A digital media art teaching system constructed by fusing virtual reality technology

Abstract

The application relates to the technical field of digital media art teaching, and discloses a digital media art teaching system constructed by fusing virtual reality technology. A teaching state evaluation unit of the system comprises a plurality of student creation state acquisition subunits corresponding to student terminals, can synchronously acquire multidimensional state information in a plurality of student creation processes, and a teaching balance degree judgment unit determines the teaching balance degree based on acquisition results; a virtual creation environment construction unit determines whether to start three-dimensional scene acquisition according to the teaching balance degree, and acquires adaptive virtual creation scene data when it is determined to start; and a teaching strategy decision unit generates an optimal teaching path plan based on the virtual creation scene data, and adjusts virtual teaching scene parameters according to the optimal teaching path plan. The system can realize accurate monitoring of student creation states, targeted construction of virtual environments and dynamic optimization of teaching strategies, and helps improve the quality of digital media art teaching and the creation ability of students.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital media art teaching, in particular to a digital media art teaching system constructed by integrating virtual reality technology. BACKGROUND

[0002] With the rapid development of the digital media art industry, the market demand for professionals with innovation ability and practical skills is increasing, and the teaching mode is gradually transforming towards digitization and intelligentization. Virtual reality technology, with its immersive and interactive characteristics, has shown broad application prospects in digital media art teaching, breaking through the limitations of space and time in traditional teaching, allowing students to carry out creative practice in virtual scenes and improving learning experience and creative efficiency.

[0003] Current solutions that integrate virtual reality technology into digital media art teaching still have many shortcomings. In terms of student creation state monitoring, most teaching systems can only collect single-dimensional student learning data such as learning duration and simple operation records, and cannot fully capture the details of students' creation process, including changes in creation ideas, adjustments in operation methods, and technical difficulties encountered. Due to the lack of synchronous and multi-dimensional collection of multiple student creation states, teachers cannot accurately grasp the learning progress and ability level of each student, and thus cannot judge whether the teaching process is balanced - some students may not be able to keep up with the teaching pace due to weak foundation, while another group of students may not be able to further improve their ability due to the teaching content being too basic, and teaching resources are difficult to achieve precise allocation.

[0004] In terms of virtual creation environment construction, most existing systems use fixed three-dimensional scene templates, regardless of the students' creation state and whether the teaching balance is met, the scene collection and loading are started according to the unified process, lacking of pertinence and flexibility. Under this mode, the virtual scene is out of touch with the actual needs of students, when students are in different creation stages or have different ability basis, the fixed virtual scene cannot match their individualized learning needs, which may lead to students' difficulty in fully tapping their creative potential in the virtual environment, and even reduce their learning enthusiasm due to the mismatch between the scene and their own ability.

[0005] At the level of teaching strategy adjustment, due to the lack of deep analysis of student creation state data and effective combination with virtual creation scene data, the teaching strategy often depends on the experience of teachers to formulate, and it is difficult to form a scientific and systematic optimal teaching path planning. Even if part of the system tries to adjust the teaching strategy, it often stays at the level of simply increasing or decreasing the content, and cannot adjust the scene parameters such as scene complexity, interaction mode and task difficulty in real time according to the actual feedback of students in the virtual creation scene, resulting in the disconnection between the teaching strategy and the virtual teaching scene, and the inability to form a closed loop of "state collection-environment adaptation-strategy adjustment", which affects the teaching quality and student learning effect. The existence of these problems restricts the depth and breadth of the application of virtual reality technology in digital media art teaching, and it is difficult to meet the needs of high-quality talent training at present. SUMMARY

[0006] The purpose of the present application is to provide a digital media art teaching system constructed by integrating virtual reality technology to solve the problems raised in the above background technology.

[0007] To achieve the above purpose, the present application provides a digital media art teaching system constructed by integrating virtual reality technology, which comprises:

[0008] A teaching state evaluation unit, comprising a plurality of student creation state collection sub-units and a teaching balance degree judgment unit, the plurality of student creation state collection sub-units correspond to a plurality of student terminals respectively, and the teaching balance degree judgment unit is used to determine the teaching balance degree based on the output of the student creation state collection sub-unit;

[0009] A virtual creation environment construction unit is used to determine whether to start three-dimensional scene collection based on the teaching balance degree of each student creation state collection sub-unit, and collect virtual creation scene data when it is determined to start three-dimensional scene collection;

[0010] A teaching strategy decision unit is used to generate an optimal teaching path planning based on the virtual creation scene data, and adjust the virtual teaching scene parameters based on the optimal teaching path planning.

[0011] Preferably, the student creation state collection sub-unit comprises a state monitoring panel arranged on the interface of the student terminal, and the state monitoring panel is synchronously displayed with the student creation interface.

[0012] A plurality of operation trajectory capture devices are arranged on the state monitoring panel.

[0013] A virtual teaching marker is arranged on the state monitoring panel.

[0014] Preferably, the teaching balance degree judgment unit determines the teaching balance degree based on the output of the student creation state collection sub-unit, which comprises:

[0015] For each of the operation trajectory capture devices, modal decomposition is performed on the input signals of the operation trajectory capture devices in a continuous teaching period to generate characteristic modal components and trend components;

[0016] A signal correction model is used to generate corrected continuous teaching period input signals based on the characteristic modal components and trend components;

[0017] A teaching balance degree is determined based on the corrected continuous teaching period input signals corresponding to the plurality of operation trajectory capture devices.

