Handwriting display method and system and storage medium

Abstract

The invention discloses a handwriting display method and system and a storage medium, and relates to the field of data processing, and the method comprises the steps: obtaining a physical change feature and a style feature of a target handwriting from handwriting information through employing a preset style feature extraction model, inputting the handwriting information, the physical change feature and the style feature into a preset generative adversarial network model, and obtaining texture features of the target handwriting, and sending the features and information to a display terminal, so that the display terminal performs rendering operation on the handwriting trajectory, and displays the restored target handwriting. According to the method, the physical change feature of the target handwriting and the style feature representing the writing style are extracted through configuration, and the texture feature of the handwriting generated by the corresponding writing tool under the corresponding writing style is extracted. And finally, the handwriting track is rendered based on the physical change feature, the style feature and the texture feature through the configuration display terminal, so that the restored target handwriting carries the style feature and the texture feature, and the handwriting restoration precision is improved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular, to a handwriting display method, system, and storage medium. Background Art

[0002] With the rise of webcasting, performers will interact with the audience by writing blessings, thereby increasing user stickiness. The existing display method for written content is to collect the handwriting of the written text based on a handwriting board or a camera, and based on the recognition and comparison of the collected handwriting, the text corresponding to the collected handwriting in the database is displayed.

[0003] However, there are obvious differences in the writing styles of different performers. For example, when different performers use different writing tools (such as pens and writing brushes) to write the same content in the "cursive" font, there are significant differences in writing styles such as connected strokes, pauses, and stroke angles among different performers, resulting in obvious writing style differences in the text written by different performers. For the text pre-stored in the database, its writing style is generated based on the same standard and there are no significant differences. This makes the text written by different performers have a unified writing style when being displayed, reducing the reduction accuracy of the written text of the performers during display. In addition, there are also problems of low reduction accuracy of the written text in scenarios such as online calligraphy teaching and electronic signatures. Summary of the Invention

[0004] In view of the above problems, the present application provides a handwriting display method, system, and storage medium to achieve the purpose of improving the reduction accuracy of written text. The specific solutions are as follows:

[0005] The first aspect of the present application provides a handwriting display method, including:

[0006] Obtaining the handwriting information of the target handwriting sent by the handwriting acquisition device;

[0007] Inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting;

[0008] Inputting the handwriting information, the physical change features, and the style features into a preset generative adversarial network model to obtain the texture features of the target handwriting;

[0009] Sending the physical change features, the style features, the texture features, and the handwriting information to a display terminal, so that the display terminal performs a rendering operation on the handwriting trajectory in the handwriting information based on the physical change features, the style features, and the texture features, restores the target handwriting, and displays the target handwriting.

[0010] In a possible implementation, the preset style feature extraction model includes a preset convolutional neural network model and a preset recurrent neural network model, and the training process of the preset style feature extraction model includes:

[0011] Obtaining handwriting information of a plurality of real handwritings, and adding a label to the handwriting information of each of the real handwritings, wherein the label content is a description parameter of the style feature;

[0012] The handwriting information with labels added are respectively input into the initial convolutional neural network model and the initial recurrent neural network model for parameter adjustment to obtain the preset style feature extraction model including the preset convolutional neural network model and the preset recurrent neural network model, wherein the input of the preset convolutional neural network model is the handwriting information of the handwriting, and the output is the physical change characteristics of the handwriting, and the input of the preset recurrent neural network model is the handwriting information of the handwriting, and the output is the style characteristics of the collected handwriting.

[0013] In a possible implementation, the preset generative adversarial network model includes a preset generative network model and a preset discriminative network model, and the inputting of the handwriting information, the physical change feature, and the style feature into the preset generative adversarial network model to obtain the texture feature of the target handwriting includes:

[0014] Inputting the handwriting information, the physical change feature and the style feature into the preset generation network model to obtain the initial texture feature of the target handwriting output by the preset generation network model;

[0015] The handwriting information, the physical change characteristics, the style characteristics and the initial texture characteristics are input into the preset discriminant network model to obtain a discrimination result of the accuracy of the initial texture characteristics output by the preset discriminant network model, and when the content of the discrimination result is accurate, the initial texture characteristics are determined as the texture characteristics of the target handwriting.

[0016] In a possible implementation, the training process of the preset generative adversarial network model includes:

[0017] Acquire multiple test sample data of real handwriting, wherein the test sample data of the real handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics, style characteristics and real texture characteristics of the real handwriting; obtain multiple test sample data of test handwriting, wherein the test sample data of the test handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics and style characteristics of the test handwriting;

[0018] Fix the parameters of the initial generation network model, input the test sample data of the real handwriting into the initial generation network model, and obtain the test texture features of the real handwriting output by the initial generation network model; input the test texture features of the real handwriting, the handwriting trajectory, the physical change features, the writing tool features, the style features and the real texture features of the real handwriting into the initial discriminant network model, so that the initial discriminant network model outputs a discrimination result for the test texture features of the real handwriting, and adjust the parameters of the initial discriminant network model based on the discrimination result and the real texture features of the real handwriting;

[0019] Fix the parameters of the initial discriminant network model after parameter adjustment, input the test sample data of the real handwriting into the initial generation network model, and input the test texture features of the real handwriting, the handwriting trajectory, the physical change features, the writing tool features and the style features output by the initial generation network model into the initial discriminant network model, and adjust the parameters of the initial generation network model based on the discrimination result output by the initial discriminant network model;

[0020] Use the test sample data of each test handwriting to jointly train the initial generation network model and the initial discriminant network model after parameter adjustment, and obtain the preset generative adversarial network model including the preset generation network model and the preset discriminant network model. The input of the preset generative adversarial network model is the handwriting information of the collected handwriting, and the output is the texture features of the collected handwriting, where the handwriting information includes the handwriting trajectory, physical change features, writing tool features and style features of the collected handwriting.

[0021] In a possible implementation, obtaining the handwriting information of the target handwriting sent by the handwriting acquisition device includes:

[0022] Obtain the handwriting information of the new handwriting of the target handwriting by the handwriting acquisition device in the current acquisition cycle, where the new handwriting is the new part of the target handwriting in the current acquisition cycle compared with the target handwriting in the historical acquisition cycle, and the end time of the historical acquisition cycle is the start time of the current acquisition cycle.

[0023] In a possible implementation, before inputting the handwriting information, the physical change features and the style features into the preset generative adversarial network model to obtain the texture features of the target handwriting, the handwriting display method further includes:

[0024] Use a preset interpolation algorithm to perform trajectory smoothing processing on the handwriting trajectory in the handwriting information according to the physical change features.

[0025] In a possible implementation, before inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting, the handwriting display method further includes:

[0026] The handwriting trajectory, inertia parameter, pressure parameter and pen tip tilt angle in the handwriting information are input into a preset trajectory correction algorithm, so that the preset trajectory correction algorithm can complete and correct the missing trajectory points in the handwriting trajectory.

