Calligraphy practice method and system

Abstract

The invention provides a calligraphy practicing method and system, and relates to the technical field of calligraphy teaching and practicing, and the method comprises the steps: obtaining an initial practicing video of a user side through collection equipment, carrying out the preprocessing of the initial practicing video based on an image processing algorithm, and generating a calligraphy practicing video corresponding to the user side; performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features comprise stroke space features and stroke time features; processing the stroke features through a calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label; and generating a personalized feedback result according to the calligraphy practice result and the practice level label, and displaying the personalized feedback result to a user side through a user interface, thereby helping a practicer to quickly find defects and perform targeted improvement without on-site guidance of a teacher, reducing the teaching cost, and meanwhile, being capable of stimulating the learning interest and motivation of the practicer. And a more convenient and efficient practice way is provided for calligraphy lovers.

Description

Technical Field

[0001] The present invention relates to the technical field of calligraphy teaching and practice, and in particular to a calligraphy practice method and system. Background Art

[0002] Calligraphy refers to the skill of writing Chinese characters using specific strokes. It conveys the beauty of words, and displays personalized style and artistic rhythm through the beauty of lines, the coordination of stroke structure and proportion, the use of ink color, and the white space treatment of the picture. With the development of science and technology, calligraphy teaching has gradually integrated digital and intelligent elements. Through modern technical means such as video analysis, high-precision image recognition technology, and artificial intelligence algorithms, calligraphy learning has become more convenient, efficient, and personalized.

[0003] In the prior art, traditional calligraphy practice teaching is highly dependent on the teacher's immediate demonstration and comments, as well as the repeated practice of the practitioners. This model is not only strictly restricted by physical space and time, but also difficult to ensure the uniformity and efficiency of teaching quality due to the subjectivity of the evaluation criteria, which leads to relatively low efficiency of calligraphy practice. It is difficult for practitioners to discover deficiencies and make targeted improvements, which reduces the practitioners' interest in learning. At the same time, it greatly increases the teaching cost and fails to provide practitioners with a convenient and efficient way to practice calligraphy.

[0004] Therefore, it is necessary to provide a method and system for calligraphy practice to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a calligraphy practice method and system for solving the problem that traditional calligraphy practice teaching is highly dependent on the teacher's demonstration and comments, as well as the repeated practice of the practitioners, which is not only strictly restricted by physical space and time, but also the evaluation standards are relatively subjective, making it difficult to ensure the uniformity of teaching quality, thereby resulting in relatively low efficiency of calligraphy practice, making it difficult for practitioners to discover deficiencies and make targeted improvements, reducing the practitioners' interest in learning, and greatly increasing the teaching costs, making it impossible to provide practitioners with a convenient and efficient way to practice calligraphy.

[0006] The present invention provides a method and system for practicing calligraphy, the method comprising: Acquire the initial practice video of the user terminal through the acquisition device, and pre-process the initial practice video based on the image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; Performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; Retrieving a calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal; A personalized feedback result corresponding to the user terminal is generated according to the calligraphy practice result and the practice level label, and is displayed to the user terminal through the user interface.

[0007] Preferably, the preprocessing of the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal specifically includes: Calling an image processing tool, and inputting the initial practice video into the image processing tool; Based on the image processing tool, denoising the initial practice video to generate an intermediate practice video; The contrast setting of the intermediate practice video is adjusted until an optimal contrast is reached to generate the calligraphy practice video.

