JP3817281B2 - Information processing apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an information processing apparatus and method, and more specifically, information processing that extracts a feature point in a point data group constituting the stroke from an input handwritten stroke and passes information based on the feature point to a lower-level process The present invention relates to an apparatus and a method.
[0002]
[Prior art]
Conventionally, a number of methods have been devised for character recognition methods for recognizing input handwriting and converting it into character codes. Some of these methods are called vector matching methods. This method is as follows.
[0003]
First, a coordinate point sequence delimited by pen-down information and pen-up information is called one stroke for the input handwriting, and the stroke is vectorized and expressed as a stroke vector. Next, each of the plurality of stroke vectors is compared with a standard stroke vector stored in advance in the recognition dictionary, and the degree of difference is calculated. In addition, the recognition dictionary defines several standard vectors that make up the character. By referring to the defined standard stroke vector set, the degree of coincidence with the input stroke vector set is calculated, and the most similar A character that is determined to be a character is output as a recognition result.
[0004]
Here, the method of vectorizing the stroke is generally to create an n number of vectors by dividing the entire stroke line segment into n after connecting the input coordinate points from the start point to the end point to make the stroke. Yes. This vector sequence is extracted as a stroke feature. The n-division method divides the entire length of the stroke by n and divides the stroke at equidistant intervals.
[0005]
[Problems to be solved by the invention]
However, in the conventional handwritten character recognition method as in the above-described conventional example, the division point is determined based on the distance obtained by dividing the total length of the stroke by n, but similar characters such as “RU”, “RO”, “N” If the character of the character is near the final point of the stroke, such as There was a problem that it was difficult.
[0006]
[Means for Solving the Problems]
The present invention has been made in view of the above problems, and it is an object of the present invention to provide an information processing apparatus and method that extract accurate feature information of a stroke from input handwritten stroke information and pass it to lower processing. is there.
[0007]
In order to solve this problem, for example, an information processing apparatus of the present invention has the following configuration. That is,
A feature point is extracted from a point data group P (0),..., P (k) constituting the stroke from a handwritten stroke input from a predetermined input device, and information based on the feature point is passed to a lower-level process. An information processing apparatus,
Stroke storage means for storing point data P (0),..., P (k) which is handwritten stroke information input from the input device;
In the point data stored in the stroke storage means,
Starting point P (i) of the point data (initial value i = 0),
Reference point P (i + 1 + j) (initial value j = 1)
When a vector defined by the starting points P (i) and P (i + 1) is defined as a reference vector, and a vector defined by P (i + 1) and the reference point P (i + 1 + j) is defined as a reference vector,
A comparison means for comparing an angle θ between the reference vector and the reference vector with a predetermined threshold;
If the angle θ is equal to or smaller than a predetermined threshold as a result of comparison by the comparison means, the point P (i + j) is set as a deletion target, and the variable j is increased by “1” and the comparison means is re-executed. And
As a result of comparison by the comparison means, when the angle θ exceeds a predetermined threshold value, the point P (i + j) is determined as a feature point, the variable i is updated based on the variable j at that time, and the variable j is "1" re-initialized, a first control means for re-executing said comparing means,
Discriminating means for discriminating whether or not the number of feature points obtained by the first control means has reached a number that can specify n consecutive vector sequences with respect to the target stroke;
A second control unit that executes the process by the first control unit again after reducing the predetermined threshold when the determination unit determines that the number of n consecutive vector sequences is less than the number that can be specified; With .
[0009]
In this case, when the number of consecutive vectors exceeds the n, the angle θ among the feature points obtained by the first control unit is further determined. It is desirable to provide an extraction means for extracting the top n + 1 feature point data from the one with the largest. As a result, n vector data can always be extracted.
[0010]
Further, it is desirable that the lower-order process is a character recognition process. As a result, the character recognition rate can be improved.
