KR100290606B1 - Method for recognizing cursive of hangul in online system - Google Patents

Abstract

PURPOSE: An online Hangul cursive recognition method is provided to designate a start and final point of a vowel in a Hangul character in advance, to extract a boundary of the vowel from the designated area, and to automatically determine an area of an initial consonant and a final consonant based on the extracted area so that it can enhance a recognition speed and efficiency in a PDA or a Hand Held PC. CONSTITUTION: The method comprises steps of a microprocessor inputting 2 dimensional coordinate points along a written trajectory of a Hangul character input via an input device(200), the microprocessor preprocessing the input Hangul character(202), the microprocessor determining a vertical or horizontal segment by tracking preprocessed segment sets in written order, and classifying segments satisfying a condition into a vertical or horizontal segment(204), the microprocessor determining candidate vowel areas from the vertical or horizontal segments according to the vowel type defined in advance(206), the microprocessor recognizing the vowel from the candidate vowel areas by using an elastic matching method(208), the microprocessor recognizing an initial and a final consonant on a basis of the recognized vowel area(210), and the microprocessor recognizing a character by combining the initial consonant, the vowel and the final consonant(212).

Description

Online Hangul Handwritten Character Recognition Method

The present invention relates to a method for recognizing handwritten characters, and more particularly, to a method for recognizing on-line Hangul handwritten characters.

In recent years, the use of personal information terminal devices such as PDA (Personal Digital Assistant or Personal Digital Accessory), HPC (Hand-held Personal Computer), and the like are increasing rapidly. These personal digital assistants are very small compared to existing desktop PCs and laptop PCs. Accordingly, an input method through handwriting, which is an input method suitable for a small PC, has emerged as an important method.

As described above, in order to enable free handwriting input in the input method through handwriting, many methods for connected online Hangul handwritten character recognition have been studied. However, connected Hangul recognition requires more memory area and recognition time than separate Hangul recognition.

As described above, in the on-line Korean handwritten character recognition, the connection type requires more memory area than the separate type and has a problem that it is difficult to apply as the recognition time is required.

Accordingly, an object of the present invention is to provide a method for recognizing on-line Hangul handwritten characters connected in a short time while using less memory.

1 is a block diagram of an apparatus for recognizing a Korean handwritten character according to an embodiment of the present invention;

2 is a flowchart illustrating a Hangul handwritten character recognition process according to an embodiment of the present invention;

3 is a processing flowchart of the preprocessing process of FIG. 2;

4 is a flowchart illustrating a process of determining vertical and horizontal segments in FIG.

5 is a flowchart illustrating an example of determining a candidate region for each neutral type of FIG. 2;

6A to 6D illustrate a form according to a process of inputting a Hangul character ″ station ″ according to an embodiment of the present invention. FIG. 6A illustrates an input form, and FIG. 6B illustrates a preprocessing form. FIG. 6C illustrates a form in which regions are divided and horizontal and vertical segments are determined, and FIG. 6D illustrates a form in which neutral ″ TT ″ regions are determined.

The present invention for achieving the above object is a process of inputting and preprocessing the Hangul characters according to the written trajectory and outputting a segment set, and tracking the preprocessed segment set in writing order, while vertically vertical segment set and horizontal The process of determining and determining the segment set, and the process of determining the candidate area for each neutral type based on the position of the start point and the end point and the direction of the end segment by using the characteristics of each neutral of the Hangul characters, and for each neutral type. Hangul characters by the process of recognizing neutrality for each candidate area, recognizing initial and finality for the initial and final areas based on the recognized neutral area, and combining recognized initial, neutral, and finality Characterized in that it comprises a process of recognizing.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and the annexed drawings, numerous specific details are set forth in order to provide a more thorough understanding of the present invention, such as examples of processing for specific character recognition or processing flow. It will be apparent to those skilled in the art that the present invention may be practiced without these specific details. And a detailed description of known functions and configurations that may unnecessarily obscure the subject matter of the present invention will be omitted.