[0018] Preferably, the teaching balance degree determination unit determines the teaching balance degree based on the corrected continuous teaching period input signals corresponding to the plurality of operation trajectory capture devices, including:

[0019] The corrected input signals of each of the operation trajectory capture devices are sampled by period;

[0020] Input signal values of a plurality of sampled teaching moments are obtained;

[0021] Single-point teaching balance degree parameters are calculated based on the input signal values of the plurality of operation trajectory capture devices at each of the sampled teaching moments;

[0022] The teaching balance degree is determined based on the single-point teaching balance degree parameters of a plurality of sampled teaching moments.

[0023] Preferably, the virtual creation environment construction unit determines whether to start three-dimensional scene acquisition based on the teaching balance degree, including:

[0024] Teaching balance degree difference values of any two student creation state acquisition sub-units are calculated;

[0025] A teaching balance degree dispersion parameter is calculated based on all teaching balance degree difference values;

[0026] When the teaching balance degree dispersion parameter is less than a preset dispersion threshold value and there is at least one teaching balance degree greater than a preset balance threshold value, it is determined to start three-dimensional scene acquisition.

[0027] Preferably, the virtual creation environment construction unit acquires virtual creation scene data, including acquiring scene data at a plurality of observation poses;

[0028] The teaching strategy decision unit generates an optimal teaching path plan based on the virtual creation scene data, including:

[0029] For each of the observation poses, a spatial relationship between the virtual teaching marker and a scene reference point is determined;

[0030] Marker spatial distribution features are generated based on the spatial relationships of a plurality of observation poses;

[0031] generate an optimal teaching path plan based on the spatial distribution characteristics of the markers.

[0032] Preferably, the teaching strategy decision unit comprises:

[0033] generating a plurality of candidate teaching paths based on the spatial distribution characteristics of the markers;

[0034] obtaining interaction response data of the student terminal in a continuous teaching period;

[0035] obtaining environmental interference parameters corresponding to the teaching period;

[0036] screening the plurality of candidate teaching paths based on the interaction response data and the environmental interference parameters to determine an optimal teaching path plan.

[0037] Preferably, the teaching dynamic compensation unit is configured to generate a teaching parameter compensation scheme based on the interaction response data, the environmental interference parameters and the optimal teaching path plan.

[0038] Preferably, the teaching strategy decision unit adjusts the virtual teaching scene parameters based on the optimal teaching path plan comprises:

[0039] generating a preset teaching strategy based on historical teaching records, the preset teaching strategy including time-periodic scene parameter configurations;

[0040] when the deviation degree of the optimal teaching path plan from the preset teaching strategy of the corresponding period exceeds a preset deviation threshold, adjusting the virtual teaching scene parameters according to the optimal teaching path plan and updating the preset teaching strategy.

[0041] Preferably, the teaching dynamic compensation unit generates a teaching parameter compensation scheme comprises:

[0042] performing secondary calibration on the teaching strategy decision space based on predicted interaction response data and predicted environmental interference parameters of a predicted teaching period;

[0043] generating a teaching compensation decision feature space;

[0044] performing parameter compensation optimization on the optimal teaching path plan based on the teaching compensation decision feature space.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] The digital media art teaching system constructed by the fusion virtual reality technology realizes comprehensive and accurate collection of multiple student creation states and scientific judgment of teaching balance degree through the setting of a teaching state evaluation unit. The multiple student creation state collection subunits correspond to multiple student terminals and can synchronously capture multi-dimensional information of each student in the creation process, covering creation operation details, thought expression, problem feedback and the like, and are no longer limited to a single data dimension. The teaching balance degree judgment unit can objectively analyze the learning progress difference and ability level gap between different students based on the collected information, clearly present whether there are problems such as uneven resource allocation and unbalanced teaching rhythm in the teaching process, and enable teachers to intuitively understand the overall teaching situation and provide a clear direction for subsequent teaching adjustment.

[0047] The virtual creation environment construction unit decides whether to start three-dimensional scene collection based on the teaching balance degree judgment result, changes the mode of fixedly starting a scene of the traditional system, and makes the construction of the virtual creation scene more targeted. When the teaching balance degree is determined to need adjustment, three-dimensional scene collection is started and virtual scene data that fits the current teaching needs and student creation states is obtained, avoiding the problem that a fixed scene template is out of touch with the actual needs of students. If the teaching balance degree is within a reasonable range, scene collection does not need to be additionally started, unnecessary resource consumption is reduced, and it is ensured that the virtual creation environment can meet the individualized creation needs of students and realize efficient use of teaching resources, so that students can carry out creation in a virtual scene that adapts to their own abilities, fully stimulate their creation potential, and improve the immersive learning experience.