[0027] In a possible implementation, after obtaining the texture feature of the target handwriting, the handwriting display method further includes:

[0028] Obtaining a custom character sent by a user through a terminal and an identifier of the target handwriting corresponding to a display style of the custom character selected by the user, and extracting the physical change feature, the style feature and the texture feature of the target handwriting based on the identifier of the target handwriting;

[0029] The custom character, the physical change feature, the style feature and the texture feature of the target handwriting are sent to the display terminal, so that the display terminal performs a rendering operation on the custom character based on the physical change feature, the style feature and the texture feature to obtain a rendered character.

[0030] A second aspect of the present application provides a handwriting display system, the handwriting display system comprising:

[0031] An information acquisition module, used for acquiring handwriting information of a target handwriting sent by a handwriting acquisition device;

[0032] A first feature extraction module, used for inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting;

[0033] A second feature extraction module is used to input the handwriting information, the physical change feature and the style feature into a preset generative adversarial network model to obtain the texture feature of the target handwriting;

[0034] An information sending module is used to send the physical change characteristics, the style characteristics, the texture characteristics and the handwriting information to a display terminal, so that the display terminal can render the handwriting track in the handwriting information based on the physical change characteristics, the style characteristics and the texture characteristics, restore the target handwriting, and display the target handwriting.

[0035] In a possible implementation, the handwriting display system further includes a first model training module, and the first model training module is configured to:

[0036] Obtaining handwriting information of a plurality of real handwritings, and adding a label to the handwriting information of each of the real handwritings, wherein the label content is a description parameter of the style feature;

[0037] The handwriting information with labels added are respectively input into the initial convolutional neural network model and the initial recurrent neural network model for parameter adjustment to obtain the preset style feature extraction model including the preset convolutional neural network model and the preset recurrent neural network model, wherein the input of the preset convolutional neural network model is the handwriting information of the handwriting, and the output is the physical change characteristics of the handwriting, and the input of the preset recurrent neural network model is the handwriting information of the handwriting, and the output is the style characteristics of the collected handwriting.

[0038] In a possible implementation, the second feature extraction module is configured to:

[0039] Inputting the handwriting information, the physical change feature and the style feature into a preset generative network model to obtain the initial texture feature of the target handwriting output by the preset generative network model, wherein the preset generative adversarial network model includes the preset generative network model and a preset discriminative network model;

[0040] The handwriting information, the physical change characteristics, the style characteristics and the initial texture characteristics are input into the preset discriminant network model to obtain a discrimination result of the accuracy of the initial texture characteristics output by the preset discriminant network model, and when the content of the discrimination result is accurate, the initial texture characteristics are determined as the texture characteristics of the target handwriting.

[0041] In a possible implementation, the handwriting display system further includes a second model training module, and the second model training module is configured to:

[0042] Acquire multiple test sample data of real handwriting, wherein the test sample data of the real handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics, style characteristics and real texture characteristics of the real handwriting; obtain multiple test sample data of test handwriting, wherein the test sample data of the test handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics and style characteristics of the test handwriting;

[0043] Fix the parameters of the initial generation network model, input the test sample data of the real handwriting into the initial generation network model, and obtain the test texture features of the real handwriting output by the initial generation network model; input the test texture features, the handwriting trajectory, the physical change features, the writing tool features, the style features and the real texture features of the real handwriting into the initial discriminant network model, so that the initial discriminant network model outputs a discrimination result for the test texture features of the real handwriting, and adjust the parameters of the initial discriminant network model based on the discrimination result and the real texture features of the real handwriting;

[0044] Fix the parameters of the initial discriminant network model after parameter adjustment, input the test sample data of the real handwriting into the initial generation network model, and input the test texture features, the handwriting trajectory, the physical change features, the writing tool features and the style features of the real handwriting output by the initial generation network model into the initial discriminant network model, and adjust the parameters of the initial generation network model based on the discrimination result output by the initial discriminant network model;

[0045] Use the test sample data of each test handwriting to jointly train the initial generation network model and the initial discriminant network model after parameter adjustment, and obtain the preset generative adversarial network model including the preset generation network model and the preset discriminant network model. The input of the preset generative adversarial network model is the handwriting information of the collected handwriting, and the output is the texture features of the collected handwriting, where the handwriting information includes the handwriting trajectory, physical change features, writing tool features and style features of the collected handwriting.

[0046] In a possible implementation, the information acquisition module is configured to:

[0047] Obtain the handwriting information of the new handwriting of the target handwriting in the current collection period by the handwriting collection device, where the new handwriting is the new part of the target handwriting in the current collection period compared with the target handwriting in the historical collection period, and the end time of the historical collection period is the start time of the current collection period.

[0048] In a possible implementation, the handwriting display system further includes:

[0049] A smoothing processing module, configured to perform trajectory smoothing processing on the handwriting trajectory in the handwriting information according to the physical change features by using a preset interpolation algorithm before inputting the handwriting information, the physical change features and the style features into the preset generative adversarial network model to obtain the texture features of the target handwriting.

[0050] In a possible implementation, the handwriting display system further includes:

[0051] A trajectory correction module, configured to input the handwriting trajectory, inertial parameters, pressure parameters, and nib tilt angle in the handwriting information into a preset trajectory correction algorithm before inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting, so that the preset trajectory correction algorithm complements and corrects the missing trajectory points in the handwriting trajectory.

[0052] In a possible implementation, the handwriting display system further includes:

[0053] A custom information sending module, configured to obtain the custom characters sent by the user through the terminal and the identifier of the target handwriting corresponding to the display style of the custom characters selected by the user after obtaining the texture features of the target handwriting, and extract the physical change features, the style features, and the texture features of the target handwriting based on the identifier of the target handwriting; send the custom characters, the physical change features, the style features, and the texture features of the target handwriting to the display terminal, so that the display terminal performs a rendering operation on the custom characters based on the physical change features, the style features, and the texture features to obtain a rendered character.

[0054] The third aspect of the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can perform the handwriting display method according to the first aspect or any implementation manner of the first aspect.

[0055] With the above technical solutions, a handwriting display method, system, and storage medium provided by the present application extract data for generating style features by configuring a handwriting acquisition device to acquire the above handwriting information. Moreover, by configuring the above preset style feature extraction model to extract features from the handwriting information, the physical change features and style features of the target handwriting are separated from the handwriting information, thereby improving the extraction accuracy of the physical change features representing the change law of the handwriting in space and the style features representing the writing style. Subsequently, by configuring a preset generative adversarial network model to extract texture features based on the handwriting information, physical change features, and style features, and taking advantage of the high feature extraction accuracy of the generative adversarial network for complex data distributions, the obtained texture features accurately represent the display effect of the target handwriting in a real writing environment. Finally, by configuring to send the physical change features, style features, texture features, and handwriting information to a display terminal, the display terminal performs a rendering operation on the handwriting trajectory in the handwriting information based on the physical change features, style features, and texture features, restores the target handwriting, and displays the target handwriting, such that the displayed target handwriting accurately represents the display effect of real handwriting, and the restoration accuracy of the target handwriting is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In combination with the accompanying drawings and with reference to the following specific embodiments, the above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the original elements and elements are not necessarily drawn to scale.