[0008] Preferably, the stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, and the stroke features include stroke space features and stroke time features, specifically including: Performing stroke recognition on each frame of the calligraphy practice video based on a convolutional neural network algorithm to obtain the stroke space feature includes the following steps: Each frame image in the calligraphy practice video is input into the convolution layer in the convolutional neural network algorithm, the features of each frame image in the calligraphy practice video are extracted, and the corresponding output spatial features are generated. The expression of the output spatial features is: ; In the formula, Represents the output spatial features; represents the convolution kernel; Represents the convolution operation; Represents each frame of the calligraphy practice video; represents the bias term; The output spatial feature is input into the activation layer in the convolutional neural network algorithm to generate the corresponding activation spatial feature, and the expression of the activation spatial feature is: ; In the formula, represents the activation space feature; Indicates the maximum value operation; Represents the output spatial features; The activation space feature is input into the pooling layer in the convolutional neural network algorithm to reduce the dimension of the activation space feature, the dimension includes height and width, and the activation space feature is subjected to maximum pooling processing to generate the stroke space feature, the expression of the stroke space feature is: ; In the formula, Indicates position index The stroke space characteristics at Indicates position index The activation space features at ; Indicates the maximum value operation; Represents the position index in the stroke space feature The corresponding pooling window area on the activation space feature.

[0009] Preferably, the stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, and the stroke features include stroke space features and stroke time features, specifically including: Processing the stroke space feature based on a recurrent neural network to obtain the stroke time feature includes the following steps: Update the hidden state of the stroke space feature: ; In the formula, Indicates time The hidden state of Indicates time The hidden state of represents the activation function; , represents the weight matrix; Indicates time The stroke space characteristics of represents the bias term; Output the stroke time feature according to the hidden state: ; In the formula, Indicates time The stroke time characteristics of represents the weight matrix; Indicates time The hidden state of Represents the bias term.

[0010] Preferably, the step of retrieving a calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal specifically includes: Set the set corresponding to the stroke feature to: ; In the formula, Represents a set of stroke features; Indicates Stroke features; Set The default standard is , the calligraphy evaluation standard model is: ; ; ; In the formula, Indicates the result of calligraphy practice; Indicates The weight of each stroke feature; Indicates Stroke features; Indicates Preset standards; Indicates The dot product of the stroke features and the preset standard; Indicates A model of stroke features; Indicates A preset standard model; represents the vector dimension; ∑ represents the summation symbol; Based on the support vector machine, the optimization problem corresponding to the calligraphy practice result is obtained: ; In the formula, Indicates the minimum value operation; represents the hyperplane normal vector; represents the hyperplane intercept; represents the square of the norm of the hyperplane normal vector; Indicates constraints; Indicates the practice level label; Indicates the result of calligraphy practice; The practice level label is obtained by solving the optimization problem corresponding to the calligraphy practice result.

[0011] Preferably, the method generates personalized feedback results corresponding to the user terminal based on the calligraphy practice results and practice level labels, and displays them to the user terminal through the user interface, and also includes adjusting the preset standards of the calligraphy evaluation standard model based on the historical calligraphy practice results and practice level labels of the user terminal.

[0012] A calligraphy practice system, comprising: An acquisition module, used to acquire an initial practice video of a user terminal through an acquisition device, and pre-process the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; A recognition module, used for performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; An evaluation module, used to retrieve a calligraphy evaluation standard model, process the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generate a calligraphy practice result and a practice level label corresponding to the user terminal; A display module is used to generate a personalized feedback result corresponding to the user terminal according to the calligraphy practice result and the practice level label, and display it to the user terminal through the user interface.

[0013] Compared with the related art, the calligraphy practice method and system provided by the present invention have the following beneficial effects: The present invention obtains the initial practice video of the user end through the acquisition device, and pre-processes the initial practice video based on the image processing algorithm to generate the calligraphy practice video corresponding to the user end; based on the deep learning algorithm, the calligraphy practice video is subjected to stroke recognition to generate stroke features, and the stroke features include stroke space features and stroke time features; the calligraphy evaluation standard model is retrieved, and the stroke features are processed by the calligraphy evaluation standard model based on the machine learning algorithm to generate the calligraphy practice results and practice level labels corresponding to the user end; the personalized feedback results corresponding to the user end are generated according to the calligraphy practice results and practice level labels, and are displayed to the user end through the user interface. The present invention can help practitioners quickly discover deficiencies and make targeted improvements through automated evaluation and real-time feedback technology, and at the same time, no professional teacher is required to provide on-site guidance, which reduces the burden of labor costs and teaching costs. Through interactive teaching and personalized feedback, the learning interest and motivation of practitioners can be stimulated, and a more convenient and efficient practice path can be provided for calligraphy enthusiasts, which is conducive to the inheritance and development of calligraphy culture. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a calligraphy practice method of the present invention; Figure 2 is a schematic diagram of a user interface of the present invention; Figure 3 The system block diagram of a calligraphy practice system of the present invention. DETAILED DESCRIPTION

