is obtained as a final recognition result.
[0055]
FIG. 6 shows an example of entry in a case where, for example, “45” (each of which is generally written with two strokes) is input as a two-digit number by combining two strokes into the entry frame. In this case, four strokes are used. Is completed. When the stroke is decomposed for each stroke and the stroke of stroke 1 is recognized, the character is determined to be a character, but its accuracy is considerably low. (If the accuracy is too low, it may be rejected without assigning any character.) When the stroke of stroke 2 is recognized, it is probably recognized as "4" with high accuracy. When the writing trajectory of the stroke 3 is recognized, it is determined as a character, but the accuracy is considerably low. When the writing trajectory of the stroke 4 is recognized, it is determined as some character, but the accuracy is considerably low. When the accuracy of the recognition result of the writing trajectory of the stroke 1, the stroke 2, the stroke 3, and the stroke 4 is compared, the recognition result of the writing trajectory of the stroke 2 is the highest. Since the writing trajectory of the stroke 2 is a part of the writing trajectory in this input, the recognition result “4” of the stroke 2 is the tenth place of the two-digit number. Only the remaining handwriting trajectory is extracted and recognized, and "5" is obtained as the first place of the two-digit number. As a result, a two-digit number “45” is obtained as a final recognition result.
[0056]
Here, a processing procedure when a two-digit number is input by a combination of two-stroke numbers such as “45” will be described in another way.
With the first pen up
Recognition of the writing trajectory of stroke 1 (recognition result 1),
In the second pen-up
Recognition of a combination of writing trajectories of strokes 1 and 2 (recognition result 12),
In the third pen-up
Recognition of a combination of writing trajectories of strokes 1 to 3 (recognition result 123),
In the fourth pen-up
Recognition of combinations of writing trajectories of strokes 1 to 4 (recognition result 1234)
And four times of recognition are performed, and a recognition result and accuracy (the accuracy increases as the recognition distance decreases and decreases as the recognition distance increases) are obtained. The recognition result with the highest accuracy among the accuracy is adopted. In this case, (recognition result 12) is adopted. If the recognition result of the first character is determined, the writing trajectory of the second character part is uniquely determined, so the remaining
Recognition of combination of writing trajectories of strokes 3 and 4 (recognition result 34)
I do. The final recognition result is obtained in a total of five recognition processes.
[0057]
In the above embodiment, in cutting out a character which is a problem in recognition of two characters, the start point of the first character is the start point of the entire writing trajectory, and the end point of the second character is the ending point of the entire writing trajectory. However, only the end point of the first character and the start point of the second character, that is, the boundary between the first character and the second character are unknown information. Since the starting point of the first character is known to obtain this unknown information, recognition is performed by assuming the search for the end point in order from the end point of the first stroke to the end point of the final stroke, and the accuracy of the recognition is compared. Thus, the end point of the stroke assumed in the recognition with the highest accuracy is determined as the end point of the first character. The start point of the second character can be automatically determined from the end point of the first character, and the start point of the next stroke after the stroke that is the end point of the first character is the start point of the second character. Since the starting point and the ending point of the second character are already known, it is not necessary to perform a search, and ordinary one-character recognition is sufficient.
[0058]
Also, the identification of whether the input is two characters or only one character can be considered as a special case of the processing procedure for cutting out two characters. This is the case where the search for the end point of the first character has the highest accuracy in the recognition up to the end point of the final stroke. In this case, the second character need not be recognized.
[0059]
Further, in order to search for the boundary between the first character and the second character, there is also a method of recognizing and searching for the starting point of the second character in order from the last stroke in reverse to the above procedure. However, the same effect as in the above embodiment can be obtained in other respects, although there is a disadvantage that the entire recognition time is delayed because the above process cannot be performed.