First, in order to solve the problem in the connected type of Korean Hangul handwriting character recognition as described above, the present invention uses the characteristics of the Hangul to recognize the neutral region in the character by separating the online connected Hangul handwriting characters in a short time using less memory. It suggests ways to be recognized. Hangeul is composed of letters consisting of initial and neutral characters, and letters consisting of initial and neutral characters. Therefore, if the neutral region is properly distinguished in the character, the character recognition may be possible only by the respective phoneme recognition and combination. In particular, since neutrality is fixed in the direction of handwriting or the position in the letter, unlike the initiality and the finality, the neutrality can be efficiently separated using this information. With this in mind, the present invention divides each neutral boundary around its start and end positions, and finds a segment combination within the boundary to achieve the neutrality. Accordingly, it is possible to distinguish a simple but efficient neutral region. The phoneme is recognized using the divided neutral regions. First, neutrality is recognized for the neutral region, and then initiality and species are recognized based on the recognized neutrality. In this way, by using the connected on-line Hangul handwritten character recognition using the neutral separation, as well as using less memory than the conventional it is possible to shorten the time required for recognition.

1 is a block diagram of an apparatus for recognizing connected online Hangul handwritten characters according to an embodiment of the present invention. The hardware is an input device 100 and a microprocessor like a conventional character recognition system. 102, a memory 104 and an output device (106). The input device 100 includes an electronic pen 108, and generates a stream of x, y coordinate points by a two-dimensional coordinate system for points along a trajectory that a user writes using the electronic pen 108. . As such an input device 100, a tablet, a touch-sensitive screen panel (TSP), or the like is usually used. In the present invention, the characters input through the input device 100 are targeted for the connected online Hangul handwritten characters. The microprocessor 102 receives the output of the input device 100 and recognizes the connected online Hangul handwritten characters by the character recognition process according to an embodiment of the present invention as described below and outputs the recognition result to the output device 106. do. The memory 104 includes a read only memory (ROM) and a random access memory (RAM). The ROM stores a character recognition program of the microprocessor 102 and various reference data for character recognition processing according to an embodiment of the present invention. The RAM temporarily stores data according to the character recognition process of the microprocessor 102. The output device 106 may be, for example, a liquid crystal display device, and outputs a character recognition result by the microprocessor 102 so that a user can identify it.

2 is a flowchart illustrating a Hangul handwritten character recognition process according to an exemplary embodiment of the present invention. The process performed by the microprocessor 102 of FIG. 1 is shown in steps 200 through 212. First, when the user writes Hangul characters using the electronic pen 108 of the input device 100, the microprocessor 102 inputs the Hangul characters from the input device 100 according to the written trajectory in step 200. . At this time, the input device 100 generates a series of x, y coordinate points by a two-dimensional coordinate system and applies them to the microprocessor 102 for the points along the trajectory that the user writes using the electronic pen 102. Reference numeral 102 samples and outputs an output of the input apparatus 100. 6A illustrates an example of a character that is handwritten and input to the input apparatus 100. FIG. 6A shows an input form when, for example, a ″ national ″ character is input.

Next, the microprocessor 102 preprocesses the input Hangul character in step 202. This preprocessing process is basically the same as the preprocessing process performed essentially before general character recognition.

3 shows a process flow diagram of the preprocessing process performed in step 202 described above in steps 300 to 302. First, the microprocessor 102 normalizes the horizontal size × vertical size, for example, 72 × 72, with respect to the input character in step 300, and in step 302, the distance between the two points is greater than or equal to a predetermined interval. For example, resampling to 9 or more. A section consisting of two points is called a `` segment ''. Therefore, this preprocessing results in a segment set consisting of two points as shown in FIG. 6B. In FIG. 6B, reference numeral 600 denotes one of the resampled points, and 602 denotes one of the segments.