[0015] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0016] Embodiment 1 like Figure 1 As shown, a method for practicing calligraphy, the method comprising: S1, acquiring an initial practice video of a user terminal through a collection device, and preprocessing the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; In modern calligraphy teaching systems, highly integrated acquisition devices, such as high-definition cameras or professional calligraphy practice boards, can be used to accurately capture the initial practice video of the user during the practice process. Using image processing algorithms, the collected initial practice video can be pre-processed, including denoising, contrast enhancement, edge detection, and video frame stabilization.

[0017] Through the above method, the clarity and integrity of the video data can be ensured, the video quality can be optimized, a clear and stable calligraphy practice video can be generated, and the analyzability of the video can be improved to enhance the user experience.

[0018] S2, performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; It is understandable that after the video preprocessing is completed, the method of combining convolutional neural network (CNN) and recurrent neural network (RNN) in the deep learning algorithm can be used to accurately recognize the strokes of the calligraphy practice video.

[0019] Specifically, accurate stroke recognition can be performed on each frame of the video frame, and by extracting stroke spatial features and temporal features, where stroke spatial features can be the shape, direction, length, curvature, etc. of the stroke, and stroke temporal features can be the writing speed, acceleration, pause point, etc. of the stroke, a stroke feature set that fully reflects the user's writing habits can be constructed. These features can reflect the physical form and dynamic writing process of the user's writing, which is convenient for subsequent calligraphy evaluation.

[0020] S3, calling a calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal; In practical applications, a pre-trained calligraphy evaluation standard model can be called to train and optimize stroke features through machine learning algorithms, such as support vector machines (SVM), and accurately evaluate the level of calligraphy practice.

[0021] Specifically, the calligraphy evaluation standard model can conduct in-depth analysis of the extracted stroke features, comprehensively consider the spatial and temporal characteristics of the strokes, generate corresponding calligraphy practice results for the user, such as overall scores, specific stroke evaluations, etc., and determine the user's practice level label, such as beginner, intermediate, advanced, etc.

[0022] S4, generating a personalized feedback result corresponding to the user terminal according to the calligraphy practice result and the practice level label, and displaying it to the user terminal through the user interface.

[0023] Among them, personalized feedback results can be intelligently generated based on calligraphy practice results and practice level labels. These personalized feedback results include objective evaluations of the user's current calligraphy level, as well as targeted suggestions, such as strengthening the practice of a certain stroke, adjusting the writing speed, etc. Through the user interface, these personalized feedback results can be intuitively displayed to the user end, thereby helping users to identify their own shortcomings, develop more effective learning plans, and improve their calligraphy level.

[0024] In a specific implementation process, the initial practice video is preprocessed based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal, specifically including: Calling an image processing tool, and inputting the initial practice video into the image processing tool; Based on the image processing tool, denoising the initial practice video to generate an intermediate practice video; The contrast setting of the intermediate practice video is adjusted until an optimal contrast is reached to generate the calligraphy practice video.

[0025] It is understandable that in order to optimize the calligraphy learning experience, based on image processing technology, high-performance image processing tools can be accurately selected, which can perform fine manipulation and efficient processing of video content, and then the initial practice video to be processed can be accurately loaded into the image processing tool through a standard file import process.

[0026] After the video is loaded, the advanced denoising function module built into the image processing tool can be used to perform deep denoising on the initial practice video. This process involves multiple filtering algorithms, such as Wiener filtering and wavelet transform, which can effectively eliminate noise interference in the video, such as background noise and image graininess, retain the clarity and details of the calligraphy practice movements, and generate an intermediate practice video.