[0057] Figure 1 It is a flowchart of a handwriting display method provided by the present application;

[0058] Figure 2 It is a flowchart of a handwriting display method provided by a possible implementation of the present application;

[0059] Figure 3 It is a block diagram of a handwriting display system provided by the present application;

[0060] Figure 4 It is a schematic structural diagram of an electronic device provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] The following describes the embodiments of the present application in combination with the accompanying drawings in the embodiments of the present application. The terms used in the embodiments of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0062] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0063] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0064] The first aspect of the present application provides a handwriting display method, such as Figure 1 As shown, the handwriting display method includes:

[0065] S101, obtaining handwriting information of a target handwriting sent by a handwriting collection device.

[0066] It should be noted that, in actual application scenarios, the above handwriting information can be a data set including each trajectory point of the target handwriting, the pressure value of each trajectory point, the acceleration of the pen tip, the acceleration of the pen tip and the tilt angle of the pen tip. According to the research of the technical personnel of this application, the handwriting information of handwritings with different writing styles is quite different. Taking the last hook of the word "乐" as an example, assuming that performer A has a brisk writing style, the handwriting information of the last hook is presented as follows: the distribution of the trajectory points at the last hook is sparse, the pressure value is reduced, the acceleration of the pen tip is increased, the angular velocity of the pen tip is increased, and the tilt angle of the pen tip is increased. For performer B who presents a writing style that changes steadily and slowly, the handwriting information of the last hook is presented as follows: the distribution of the trajectory points at the last hook is dense, the pressure value is increased, the acceleration of the pen tip is reduced, the angular velocity of the pen tip is reduced, and the tilt angle of the pen tip is reduced. It can be seen that the above handwriting information can characterize different writing styles. Therefore, this application obtains the handwriting information of the target handwriting sent by the above handwriting acquisition device through configuration, thereby providing basic data for the subsequent extraction of style features.

[0067] It should be noted that in actual application scenarios, the above handwriting acquisition devices include, but are not limited to: near-infrared cameras, pressure-sensitive films, inertial sensors, etc. Among them, the above infrared camera is used to collect the trajectory and pen tip tilt angle of the target handwriting, the above pressure-sensitive film is used to collect the pressure of each trajectory point of the target handwriting and the trajectory of the target handwriting, and the above inertial sensor is used to collect the pen tip angular velocity and pen tip acceleration. By configuring the handwriting acquisition device to collect the above handwriting information, the present application realizes the extraction of data for generating style features.

[0068] In a possible implementation, the above pressure value, pen tip angular velocity, pen tip acceleration, and pen tip tilt angle can be associated with the coordinates of each trajectory point through a timestamp.

[0069] S102. Input the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting.

[0070] It should be noted that since the above handwriting information includes a large number of trajectory points that make up the target handwriting, as well as the corresponding pressure values, pen tip accelerations, pen tip angular velocities, pen tip tilt angles, and association relationships of each trajectory point, the dimension is relatively complex and interference is likely to occur between them. Therefore, the present application configures the above preset style feature extraction model to extract features from the handwriting information, so as to separate the physical change features and style features of the target handwriting from the handwriting information, thereby improving the extraction accuracy of the physical change features that characterize the change law of the handwriting in space and the style features that characterize the writing style.

[0071] It should be noted that in actual application scenarios, the above physical change features are feature data that characterize the change law of the target handwriting in space. For example: the speed curve of the target handwriting, pressure distribution, acceleration change curve, etc.

[0072] It should be noted that in actual application scenarios, the above style features are feature data that characterize the writing style of the target handwriting, such as: the connected stroke feature of the target handwriting, the handwriting change frequency feature, the handwriting direction change feature, etc.

[0073] S103. Input the handwriting information, physical change features, and style features into a preset generative adversarial network model to obtain the texture features of the target handwriting.

[0074] It should be noted that in the actual application scenario, the above-mentioned textural feature of the target handwriting is the feature data representing the display effect of the target handwriting on the writing surface. Through a large number of studies by the R & D personnel of this application, it is found that the above-mentioned physical change features and style features both affect the textural feature. Specifically, assuming that the writing style of performer A is slow and heavy-handed, when performer A uses a brush to write the target handwriting, the target handwriting has at least the following characteristics: the number of connected strokes of the target handwriting is small, the frequency of handwriting changes is low, the rate of change of handwriting direction is low, etc. It can be seen that there is a correlation between the above-mentioned style feature and the textural feature. And when using a brush to write, affected by the contact area between the brush and the writing surface, the pressure exerted by the brush on the writing surface, the adsorption force of the brush on the ink, and the acceleration of the brush on the ink, the textural feature when the target handwriting is displayed has the following characteristics: the accumulation amount of ink at the starting end and the ending end of the target handwriting is higher than that in the middle area between the starting end and the ending end, and the width of the target handwriting is higher. It can be seen that there is a correlation between the above-mentioned physical change feature and the textural feature. Therefore, this application configures a preset generative adversarial network model to extract the textural feature according to the handwriting information, physical change feature, and style feature, so that the obtained textural feature can represent the display effect of the target handwriting in the real writing environment, thereby improving the reduction accuracy of the target handwriting.

[0075] It should be noted that in the actual application scenario, the above-mentioned preset generative adversarial network model is a model constructed based on the Generative Adversarial Networks (GAN). Since the data distribution complexity in the above-mentioned handwriting information, physical change feature, and style feature is relatively high, therefore, this application configures a preset generative adversarial network model to extract the textural feature according to the handwriting information, physical change feature, and style feature, and utilizes the characteristic that the generative adversarial network has a high feature extraction accuracy for complex data distributions to improve the accuracy of the obtained textural feature.

[0076] S104. Send the physical change feature, style feature, textural feature, and handwriting information to the display terminal, so that the display terminal performs a rendering operation on the handwriting trajectory in the handwriting information based on the physical change feature, style feature, and textural feature, restores the target handwriting, and displays the target handwriting.

[0077] Those skilled in the art can understand that, in actual application scenarios, the above rendering operation can be implemented by a variety of rendering tools, such as: physically-based rendering (Physically-Based Rendering) engine, OpenGL ES graphics rendering tool (OpenGL for Embedded Systems, OpenGL ES), handwriting generation tools (such as TensorFlow), etc. This application does not make too many restrictions and redundant descriptions on the specific implementation process of the above rendering operation and the selected rendering tools.

[0078] In a possible implementation, since the physical change characteristics, style characteristics, texture characteristics and handwriting information are sent to the display terminal through a data transmission network, in order to prevent the above characteristics and information from being maliciously tampered with during the transmission process, the physical change characteristics, style characteristics, texture characteristics and handwriting information can be encrypted using a key, and the display terminal can decrypt them using the key before rendering, thereby improving the security of data transmission.