After performing the preprocessing process as described above in step 202, the microprocessor 100 performs the vertical and horizontal segment determination process in step 204. The condition is satisfied while tracking the preprocessed segment set in writing order. The segments are classified into one of the vertical segment and the horizontal segment, respectively. That is, as described later, vertical and horizontal segments to be used for determining neutral regions are determined.

4 is a flowchart illustrating the vertical and horizontal segment determination process performed in step 204 as shown in steps 400 and 408, and is repeated for each segment as an example of processing for one segment. Is processed. First, the microprocessor 102 inputs a segment in operation 400 and compares a change in x-coordinate value and a change in y-coordinate value between two points constituting the segment in operation 402 to 404. At this time, if the change in the x-coordinate value is larger than the change in the y-coordinate value by a predetermined value or more, in step 408, the corresponding segment is determined and registered as the horizontal segment. On the contrary, if the change in the y-coordinate value is larger than the change in the x-coordinate value by more than a predetermined value, the corresponding segment is determined as a vertical segment and registered in step 406. If there is no difference more than the predetermined value between the change of the x coordinate value and the change of the y coordinate value, the segment is ignored and ends. At this time, the constant value is set to 2, for example. As a result, the segment in the direction in which the stroke travels ″ down ″ is determined as the vertical segment, and the segment in the direction in which the stroke travels ″ right ″ is determined as the horizontal segment. By the vertical and horizontal segment determination process, a vertical and horizontal segment set is obtained as shown in FIG. 6C. In FIG. 6C, segments denoted by V represent vertical segments and segments denoted by H denote horizontal segments. Therefore, in the following description, the vertical segment is referred to as ″ V ″ and the horizontal segment is referred to as ″ H ″ for convenience.

Thereafter, the microprocessor 100 finds a neutral region for the H and V sets for each neutral type defined in operation 206 and determines it as a candidate region for each neutral type. The neutral type is based on the position of the start point and the end point and the direction of the start and end segments belonging to the unit areas in which one character area is equally divided in the x and y axis directions of the two-dimensional coordinate system for each neutral character of the Hangul character. Separate. This will be described in more detail.

Hangul is composed of initial, neutral, or first, neutral, and final, so it must include neutral, where the beginning and end of neutral has a fixed position within the character. In consideration of this, as shown in FIG. 6C, an area of a character is divided into three portions in the x-axis direction and the y-axis direction, respectively, and divided into left, center, and right in the x-axis direction, and up, center, and down in the y-axis. The whole area is divided into 9 equal parts. In addition, each neutral segment has a consistent writing direction at the beginning and end segments. Accordingly, in the present invention, the writing direction is divided into ″ H ″ and the lower side, ″ V ″. In addition, according to each position, as shown in Table 1, it is divided into ″ H1 ″, ″ H2 ″ and ″ V1 ″, ″ V2 ″. Thus, each neutral is divided into nine neutral types as shown in Table 1 on the basis of the position of the start point, the direction of the start segment, the position of the end point, and the direction of the segment.

Beginning End Bottom right, down (VI) Center down, down (V2) Right center, right (H1) Right area, right (H2) Right upper and lower (V1) ㅣ, ㅐ, ㅒ ㅏ, ㅑ Center, Down (V2) ㅚ, ㅙ ㅘ ㅗ, ㅛ Center, right (H1) ㅓ, ㅕ, ㅔ, ㅖ Left, right (H2) ㅢ, ㅞ, ㅟ, ㅝ ㅜ, ㅠ ㅡ

In Table 1, the starting part indicates the position of the starting point and the direction of the starting segment. For example, ″ upper right, down ″ to ″ upper right ″ indicate that the position of the starting point is ″ upper right ″ among the nine areas shown in FIG. 6c, and ″ down ″ indicates the direction of the starting segment starting therefrom. . In case of ″ upper right and down ″ at the beginning, it is classified as ″ V1 ″, and in case of ″ lower right and down ″ at the end, it is classified as ″ V1 ″. The same holds true for the remaining cases. Similarly, the end points indicate the position of the end point and the direction of the end segment.