[0027] Color management technology can be used to fine-tune the contrast settings of the intermediate practice videos. This process relies on image histogram analysis, and by dynamically adjusting the distribution of light and dark parts in the image, the best visual balance can be achieved. At the same time, the preview function can be used to view the adjustment effect in real time to ensure that the contrast is appropriate, and the level and strength changes of the calligraphy strokes can be clearly displayed, and the calligraphy practice video can be generated.

[0028] The stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, wherein the stroke features include stroke space features and stroke time features, specifically including: A convolutional neural network (CNN) usually contains multiple layers, namely a convolutional layer, an activation layer, and a pooling layer.

[0029] Performing stroke recognition on each frame of the calligraphy practice video based on a convolutional neural network algorithm to obtain the stroke space feature includes the following steps: Each frame image in the calligraphy practice video is input into the convolution layer in the convolutional neural network algorithm, the features of each frame image in the calligraphy practice video are extracted, and the corresponding output spatial features are generated. The expression of the output spatial features is: ; In the formula, Represents the output spatial features; represents the convolution kernel; Represents the convolution operation; Represents each frame of the calligraphy practice video; represents the bias term; The output spatial feature is input into the activation layer in the convolutional neural network algorithm to generate the corresponding activation spatial feature, and the expression of the activation spatial feature is: ; In the formula, represents the activation space feature; Indicates the maximum value operation; Represents the output spatial features; The activation space feature is input into the pooling layer in the convolutional neural network algorithm to reduce the dimension of the activation space feature, the dimension includes height and width, and the activation space feature is subjected to maximum pooling processing to generate the stroke space feature, the expression of the stroke space feature is: ; In the formula, Indicates position index The stroke space characteristics at Indicates position index The activation space features at ; Indicates the maximum value operation; Represents the position index in the stroke space feature The corresponding pooling window area on the activation space feature.

[0030] The stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, wherein the stroke features include stroke space features and stroke time features, specifically including: Processing the stroke space feature based on a recurrent neural network to obtain the stroke time feature includes the following steps: Update the hidden state of the stroke space feature: ; In the formula, Indicates time The hidden state of Indicates time The hidden state of represents the activation function; , represents the weight matrix; Indicates time The stroke space characteristics of represents the bias term; Output the stroke time feature according to the hidden state: ; In the formula, Indicates time The stroke time characteristics of represents the weight matrix; Indicates time The hidden state of Represents the bias term.

[0031] The step of retrieving the calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal specifically includes: Based on deep learning algorithms, we can perform stroke recognition on calligraphy practice videos to generate stroke features, and these stroke features can be expressed in the form of vectors to facilitate subsequent result evaluation and analysis. These stroke feature vectors may include the length, direction, and curvature of the strokes.

[0032] Set the set corresponding to the stroke feature to: ; In the formula, Represents a set of stroke features; Indicates Stroke features; Set The default standard is , the calligraphy evaluation standard model is: ; ; ; In the formula, Indicates the result of calligraphy practice; Indicates The weight of each stroke feature; Indicates Stroke features; Indicates Preset standards; Indicates The dot product of the stroke features and the preset standard; Indicates A model of stroke features; Indicates A preset standard model; represents the vector dimension; ∑ represents the summation symbol; In the calligraphy evaluation standard model, stroke features and preset standards are calculated in the form of vectors, and cosine similarity It can measure the similarity between stroke features and preset standards in direction. The value range is [-1,1]. When the directions of the two vectors of the stroke feature and the preset standard are exactly the same, the cosine similarity is 1; when the directions of the two vectors are completely opposite, the cosine similarity is -1; when the two vectors are orthogonal, the cosine similarity is 0.

[0033] In a support vector machine (SVM), by determining the hyperplane normal vector and the hyperplane intercept, the interval between different practice level labels can be maximized. After determining the optimization problem corresponding to the calligraphy practice result, the practice level label corresponding to the calligraphy practice result can be obtained by solving the optimization problem.

[0034] Based on the support vector machine, the optimization problem corresponding to the calligraphy practice result is obtained: ; In the formula, Indicates the minimum value operation; represents the hyperplane normal vector; represents the hyperplane intercept; represents the square of the norm of the hyperplane normal vector; Indicates constraints; Indicates the practice level label; Indicates the result of calligraphy practice; The practice level label is obtained by solving the optimization problem corresponding to the calligraphy practice result.