[0048] The teaching strategy decision unit generates an optimal teaching path plan based on virtual creation scene data, breaking the limitation of traditional strategies that rely on teacher experience. Through analysis of virtual scene data, a teaching path that meets the overall teaching goal and takes into account individual differences can be developed based on the creation state of students and the teaching balance degree, with clear teaching focuses, practical tasks and guidance directions at different stages. At the same time, based on the optimal teaching path plan, virtual teaching scene parameters are adjusted to realize the deep integration of teaching strategies and virtual scenes: according to the advancement of the teaching path, the complexity, interaction logic, resource allocation and the like of the scene are optimized in real time, so that the virtual teaching scene is always matched with the teaching progress and student ability. For example, when students need to improve their complex scene modeling ability in the teaching path plan, the system can automatically adjust the richness and difficulty of the modeling elements of the virtual scene; when it is found that some students have difficulties in a specific operation link, more detailed guidance can be provided by adjusting the scene interaction mode. This dynamic adjustment mechanism forms a complete closed loop in the teaching process, enabling each student to learn in a virtual environment that adapts to the teaching rhythm, effectively solving the progress imbalance problem that may occur in teaching, helping students gradually improve their digital media art creation ability, and promoting the overall improvement of teaching quality.

[0049] The whole system cooperates with each unit to form a complete teaching support system from student state monitoring to environment construction to strategy adjustment. The teaching state evaluation unit provides basis for virtual creation environment construction, ensures the rationality of scene construction; virtual creation scene data provides support for teaching strategy decision, makes strategy adjustment more scientific; and the adjustment of teaching strategy in turn optimizes the subsequent student state monitoring focus and virtual scene construction direction, the three promote each other, continuously improve the accuracy and effectiveness of teaching. In actual teaching application, the system can adapt to different scale teaching scenes, whether it is small class precise teaching or large class collective teaching, can ensure the fairness and efficiency of the teaching process through the attention to the creation state of each student and the dynamic allocation of teaching resources, and provides a feasible solution for the innovation of digital media art teaching mode. BRIEF DESCRIPTION OF DRAWINGS DETAILED DESCRIPTION OF THE INVENTION BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The timing diagram of the digital media art teaching system constructed by fusing virtual reality technology according to the present application;

[0051] Figure 2 The working principle flow chart for the student creation state acquisition subunit;

[0052] Figure 3 The working principle flow chart for determining the teaching balance degree of the teaching balance degree judgment unit;

[0053] Figure 4 The working principle flow chart for virtual creation scene data acquisition and optimal teaching path planning generation. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0055] Please refer to Figure 1 The present application provides a digital media art teaching system constructed by fusing virtual reality technology, which comprises:

[0056] By monitoring the student creation state in real time and dynamically adjusting the teaching strategy, personalized teaching is realized. The system includes a teaching state evaluation unit, a virtual creation environment construction unit and a teaching strategy decision unit. The teaching state evaluation unit obtains student creation data through multiple student creation state acquisition sub-units, and analyzes the teaching balance state through a teaching balance degree judgment unit. The virtual creation environment construction unit decides whether to start three-dimensional scene acquisition according to the teaching balance degree judgment result, and acquires virtual creation scene data. The teaching strategy decision unit generates an optimal teaching path plan based on the collected scene data, and adjusts the virtual teaching scene parameters to optimize the teaching effect.

[0057] Embodiment 1: refer to Figure 2 In the actual deployment of the digital media art teaching system, the specific implementation mode of the student creation state acquisition sub-unit described in embodiment 1 is as follows. The student terminal interface integrates a dynamically rendered state monitoring panel, which is strictly synchronized with the display of the digital media creation software interface currently used by the student. This synchronization is achieved through the frame buffer sharing mechanism of the underlying graphics processing unit. When the student is drawing, modeling or editing animation in the main creation area, the state monitoring panel displays the same visual content in real time, including canvas size, color space, layer structure and other core parameters. The panel uses a semi-transparent overlay layer technology, and its transparency can be dynamically adjusted within the range of 20% to 80%, avoiding blocking of key content in the creation area.

[0058] The edge area of the state monitoring panel is embedded with six groups of high-sensitivity operation trajectory capture devices. Four groups of capacitive touch sensors are distributed at the four corners of the panel, which can collect the contact pressure data of the student's fingers or touch pen in real time, with a pressure sensing accuracy of 0.1 newton level. The optical gesture recognition module is integrated at the bottom center of the panel, which can capture the three-dimensional motion trajectory when the palm is suspended through the infrared laser grid. The electromagnetic induction area is set at the right edge of the panel, which can record the pen tip inclination angle and rotation azimuth data at a sampling rate of 200Hz when the student uses a digital pen for creation. The original signals generated by all capture devices are transmitted to the preprocessing module through a special data channel, and are converted into structured trajectory data packets containing timestamp, coordinate value and operation type.

[0059] The virtual teaching marker system adopts a layered rendering architecture. The base layer contains static positioning markers such as golden section line grids, perspective auxiliary frames, and other geometric reference systems. These markers are presented as semi-transparent green lines with a line width of 0.5 pixels. The dynamic layer is automatically activated according to the teaching progress. For example, when students are doing color composition exercises, a ring-shaped color wheel marker appears at the edge of the panel. The diameter of the ring dynamically scales with the frequency of palette usage. Interactive markers are represented as draggable three-dimensional controls, such as virtual light source spheres. Students can adjust their spatial position through multi-point touch control. The surface of the sphere displays the light intensity value in real time. The visibility of all markers is controlled by the teaching strategy engine. When the system detects that the students have stayed in the material mapping stage for more than a certain period of time, it automatically superimposes a UV unwinding diagram marker in the corresponding area.