Therefore, the neutral types having neutral phonemes that are actually classified except the blank in Table 1 are ″ V1V1 ″, ″ V1H1 ″, ″ V2V1 ″, ″ V2H1 ″, ″ V2H2 ″, ″ H1V1 ″, ″ H2V1 ″, and ″ H2V2. 9 kinds of ″, ″ H2H2 ″. Here, for example, the form ″ V1V1 ″ refers to the first ″ V1 ″ at the beginning and the second ″ H1 ″ to the end.

Referring to Table 1 above, ″ V1V1 ″ has the start point position ″ upper right ″, the start segment direction ″ down ″, the end point position ″ lower right ″ and the end segment direction ″ down ″ The type corresponds to ″ ㅣ, ㅐ, ㅒ ″. ″ V1H1 ″ is the type where the starting point is ″ upper right ″, the starting segment is ″ down ″, the ending point is ″ right center ″, and the ending segment is ″ right ″. ″ Is applicable. ″ V2V1 ″ is a type where the start point is ″ center ″, the start segment is ″ down ″, the end point is ″ lower right ″, and the end segment is ″ down ″. Yes. ″ V2H1 ″ is the type where the start point is ″ center ″, the start segment is ″ down ″, the end point is ″ right center ″, and the end segment is ″ right ″. do. ″ V2H2 ″ is the type where the start point is ″ center ″, the start segment is ″ down ″, the end point is ″ all right area ″, and the end segment is ″ right ″. ″ Is applicable. ″ H1V1 ″ is the type where the start point is ″ center ″, the start segment is ″ right ″, the end point is ″ bottom right ″, and the end segment is ″ down ″. ㅔ, ㅖ ″ is applicable. ″ H2V1 ″ is the type where the start point is ″ left ″, the start segment is ″ right ″, the end point is ″ bottom right ″, and the end segment is ″ down ″. ㅟ, ㅝ ″ is applicable. ″ H2V2 ″ is the type where the start point is ″ left ″, the start segment is ″ right ″, the end point is ″ down the center ″, and the end segment is ″ down ″. Is applicable. ″ H2H2 ″ is the type where the starting point is ″ left ″, the starting segment is ″ right ″, the ending point is ″ all right area ″, and the ending segment is ″ right ″, where ″ ㅡ ″ is Yes.

The neutrality of Hangul can represent each characteristic only by the combination of ″ H ″, ″ V ″ shown in Table 1 above. In other words, unlike initial and final, neutrality is weak in the degree of deformation and consistent in the direction of stroke. Therefore, each neutral may be represented by a combination of ″ H ″ and ″ V ″ as shown in Table 2 below. In Table 2 below, ″ S ″ represents any segment.

neutrality H, V combination ㅣ V1V * V1 ㅡ H2H * H2 ㅏ, ㅑ V1V * H * H1 ㅗ, ㅛ V2V * H * H2 ㅐ, ㅒ V1V * H + V * V1 ㅓ, ㅕ H1H * V * V1 ㅔ, ㅖ H1H * V + V * V1 ㅜ, ㅠ H2H * V * V2 ㅚ V2V * H + V * V1 ㅙ V2V * H + V + V * V1 ㅘ V2V * H + V + H * H1 ㅢ, ㅟ, ㅝ, ㅞ H2H * SV * V1

After determining the total area by starting and ending positions, starting segments, and ending segments of each neutral, a corresponding neutral region can be found in a character by finding a combination of the specified segments as shown in Table 2 above. In this way, when the neutral region is determined in the character, the previous region becomes the initial, and the subsequent region becomes the final based on the neutral, so that the segmentation of the phoneme is automatically performed. When the regions between the phonemes are divided as described above, after each phoneme is recognized, a code of the recognized character can be obtained using the combination.

