An electronic writing practice method based on convolutional neural network

Abstract

The present invention discloses an electronic writing practice method based on a convolutional neural network, which belongs to the field of calligraphy. The electronic writing practice method based on a convolutional neural network is to learn the real writing effect based on the pen movement data of the result obtained by traditional electronic writing, establish the connection between the real pen and the capacitive pen, and apply the pen movement data to the traditional electronic writing method, so that the touch device can reflect more authentic writing content, thereby solving the problem that the writing of the capacitive pen cannot provide the effect brought by the writing of the real pen. The method can help the practitioner better understand his or her own writing practice level, master the pen movement rules, improve the writing method, and thus improve the calligraphy level.

Description

Technical Field

[0001] The present invention relates to the field of calligraphy, and in particular to an electronic writing practice method based on convolutional neural network. Background Art

[0002] Calligraphy culture, calligraphy is a treasure of Chinese classical art. Among the written languages ​​of all countries in the world, no other written language has developed into a unique art form like the writing of Chinese characters, and it has a long history. China's 5,000 years of brilliant civilization and incomparable rich written records have been recognized by the world.

[0003] Practicing calligraphy is undoubtedly a very important part of learning Chinese culture. Each writing technique in hard pen calligraphy, such as hiding the tip of the pen, revealing the tip of the pen, reversing the tip of the pen, centering the tip of the pen, turning the tip of the pen, returning the tip of the pen, lifting, pressing, pausing, stopping, blunting and turning, will have different effects on the writing effect.

[0004] Due to the influence of the environment and venue, brush calligraphy cannot be practiced anytime and anywhere. It is necessary to prepare tools such as brushes, ink, rice paper, inkstone and brush washer. The same is true for hard pen calligraphy. Carrying practice tools with you also brings great inconvenience to daily training. The time and venue for training are limited, which causes certain troubles and financial pressure for calligraphy enthusiasts to improve their skills.

[0005] For this problem, the existing technology can already solve the problems of venues and hardware equipment for writing practice. The use of touch devices and capacitive pens can meet the needs of practitioners to practice anytime and anywhere. However, the existing technology uses the tip of the capacitive pen to contact the touch device to display the coloring at the corresponding position through electromagnetic reaction. This method lacks authenticity for real calligraphy practice. For example, in hard pen calligraphy, the stroke "vertical" can be further subdivided into hanging needle vertical or falling vertical. The two writing methods are obviously different under the effects of different pen-handling methods such as writing strength and writing speed of the practitioner. However, traditional calligraphy practice tools can only display it as "vertical" in a specified style. Therefore, the existing electronic calligraphy practice method lacks the connection with authenticity. It is not easy to feedback the real effect of the writing content by using the traditional electronic calligraphy practice method for writing practice, and thus it is impossible to provide better practice suggestions for the practitioner. Summary of the invention

[0006] In order to solve the problems existing in the above-mentioned traditional electronic writing devices, the present invention proposes an electronic writing practice method based on a convolutional neural network, which studies the real writing effect based on the pen movement data of the results obtained from traditional electronic writing, establishes a connection between the real pen and the capacitive pen, and applies the pen movement data to the traditional electronic writing method, so that the touch device can reflect more realistic writing content.

[0007] In view of the above situation, the present invention proposes an electronic writing practice method based on a convolutional neural network, comprising the following steps:

[0008] Step 1: Collect the first sample written in the usual way,

[0009] Step 2: Obtain a second sample in writing mode using a touch device.