[0079] In a possible implementation, since the handwriting information includes a large number of track points of the target handwriting, there is a risk of missing track points during data transmission. Therefore, after the display terminal receives the handwriting information, the display terminal can be configured to perform a completion operation on the track points in the handwriting information, and the completion operation can include the following steps A1 to A6.

[0080] Step A1, traverse each track point to obtain the distance between adjacent track points, and trigger step A2.

[0081] Step A2, determine whether there is a distance greater than a preset threshold. If yes, trigger step A6. If no, trigger step A3.

[0082] Step A3, determining whether two adjacent track points corresponding to a distance greater than the preset threshold both have key track tags. If not, triggering step A4, and if so, triggering step A5.

[0083] In a possible implementation, the above-mentioned key trajectory label can be a label that characterizes the trajectory point as a trajectory point that constitutes a key stroke. Taking the hook at the end of the word "乐" as an example, the trajectory point at the corner of the hook is the key trajectory point with the key trajectory label added, and the other trajectory points except the ends and corners of the hook are non-key trajectory points without adding key trajectory labels. The above-mentioned key trajectory labels can be added manually or based on a preset script. This application does not make too many restrictions and redundant descriptions on the method of adding the above-mentioned key trajectory labels.

[0084] Step A4: Invoke the first preset completion algorithm to complete the trajectory points based on at least two adjacent trajectory points in Step A3, and update the supplemented trajectory points to the handwriting information. Then trigger Step A2.

[0085] It should be noted that since at least one of the two adjacent trajectory points used for trajectory point completion in Step A4 above is not a key trajectory point, in order to reduce the memory occupancy of the display terminal, the first preset completion algorithm with lower computing power requirements can be used to complete the missing trajectory points. In actual application scenarios, there can be various types of the above first preset completion algorithms, including but not limited to: Linear Interpolation algorithm, Inverse Distance Weighting algorithm, Mean Interpolation algorithm, etc. This application does not overly limit and elaborate on the specific type of the above first preset completion algorithm and the process of completing the trajectory points.

[0086] In a possible implementation, in addition to using the above first preset completion algorithm, in order to avoid the missing trajectory points affecting the restoration accuracy of the target handwriting, it can also be processed by performing deferred shading on the non-key trajectory points. This application does not overly elaborate on the specific execution process of deferred shading.

[0087] Step A5: Invoke the second preset completion algorithm to complete the trajectory points based on at least two adjacent trajectory points in Step A3, and update the supplemented trajectory points to the handwriting information. Then trigger Step A2.

[0088] Step A6: Output the discrimination result that no completion is required.

[0089] It should be noted that since both of the two adjacent trajectory points used for trajectory point completion in Step A5 above are key trajectory points, the trajectory points to be completed are also key trajectory points. Therefore, in order to improve the accuracy of the restored target trajectory, the accuracy of the completed trajectory points needs to be improved. The above second preset completion algorithm can output an algorithm with higher accuracy. The types of the above second preset completion algorithms include but not limited to: Cubic Spline Interpolation algorithm, Kalman Filter algorithm, Bézier Curve Fitting algorithm, etc. This application does not overly limit and elaborate on the specific type of the above second preset completion algorithm and the process of completing the trajectory points.

[0090] In a possible implementation, the above display terminal can also perform rendering configuration based on anti-aliasing technology to avoid the jagged effect of the restored target handwriting and improve the display clarity of the target handwriting. Moreover, by configuring the display terminal to perform normalized coordinate mapping on each trajectory point of the target handwriting during the rendering process, the displayed target handwriting can be adapted to different monitor sizes, thereby improving the user experience.

[0091] In this application, by configuring the handwriting acquisition device to collect the above handwriting information, the extraction of data for generating style features is realized. Moreover, by configuring the above preset style feature extraction model to extract features from the handwriting information, the physical change features and style features of the target handwriting are separated from the handwriting information, thereby improving the extraction accuracy of the physical change features representing the change law of handwriting in space and the style features representing the writing style. Subsequently, by configuring the preset generative adversarial network model to extract texture features according to the handwriting information, physical change features, and style features, and taking advantage of the high feature extraction accuracy of the generative adversarial network for complex data distributions, the obtained texture features accurately represent the display effect of the target handwriting in the real writing environment. Finally, by configuring to send the physical change features, style features, texture features, and handwriting information to the display terminal, the display terminal performs a rendering operation on the handwriting trajectory in the handwriting information based on the physical change features, style features, and texture features, restores the target handwriting, and displays the target handwriting, so that the displayed target handwriting accurately represents the display effect of the real handwriting, and the restoration accuracy of the target handwriting is improved.

[0092] In a possible implementation, the above preset style feature extraction model includes a preset convolutional neural network model and a preset recurrent neural network model. The training process of the preset style feature extraction model includes:

[0093] Obtain the handwriting information of multiple real handwritings, and add labels to the handwriting information of each real handwriting. The label content is the description parameters of the style features;

[0094] Input the labeled handwriting information into the initial convolutional neural network model and the initial recurrent neural network model respectively for parameter adjustment to obtain a preset style feature extraction model including a preset convolutional neural network model and a preset recurrent neural network model. The input of the preset convolutional neural network model is the handwriting information of the handwriting, and the output is the physical change features of the handwriting. The input of the preset recurrent neural network model is the handwriting information of the handwriting, and the output is the style features of the collected handwriting.

[0095] It should be noted that in the actual application scenario, the above tag content can be descriptive parameters determined by developers based on the analysis of the handwriting information of each real handwriting, so as to describe the style characteristics corresponding to the handwriting information. For example, heavy pressure and pauses, rapid flicks, long-distance connected strokes, etc. By configuring the above tag content to be added to the handwriting information of each real handwriting, the initial recurrent neural network model of the present application can achieve the extraction accuracy of the common parameters of different types of style characteristics, thereby improving the accuracy of obtaining the style characteristics output by the preset recurrent neural network model.

[0096] It should be noted that in the actual application scenario, the above initial convolutional neural network model can be a model constructed based on a Convolutional Neural Network (CNN), and the above initial recurrent neural network model can be a model constructed based on a Recurrent Neural Network (RNN). Since the above handwriting information includes both spatial structure dimension data of multiple independent trajectory points (such as trajectory point coordinates, trajectory point pressure direction, pen tip acceleration direction), and time dimension sequence data associated with the trajectory point timestamps (such as trajectory point pressure, pen tip acceleration, pen tip angular velocity, and pen tip tilt angle). And the above spatial dimension data has a relatively high correlation with physical change characteristics, and the above time dimension data has a relatively high correlation with style characteristics. Therefore, by configuring the preset style feature extraction model of the present application to include a preset convolutional neural network model and a preset recurrent neural network model, and utilizing the characteristics that the convolutional neural network has high extraction accuracy for spatial structured data and the recurrent neural network has high extraction accuracy for sequence data, the high-precision extraction of physical change characteristics and style characteristics is realized.