[0060]
In the above embodiment, if the recognition means employs a feature extraction method of extracting features by integrating all the writing trajectories described in the related art, for example, the writer can obtain a larger number of strokes than the number of strokes required for normal writing. Recognition of up to two characters is performed without any problem even when writing with a small number of strokes. In other words, if the character recognition method with the number of strokes is adopted in character recognition of one character, the feature of the number of strokes can be maintained even in character recognition of up to two characters. In addition, blurring or the like caused by the contact state between the plane input medium (tablet) and the writing means (pen) may result in pen-up unintended by the writer, resulting in a writing trajectory having a larger number of strokes than actually. However, even in this case, recognition of up to two characters is performed without any problem.
[0061]
In the above-mentioned embodiment, the month, day, hour, minute, and second such as date, time, etc. are described as numbers expressed in a maximum of two digits. The form of expression is widely used, and it is applicable to the era calendar, including Meiji, Taisho, Showa, and Heisei at the present time. Also, in Japan, TV channels and the like can be generally represented by a maximum of two digits even in consideration of CATV, and are applicable. The present invention is not limited to these, and can be applied to a meaningless combination of numbers, and the operation is exactly the same.
[0062]
Although the above embodiment has mainly been described with reference to a two-digit number, the present invention is not particularly limited to a number, but includes any two-character idioms, hiragana, katakana, kanji, alphabetic characters, symbols, etc. It can be applied to the two combinations of characters by associating the character range stored in the character dictionary storage means with the same operation.
[0063]
【The invention's effect】
As described above, in the character recognition device and the character recognition method of the present invention, the recognition result with the highest accuracy is used from among the recognition results obtained for each pen-up, and the adopted recognition result is the last pen-up. At Accumulated by All of When the recognition result is obtained from the coordinate value series, the recognition result is output as one character, and the adopted recognition result is Obtained from all coordinate value series accumulated up to the point of pen-up in the middle In this case, the recognition result is output as the first character of the two characters, and the remaining coordinate value series not corresponding to the adopted recognition result is recognized and output as the second character of the two characters. It can be entered as a single information, whether the information of up to two characters is one character or two characters, and the user can input naturally according to the human thought process without being confused by the entry method or trial and error There is an effect that can be performed.
[0064]
Further, since the clipping between two characters is performed based on the accuracy of recognition performed for each pen-up, the pen-up time, the pen moving direction at the time of pen-up, and the pen moving distance at the time of pen-up As compared with a method performed based on such parameters as described above, there is an effect that the possibility of malfunction due to a difference in writing style of the user or the like is reduced.
[0065]
Also, since one information is limited to a maximum of two characters and the combination of strokes is verified only for the search for the end point of the first character, the search is not necessary for the subsequent points. The number of individual recognition processes is smaller and the recognition distance of the character combination is smaller than in the case where the method of verifying all combinations of strokes, searching for the optimum character combination and obtaining a recognition result is limited to a maximum of two characters. There is no need to perform the summation, and the processing procedure can be easily implemented in a smaller scale by hardware or software, and the overall recognition speed is increased.
[0066]
In addition, since information on the maximum number of strokes of the character to be recognized is not used for cutting out characters, even if the number of strokes is larger than the number of strokes required for normal writing, even if the number of strokes is increased due to blurring, there is no problem. Has a recognizable effect.
[0067]
Also, If you limit the character type to numbers, The configuration of the character dictionary storage means is easy, the recognition speed is fast, and there is an effect that erroneous recognition is reduced.
[0068]
In addition, since the information of up to two digits can be entered as a single information regardless of whether it is a one-digit number or a two-digit number, the year, month, day, hour, such as date and time, There is an effect that a user who sets information of up to two digits such as minutes, seconds, TV channels, etc. can perform a natural input according to a human thinking process without being confused by an entry method or trial and error.
[Brief description of the drawings]
FIG. 1 is a block circuit diagram of a character recognition device according to a first embodiment of the present invention.
FIG. 2 is a flowchart illustrating an operation of the character recognition device according to the first exemplary embodiment.
FIG. 3 is a diagram for explaining the operation of the character recognition device according to the first embodiment.
FIG. 4 is a diagram for explaining the operation of the character recognition device according to the first embodiment.
FIG. 5 is a diagram for explaining the operation of the character recognition device according to the first embodiment.
FIG. 6 is a diagram for explaining the operation of the character recognition device according to the first embodiment.