FIG. 5 is a flowchart showing an example of determining a candidate region for each neutral type performed in step 206. For example, ″ ㅘ ″ in which neutrality is included in ″ V2H1 ″ among the neutral types shown in Table 1 above. In the case of (500) to (522) is shown. ″ ㅘ ″ has a segment combination of ″ V2V * H + V + H * H1 ″ as shown in Table 2 above. First, the microprocessor 100 inputs a set of V and H in operation 500 to select an intermediate section for each neutral type in operation 502. Next, in step 504, the segment is tracked to find a start segment defined for each neutral type. At this time, since the neutral type currently being processed is ″ V2H1 ″, check whether the starting point of the segment is ″ V2 ″. If the start point of the segment is not ″ V2 ″, the process of determining the ″ V2H1 ″ candidate area is terminated.

If a start segment defined for each neutral type, that is, ″ V2 ″ is found in step 504, in step 506 to 518, a defined combination consisting of V and H segments for achieving a neutral of the type is found. At this time, in step 506, the process proceeds until the next segment is not ″ V ″, and when ″ Va ″ ends, the next ″ Ha ″ is found. Where ″ Va ″ means the last ″ V ″ in a series of ″ V ″ s, and ″ Ha ″ means the first ″ H ″ in a series of ″ H ″ s. In step 508, it is checked whether the positional relationship between the end point of ″ Va ″ and the start point of ″ Ha ″ satisfies the condition of the segment combination ″ V2V * H + V + H * H1 ″. If the condition is not satisfied, the ″ V2H1 ″ candidate area determination process is terminated.

If the condition is satisfied in step 508, the process proceeds until the next segment is not ″ H ″ in step 510. When ″ Hb ″ ends, the next ″ Vb ″ is found. Where ″ Vb ″ means the first ″ V ″ in the second consecutive ″ V ″ set, and ″ Hb ″ means the last ″ H ″ in the first consecutive ″ H ″ set. In step 512, it is checked whether the positional relationship between the end point of ″ Hb ″ and the start point of ″ Vb ″ satisfies the condition of the segment combination ″ V2V * H + V + H * H1 ″. If the condition is not satisfied, the ″ V2H1 ″ candidate area determination process is terminated.

If the condition is satisfied in step 512, the process proceeds until the next segment is not ″ V ″ in step 514. When ″ Vc ″ ends, the next ″ Hc ″ is searched for. Where ″ Vc ″ means the last ″ V ″ in the second consecutive ″ V ″ set, and ″ Hc ″ means the first ″ H ″ in the second consecutive ″ H ″ set. In step 516, it is checked whether the positional relationship between the end point of ″ Hc ″ and the start point of ″ Vc ″ satisfies the condition of the segment combination ″ V2V * H + V + H * H1 ″. If the condition is not satisfied, the ″ V2H1 ″ candidate area determination process is terminated.

If the condition is satisfied in step 516, the process proceeds until the next segment is not ″ H ″ in step 518, and then checks whether the last ″ H ″ is ″ H1 ″ in step 520. In other words, if a segment of the defined combination is found, the end segment is checked. At this time, if the end segment is ″ H1 ″ of the corresponding type, in step 522, the area from the start segment to the end segment is determined as a neutral candidate area of the ″ V2H1 ″ of the type and registered. For example, in the case of the ″ station ″ as shown in Figs. 6A to 6C, the neutral ″ TT ″ area is determined as shown in Fig. 6D. The segments between the points marked with ″ _. ″ In FIG. 6D are neutral regions.

As described above, the process of determining the candidate region for each neutral type is processed corresponding to each neutral type to determine the candidate region for each neutral type. That is, for each of the nine neutral types shown in Table 1, the processing as shown in FIG. 5 is performed corresponding to the segment combination of the corresponding neutral type.