[0097] In a possible implementation, the preset generative adversarial network model includes a preset generator network model and a preset discriminator network model. Inputting the handwriting information, physical change characteristics, and style characteristics into the preset generative adversarial network model to obtain the texture characteristics of the target handwriting, including:

[0098] Inputting the handwriting information, physical change characteristics, and style characteristics into the preset generator network model to obtain the initial texture characteristics of the target handwriting output by the preset generator network model;

[0099] Inputting the handwriting information, physical change characteristics, style characteristics, and initial texture characteristics into the preset discriminator network model to obtain the discrimination result of the accuracy of the initial texture characteristics output by the preset discriminator network model, and when the content of the discrimination result is accurate, determining the initial texture characteristics as the texture characteristics of the target handwriting.

[0100] It should be noted that in the actual application scenario, the principle of the above preset generative adversarial network model for obtaining the texture features of the target handwriting is as follows: Configure the preset generative network model to generate initial texture features close to the real target handwriting according to the handwriting information, physical change features, and style features, so as to deceive the preset discriminative network model into outputting a discriminative result that is accurate for the content of the initial texture features. Configure the preset discriminative network model to discriminate the initial texture features according to the handwriting information, physical change features, and style features, and its goal is to output a discriminative result with inaccurate content. Through the confrontation between the preset generative network and the preset discriminative network model in this application, the accuracy of the texture features of the finally determined target handwriting is improved.

[0101] In a possible implementation, when the discriminative result content output by the preset discriminative network model is inaccurate, the handwriting information, physical change features, and style features will be configured to be input into the preset generative network model again, and the preset discriminative network model will be configured to discriminate the initial texture features of the target handwriting output by the preset generative network model again until the discriminative result output by the preset discriminative network model is accurate.

[0102] In a possible implementation, the training process of the preset generative adversarial network model includes:

[0103] Obtain test sample data of multiple real handwritings. The test sample data of real handwritings includes the handwriting trajectory, physical change features, writing tool features, style features, and real texture features of the real handwriting; obtain test sample data of multiple test handwritings. The test sample data of test handwritings includes the handwriting trajectory, physical change features, writing tool features, and style features of the test handwriting;

[0104] Fix the parameters of the initial generative network model, input the test sample data of real handwritings into the initial generative network model, and obtain the test texture features of real handwritings output by the initial generative network model; input the test texture features of real handwritings, handwriting trajectory, physical change features, writing tool features, style features, and real texture features into the initial discriminative network model, so that the initial discriminative network model outputs a discriminative result for the test texture features of real handwritings, and adjust the parameters of the initial discriminative network model based on the discriminative result and the real texture features of real handwritings;

[0105] Fix the parameters of the initial discriminative network model after parameter adjustment, input the test sample data of real handwritings into the initial generative network model, input the test texture features of real handwritings, handwriting trajectory, physical change features, writing tool features, and style features output by the initial generative network model into the initial discriminative network model, and adjust the parameters of the initial generative network model based on the discriminative result output by the initial discriminative network model;

[0106] The initial generated network model with adjusted parameters and the initial discriminant network model with adjusted parameters are jointly trained using the test sample data of each test handwriting to obtain a preset generative adversarial network model including a preset generated network model and a preset discriminant network model. The input of the preset generative adversarial network model is the handwriting information of the collected handwriting, and the output is the texture feature of the collected handwriting. Among them, the handwriting information includes the handwriting trajectory, physical change feature, writing tool feature, and style feature of the collected handwriting.

[0107] It should be noted that in the actual application scenario, the present application inputs the test texture feature, handwriting trajectory, physical change feature, writing tool feature, style feature, and real texture feature of the real handwriting into the initial discriminant network by configuration to train the initial discriminant network based on the difference between the test texture feature and the real texture feature, so as to improve the sensitivity of the initial discriminant network to the real texture feature, thereby improving the output accuracy of the obtained preset discriminant network model.

[0108] It should be noted that in the actual application scenario, the present application fixes the parameters of the initial discriminant network model with adjusted parameters by configuration, inputs the test sample data of the real handwriting into the initial generated network model, and inputs the test texture feature, handwriting trajectory, physical change feature, writing tool feature, and style feature of the real handwriting output by the initial generated network model into the initial discriminant network model, and adjusts the parameters of the initial generated network model based on the discriminant result output by the initial discriminant network model, so as to train the initial generated network model using the test sample data of the real handwriting on the premise of avoiding the influence of the initial generated network model on the accuracy of the initial discriminant network model, thereby improving the output accuracy of the obtained preset generated network model.

[0109] It should be noted that in the actual application scenario, the present application jointly trains the initial generated network model with adjusted parameters and the initial discriminant network model with adjusted parameters using the test sample data of each test handwriting, so that the obtained preset generative adversarial network model reaches the Nash equilibrium, avoids the preset generative adversarial network model falling into an infinite loop, and improves the operation reliability of the preset generative adversarial network model.

[0110] In a possible implementation, obtaining the handwriting information of the target handwriting sent by the handwriting acquisition device includes:

[0111] Obtaining the handwriting information of the new handwriting of the target handwriting by the handwriting acquisition device in the current acquisition cycle, where the new handwriting is the new part of the target handwriting in the current acquisition cycle compared with the target handwriting in the historical acquisition cycle, and the end time of the historical acquisition cycle is the start time of the current acquisition cycle.

[0112] It should be noted that in actual application scenarios, in order to restore the writing process to further improve the interactivity with the user, this application obtains the handwriting information of the newly added handwriting of the target handwriting in the current collection cycle by configuring the handwriting collection device, and restores the newly added handwriting based on the handwriting information of the newly added handwriting, so as to realize the dynamic restoration and display of the target handwriting in the writing process, and further improve the user experience. And compared with the method of transmitting and restoring the target handwriting after it is completely written, the display delay time is shortened.

[0113] It should be noted that, in actual application scenarios, there are many implementation methods for the handwriting collection device to collect the handwriting information of the above-mentioned newly added handwriting, and an exemplary implementation is provided here, including the following operation steps B1 to B2.

[0114] Step B1, the handwriting collection device determines whether the current collection time is the end time of the current detection cycle. If yes, step B2 is triggered. If no, step B1 is triggered.

[0115] Step B2, the handwriting collection device extracts each track point and its associated information in the local cache corresponding to the time stamp in the current detection cycle according to the timestamp of each track point, and obtains the handwriting information of the newly added handwriting of the target handwriting in the current collection cycle.

[0116] In a possible implementation, when the handwriting information of the newly added handwriting is restored based on the target handwriting, in order to avoid the delay caused by data transmission affecting the fluidity of the target handwriting display, the handwriting information of different newly added handwritings can be dynamically encoded and transmitted. Specifically: for the newly added handwriting including key track points, the encoding rate of the handwriting information of the newly added handwriting can be increased, and the transmission priority of the handwriting information of the newly added handwriting can be increased. For the newly added handwriting that does not include key track points, the encoding rate under the normal state can be maintained, and the transmission priority can be adjusted according to the order of collection.