FIG. 7 is a block circuit diagram showing an example of a conventional character recognition device.
FIG. 8 is a block circuit diagram for explaining the operation of a recognition unit of a conventional character recognition device.
FIG. 9 is a diagram showing an entry example for explaining the operation of the conventional character recognition device.
FIG. 10 is a diagram showing an entry example for explaining the operation of a conventional character recognition device.
[Explanation of symbols]
REFERENCE SIGNS LIST 1 input medium, 2 coordinate value detecting means, 3 pen information extracting means, 4 coordinate value storing means, 5 coordinate value extracting means, 6 first switching means, 7 recognition means, 8 second switching means, 9 accuracy comparing means , 10 storage means, 11 third switching means, 12 fourth switching means, 13 10's position display means, 14's position display means, 15 area identification means, 16 clocking means, 17 input end detection means, 18 counting Means, 19 digit number determining means.

Claims (6)

A plane input medium including coordinate detection means for sequentially detecting coordinate values where the writing means is positioned according to a writing operation, and accumulation means for sequentially accumulating a coordinate value series written within a predetermined time in the area. a contact state notification means for notifying the state of contact between the flat input medium and the writing means, the pen up from the contact notification means until it each time (from flat input medium said writing means leaves) state is notified Recognition means for recognizing a writing trajectory written on the plane input medium based on all accumulated coordinate value series, and recognition with the highest accuracy among the recognition results obtained by the recognition means obtained for each pen-up and a probability comparing means for adopting the result, adopted recognition result and one character the recognition result if obtained from all of the coordinate value sequence stored up to the time of the last pen-up Output Te, if adopted recognition results have been obtained from all of the coordinate value sequence stored in up to the point of the middle of the pen-up outputs the recognition result as the first character of the two letters together, A character recognition device for recognizing and outputting a remaining coordinate value series that does not correspond to an adopted recognition result as a second character of two characters. The recognition unit includes a feature extraction unit that extracts a feature amount from an input coordinate value series, a character dictionary storage unit that stores a feature amount of a predetermined character, and a feature amount output from the feature extraction unit. Comparison means for sequentially comparing the characteristic amounts of the characters stored in the character dictionary storage means to obtain a difference between the characteristic amounts; and the character dictionary storage means having the smallest difference between the characteristic amounts obtained by the comparison means 2. The character recognition device according to claim 1, further comprising: an identification unit that uses the character stored in the storage unit as a recognition result. 3. The character recognition apparatus according to claim 2, wherein said character dictionary storage means is a number dictionary storage means that stores only a feature amount of a one-digit number. Sequentially detecting the coordinate values at which the writing means on the plane input medium is located according to the writing operation; sequentially accumulating a series of coordinate values written within a predetermined time in the area; A step of notifying a contact state between the medium and the writing means, and a step of notifying a pen-up state (the separation of the writing means from the plane input medium) based on all coordinate value series accumulated up to that time. A step of recognizing a writing trajectory written on the plane input medium, a step of employing a recognition result with the highest accuracy among recognition results obtained for each pen-up, and a step of recognizing the adopted pen-result at the last pen-up . and outputting the recognition result as one character if they are derived from accumulated all coordinate value sequence was up, when adopted recognition result in the middle of the pen-up If you are obtained from all of the coordinate value sequence stored until, and outputs the recognition result as the first character of two characters, the remaining coordinate value sequence that does not correspond to the adopted recognition result 2 Recognizing and outputting the second character of the characters. The step of adopting the recognition result with the highest accuracy among the recognition results obtained for each pen-up includes the steps of extracting a feature amount from the input coordinate value sequence, and the feature amount extracted from the input coordinate value sequence. to determining a difference between the feature quantity and a feature quantity of characters stored in the character dictionary compared sequentially, the character is the difference between the feature amounts are remembers the smallest the character dictionary as a recognition result 5. The character recognition method according to claim 4, comprising a step. 6. The character recognition method according to claim 5, wherein the character dictionary is a number dictionary that stores only a feature amount of a one-digit number.

Images