[0011]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
[0012]
<First Embodiment>
FIG. 1 is a block diagram showing a functional configuration of the online handwritten character recognition apparatus according to the first embodiment. In the figure, reference numeral 101 denotes a coordinate input device such as a doublet or a digitizer that inputs coordinates by operating a coordinate input pen. In this embodiment, 101 is a character input unit that acquires input character data. A stroke storage unit 102 stores a set of coordinate points between pen-down information and pen-up information as a single stroke using the coordinate points obtained by the character input unit 101. Reference numeral 103 denotes a feature extraction unit, which extracts data as feature points from the stroke data in the stroke storage unit 102. Reference numeral 104 denotes a feature point erasure unit, which generates n + 1 point data for constructing n vectors from the feature point data obtained by the feature point extraction unit 104, so that unnecessary data is erased. It is a point erasure unit. Reference numeral 105 denotes a vectorization unit, which generates n vector data based on n + 1 point data obtained from the feature point erasure unit 105. As described above, when one handwritten character is input by the character input unit 101, stroke information from pen down to pen up is sequentially extracted, and the vectorization unit 105 stores n vectors for each stroke. Is done.
[0013]
A recognition unit 106 receives vector data for each stroke from the vectorization unit 105 and performs a comparison operation with a basic vector stored in the recognition dictionary unit 108 to recognize characters. Reference numeral 107 denotes a result output unit that outputs the recognition result (character code) output from the recognition unit 106. In the character recognition process, the most likely candidate is normally output as the first candidate, and then output as the second candidate, the third candidate,. This embodiment also complies with this.
[0014]
The processing contents of the feature extraction unit 103 in the above configuration will be described with reference to the flowcharts of FIGS.
[0015]
FIG. 2 is a diagram of the strokes that have been entered (handwriting between pen-down and pen-up), and illustrates the coordinate points that make up the stroke. The coordinate point P0 is the start point, and the coordinate point Pk is the end point. Accordingly, the total number of coordinate points is k + 1.
[0016]
FIG. 3 is a flowchart showing processing contents for extracting feature points of a stroke. In the following description, a vector from the point Pn to the point P (n + 1) is represented as V (PnP (n + 1)).
[0017]
First, the start point counter i is set to 0 in step S301. In step S302, the target point counter j is set to 1. In step S303, it is determined whether or not the total value of the variables i and j is equal to or less than k, that is, whether or not the present process is completed. If not, the process proceeds to the following step S304.
[0018]
In step S304, whether or not the angle θ formed by the vector V (PiP (i + 1)) and the vector V (P (i + 1) P (i + 1 + j)) is equal to or smaller than a preset threshold angle θt. Determine whether. If θ ≦ θt, the process proceeds to step S305, where the coordinate point P (i + j) is marked as a deletion target point. In step S306, the target point counter j is incremented to check the deletion of the next coordinate point. Return to.
[0019]
Hereinafter, this process is repeated. If it is determined that the relationship between the angles θ and θt formed by the two vectors is such that θ> θt during this period, the process proceeds to step S307, and the variable i is updated with “i + 1 + j”, and the process returns to step S302.
[0020]
Now, when all the coordinate points have been checked in step S305, this processing is completed. After this processing is completed, the marked coordinate points are continuous coordinate points with little change in angle, and what remains after deleting this point group becomes the coordinate point of the portion with a large angle. Note that both the start point P0 and the end point Pk are regarded as feature points. The threshold angle θt in the above process is parameterized and is variable. The reason will be described later.
[0021]
The processing contents will be described in more detail with reference to FIG. Assume that the point Pi is the base point and j = 1. In this case, the two vectors are
V (PiP (i + 1)), V (P (i + 1) P (i + 2))
It becomes.
[0022]
If the angle θ1 formed by these two vectors is less than or equal to the threshold angle θt, one of the end points P (i + 2) of the second vector V (P (i + 1) P (i + 2)) The previous point P (i + 1) is set as a deletion target. As a result, the variable j is incremented by 1 (j = 2).