[0060] The data fusion of the operation trajectory capture device uses a space-time alignment algorithm. The original data collected by each sensor is first synchronized in time. The main clock of the teaching system is used as the reference, and the alignment accuracy is controlled within 5 milliseconds. Touch pressure data and electromagnetic pen trace data are unified through a coordinate transformation matrix into the panel coordinate system, forming a three-dimensional trajectory point cloud containing X / Y coordinates, pressure values, and timestamps. Gesture recognition data is converted into control vectors of virtual joysticks through a skeletal joint mapping algorithm. The final generated operation trajectory feature vector contains 12-dimensional feature indicators such as displacement variance of operation points per unit time, pressure change gradient, and interactive event trigger interval.

[0061] The dynamic response mechanism of the virtual teaching marker includes three layers of feedback logic. The basic response layer automatically switches the marker type according to the current operation tool of the student, such as displaying three-dimensional grid density reference lines when switching to a carving knife tool. The advanced response layer analyzes the characteristics of the creation content. When it detects that the scene contains a large area of water, it automatically loads a demonstration model of Fresnel reflection principles. The abnormal response layer monitors operation trajectory abnormalities. When it detects continuous rapid erasing actions exceeding a threshold, it superimposes a historical operation playback window marker in the corresponding area. The color coding of all markers follows the teaching semantic rules. Red markers represent operation warnings, blue markers represent knowledge point prompts, and green markers indicate correct operation paths.

[0062] When a certain operation trajectory capture device continuously loses data for more than 3 seconds, the data interpolation compensation algorithm of the adjacent sensor is automatically enabled. If the panel rendering appears a delay of more than 100 milliseconds, the system immediately starts the degradation mode, switching the marker rendering precision from vector graphics to bitmap caching, maintaining the minimum teaching guidance function. The synchronization status of the panel and the creation interface is detected every 16 milliseconds. When the frame rate difference exceeds 5%, the rendering pipeline resource allocation priority is automatically adjusted.

[0063] In the three-dimensional modeling teaching scenario, when the student is modeling the head of the character, the muscle structure dissection markers appear on the right side of the state monitoring panel. The markers are displayed with an X-ray perspective effect, and the student can adjust the transparency of the markers from 0% (fully transparent) to 100% (opaque) continuously through a two-finger pinch gesture. At the same time, the operation track capture device at the edge of the panel records the time distribution of the student's observation of the dissection structure. When it is detected that the observation time in the zygomatic region is insufficient, a high-light flashing marker is automatically superimposed in that region. During the modeling process, the electromagnetic induction area continuously collects the carving knife pressure data. When the pressure value continuously exceeds the set threshold, a pressure histogram warning marker is displayed next to the tool icon.

[0064] At the first start of each day, the touch sensor reference point calibration program is automatically executed, and a nine-square calibration pattern is displayed on the panel. During the weekly deep maintenance, the distortion correction process of the optical gesture recognition module is activated, and the camera parameters are updated by recognizing ten standard gestures. All calibration data are stored in the calibration history database to establish a device state degradation model. When a downward trend in sensor accuracy is detected, a maintenance warning notification is triggered in advance.

[0065] Embodiment 2: see Figure 3 During the operation of the digital media art teaching system, the teaching balance judgment unit described in embodiment 2 realizes signal processing and balance analysis in the following way. When the student is creating a virtual sculpture in the three-dimensional modeling course, the six operation track capture devices continuously generate data streams. The touch sensors located at the four corners of the state monitoring panel record the force variation curve of the student's finger pressing, the gesture recognition module at the bottom captures the dynamic track of the student's hands pinching and scaling the model in the air, and the electromagnetic induction area on the right transmits the time sequence data of the tilt angle of the digital pen. These raw signals are transmitted to the signal preprocessing module at a rate of 200 frames per second through a dedicated data bus.

[0066] The signal mode decomposition process is executed on a dedicated hardware acceleration card. Taking the tilt angle signal of the electromagnetic pen as an example, the signal first enters a three-stage cascaded filter group. The first stage high-pass filter strips the device inherent vibration noise, the second stage band-pass filter separates the wrist tremor component, and the third stage low-pass filter extracts the overall motion trend of the arm. The decomposed feature mode components include three typical patterns: the 0.5-2Hz frequency band reflects the fine control during local detail carving, the 2-5Hz frequency band corresponds to medium-scale shaping actions, and the components above 5Hz record the burst operations of rapid tool switching. The trend component is calculated by sliding time window integration, and the average tilt azimuth angle of the pen is updated every 30 seconds.

[0067] When the characteristic modal components detect continuous high-frequency jitter (such as tremors caused by student hand fatigue), the model automatically injects a smoothing filter algorithm to eliminate abnormal fluctuations while preserving the intended operation. For touch pressure signals, when the trend component shows that the pressure value is consistently below 20% of the historical average, the correction model initiates sensitivity gain compensation to amplify the weak touch signal to a recognizable range. All correction operations are performed in a dedicated digital signal processor, with processing latency controlled within 8 milliseconds to ensure real-time performance.

[0068] The system uses 5-minute intervals as the basic teaching unit, performing three random samplings within each unit. Sampling is driven by a teaching clock, avoiding breaks and prioritizing key teaching points (such as practical sessions after teacher demonstrations). In a sculpture lesson, at the 25-minute sampling point, the system acquires corrected signal values ​​from six student terminals: terminal A's pen pressure is 0.73 N (Newtons), terminal B's gesture zoom speed is 15 mm / s, terminal C's touch trajectory curvature is 0.18, and the data from the other terminals is recorded similarly. These values ​​are then fed into the real-time calculation engine.