[0010] Step 3: Obtain the pen movement data in the writing mode of the touch device described in step 2;

[0011] Step 4: Use the convolutional neural network to train the samples and data from step 1 to step 3 to obtain the corresponding relationship. The specific training method is:

[0012] Step 401: preprocessing the first sample and the second sample of the same content, the preprocessing includes: a. storing the first sample and the second sample according to the same image size, selecting the same significant feature point of the two samples and marking it as the origin, and then obtaining the coordinates of each feature point in the first sample and the second sample respectively; b. obtaining the trajectory coordinates of the writing paths of the first sample and the second sample to determine the direction of the pen tip; the motion trajectory of the first sample is a trajectory formed by the midpoint coordinates of the coordinates of the corresponding points on both sides of the same stroke; for example, for a certain horizontal stroke, the coordinates of the two points corresponding to any point in the trajectory of the first sample are (x1, y1) and (x1, y2), respectively, then the midpoint coordinates are (x1, (y1+y2) / 2), the motion trajectory of the first sample is the curve formed by connecting all the midpoint coordinate values, and the motion trajectory of the pen tip of the second sample is the motion trajectory of the pen tip; c. obtaining the deformation difference of the first sample and the second sample at the same point of the pen trajectory; the deformation difference of the first sample and the second sample at the same point of the pen trajectory is that the center-point pen stroke is only affected by the pen data;

[0013] Step 402: training the deformation difference in the pen direction by pen tip pressure and writing speed;

[0014] Step 403: Since the second sample at a different position of the pen trajectory can reflect the influence of the reverse stroke starting and the reverse stroke ending, the first sample and the second sample at the different trajectory coordinates in step 401 are analyzed, the first sample is overlaid on the second sample, the path traversal number at the different trajectory coordinates is obtained, and the color depth of the path traversal number is superimposed on the second sample;

[0015] Take the color depth of any two different track coordinates with the same pen speed (defined as v0) and calculate the average value, and define the average value as the basic color depth d0. Continue to obtain the pen speed v1 at the remaining track coordinates. At this time, the color depth d at any track of the entire sample is defined as:

[0016] d=α*d0*(v0 / v1)

[0017] Where α represents the color depth relationship coefficient;

[0018] Step 5: Generate a writing practice mapping according to the correspondence described in step 4.

[0019] Preferably, the conventional writing method described in step 1 includes but is not limited to hard pen writing and brush writing, the first sample in step 1 is a picture sample under the conventional writing method, the second sample in step 2 is a picture sample under the writing method of a touch device, and the pen movement data in step 3 includes but is not limited to the writing speed, writing time, pen tip pressure and writing path obtained and fed back by the touch device. For each user, the difference in handwriting depends on the difference in writing habits, and the writing habits are determined by the pen movement data of each user. Referring to the handwriting identification method, the first sample and the second sample involved in the present invention are picture samples written by the same user in different forms, so the first sample and the second sample written by the same user are consistent.

[0020] Preferably, the corresponding relationship in step 3 is the difference relationship between the second sample in step 2 generated under the pen movement data in step 3 and the first sample in step 1, and the mapping in step 4 is the mapping of the second sample in step 2 to the first sample in step 1 generated by the method generated in step 3. When the difference relationship between the first sample and the second sample is calculated by the above method and the convolutional neural network is trained with the pen movement data, the first sample corresponding to any second sample under different pen movement data conditions can be obtained.

[0021] Compared with the prior art, the present invention has the following beneficial effects: the results obtained by the traditional electronic writing method are trained through the pen movement data, and the corresponding relationship between the results obtained by the traditional electronic writing method and the results obtained by real writing is established through learning, and after the mapping is formed, the feedback is fed back to the touch device to optimize the results obtained by the traditional electronic writing method, thereby establishing a connection between electronic writing and real writing, helping practitioners to learn calligraphy more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Attached Figure 1 Flow chart of the writing practice method proposed by the present invention

[0023] Attached Figure 2 Flow chart of the training method proposed by the present invention DETAILED DESCRIPTION

[0024] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0025] like Figure 1-2The electronic writing practice method based on a convolutional neural network proposed by the present invention comprises the following steps:

[0026] Step 1: Collect the first sample written by brush writing.

[0027] Step 2: Get the second sample of the capacitive pen writing on the capacitive screen.