[0035] The method of generating a personalized feedback result corresponding to the user terminal based on the calligraphy practice results and the practice level label and presenting it to the user terminal through the user interface also includes adjusting the preset standard of the calligraphy evaluation standard model based on the historical calligraphy practice results and the practice level label of the user terminal.

[0036] like Figure 2 As shown, through an efficient and personalized feedback mechanism, based on the calligraphy practice results of the user end and its corresponding practice level label, customized feedback results can be accurately generated and displayed to the user end through the user interface on the terminal device.

[0037] In addition, the user's historical calligraphy practice results and their corresponding practice level changes can be continuously obtained, and the user's progress trajectory and learning preferences can be dynamically analyzed to achieve self-optimization and adaptive adjustment of the system. Based on the data analysis results, the preset standards of the calligraphy evaluation standard model can be adjusted in real time to ensure that the accuracy and pertinence of the feedback are gradually enhanced as the user's ability improves.

[0038] Through the above method, the dynamic adjustment mechanism is adopted to promote the continuous optimization of user-side learning effects and improve the flexibility and pertinence of the calligraphy learning system in the field of personalized practice teaching.

[0039] Embodiment 2 like Figure 3 As shown, a system for practicing calligraphy, the system comprising: An acquisition module, used to acquire an initial practice video of a user terminal through an acquisition device, and pre-process the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; A recognition module, used for performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; An evaluation module, used to retrieve a calligraphy evaluation standard model, process the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generate a calligraphy practice result and a practice level label corresponding to the user terminal; A display module is used to generate a personalized feedback result corresponding to the user terminal according to the calligraphy practice result and the practice level label, and display it to the user terminal through the user interface.

[0040] Through the introduction of the above embodiments, the present invention uses a calligraphy practice method and system to obtain the initial practice video of the user end through a collection device, and pre-process the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user end; based on a deep learning algorithm, the calligraphy practice video is subjected to stroke recognition to generate stroke features, and the stroke features include stroke space features and stroke time features; the calligraphy evaluation standard model is retrieved, and the stroke features are processed through the calligraphy evaluation standard model based on a machine learning algorithm to generate calligraphy practice results and practice level labels corresponding to the user end; the personalized feedback results corresponding to the user end are generated according to the calligraphy practice results and practice level labels, and are displayed to the user end through the user interface. The present invention can help practitioners quickly discover deficiencies and make targeted improvements through automated evaluation and real-time feedback technology; at the same time, no professional teacher is required to provide on-site guidance, which reduces the burden of labor costs and reduces teaching costs; through interactive teaching and personalized feedback, the learning interest and motivation of practitioners can be stimulated; and a more convenient and efficient practice path can be provided for calligraphy enthusiasts, which is conducive to the inheritance and development of calligraphy culture.

[0041] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0042] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0043] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A method for practicing calligraphy, characterized in that: The method comprises: Acquire the initial practice video of the user terminal through the acquisition device, and pre-process the initial practice video based on the image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; Performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; Retrieving a calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal; A personalized feedback result corresponding to the user terminal is generated according to the calligraphy practice result and the practice level label, and is displayed to the user terminal through the user interface.

2. A calligraphy practice method according to claim 1, characterized in that: The preprocessing of the initial practice video based on the image processing algorithm to generate a calligraphy practice video corresponding to the user terminal specifically includes: Calling an image processing tool, and inputting the initial practice video into the image processing tool; Based on the image processing tool, denoising the initial practice video to generate an intermediate practice video; The contrast setting of the intermediate practice video is adjusted until an optimal contrast is reached to generate the calligraphy practice video.