V (PiP (i + 1)), V (P (i + 1) P (i + 3))
Is compared with the threshold angle θt. Here, when θ2 is also equal to or smaller than the threshold angle θt, the point P (i + 2) is set as a deletion target, and the variable j is incremented (j = 3). As a result, the nose angle θ3 of the next two vectors is compared with the threshold angle θt. That is,
V (PiP (i + 1)), V (P (i + 1) P (i + 4))
When this θ3 is larger than the threshold angle θt, the point P (i + 3) is set as a feature point (singular point).
[0023]
As a result, when both θ1 and θ2 are equal to or smaller than the threshold angle θt and θ3 is larger than the threshold angle θt, the points P (i + 1) and P (i + 2) are set as deletion targets. . In other words, when the degree of bending of the coordinate points constituting the stroke is small, each point is set as a deletion target.
[0024]
In step S307, the variable i is updated with “i + 1 + j”, and in step S302, the variable j is initialized with “1”. Therefore, the point Pi moves to the position of the point “P (i + 4)” in FIG. Then, the processing described above is repeatedly executed from there.
[0025]
In order to divide a stroke into n (to extract n vectors from one stroke), the number of feature points is n + 1. However, when the above processing is performed, the number obtained depending on the threshold angle θt and the stroke of interest is indefinite, and n + 1 points cannot be obtained uniformly. Therefore, in the present embodiment, the threshold value θt when performing the processing of FIG. 3 is made variable so that an appropriate number of feature points can be extracted. This process will be described with reference to the flowchart of FIG.
[0026]
First, in step S501, the threshold angle θt used in the feature extraction process is set to 180 ° as an initial value. In step S502, the feature extraction process described with reference to FIG. 3 is performed using the set threshold angle θt.
[0027]
Thereafter, the process proceeds to step S503, where the number of feature points obtained in step S502 is checked to determine whether or not the number of feature points is n + 1 or more. If the number of extracted feature points is less than n + 1, the process proceeds to step S504, the threshold angle θt is decreased by a predetermined angle (for example, 1 °), and the process returns to step S502.
[0028]
Hereinafter, when the threshold angle θt is sequentially decreased and the processing by the feature point extraction unit 103 is performed, the number of obtained feature points increases, and finally n + 1 or more feature points are obtained. At this time, the process proceeds from step S503 to step S505, the obtained feature point data is output to the feature point erasing unit 104, and this process ends.
[0029]
Next, the processing operation of the feature point erasing unit 145 will be described with reference to the flowchart of FIG.
[0030]
As described above, at least n + 1 feature points are output from the feature point extraction unit 103 (n + 1 or more feature points are obtained). However, when n vectors are extracted from the target stroke, if there are more than n + 1 feature points, it is desirable to remove the corresponding feature points. Therefore, the feature point deletion unit 104 deletes the extra feature points by performing processing according to the flowchart shown in FIG.
[0031]
First, in step S601, the number of feature points is checked, and the process ends when the number of feature points reaches n + 1. If the number of feature points is greater than n + 1, in step S602, processing is performed to remove the point with the smallest angle change between adjacent feature points from the feature point. This process may use the technique used in the feature point extraction process.
[0032]
As a result of the above, from the feature point erasure unit 104, the points in the vicinity where the curvature of the trajectory represented by the stroke data is large remain finally, the points having a small curvature are erased, and finally, n vector data Only data for generating is generated.
[0033]
Next, the processing operation of the vectorization unit 105 will be described with reference to the flowchart of FIG.
[0034]
In step S701, the counter i is initialized to “0”. In step S702, it is checked whether vector processing has been performed for all n + 1 feature points. If this check result is negative, the following steps S703 to S705 are performed.
[0035]
First, in step S703, a direction vector of a straight line passing through two adjacent feature points is obtained. The direction vector is an eight-direction vector shown in FIG. The obtained direction vector is stored as a vector value in step S704. In step S705, a process for moving to the next feature point is performed, and the process returns to step S02.
[0036]
When vector values are obtained for all feature points in this way, n vector values are generated in step S706.