[0028] Step 3: Obtain the writing speed, writing time, pen tip pressure and writing path of the second sample in step 2; Step 4: Use a convolutional neural network to train the samples and data of steps 1 to 3 to obtain the corresponding relationship, specifically:

[0029] Step 401: preprocessing the first sample and the second sample of the same content, the preprocessing includes: a. storing the first sample and the second sample according to the same image size, selecting the same significant feature point of the two samples and marking it as the origin, and then obtaining the coordinates of each feature point in the first sample and the second sample respectively; b. obtaining the trajectory coordinates of the writing paths of the first sample and the second sample to determine the direction of the pen tip; the motion trajectory of the first sample is a trajectory formed by the midpoint coordinates of the coordinates of the corresponding points on both sides of the same stroke; for example, for a certain horizontal stroke, the coordinates of the two points corresponding to any point in the trajectory of the first sample are (x1, y1) and (x1, y2), respectively, then the midpoint coordinates are (x1, (y1+y2) / 2), the motion trajectory of the first sample is the curve formed by connecting all the midpoint coordinate values, and the motion trajectory of the pen tip of the second sample is the motion trajectory of the pen tip; c. obtaining the deformation difference of the first sample and the second sample at the same point of the pen trajectory, the deformation difference of the first sample and the second sample at the same point of the pen trajectory, that is, the center-point pen stroke is only affected by the pen data;

[0030] Step 402: training the deformation difference in the pen direction by pen tip pressure and writing speed;

[0031] Step 403: Since the second sample at a different location of the pen trajectory can reflect the influence of starting the pen with the tip of the brush in reverse and ending the pen with the tip of the brush in reverse, analyze the first sample and the second sample at a location where the trajectory coordinates are different in step 401, overwrite the second sample with the first sample, obtain the path traversal number at a location where the trajectory coordinates are different, and perform color depth superposition of the path traversal number on the second sample.

[0032] Step 5: Generate a writing practice mapping based on the correspondence described in step 4.

[0033] When the difference relationship between the first sample and the second sample is calculated by the above method and the convolutional neural network is trained through the pen movement data, the first sample corresponding to any second sample under different pen movement data conditions can be obtained.

[0034] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An electronic writing practice method based on convolutional neural network, characterized in that: The following steps are involved: Step 1: Collect the first sample written in a conventional manner; Step 2: Obtain a second sample in a writing mode using a touch device; Step 3: Obtain the pen movement data in the writing mode of the touch device described in step 2; Step 4: Use a neural network to train the samples and data of steps 1 to 3 to obtain corresponding relationships. The specific steps of the training are: Step 401: Preprocessing a first sample and a second sample of the same content, wherein the preprocessing includes: a. Store the first sample and the second sample in the same image size. b. Obtaining the trajectory coordinates of the writing paths of the first sample and the second sample to determine the direction of the pen tip, c. Obtaining the deformation difference between the first sample and the second sample at the same position of the pen trajectory; Step 402: training the deformation difference in the pen direction by pen tip pressure and writing speed; Step 403: Analyze the first sample and the second sample at the different positions of the trajectory coordinates in step 401, cover the second sample with the first sample, obtain the path traversal number at the different positions of the trajectory coordinates, and perform color depth superposition of the path traversal number on the second sample; Step 5: Generate a writing practice mapping according to the correspondence described in step 4.

2. The electronic writing practice method based on convolutional neural network according to claim 1, characterized in that: The conventional writing method described in step 1 includes but is not limited to hard pen writing and brush writing. The first sample in step 1 is a picture sample under the conventional writing method. The second sample in step 2 is a picture sample under the writing method of a touch device. The pen movement data in step 3 includes but is not limited to the writing speed, writing time, pen tip pressure and writing path obtained and fed back by the touch device.

3. The electronic writing practice method based on convolutional neural network according to claim 1, characterized in that: The corresponding relationship in step 3 is the difference relationship between the second sample in step 2 generated under the pen data in step 3 and the first sample in step 1, and the mapping in step 4 is the mapping of the second sample in step 2 to generate the first sample in step 1 under the method generated in step 3.

4. The electronic writing practice method based on convolutional neural network according to claim 1, characterized in that: The first sample of step 1 and the second sample of step 2 are synchronously input through the same clamping component.

Images