[0117] In a possible implementation, before the handwriting information, physical change features and style features are input into the preset generative adversarial network model to obtain the texture features of the target handwriting, the above Figure 1 The handwriting display method shown also includes:

[0118] The handwriting trajectory in the handwriting information is smoothed by using a preset interpolation algorithm according to the physical change characteristics.

[0119] It should be noted that in actual application scenarios, due to the inevitable existence of noise points or offset points that deviate far from the handwriting trajectory in the above-mentioned handwriting trajectory, if the trajectory smoothing is not performed, the texture features obtained will easily produce breakpoints or hard edges due to the above-mentioned noise points or offset points, resulting in breaks in the restored target handwriting or abrupt handwriting transitions, thereby reducing the accuracy of the restored target handwriting.

[0120] In a possible implementation, before the handwriting information is input into the preset style feature extraction model to obtain the physical change features and style features of the target handwriting, the above Figure 1 The handwriting display method shown also includes:

[0121] The handwriting trajectory, inertia parameter, pressure parameter and pen tip tilt angle in the handwriting information are input into a preset trajectory correction algorithm, so that the preset trajectory correction algorithm can complete and correct the missing trajectory points in the handwriting trajectory.

[0122] It should be noted that in actual application scenarios, the handwriting trajectory in the above handwriting information can be a set of track base points extracted from the image collected by the near-infrared camera. In a real writing scene, due to the influence of writing tools, performers or light at the writing scene, there is a risk of missing track points when obtaining the track point set. If physical change features and style features are extracted based on handwriting information with missing track points, the accuracy of obtaining physical change features and style features will be reduced. Therefore, the present application configures the handwriting trajectory, inertia parameters, pressure parameters and pen tip tilt angle in the handwriting information to be input into a preset trajectory correction algorithm, so that the preset trajectory correction algorithm can complete and correct the missing trajectory points in the handwriting trajectory, thereby improving the completeness of the handwriting trajectory, and then improving the accuracy of obtaining physical change features and style features.

[0123] It should be noted that, in actual application scenarios, the above-mentioned preset trajectory correction algorithm can be an interpolation algorithm or an algorithm built based on a convolutional neural network. Among them, the implementation method of the above-mentioned preset trajectory correction algorithm built based on a convolutional neural network can be: obtain handwriting information of multiple real handwritings, and the handwriting information of the above-mentioned real handwritings includes the handwriting trajectory, inertia parameters, pressure parameters and pen tip tilt angle of the real handwriting. Use a noise reduction algorithm to perform noise reduction on the handwriting trajectory of each real handwriting; use the handwriting information including the handwriting trajectory after noise reduction to train the initial convolutional neural network to obtain a preset convolutional neural network. After establishing the data transmission relationship between the noise reduction algorithm and the preset convolutional neural network, the construction of the preset trajectory correction algorithm is completed.

[0124] In a possible implementation, the above handwriting acquisition device can be configured to dynamically adjust the sampling density of trajectory points based on the change rate of inertial parameters. When the change rate shows positive growth, the number of acquired trajectory points is increased; when the change rate shows negative growth, the number of sampled trajectory points is decreased, thereby reducing the redundancy of the obtained handwriting trajectory and improving the subsequent data processing speed.

[0125] In a possible implementation, after obtaining the texture features of the target handwriting, the above Figure 1 handwriting display method as shown further includes:

[0126] Obtaining the custom characters sent by the user through the terminal and the identifier of the target handwriting corresponding to the display style of the custom characters selected by the user, and extracting the physical change features, style features, and texture features of the target handwriting based on the identifier of the target handwriting;

[0127] Sending the custom characters, the physical change features, style features, and texture features of the target handwriting to the display terminal, so that the display terminal performs a rendering operation on the custom characters based on the physical change features, style features, and texture features to obtain a rendered character.

[0128] It should be noted that in an actual application scenario, the above custom characters can be characters or character strings input by the user based on their own customization requirements, and the characters or character strings can be user-defined text, names, or marking symbols.

[0129] It should be noted that in an actual application scenario, in this application, by configuring the identifier of the target handwriting corresponding to the display style of the custom characters selected by the user, extracting the physical change features, style features, and texture features of the target handwriting, and configuring the display terminal to perform a rendering operation on the custom characters based on the physical change features, style features, and texture features, a rendered character that meets the user's customization requirements is obtained, enhancing the interaction with the user and improving the user experience.

[0130] In a possible implementation, after sending the custom characters, the physical change features, style features, and texture features of the target handwriting to the display terminal, the above handwriting display method further includes:

[0131] Obtaining the custom display attribute parameters sent by the user through the terminal that meet the preset format, and importing the custom attribute parameters into the preset display attribute template;

[0132] Sending the preset display attribute template after importing the parameters to the display terminal, so that the display terminal performs display background rendering based on the preset display attribute template after importing the parameters, and displays the rendered character in front of the rendered display background.

[0133] This application displays the rendered characters in front of the rendered display background through the configured display terminal, further meeting the customized needs of users and improving the user experience. At the same time, since the rendered characters and the display background are rendered separately, the display background will not affect the display of the rendered characters.

[0134] For the convenience of understanding the above-mentioned Figure 1 handwriting display method as shown, a possible implementation of this application is specifically described herein:

[0135] As Figure 2 shown, it is a flowchart of a handwriting display method, and the specific operation steps are as follows:

[0136] Step S201, obtain the handwriting information of the target handwriting sent by the handwriting acquisition device, and trigger step S202.

[0137] Step S202, input the handwriting trajectory, inertial parameters, pressure parameters, and pen tip tilt angle in the handwriting information into the preset trajectory correction algorithm, so that the preset trajectory correction algorithm complements and corrects the missing trajectory points in the handwriting trajectory, and trigger step S203.

[0138] Step S203, input the handwriting information into the preset style feature extraction model to obtain the physical change features and style features of the target handwriting, and trigger step S204.

[0139] Step S204, use the preset interpolation algorithm to perform trajectory smoothing processing on the handwriting trajectory in the handwriting information according to the physical change features, and input the handwriting information, physical change features, and style features after trajectory smoothing processing into the preset generative adversarial network model, and trigger step S205.

[0140] Step S205, input the handwriting information, physical change features, and style features into the preset generation network model to obtain the initial texture features of the target handwriting output by the preset generation network model, and trigger step S206.

[0141] Step S206, input the initial texture features of the target handwriting into the preset discriminant network model by inputting the handwriting information, physical change features, style features, and initial texture features, and obtain the discriminant result output by the preset discriminant network model, and trigger step S207.

[0142] Step S207, determine whether the content of the discriminant result is accurate. If so, trigger step S208; if not, trigger step S205.

[0143] Step S208: Send the physical change features, style features, texture features, and handwriting information to the display terminal, so that the display terminal renders the handwriting trajectory in the handwriting information based on the physical change features, style features, and texture features, restores the target handwriting, and displays the target handwriting.