[0037]
The character recognition unit 106 performs character recognition processing based on all vector data sent from the vectorization unit 105 when input of all strokes for one character input is completed. That is, the input vector data of each stroke is compared with the standard vector data in the dictionary unit 108, and the first candidate, the second candidate,. Send to part 107.
[0038]
In addition, although explanation is mixed, when the time series coordinate data which constitutes the input stroke data are arranged almost on a straight line (for example, when a straight line or a line close to it is input by handwriting), both end points are features. Although it is processed as a point, it is difficult to specify a feature point among points located in the middle. However, as the vector data (see FIG. 8) for the target stroke output from the vectorization unit 105, the same vector code is continuous, so the result is the same regardless of which point is the feature point. Therefore, in the process of FIG. 5, many feature points are extracted at a stroke with one threshold angle θt (a value close to 0 °) and passed to the process of FIG. 6, but there are a plurality of feature points having the same angle in step S602. In such a case, it can be dealt with by simply determining any one as a feature point, so there is no problem.
[0039]
In the above embodiment, the result output unit 107 performs a process of passing the result to the lower processing unit as a character recognition result. As the lower-level processing, for example, there is document processing. In short, any processing is possible as long as processing is performed based on the result obtained by the character recognition processing.
[0040]
As described above, with respect to stroke data (coordinate data group) having a portion with a large degree of bending or a portion with a large degree of bending, the coordinate data in the vicinity where the curvature is large is positively processed as an effective point. So, for example, if you input a flat “RU”, you can always get a feature vector with good accuracy of handwritten characters, because the last part, as well as the part where the vector direction is large, will be held reliably. As a result, the recognition rate can be increased.
[0041]
In the present embodiment, feature points are extracted using vector angles for feature extraction processing. However, the present invention is not limited to angles, and it is used in combination with a process of erasing coordinate points within a certain distance and thinning out coordinate points. It doesn't matter.
[0042]
<Second Embodiment>
In the above-described embodiment (first embodiment), the method of performing n-division using the extraction process for extracting the feature of the stroke has been described. However, in the second embodiment, the coordinate point density within a certain range in the stroke is calculated. A method of calculating and dividing n by changing the number of divisions according to the density will be described.
[0043]
FIG. 9 is a block diagram showing a functional configuration of the online handwritten character recognition method in the second embodiment. In FIG. 9, 901, 902, 906, 907, and 908 perform substantially the same processing as 101, 102, 106, 107, and 108 in FIG.
[0044]
The difference is that a point density calculation unit 903 is newly provided. Hereinafter, processing contents of the point density calculation unit 903 will be described with reference to the flowchart of FIG.
[0045]
First, in step S1001, the total length of the stroke is calculated. The total length of the stroke can be obtained by calculating the Euclidean distance between two adjacent coordinates constituting the stroke, and starting from the start point coordinates and adding to the end point coordinates. Then, the obtained total length is equally divided into n, and the value is set to l.
[0046]
In step S1002, a coordinate point having a length l on the stroke is mapped, and this coordinate point is called an n-divided point Pi (0 ≦ i ≦ n + 1). Here, “on the stroke” refers to a straight line formed by connecting coordinate points (see FIG. 11).
[0047]
Next, in step S1003, the interval between the first n dividing point Pi and the second n dividing point Pi + 1 (0 ≦ i ≦ n + 1) is Rj (0 <= j <= n), and the number of coordinate points in the interval Rj is set. Count. The number of coordinate points counted is the point density. FIG. 12 and FIG. 13 show a state of division into four (n is 4) and a point density for each section for a certain stroke.
[0048]
Next, the operation of the dividing point determination unit 904 will be described with reference to FIG.
[0049]
First, in step S1401, the section having the maximum point density (the number of sections in the section relative to the total number of points) is searched from each section obtained by the point density calculation unit. If it is determined in step S1402 that the point density of the section is larger than a predetermined threshold value, the process proceeds to step S1403.