As described above, after obtaining the neutral region for each type, the microprocessor 100 attempts the neutral recognition on the obtained neutral region in step 208, which is the same as a normal phoneme recognition process. The phoneme recognition selects a phoneme model closest to the input phoneme as a candidate through comparison with a phoneme model registered in advance using elastic matching. At this time, only models having a defined phoneme type among the candidate models are selected as candidates. For example, up to 30 candidate regions are selected in order of recognition scores.

Next, the microprocessor 100 recognizes the initial and finality of the initial and final regions based on the neutral region recognized as described above at step 210. In this case, the initial and final recognition is also the same as the conventional phonemic recognition. And, for example, each has a maximum of three candidates.

When all of the initial, neutral, and final characters are recognized by the above process, the Hangul character is recognized by the combination of the initial, neutral, and final characters in step 212. The character recognition is also the same as the conventional character recognition. That is, after the recognition of the phoneme is completed, in the case of the combination where the code of the completed type is generated after the combination, a size penalty is applied to each phoneme, and then a recognition score for the character is calculated. Finally, the character code generated by the phoneme combination having the highest recognition score is output as a recognition result and notified through the output device 106.

The present invention described above shows a particularly high recognition rate in the case of the connected type even during online Hangul handwritten character recognition. In order to recognize the present invention, the recognition test was performed in comparison with the existing method of registering and recognizing the phoneme connected with the initial, neutral, and final forms. As a result, the error rate was reduced by more than 50% and fast recognition result was obtained by recognizing 2000 characters per second on a 90MHz Pentium PC.

Therefore, after determining the start and end positions of the neutral in the Hangul characters in advance, the neutral boundary is extracted around the region, and if the regions of the initial and final species are automatically determined based on the extracted neutral region, By using the phoneme recognition, it is possible to use fast on-line handwritten Hangul character recognition while using less memory.

Meanwhile, in the above description of the present invention, specific embodiments have been described, but various modifications can be made without departing from the scope of the present invention. Particularly, in the embodiment of the present invention, an example of dividing the character area into nine is given, but may be divided and used as necessary. In addition, the normalization size and resampling interval in the preprocessing may vary. Therefore, the scope of the invention should not be defined by the described embodiments, but should be defined by the equivalent of claims and claims.

As described above, the present invention can recognize the connected online Hangul handwritten characters in a short time while using less memory, and thus can be efficiently used in personal information terminal devices such as PDAs and HPCs.

Claims (10)