[0069] First, the arithmetic mean of the six signal values ​​is calculated; for example, the average pen pressure is 0.68N. Then, the standard deviation is calculated; in this sample, the standard deviation is 0.12N. The balance parameter is generated by converting the coefficient of variation (standard deviation / mean). The conversion formula sets the coefficient of variation to 0-0.3, corresponding to a balance score of 100-80, and 0.3-0.6, corresponding to 80-60. In this sample, the coefficient of variation is 0.18, resulting in a single-point balance parameter of 86. Simultaneously, an operation type consistency factor is introduced; when 80% of students are detected performing similar operations (e.g., all performing surface smoothing), the balance parameter increases by 5%.

[0070] Sampling data from three consecutive teaching units (15 minutes each) were used to form a time series. The balance score was 82 points for Unit 1 (0-5 min), 79 points for Unit 2 (5-10 min), and 85 points for Unit 3 (10-15 min). The final teaching balance score was calculated using an exponentially weighted moving average, assigning higher weight to the most recent time period. With a decay factor of 0.6, the final balance score was calculated as follows: 85 × 0.6 + 79 × 0.24 + 82 × 0.16 = 83.2 points. This value was quantified as a percentage and stored in the teaching status database.

[0071] If the teaching balance score is consistently below 75 points for the first 10 minutes, the sampling frequency is automatically increased to once per minute. In one sampling, an abnormal deviation of the character skeleton rotation angle signal from the group mean was detected on terminal D. The system immediately initiated a signal source investigation, discovering a constant deviation of 0.5 degrees in the gyroscope of the terminal's digital pen. The teaching balance judgment unit automatically marked this terminal's data as a special sample, excluding the influence of outliers when calculating the group dispersion, thus avoiding misjudgments of teaching imbalance.

[0072] The mode decomposition filter bank undergoes phase calibration monthly, and the separation accuracy of each frequency band is verified using standard test signals. The parameters of the signal correction model are updated quarterly, and new compensation coefficients are trained based on historical data. The teaching balance algorithm adjusts weight allocation each semester, optimizing conversion rules based on teacher evaluation feedback. All adjustments are completed through a remote configuration center, without interrupting on-site teaching activities.

[0073] This implementation method constructs a balanced evaluation system adapted to different teaching scenarios through multi-level signal processing and dynamic sampling mechanisms. The conversion process from the original operation signal to the teaching balance not only preserves individual operation characteristics but also effectively extracts the commonalities of group learning states, providing an objective basis for the dynamic adjustment of the virtual teaching environment.

[0074] Example 3: See Figure 4 In the 3D scene acquisition decision-making process of the digital media art teaching system, the teaching balance discrete parameter calculation and scene acquisition triggering mechanism described in Example 3 are implemented as follows: The system continuously monitors the teaching balance data stream from different student terminals. For example, in a virtual scene design course, the balance score of eight student terminals is updated every 3 minutes, with a score range of 0-100. This data is transmitted to the balance analysis engine in real time through a distributed message queue. The engine maintains the teaching balance time series of the most recent 15 minutes for each terminal.

[0075] The calculation of the difference in teaching balance uses a dynamic matching algorithm based on time alignment. When the balance sequence of terminal A is [82,85,79,83,80] and the sequence of terminal B is [78,76,81,79,77], the system first performs dynamic time warping on the two sequences to find the optimal alignment path. The aligned sequence point pairs are (82,78), (85,76), (79,81), (83,79), and (80,77). The difference value is calculated using a modified Manhattan distance formula.

[0076] ;

[0077] in: This represents the difference in balance between terminals A and B, where n is the length of the aligned sequence (n=5 in this example). and These represent the i-th balance values ​​of terminals A and B after alignment. The time decay weighting coefficient is 0.9^i (where i ranges from 0 to n-1). This formula calculates the difference between terminals A and B to be 3.72. The system performs the same calculation on all terminal combinations (in this example, C(8,2)=28 pairs) to generate a difference matrix.

[0078] The calculation of the discrete parameters for teaching equilibrium employs a multi-stage aggregation strategy. First, the difference matrix is ​​normalized so that the sum of the differences between each terminal and other terminals is 1. Then, the Frobenius norm of the matrix is ​​calculated as an initial dispersion index. Finally, a course stage adjustment factor is introduced. (Values ​​range from 0.7 to 1.3, varying linearly with the course progress), resulting in the final discrete parameters:

[0079] ;

[0080] Where: D is the discrete parameter for teaching balance, and m is the total number of student terminals. This is the normalized difference value. When this parameter is below the preset threshold of 0.35 for two consecutive detection cycles, and at least one terminal's equalization score exceeds 85, the 3D scene acquisition process is triggered.

[0081] In the virtual character design course, when the system determines that the acquisition conditions are met, it first sends an acquisition preparation command to the target terminal (the terminal with the highest balance). The 6DoF positioning system built into the VR headset of this terminal then begins to initialize the environment map. Four wide-angle RGB-D cameras simultaneously acquire scene data from different directions (front, left, upper right, and lower rear), with each camera recording 1280×720 resolution depth and color images at 30fps. The spatial registration algorithm aligns the four video streams to a unified coordinate system, generating a sparse point cloud containing 32,000 feature points.