3. A calligraphy practice method according to claim 1, characterized in that: The stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, wherein the stroke features include stroke space features and stroke time features, specifically including: Performing stroke recognition on each frame of the calligraphy practice video based on a convolutional neural network algorithm to obtain the stroke space feature includes the following steps: Each frame image in the calligraphy practice video is input into the convolution layer in the convolutional neural network algorithm, the features of each frame image in the calligraphy practice video are extracted, and the corresponding output spatial features are generated. The expression of the output spatial features is: ； In the formula, Represents the output spatial features; represents the convolution kernel; Represents the convolution operation; Represents each frame of the calligraphy practice video; represents the bias term; The output spatial feature is input into the activation layer in the convolutional neural network algorithm to generate the corresponding activation spatial feature, and the expression of the activation spatial feature is: ； In the formula, represents the activation space feature; Indicates the maximum value operation; Represents the output spatial features; The activation space feature is input into the pooling layer in the convolutional neural network algorithm to reduce the dimension of the activation space feature, the dimension includes height and width, and the activation space feature is subjected to maximum pooling processing to generate the stroke space feature, the expression of the stroke space feature is: ； In the formula, Indicates position index The spatial characteristics of the strokes at Indicates position index The activation space features at ; Indicates the maximum value operation; Represents the position index in the stroke space feature The corresponding pooling window area on the activation space feature.

4. A calligraphy practice method according to claim 3, characterized in that: The stroke recognition based on the deep learning algorithm is performed on the calligraphy practice video to generate stroke features, wherein the stroke features include stroke space features and stroke time features, specifically including: Processing the stroke space feature based on a recurrent neural network to obtain the stroke time feature includes the following steps: Update the hidden state of the stroke space feature: ； In the formula, Indicates time The hidden state of Indicates time The hidden state of represents the activation function; , represents the weight matrix; Indicates time The stroke space characteristics of represents the bias term; Output the stroke time feature according to the hidden state: ； In the formula, Indicates time The stroke time characteristics of represents the weight matrix; Indicates time The hidden state of Represents the bias term.

5. The method for practicing calligraphy according to claim 1, characterized in that: The step of retrieving the calligraphy evaluation standard model, processing the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generating a calligraphy practice result and a practice level label corresponding to the user terminal specifically includes: Set the set corresponding to the stroke feature to: ； In the formula, Represents a set of stroke features; Indicates Stroke features; Set The default standard is , the calligraphy evaluation standard model is: ； ； ； In the formula, Indicates the result of calligraphy practice; Indicates The weight of each stroke feature; Indicates Stroke features; Indicates Preset standards; Indicates The dot product of the stroke features and the preset standard; Indicates A model of stroke features; Indicates A preset standard model; represents the vector dimension; ∑ represents the summation symbol; Based on the support vector machine, the optimization problem corresponding to the calligraphy practice result is obtained: ； In the formula, Indicates the minimum value operation; represents the hyperplane normal vector; represents the hyperplane intercept; represents the square of the norm of the hyperplane normal vector; Indicates constraints; Indicates the practice level label; Indicates the result of calligraphy practice; The practice level label is obtained by solving the optimization problem corresponding to the calligraphy practice result.

6. A calligraphy practice method according to claim 1, characterized in that: The method of generating a personalized feedback result corresponding to the user terminal based on the calligraphy practice results and the practice level label and presenting it to the user terminal through the user interface also includes adjusting the preset standard of the calligraphy evaluation standard model based on the historical calligraphy practice results and the practice level label of the user terminal.

7. A calligraphy practice system, applied to a calligraphy practice method as claimed in any one of claims 1 to 6, the system comprising: An acquisition module, used to acquire an initial practice video of a user terminal through an acquisition device, and pre-process the initial practice video based on an image processing algorithm to generate a calligraphy practice video corresponding to the user terminal; A recognition module, used for performing stroke recognition on the calligraphy practice video based on a deep learning algorithm to generate stroke features, wherein the stroke features include stroke space features and stroke time features; An evaluation module, used to retrieve a calligraphy evaluation standard model, process the stroke features through the calligraphy evaluation standard model based on a machine learning algorithm, and generate a calligraphy practice result and a practice level label corresponding to the user terminal; A display module is used to generate a personalized feedback result corresponding to the user terminal according to the calligraphy practice result and the practice level label, and display it to the user terminal through the user interface.

Images