Digital interactive exhibition method and system based on VR technology

Abstract

The invention discloses a digital interactive exhibition method and system based on a VR technology, and the method comprises the steps: generating a virtual exhibition hall topology with a curvature attribute based on a nonlinear manifold learning algorithm, and enabling the virtual exhibition hall topology to comprise exhibit node clusters which can be dynamically recombined, and an elastic path network which is connected with the exhibit node clusters, the exhibit node clusters preload spatial semantic codes; the method comprises the following steps: capturing pose transformation data and biological characteristic signals of a user in a virtual exhibition hall topology in real time through an embedded sensor array of head-mounted display equipment, and mapping the pose transformation data and the biological characteristic signals to a topology coordinate system of the virtual exhibition hall topology by utilizing a space-time alignment module; according to the method, multi-modal sensory splitting is eliminated through a curvature manifold space registration engine, cognitive loads are dynamically optimized in combination with a neurocognitive topological regulation mechanism, and a physical space and physiological response self-adaptive intelligent system is constructed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital interactive exhibition, and particularly relates to a digital interactive exhibition method and system based on VR technology. BACKGROUND

[0002] With the development of VR technology, users can feel all aspects of exhibits without obstacles and more intuitively, but the current virtual reality digital exhibition system has technical bottlenecks, mainly embodied in the following two aspects: The traditional architecture adopts a rigid Euclidean space mapping mechanism, which causes the visual focus, tactile feedback and spatial sound field to be difficult to accurately match, especially in a dynamic environment reorganization scene (such as path network topological splitting), the spatial reference of each sensory channel produces systematic deviation, causing user perception to be fragmented and dizzy, and the existing technology lacks a unified mathematical framework for cross-modal space registration, and cannot guarantee the spatial consistency of multi-sensory feedback.

[0003] The fixed exhibit layout and information delivery strategy cannot adapt to the real-time cognitive state of users, when high-density crowds gather or complex knowledge paths are activated, the user's neurophysiological indicators show significant load characteristics (such as abnormal fluctuations in brain rhythm), and the static system cannot perceive changes in cognitive state, and lacks the ability to dynamically optimize topology based on physiological signals, resulting in a high rate of interruption of the visit experience. SUMMARY

[0004] In order to solve the above problems, the embodiments of the present application provide a digital interactive exhibition method based on VR technology, the method comprising: Generating a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, the virtual exhibition hall topology comprising a dynamically reconfigurable exhibit node cluster and an elastic path network connecting the exhibit node cluster, the exhibit node cluster being preloaded with spatial semantic encoding; Real-time capturing of user pose transformation data and biological feature signals in the virtual exhibition hall topology through an embedded sensor array of a head-mounted display device, and mapping the pose transformation data and biological feature signals to a topological coordinate system of the virtual exhibition hall topology using a space-time alignment module; According to the user's current topological region coordinates and gaze focus coordinates in the virtual exhibition hall topology, activating the spatial semantic encoding of the corresponding exhibit node cluster, and generating a cognitive intention vector by fusing the biological feature signals through a semantic association strength matrix; Retrieving associated attributes in a knowledge graph based on the cognitive intention vector, converting physical characteristic parameters of virtual exhibits into tactile rendering instructions through spatial registration mapping, driving a distributed tactile array to generate a force field gradient matching the current position in the virtual exhibition hall topology, and generating spatial audio matching the current position through sound field beamforming technology; Continuously collect the topological behavior trajectory and neurophysiological indicators of the user in the elastic path network, construct a cognitive load and spatial position joint probability model, and dynamically adjust the distribution density and information presentation dimension of the exhibit node cluster.

[0005] Further, the nonlinear manifold learning algorithm adopts a hybrid model of isometric feature mapping and local linear embedding, which includes: The distribution density of the exhibit node cluster is dynamically adjusted by user historical visit data; The adaptive path network is automatically split or merged according to real-time crowd density.

[0006] Further, the spatio-temporal alignment module includes a quantum particle swarm optimization algorithm for solving the timestamp drift problem of multi-modal data, and the execution method includes: Convert the eye focus coordinates into azimuth and elevation angles in the spherical coordinate system; Map the gesture space vector to the tangent space of the virtual scene through Lie group transformation.