[0144] It should be noted that in the actual application scenario, the above-mentioned step S201 as Figure 2 shown is a possible implementation of step S101 as Figure 1 shown. The above-mentioned step S203 as Figure 2 shown is a possible implementation of step S102 as Figure 1 shown. The above-mentioned steps S204 to S207 as Figure 2 shown are a possible implementation of step S103 as Figure 1 shown. The above-mentioned step S208 as Figure 2 shown is a possible implementation of step S102 as Figure 1 shown.

[0145] The second aspect of this application provides a handwriting display system, as Figure 3 shown, the handwriting display system includes:

[0146] An information acquisition module 301, configured to acquire the handwriting information of the target handwriting sent by the handwriting acquisition device;

[0147] A first feature extraction module 302, configured to input the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting;

[0148] A second feature extraction module 303, configured to input the handwriting information, physical change features, and style features into a preset generative adversarial network model to obtain the texture features of the target handwriting;

[0149] An information sending module 304, configured to send the physical change features, style features, texture features, and handwriting information to the display terminal, so that the display terminal renders the handwriting trajectory in the handwriting information based on the physical change features, style features, and texture features, restores the target handwriting, and displays the target handwriting.

[0150] In a possible implementation, the above-mentioned handwriting display system as Figure 3 shown further includes a first model training module, and the first model training module is set during the training process of the preset style feature extraction model as:

[0151] Acquire the handwriting information of multiple real handwritings, and add labels to the handwriting information of each real handwriting, where the label content is the description parameters of the style features;

[0152] The handwriting information with tags added is respectively input into the initial convolutional neural network model and the initial recurrent neural network model for parameter adjustment, to obtain a preset style feature extraction model including a preset convolutional neural network model and a preset recurrent neural network model. The input of the preset convolutional neural network model is the handwriting information of the handwriting, and the output is the physical change features of the handwriting. The input of the preset recurrent neural network model is the handwriting information of the handwriting, and the output is the style features of the collected handwriting.

[0153] In a possible implementation, the above-mentioned second feature extraction module 303 is set as:

[0154] The handwriting information, physical change features, and style features are input into the preset generation network model to obtain the initial texture features of the target handwriting output by the preset generation network model. The preset generative adversarial network model includes a preset generation network model and a preset discriminative network model;

[0155] The handwriting information, physical change features, style features, and initial texture features are input into the preset discriminative network model to obtain the discriminative result of the accuracy of the initial texture features output by the preset discriminative network model, and when the content of the discriminative result is accurate, the initial texture features are determined as the texture features of the target handwriting.

[0156] In a possible implementation, the above-mentioned such as Figure 3 The handwriting display system shown also includes a second model training module, which is set during the training process of the preset generative adversarial network model as:

[0157] Obtain the test sample data of multiple real handwritings. The test sample data of real handwritings includes the handwriting trajectory, physical change features, writing tool features, style features, and real texture features of the real handwriting; obtain the test sample data of multiple test handwritings. The test sample data of test handwritings includes the handwriting trajectory, physical change features, writing tool features, and style features of the test handwriting;

[0158] Fix the parameters of the initial generation network model, input the test sample data of real handwritings into the initial generation network model to obtain the test texture features of real handwritings output by the initial generation network model; input the test texture features of real handwritings, handwriting trajectory, physical change features, writing tool features, style features, and real texture features into the initial discriminative network model, so that the initial discriminative network model outputs the discriminative result of the test texture features of real handwritings, and adjust the parameters of the initial discriminative network model based on the discriminative result and the real texture features of real handwritings;

[0159] Fix the parameters of the initial discriminant network model after parameter adjustment, input the test sample data of real handwriting into the initial generation network model, and input the test texture features, handwriting trajectories, physical change features, writing tool features, and style features of real handwriting output by the initial generation network model into the initial discriminant network model, and adjust the parameters of the initial generation network model based on the discrimination results output by the initial discriminant network model;

[0160] Use the test sample data of each test handwriting to jointly train the initial generation network model after parameter adjustment and the initial discriminant network model after parameter adjustment to obtain a preset generative adversarial network model including a preset generation network model and a preset discriminant network model. The input of the preset generative adversarial network model is the handwriting information of the collected handwriting, and the output is the texture feature of the collected handwriting. Among them, the handwriting information includes the handwriting trajectory, physical change feature, writing tool feature, and style feature of the collected handwriting.

[0161] In a possible implementation, the above information acquisition module 301 is configured as:

[0162] Obtain the handwriting information of the new handwriting of the target handwriting by the handwriting acquisition device during the current acquisition cycle. Among them, the new handwriting is the new part of the target handwriting in the current acquisition cycle compared with the target handwriting in the historical acquisition cycle, and the end time of the historical acquisition cycle is the start time of the current acquisition cycle.

[0163] In a possible implementation, the above as Figure 3 The handwriting display system shown further includes:

[0164] A smoothing processing module, configured to perform trajectory smoothing processing on the handwriting trajectory in the handwriting information according to the physical change feature by using a preset interpolation algorithm before inputting the handwriting information, physical change feature, and style feature into the preset generative adversarial network model to obtain the texture feature of the target handwriting.

[0165] In a possible implementation, the above as Figure 3 The handwriting display system shown further includes:

[0166] A trajectory correction module, configured to input the handwriting trajectory, inertial parameter, pressure parameter, and nib tilt angle in the handwriting information into a preset trajectory correction algorithm before inputting the handwriting information into the preset style feature extraction model to obtain the physical change feature and style feature of the target handwriting, so that the preset trajectory correction algorithm completes and corrects the missing trajectory points in the handwriting trajectory.

[0167] In a possible implementation, the above as Figure 3 The handwriting display system shown further includes:

[0168] A custom information sending module, which is used to obtain the custom characters sent by the user through the terminal and the identifier of the target handwriting corresponding to the display style of the custom characters after obtaining the texture features of the target handwriting, and extract the physical change features, style features and texture features of the target handwriting based on the identifier of the target handwriting; send the custom characters, the physical change features, style features and texture features of the target handwriting to the display terminal, so that the display terminal performs a rendering operation on the custom characters based on the physical change features, style features and texture features to obtain the rendered characters.

[0169] The third aspect of this application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the handwriting display method of the first aspect or any implementation manner of the first aspect above.

[0170] The embodiments of this application also provide an electronic device. Refer to Figure 4 As shown, it shows a schematic structural diagram of an electronic device suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, notebook computers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), desktop computers, and the like. Figure 4 The electronic device shown is only an example and should not bring any limitations to the functions and usage scopes of the embodiments of this application.

[0171] As Figure 4 shown, the electronic device may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage device 408 into the random access memory (RAM) 403. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 403. The processing device 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. The input / output (I / O) interface 405 is also connected to the bus 404.

[0172] Generally, the following devices may be connected to the I / O interface 405: an input device 406 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 408 including, for example, a memory card, a hard disk, etc.; and a communication device 409. The communication device 409 can allow the electronic device to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 4An electronic device with various devices is shown, but it should be understood that it is not required to implement or have all the shown devices. Instead, more or fewer devices may be implemented or had.