[0082] The spatial relationship analysis of virtual teaching markers employs a hierarchical recognition strategy. The system first establishes three reference planes in the scene's baseline coordinate system: the ground plane (fitted using the RANSAC algorithm), the workbench plane (determined by the point cloud of the student's operating area), and the virtual canvas plane (calculated based on the position of the creation software interface). Then, the markers are categorized: red-framed markers are projected onto the ground plane to calculate their coverage area, while blue rotating arrow markers are analyzed for their angle with the workbench plane. At each observation pose, the system records the polar coordinates (ρ, θ, φ) of the marker's center point relative to the nearest reference point, where ρ represents the distance, θ is the horizontal angle, and φ is the vertical angle.

[0083] In a character rigging tutorial, data collected from four poses showed that, from the frontal view, the average distance between the two rotating arrow markers and the pelvic reference point was 1.2 meters, and from the upper left view, the vertical distribution angle deviation of the spinal marker chain was 4.5 degrees. This spatial relationship data was converted into 7-dimensional feature vectors and input into the path planning engine. The engine generated three candidate paths based on these feature vectors: path A prioritizes pelvic region operations, path B focuses on spinal adjustments, and path C balances the training time for each part. Each path is accompanied by a spatial complexity score (level 1-5) and an expected operation intensity coefficient (0.1-0.9).

[0084] When the data quality index (calculated based on the number of feature points and signal-to-noise ratio) for a certain camera pose falls below 0.6, backup data from adjacent time periods is automatically used for completion. When a registration error greater than 5 cm occurs during spatial registration, a rapid recalibration process based on IMU data is triggered. For temporarily unidentifiable markers, the system retains their original point cloud data and marks them as "objects to be analyzed," and performs secondary identification in subsequent analysis in conjunction with student operation logs.

[0085] By quantifying the discrete characteristics of the teaching state, a precise scene acquisition trigger criterion was established. Multi-view spatial data analysis not only captured the geometric characteristics of the teaching environment but also revealed the topological relationships between virtual markers, providing a spatial relational basis for generating teaching paths that conform to the laws of 3D art creation. The time decay weight and course stage factor introduced into the formula enable the system to adaptively adjust its sensitivity to historical data and teaching progress.

[0086] Example 4: In the teaching implementation of the virtual scene lighting design course, the teaching strategy decision-making unit operates according to the following process. After the system completes the 3D scene acquisition, the spatial distribution characteristics of the markers are parsed into structured data. Three types of key markers were detected in this course: red spotlight icons (12), blue floodlight markers (8), and green light and shadow attenuation indicators (5). Based on the 3D coordinates of these markers, the path planning engine generates three candidate teaching paths, as shown in Table 1.

[0087] Table 1: Evaluation Parameters for Candidate Teaching Paths

[0088]

[0089] Path A employs a "regional progression" strategy: first, the main light source group is deployed (4 red icons), then auxiliary light sources are added (blue markers), and finally, attenuation parameters are adjusted (green indicators). Path B implements a "type-centralized" approach: all spotlights are set up first, then the floodlights are addressed, and finally, attenuation is adjusted uniformly. Path C is designed as a "dynamic balance" approach: three types of light sources are arranged alternately, and attenuation is fine-tuned immediately after every two light settings are completed.

[0090] Data shows that in the last three lessons, the class's average completion rate for multi-step continuous operations was 83%, with the peak error rate in the parameter fine-tuning stage occurring 25 minutes after the start of the lesson. The environmental monitoring module reported real-time network latency fluctuations ranging from 120-180ms, classroom background noise levels maintained at around 65 decibels, and terminal GPU temperatures generally within the safe threshold of 72℃.

[0091] The multi-objective selection process operates within a decision matrix. Path A, with a spatial complexity level of 3 corresponding to medium navigation difficulty, has an operation intensity coefficient of 0.6 that matches the average operational ability of the class. Its 82% response matching rate indicates that the path's step settings align with students' operational habits. However, the interference sensitivity of 0.35 suggests that real-time rendering steps within the path may experience stuttering when network latency exceeds 150ms. Path B is marked as eliminated because its operation intensity coefficient of 0.8 exceeds the class's tolerance threshold (historical data shows that the error rate increases by 40% when it exceeds 0.75), and its high interference sensitivity (0.42) poses a risk in the current network environment. Path C, with its low spatial complexity (level 2) reducing cognitive load, an operation intensity of 0.5 allowing for adjustment, an 85% response matching rate reflecting that its step sequence matches the class's operational rhythm, and a low interference sensitivity of 0.28 demonstrating stability under the current environmental parameters, is selected as the optimal teaching path plan.

[0092] The system compares the optimal path with the preset teaching strategy library. The preset strategy configures the scene parameters for the 25-35 minute timeframe as follows: display 6 light source controls on the interface, material editor transparency is set to 50% by default, and error message frequency is set to once every 3 minutes. Path C requires displaying 9 light source controls during this timeframe, adjusting the material editor transparency to 70%, and reducing the error message frequency to once every 5 minutes. The deviation is calculated using a multi-dimensional vector angle method: preset strategy vector = [6, 50, 3]; path planning vector = [9, 70, 5]; deviation = arccos(dot product of two vectors / (product of magnitudes)) × 180 / π.