[0050]
If the process proceeds to step S1403, the image is re-divided so that the point density is equal to or less than the threshold value. For example, the length of the section is obtained and divided into two equal parts, a new section is created, and the point density of each section is calculated. The point density calculation method is the same as that of the point density calculation unit. Since the number of divisions has increased in this process (because n + 1 sections have occurred), in step S1404, a section having the smallest point density is searched for, and in step S1405, a section adjacent to it (if there is a point density on both sides) Merged with lower section). In addition, the n division points that originally divided the merged sections are deleted. As a result, the target stroke is returned to the n-division state again. The point density after the merger is obtained by calculating the point density average of both sections. Hereinafter, when such an operation is repeated and the process is completed, a newly generated n-division point is obtained. Then, the point coordinates at both ends of the dividing point are output as effective coordinate data as a vector. As a result, it is possible to cope with the fact that the writing speed is slowed at a portion where the direction of the stroke is unconsciously changed by humans, and that many coordinate points are detected at that portion, so that the division can be performed with high accuracy. Thereafter, vectorization is performed using the obtained n division points, and the same processing as in the first embodiment is performed.
[0051]
<Third Embodiment>
In the first and second embodiments, the present invention is realized by the apparatus having the configuration shown in FIG. 1 or FIG. 9, but the present invention is not limited to this, and may be realized by a program.
[0052]
FIG. 15 shows a block configuration of an apparatus according to the third embodiment. In the figure, 1 is a CPU that controls the entire apparatus, 2 is a ROM that stores a boot program and the like, and 3 is a RAM that stores programs and data processed by the CPU 1. Reference numeral 4 denotes a coordinate input board. Reference numeral 5 denotes a coordinate detection unit that detects coordinates traced on the coordinate input board 4 with an input pen (not shown) by an operator. Note that the input pen is provided with a switch for detecting whether or not the coordinate input plate 4 is pressed, and the pressing state of the switch is output to the CPU 1. Reference numeral 6 denotes a liquid crystal display (LCD) provided on the lower surface of the coordinate input plate 4. Reference numeral 7 denotes an LCD driver for controlling driving of the liquid crystal display 6.
[0053]
Here, the CPU 1 controls the LCD driver 7 so as to connect the point coordinate data sequentially input from the coordinate detection unit 5 with line segments while detecting pen-down, so that it is as if characters are written on the paper. Thus, the handwriting can be displayed on the LCD 6.
[0054]
Reference numeral 8 denotes an external storage device such as a floppy disk or a hard disk device, which stores an operating system, various application programs, and programs related to character recognition processing according to the present embodiment. An area for storing a character recognition dictionary and a recognition result (character code) to be referred to when the recognition process is performed is also secured.
[0055]
In the above configuration, when the character recognition is performed, the CPU 1 loads a program for character recognition into the RAM 3 and performs character recognition processing according to the loaded program.
[0056]
An example of a character recognition program stored in the external storage device 8 is shown in FIG. In the figure, a storage medium 160 is shown, which is compatible with the external storage device 8 in the above embodiment. Specifically, any medium that stores information, such as a floppy disk, a hard disk, or a CD-ROM, is used. Of course, anything can be used.
[0057]
In the figure, 161 is a detection module that detects the degree of change in the vector direction of the point data group of the input stroke, 162 is a comparison module that compares the detected change in vector direction with a predetermined threshold, and 163 If the change in the vector direction is greater than or equal to the threshold value as a result of the comparison, some point data constituting the vector is valid. It is a control module that removes.
[0058]
Here, the storage medium further determines whether or not the number of effective point data obtained by the control module has become a number that can specify n consecutive vector sequences with respect to the target stroke. A determination module and a second module that adjusts the threshold and executes the detection module, the comparison module, and the control module again when it is determined that a number sufficient to identify n consecutive vector sequences cannot be obtained. It is desirable to have a control module.
[0059]
Further, when the determination module determines that the number of effective point data obtained based on the threshold value adjusted by the second control module is n + 1 or more, the storage medium has the obtained feature points. It is desirable to provide a module that extracts the upper n + 1 point data from those having a large vector change.