[0173] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions run on an electronic device, the electronic device is enabled to implement any one of the handwriting display methods provided by the embodiments of the present application.

[0174] In addition, it should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationships between the modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines.

[0175] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by dedicated hardware including application-specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures for implementing the same function can also be various, such as analog circuits, digital circuits or dedicated circuits. However, for the present application, in more cases, software program implementation is a better implementation manner. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, such as a floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disc of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a training device, or a network device, etc.) to execute the methods described in various embodiments of the present application.

[0176] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product.

[0177] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partly generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a training device or a data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

Claims

1. A handwriting display method, characterized in that: include: Obtaining handwriting information of a target handwriting sent by a handwriting collection device; Inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting; Inputting the handwriting information, the physical change feature and the style feature into a preset generative adversarial network model to obtain the texture feature of the target handwriting; The physical change feature, the style feature, the texture feature and the handwriting information are sent to a display terminal, so that the display terminal renders the handwriting track in the handwriting information based on the physical change feature, the style feature and the texture feature, restores the target handwriting, and displays the target handwriting.

2. The handwriting display method according to claim 1, characterized in that: The preset style feature extraction model includes a preset convolutional neural network model and a preset recurrent neural network model. The training process of the preset style feature extraction model includes: Obtaining handwriting information of a plurality of real handwritings, and adding a label to the handwriting information of each of the real handwritings, wherein the label content is a description parameter of the style feature; The handwriting information with labels added are respectively input into the initial convolutional neural network model and the initial recurrent neural network model for parameter adjustment to obtain the preset style feature extraction model including the preset convolutional neural network model and the preset recurrent neural network model, wherein the input of the preset convolutional neural network model is the handwriting information of the handwriting, and the output is the physical change characteristics of the handwriting, and the input of the preset recurrent neural network model is the handwriting information of the handwriting, and the output is the style characteristics of the collected handwriting.

3. The handwriting display method according to claim 1, characterized in that: The preset generative adversarial network model includes a preset generative network model and a preset discriminative network model, and the handwriting information, the physical change feature, and the style feature are input into the preset generative adversarial network model to obtain the texture feature of the target handwriting, including: Inputting the handwriting information, the physical change feature and the style feature into the preset generation network model to obtain the initial texture feature of the target handwriting output by the preset generation network model; The handwriting information, the physical change characteristics, the style characteristics and the initial texture characteristics are input into the preset discriminant network model to obtain a discrimination result of the accuracy of the initial texture characteristics output by the preset discriminant network model, and when the content of the discrimination result is accurate, the initial texture characteristics are determined as the texture characteristics of the target handwriting.

4. The handwriting display method according to claim 3, characterized in that: The training process of the preset generative adversarial network model includes: Acquire multiple test sample data of real handwriting, wherein the test sample data of the real handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics, style characteristics and real texture characteristics of the real handwriting; obtain multiple test sample data of test handwriting, wherein the test sample data of the test handwriting includes the handwriting trajectory, physical change characteristics, writing tool characteristics and style characteristics of the test handwriting; The parameters of the initial generation network model are fixed, the test sample data of the real handwriting is input into the initial generation network model, and the test texture features of the real handwriting output by the initial generation network model are obtained; the test texture features, the handwriting trajectory, the physical change features, the writing tool features, the style features and the real texture features of the real handwriting are input into the initial discrimination network model, so that the initial discrimination network model outputs the discrimination result of the test texture features of the real handwriting, and the parameters of the initial discrimination network model are adjusted based on the discrimination result and the real texture features of the real handwriting; The parameters of the initial discriminant network model after parameter adjustment are fixed, the test sample data of the real handwriting is input into the initial generation network model, and the test texture feature, the handwriting trajectory, the physical change feature, the writing tool feature and the style feature of the real handwriting output by the initial generation network model are input into the initial discriminant network model, and the parameters of the initial generation network model are adjusted based on the discrimination result output by the initial discriminant network model; The initial generative network model with parameters adjusted and the initial discriminative network model with parameters adjusted are jointly trained using the test sample data of each test handwriting to obtain the preset generative adversarial network model including the preset generative network model and the preset discriminative network model, wherein the input of the preset generative adversarial network model is the handwriting information of the collected handwriting, and the output is the texture feature of the collected handwriting, wherein the handwriting information includes the handwriting trajectory, physical change features, writing tool features and style features of the collected handwriting.

5. The handwriting display method according to claim 1, characterized in that: The step of obtaining the handwriting information of the target handwriting sent by the handwriting collection device includes: The handwriting information of the newly added handwriting of the target handwriting in the current collection cycle of the handwriting collection device is obtained, wherein the newly added handwriting is the newly added part of the target handwriting in the current collection cycle compared with the target handwriting in the historical collection cycle, and the end time of the historical collection cycle is the start time of the current collection cycle.

6. The handwriting display method according to claim 1, characterized in that: Before inputting the handwriting information, the physical change feature and the style feature into a preset generative adversarial network model to obtain the texture feature of the target handwriting, the handwriting display method further includes: A preset interpolation algorithm is used to perform trajectory smoothing processing on the handwriting trajectory in the handwriting information according to the physical change characteristics.

7. The handwriting display method according to claim 1, characterized in that: Before inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting, the handwriting display method further includes: The handwriting trajectory, inertia parameter, pressure parameter and pen tip tilt angle in the handwriting information are input into a preset trajectory correction algorithm, so that the preset trajectory correction algorithm can complete and correct the missing trajectory points in the handwriting trajectory.

8. The handwriting display method according to claim 1, characterized in that: After obtaining the texture features of the target handwriting, the handwriting display method further includes: Obtaining a custom character sent by a user through a terminal and an identifier of the target handwriting corresponding to a display style of the custom character selected by the user, and extracting the physical change feature, the style feature and the texture feature of the target handwriting based on the identifier of the target handwriting; The custom character, the physical change feature, the style feature and the texture feature of the target handwriting are sent to the display terminal, so that the display terminal performs a rendering operation on the custom character based on the physical change feature, the style feature and the texture feature to obtain a rendered character.

9. A handwriting display system, characterized in that: The handwriting display system comprises: An information acquisition module, used for acquiring handwriting information of a target handwriting sent by a handwriting acquisition device; A first feature extraction module, used for inputting the handwriting information into a preset style feature extraction model to obtain the physical change features and style features of the target handwriting; A second feature extraction module is used to input the handwriting information, the physical change feature and the style feature into a preset generative adversarial network model to obtain the texture feature of the target handwriting; An information sending module is used to send the physical change characteristics, the style characteristics, the texture characteristics and the handwriting information to a display terminal, so that the display terminal can render the handwriting track in the handwriting information based on the physical change characteristics, the style characteristics and the texture characteristics, restore the target handwriting, and display the target handwriting.

10. A computer storage medium, characterized in that: The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the handwriting display method as described in any one of claims 1 to 8.

Images