[0093] The calculated deviation angle reached 54°, exceeding the preset threshold of 45°. The system then performed parameter adjustments: the rendering engine expanded the light source control group to a three-column layout, the default transparency of the material editor was increased to 70%, and the error detection module extended the alarm interval to 5 minutes. At the same time, the configuration template for this course type in the preset strategy library was updated, a new "Medium Complexity Class" parameter group was added, and the adjusted scene configuration data was recorded.

[0094] During the path execution, the system continuously monitors the implementation status. When it detects that a student has completed 20% more progress than planned in the first 15 minutes, a dynamic acceleration mechanism is automatically activated: the advanced light source type, originally scheduled for the 18th minute, is introduced at the 12th minute. If three students are detected repeatedly performing actions during the shadow softening phase, an auxiliary guidance layer is immediately inserted, displaying a light scattering diagram at the edge of the interface. All real-time adjustment data is recorded in the strategy optimization log for improving path planning for similar courses in the future.

[0095] Through multi-dimensional path evaluation and dynamic strategy adjustment, an environmental parameter control mechanism adapted to real-time teaching conditions was constructed. The tabular data-driven screening process objectively reflects the technical characteristics and implementation conditions of different path schemes, while the deviation model based on vector space calculation enables quantitative management of the teaching strategy optimization process.

[0096] Example 5: In the teaching implementation of the 3D animation rendering course, the teaching dynamic compensation unit optimizes parameters through the following mechanism. When the optimal teaching path planning enters the skinning weight adjustment stage, the system starts the predictive analysis module. This module retrieves interactive response data from the same time period in the last three class periods, including twelve indicators such as vertex operation completion rate, number of weight slider adjustments, and error rollback frequency. The environmental monitor simultaneously collects classroom light intensity change curves, air conditioner vent temperature and humidity data, and network switch port traffic statistics. These historical data are processed by a time series decomposition engine to separate daily periodic fluctuations, course progress trend items, and random disturbance components.

[0097] The system generates interactive response data for predicted teaching sessions using an adaptive sliding window algorithm. Taking vertex operation completion rate as an example, the system identifies a regular pattern of declining completion rate in the class during the first 40-50 minutes of the lesson, with a decrease of approximately 15%-20% of the baseline value. Combined with the complexity coefficient of the day's course content (the current course has 1.2 times the baseline complexity), the prediction engine outputs a curve showing the future completion rate change, marking time intervals where the rate may fall below a warning value. Environmental interference prediction analyzes the air conditioning system's operating logs, revealing a periodic decrease in cooling power between 2:00 PM and 3:00 PM daily, based on which the potential increase in rendering computation latency is estimated.

[0098] The secondary calibration process of the instructional strategy decision space is executed on distributed computing nodes. The calibration engine divides the original decision space into 1024 multi-dimensional grid cells, each corresponding to an instructional strategy under a specific parameter combination. Based on prediction data, the system dynamically adjusts the weights of the grid cells: when the prediction indicates that the network latency may exceed 150ms in a certain period, the weight of strategy cells that rely on real-time data transmission is reduced; when the predicted ambient temperature exceeds 26℃, the failure probability of GPU load-sensitive strategies is increased. The calibrated decision space forms a three-dimensional probability distribution cloud map, with high-probability regions corresponding to robust instructional strategy combinations.

[0099] The construction of the teaching compensation decision feature space employs manifold learning technology. The system maps a set of teaching parameters (including interface response speed, prompt information density, and operation fault tolerance threshold) with 38 dimensions to a three-dimensional feature space through nonlinear dimensionality reduction. The first principal dimension represents the cognitive load level, determined by the complexity of interface elements and the number of operation steps. The second principal dimension reflects the strength of technical support, depending on the amount of real-time auxiliary resources invested. The third principal dimension indicates environmental adaptability, associated with the device performance buffer coefficient. Each coordinate point in the feature space corresponds to a teaching state fingerprint; for example, coordinates (0.7, -0.3, 1.2) represent a teaching scenario with moderate cognitive load, weak technical support, and strong environmental adaptability.

[0100] The parameter compensation optimization engine runs an optimization algorithm within the feature space. For each key node in the optimal teaching path planning, the system generates a compensation sphere within the feature space, with the sphere's radius dynamically shrinking as the course progresses. During the character expression binding teaching segment, if the feature coordinates (-0.2, 0.5, -0.8) are detected as falling into a high-risk area for operational errors, the engine immediately initiates a three-level compensation: The first level adjusts the teaching pace, splitting the originally planned continuous 12 expression mixing targets into three groups, inserting a 2-minute effect preview after each group's operation; the second level adds auxiliary prompts, overlaying muscle movement vector diagrams around key control points at the corners of the eyes and mouth; the third level modifies the task difficulty, changing the precise value input to interval slider control, allowing ±10% parameter deviation. After the compensation scheme is generated, it is injected into the teaching scene via a dynamic loader, with scene switching latency controlled within 300 milliseconds.

[0101] In the cloth simulation teaching unit, when a continuous negative drift of the feature space coordinates along the second dimension is detected (due to insufficient technical support), an automatic compensation scheme iteration is triggered: First, the allocation ratio of physics engine computing resources is increased, raising the number of cloth calculation iterations from the default 30 to 45; if the drift trend does not ease, a preparatory scheme is activated, adding a real-time tension heatmap to the right side of the interface; finally, the teaching assistant virtual character is activated to demonstrate the key parameter adjustment process in the scene. All compensation operations are recorded in the decision log, marking the triggering conditions, execution intensity, and duration.