[0060]
Furthermore, it is desirable to provide a character recognition processing module as a lower level process.
[0061]
The above storage medium 160 corresponds to the first embodiment, but if applicable to the second embodiment, a point density detection module that detects the density distribution of the point data group of the input stroke, and It is easy from the above example to include a section setting module that sets sections based on the level of the obtained point density, and an extraction module that extracts both end points of each set section as feature points. I can think of it.
[0062]
In this case, the point density detection module counts the number of coordinate points included in each divided section, and a division module that equally divides the stroke length represented by the time-series point coordinate data of the input stroke. And a counting module.
[0063]
The partition setting module includes a comparison module that compares the maximum point density of each partition detected by the point density detection module with a predetermined density, and the maximum point density when the maximum point density is equal to or higher than the predetermined density. A first partition correction module that generates n + 1 partitions by dividing the point density partition into two by the partition length, and the lowest point density partition of the generated n + 1 partitions and the adjacent one. It is desirable to have a second partition correction module that combines the partitions to reset n partitions and run the comparison module again.
[0064]
Also in this case, in order to function as a character recognition device, it is desirable to include a character recognition module as a low-order process.
[0065]
As described above, according to this embodiment, when recognizing online handwritten characters, it is possible to extract highly accurate vector data from the input stroke data.
[0066]
In the embodiment, the handwritten information is directly input via the character input unit. However, the handwritten information may be input from a storage medium storing the handwritten information or via communication. The present invention is not limited to the above embodiment.
[0067]
Further, as described in the third embodiment, the present invention relates to a technique for extracting a feature point from input stroke information and passing information corresponding to the feature point to a lower-level process. As long as it does not matter. However, when character recognition processing is adopted as the lower-level processing, the recognition rate can be improved, so it is desirable to adapt to character recognition processing.
[0068]
As described above, the present invention may be applied to a system constituted by a plurality of devices or an apparatus constituted by one device. Needless to say, the present invention can also be applied to a case where the present invention is implemented by supplying a program to a system or apparatus. In this case, the storage medium storing the program according to the present invention constitutes the present invention. Then, by reading the program from the storage medium to the system or apparatus, the system or apparatus operates in a predetermined manner.
[0069]
【The invention's effect】
As described above, according to the present invention, it is possible to extract feature information with high accuracy of the stroke from the input handwritten stroke information and pass it to a lower-level process.
[0070]
[Brief description of the drawings]
FIG. 1 is a block diagram showing a functional configuration of an online handwritten character recognition method in the present embodiment.
FIG. 2 is a diagram showing an example of strokes written in the present embodiment.
FIG. 3 is a flowchart showing a feature extraction process in the first embodiment.
FIG. 4 is a diagram for explaining the principle of feature extraction processing;
FIG. 5 is a flowchart showing control of the number of feature points in the first embodiment.
FIG. 6 is a flowchart showing an operation of a feature point erasing unit in the first embodiment.
FIG. 7 is a flowchart showing the operation of the vectorization unit in the present embodiment.
FIG. 8 is a diagram showing direction vectors in the present embodiment.
FIG. 9 is a functional block diagram of an on-line handwritten character recognition apparatus according to the second embodiment.
FIG. 10 is a flowchart illustrating an operation of a point density calculation unit according to the second embodiment.
FIG. 11 is a diagram illustrating n division points according to the second embodiment.
FIG. 12 is a diagram showing a state of stroke division in the second embodiment.
FIG. 13 is a diagram showing a point density after stroke division in the second embodiment.
FIG. 14 is a flowchart illustrating an operation of a division point determination unit according to the second embodiment.
FIG. 15 is a block diagram showing a specific device configuration according to the third embodiment.
FIG. 16 is a diagram showing program modules stored in a storage medium according to the third embodiment.