In the method of recognizing connected online Hangul cursive characters, Inputting and preprocessing the Hangul characters according to a written trajectory; Determining a vertical segment set and a horizontal segment set from the segments while tracking the preprocessed segment set in writing order; Determining the candidate region for each neutral type based on the position of the start point and the end point and the direction of the end segment using the characteristics of the neutrals of the Hangul characters; Recognizing neutrality for each candidate region for each neutral type; Recognizing the initial and finality of the initial and final regions based on the recognized neutral regions; On-line Hangul handwritten character recognition method characterized in that it comprises the step of recognizing the input Hangul characters by the combination of the recognized initial, neutral, and final. The on-line Korean character according to claim 1, wherein the vertical segment is a segment in a direction in which the stroke travels ″ down ″, and the horizontal segment is a segment in a direction in which the stroke progresses ″ right ″. Handwritten character recognition method. The method of claim 2, wherein the vertical and horizontal segment determination process, Comparing the change of the x coordinate value and the change of the y coordinate value between two points forming the segment; If the change in the x coordinate value is greater than a predetermined value than the change in the y coordinate value, determining and registering the corresponding segment as a horizontal segment; Determining and registering the corresponding segment as a vertical segment when the y coordinate value change is greater than the predetermined value than the x coordinate value change, and And a step of ignoring the corresponding segment if there is no difference over the predetermined value between the change of the x coordinate and the change of the y coordinate. The method of claim 2, wherein the process of determining the candidate region for each neutral type comprises: for each neutral type; Selecting an intermediate section for each neutral type by inputting the segment set; Finding a starting segment defined for each neutral type while tracking the segment; Finding a starting segment defined for each neutral type and finding a defined combination consisting of the vertical and horizontal segments for achieving a neutral of the type; If the segment of the combination defined above is found, checking the end segment; And if the end segment is a corresponding type, determining a region from the start segment to an end segment as a candidate region for each type. In the method of recognizing connected online Hangul cursive characters, Inputting and preprocessing the Hangul characters according to a written trajectory; Determining a vertical segment set and a horizontal segment set from the segments while tracking the preprocessed segment set in writing order; Neutral type that divides one character area for each neutral of the Hangul characters based on the position of start and end points belonging to the unit areas divided in x, y axis direction of the 2D coordinate system and the direction of start and end segments Determining neutral regions for the vertical and horizontal segment sets for each candidate and determining candidate regions for each neutral type; Recognizing neutrality for each candidate region for each neutral type; Recognizing the initial and finality of the initial and final regions based on the recognized neutral regions; On-line Hangul handwritten character recognition method characterized in that it comprises the step of recognizing the input Hangul characters by the combination of the recognized initial, neutral, and final. The method of claim 5, wherein the vertical segment is a segment in a direction in which a stroke travels ″ down ″, and the horizontal segment is a segment in a direction in which a stroke progresses ″ right ″. Handwritten character recognition method. The method of claim 6, wherein the vertical and horizontal segment determination process, Comparing the change of the x coordinate value and the change of the y coordinate value between two points forming the segment; If the change in the x coordinate value is greater than a predetermined value than the change in the y coordinate value, determining and registering the corresponding segment as a horizontal segment; Determining and registering the corresponding segment as a vertical segment when the y coordinate value change is greater than the predetermined value than the x coordinate value change, and And a step of ignoring the corresponding segment if there is no difference over the predetermined value between the change of the x coordinate and the change of the y coordinate. The method of claim 5 or 7, wherein the unit areas of the candidate type determination process for each neutral type are divided into ″ left ″, ″ center ″, and ″ right ″ in the x-axis direction of the two-dimensional coordinate system. And a 9-divided area divided into ″ up ″, ″ center ″, and ″ down ″ in the y-axis direction. The method of claim 8, wherein the neutral types of the candidate region determination process for each neutral type include: The start point is ″ upright ″, the start segment is ″ down ″, the end point is ″ bottom right ″, and the end segment is ″ down ″, The start point is ″ upper right ″, the start segment is ″ down ″, the end point is ″ right center ″ and the end segment is ″ right ″, The start point is ″ center ″ and the start segment is ″ down ″, the end point is ″ bottom right ″ and the end segment is ″ down ″, The start point is ″ center ″ and the start segment is ″ down ″, the end point is ″ right center ″ and the end segment is ″ right ″, The start point is ″ center ″, the start segment is ″ down ″, the end point is ″ all right area ″ and the end segment is ″ right ″, The start point is ″ center ″ and the start segment is ″ right ″, the end point is ″ bottom right ″ and the end segment is ″ down ″, The start point is ″ left ″ and the start segment is ″ right ″, the end point is ″ bottom right ″ and the end segment is ″ down ″, The start point is ″ left ″ and the start segment is ″ right ″, the end point is ″ bottom center ″ and the end segment is ″ down ″, An online Korean cursive, characterized in that the starting point is ″ left ″ and the starting segment is ″ right ″, the ending point is ″ all right area ″ and the ending segment is ″ right ″. Character recognition method. The method of claim 9, wherein the process of determining candidate regions for each neutral type comprises, for each neutral type, Selecting an intermediate section for each neutral type by inputting the segment set; Finding a starting segment defined for each neutral type while tracking the segment; Finding a starting segment defined for each neutral type and finding a defined combination consisting of the vertical and horizontal segments for achieving a neutral of the type; If the segment of the combination defined above is found, checking the end segment; And if the end segment is a corresponding type, determining a region from the start segment to an end segment as a candidate region for each type.

Images