[0102] A dynamic compensation mechanism for the teaching process was established through multi-level prediction and feature space mapping. From extracting patterns from historical data to analyzing real-time feature coordinates, the system develops the ability to predict changes in the teaching state. The three-level response system of the compensation strategy maintains the continuity of basic teaching while providing progressive solutions for abnormal states, enabling the virtual teaching environment to have adaptive adjustment characteristics. The dimensionality reduction representation of the decision feature space effectively simplifies the optimization process of complex parameters, while the dynamic loading mechanism ensures the seamless integration of compensation measures.

[0103] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0104] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital media art teaching system constructed by fusing virtual reality technology, characterized in that, Comprise: Teaching state evaluation unit, comprising a plurality of student creation state acquisition subunits and a teaching balance degree judgment unit, the plurality of student creation state acquisition subunits correspond to a plurality of student terminals respectively, the teaching balance degree judgment unit is used for determining teaching balance degree based on the output of the student creation state acquisition subunit; Virtual creation environment construction unit, for determining whether to start three-dimensional scene acquisition based on the teaching balance degree of each student creation state acquisition subunit, collecting virtual creation scene data when determining to start three-dimensional scene acquisition; Teaching strategy decision unit, for generating optimal teaching path planning based on the virtual creation scene data, and adjusting virtual teaching scene parameters based on the optimal teaching path planning; The student creation state acquisition subunit comprises a state monitoring panel arranged on the student terminal interface, and the state monitoring panel is synchronously displayed with the student creation interface; A plurality of operation trajectory capture devices are arranged on the state monitoring panel; A virtual teaching marker is arranged on the state monitoring panel; The teaching balance degree judgment unit determines the teaching balance degree based on the output of the student creation state acquisition subunit, comprising: For each operation trajectory capture device, the input signal of the operation trajectory capture device in a continuous teaching period is modal decomposed to generate a characteristic modal component and a trend component; A corrected continuous teaching period input signal is generated based on the characteristic modal component and the trend component through a signal correction model; The teaching balance degree is determined based on the corrected continuous teaching period input signals corresponding to a plurality of operation trajectory capture devices; The teaching balance degree judgment unit determines the teaching balance degree based on the corrected continuous teaching period input signals corresponding to a plurality of operation trajectory capture devices, comprising: The corrected input signal of each operation trajectory capture device is period sampled; Input signal values of a plurality of sampling teaching moments are obtained; Single-point teaching balance degree parameters are calculated based on the input signal values of a plurality of operation trajectory capture devices at each sampling teaching moment; The teaching balance degree is determined based on single-point teaching balance degree parameters at a plurality of sampling teaching moments.

2. The digital media art teaching system according to claim 1, wherein: The virtual creation environment construction unit determines whether to start three-dimensional scene acquisition based on the teaching balance degree, comprising: A teaching balance degree difference value of any two student creation state acquisition subunits is calculated; A teaching balance degree dispersion parameter is calculated based on all teaching balance degree difference values; When the teaching balance degree dispersion parameter is less than a preset dispersion threshold value and there is at least one teaching balance degree greater than a preset balance threshold value, it is determined to start three-dimensional scene acquisition.

3. The digital media art teaching system according to claim 1, wherein: The virtual creation environment construction unit collects virtual creation scene data, comprising collecting scene data at a plurality of observation poses; The teaching strategy decision unit generates optimal teaching path planning based on the virtual creation scene data, comprising: For each observation pose, the spatial relationship between the virtual teaching marker and the scene reference point is determined; An identifier spatial distribution feature is generated based on the spatial relationship of a plurality of observation poses; Generate an optimal teaching path plan based on the spatial distribution characteristics of the markers.

4. The digital media art teaching system of claim 3, wherein: The teaching strategy decision unit generates an optimal teaching path plan based on the spatial distribution characteristics of the markers includes: Generate a plurality of candidate teaching paths based on the spatial distribution characteristics of the markers; Obtain the interactive response data of the student terminal in the continuous teaching period; Obtain the environmental interference parameters corresponding to the teaching period; Based on the interactive response data and environmental interference parameters, the plurality of candidate teaching paths are screened to determine the optimal teaching path plan.

5. The digital media arts teaching system of claim 4, wherein, Also includes: A teaching dynamic compensation unit for generating a teaching parameter compensation scheme based on the interactive response data, environmental interference parameters and optimal teaching path plan.

6. The digital media art teaching system of claim 5, wherein: The teaching strategy decision unit adjusts the virtual teaching scene parameters based on the optimal teaching path plan includes: Generate a preset teaching strategy based on historical teaching records, which contains time-periodic scene parameter configuration; When the deviation degree of the optimal teaching path plan and the corresponding period preset teaching strategy exceeds the preset deviation threshold, adjust the virtual teaching scene parameters according to the optimal teaching path plan and update the preset teaching strategy.

7. The digital media art teaching system of claim 6, wherein: The teaching dynamic compensation unit generates a teaching parameter compensation scheme includes: Based on the predicted interactive response data and predicted environmental interference parameters of the predicted teaching period, the teaching strategy decision space is calibrated again; Generate a teaching compensation decision feature space; Based on the teaching compensation decision feature space, the optimal teaching path plan is optimized for parameter compensation.

Images