[Explanation of symbols]
101, 901 Character input unit 102, 902 Stroke storage unit 103 Feature extraction unit 104 Feature point deletion unit 105, 905 Vectorization unit 106, 906 Recognition unit 107, 907 Result output unit 108, 908 Dictionary unit 903 Point density calculation unit 904 Division Point determination part

Claims (6)

A feature point is extracted from a point data group P (0),..., P (k) constituting the stroke from a handwritten stroke input from a predetermined input device, and information based on the feature point is passed to a lower-level process. An information processing apparatus,
Stroke storage means for storing point data P (0),..., P (k) which is handwritten stroke information input from the input device;
In the point data stored in the stroke storage means,
Starting point P (i) of the point data (initial value i = 0),
Reference point P (i + 1 + j) (initial value j = 1)
When a vector defined by the starting points P (i) and P (i + 1) is defined as a reference vector, and a vector defined by P (i + 1) and the reference point P (i + 1 + j) is defined as a reference vector,
A comparison means for comparing an angle θ between the reference vector and the reference vector with a predetermined threshold;
If the angle θ is equal to or smaller than a predetermined threshold as a result of comparison by the comparison means, the point P (i + j) is set as a deletion target, and the variable j is increased by “1” and the comparison means is re-executed. And
As a result of comparison by the comparison means, when the angle θ exceeds a predetermined threshold value, the point P (i + j) is determined as a feature point, the variable i is updated based on the variable j at that time, and the variable j is "1" re-initialized, a first control means for re-executing said comparing means,
Discriminating means for discriminating whether or not the number of feature points obtained by the first control means has reached a number that can specify n consecutive vector sequences with respect to the target stroke;
A second control unit that executes the process by the first control unit again after reducing the predetermined threshold when the determination unit determines that the number of n consecutive vector sequences is less than the number that can be specified; the information processing apparatus characterized in that it comprises and. Further, when the determining means determines that the number of consecutive vectors exceeds the n, among the feature points obtained by the first control means, the top n + 1 from the features having the large angle θ. the information processing apparatus according to claim paragraph 1, characterized in that it comprises an extraction means for extracting feature points data of.   The information processing apparatus according to claim 1, wherein the lower-order process is a character recognition process. A feature point is extracted from a point data group P (0),..., P (k) constituting the stroke from a handwritten stroke input from a predetermined input device, and information based on the feature point is passed to a lower-level process. An information processing method,
A step of storing point data P (0),..., P (k), which is handwritten stroke information input from the input device, in a stroke storage means;
In the point data stored in the stroke storage means,
The starting point of the point data is P (i) (initial value i = 0),
The reference point is P (i + 1 + j) (initial value j = 1),
When a vector defined by the starting points P (i) and P (i + 1) is defined as a reference vector, and a vector defined by P (i + 1) and the reference point P (i + 1 + j) is defined as a reference vector,
A comparison step of comparing an angle θ formed by the reference vector and the reference vector with a predetermined threshold;
When the comparison result of the comparison step shows that the angle θ is equal to or smaller than a predetermined threshold, the point P (i + j) is set as a deletion target, and the variable j is increased by “1” and the comparison step is re-executed. And
As a result of the comparison in the comparison step, when the angle θ exceeds a predetermined threshold value, the point P (i + j) is determined as a feature point, the variable i is updated based on the variable j at that time, and the variable j is "1" re-initialized, a first control step of re-executing said comparing step,
A discriminating step for discriminating whether or not the number of feature points obtained in the first control step has reached a number that can identify n consecutive vector sequences with respect to the target stroke;
A second control step of executing the processing by the first control means again after subtracting the predetermined threshold when it is determined by the determination step that the number of n consecutive vector sequences is less than the number that can be specified; an information processing method characterized in that it comprises and. Further, when it is determined by the determination step that the number of consecutive vectors exceeds the n, among the feature points obtained in the first control step, n + 1 high-order points from the one having the large angle θ. The information processing method according to claim 4, further comprising an extraction step of extracting the feature point data. The information processing method according to claim 4, wherein the lower-order process is a character recognition process.

Images