[0007] Further, the semantic association strength matrix is generated by pre-training with a graph convolution network, and the knowledge dynamic visualization method includes: When the user's gaze duration on a certain exhibit node is detected to exceed the threshold, activate the k-hop associated path of the node in the knowledge graph; According to the attention weight distribution, the associated path is rendered with a transparency gradient.

[0008] Further, the generation method of the force field gradient includes: Extract the surface texture feature Fourier descriptor of the virtual exhibit; Convert the descriptor into a sequence of haptic excitation signals through a waveform superposition algorithm.

[0009] Further, the sound field beamforming technology adopts a deformable microphone array, and the implementation method includes: According to the user's head azimuth, dynamically adjust the direction of the main lobe of the beam; when detecting that the number of users is greater than 1, automatically generate non-overlapping independent sound zones.

[0010] Further, the neurophysiological indicators include the θ / β band power ratio of the brain waves, and the state space definition of the cognitive load and spatial position joint probability model includes: State: User cognitive load index = f(pupil diameter change rate, interaction action frequency); Action: Discretization level set of exhibit information granularity; Reward function: The product of user stay time and knowledge absorption efficiency.

[0011] Further, when the cognitive load index exceeds the adaptive threshold, the execution method includes: Splitting the current exhibit node cluster into a sub-topology structure; Inserting a buffer node in the path network and reducing the information transmission rate.

[0012] Further, the method further comprises a virtual-real space error compensation mechanism: By data fusion of the inertial measurement unit and the external optical tracking system, a Kalman filter of the user kinematics model is constructed; When the residual error between the physical displacement and the virtual pose exceeds the tolerance, the elastic deformation compensation of the topological structure is triggered.

[0013] On the other hand, the application also provides a digital interactive exhibition system based on VR technology, which comprises: A topological generation module generates a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, the virtual exhibition hall topology comprises dynamically reconfigurable exhibit node clusters and an elastic path network connecting the exhibit node clusters, and the exhibit node clusters are preloaded with spatial semantic encoding; A sensing alignment module captures user pose transformation data and biological feature signals in the virtual exhibition hall topology in real time through an embedded sensor array of a head-mounted display device, and maps the pose transformation data and biological feature signals to a topological coordinate system of the virtual exhibition hall topology using a space-time alignment module; An intention analysis module activates the spatial semantic encoding of the corresponding exhibit node cluster according to the user's current topological area coordinates and gaze focus coordinates in the virtual exhibition hall topology, and generates a cognitive intention vector by fusing the biological feature signals through a semantic correlation strength matrix; A multi-sense feedback module retrieves associated attributes in a knowledge graph based on the cognitive intention vector, converts the physical characteristic parameters of virtual exhibits into haptic rendering instructions through spatial registration mapping, drives a distributed haptic array to generate a force field gradient matching the current position in the virtual exhibition hall topology, and generates spatial audio matching the current position through sound field beamforming technology; A dynamic tuning module continuously collects topological behavior trajectories and neurophysiological indicators of the user in the elastic path network, constructs a cognitive load and spatial position joint probability model, and dynamically adjusts the distribution density and information presentation dimension of the exhibit node cluster.

[0014] The digital interactive exhibition method based on VR technology provided by the application has the following technical effects and advantages: The application eliminates multimodal sensory fragmentation through a curvature manifold space registration engine, dynamically optimizes cognitive load by combining neural cognitive topology regulation mechanisms, and constructs an intelligent system that is self-adaptive to physical space and physiological response. Based on the dynamic registration mechanism of differential manifold, the application real-time aligns visual focus, tactile feedback and spatial sound field in the environment reorganization scene, solves the dizziness risk caused by sensory deviation; the neurophysiological driven topology regulator triggers path network splitting to reduce information density or merges nodes to improve efficiency according to real-time cognitive state, and realizes the dynamic balance of cognitive load and information architecture. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flow chart of a digital interactive exhibition method based on VR technology in embodiment one; Figure 2 A flow chart of a digital interactive exhibition method based on VR technology in embodiment two; Figure 3 A connection schematic diagram of a digital interactive exhibition system based on VR technology in embodiment three. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0017] Embodiment one: Please refer to Figure 1 The embodiment of the application provides a digital interactive exhibition method based on VR technology, and the method comprises the following steps: S1. Generating a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, wherein the virtual exhibition hall topology comprises dynamically reorganizable exhibit node clusters and an elastic path network connecting the exhibit node clusters, and the exhibit node clusters are preloaded with spatial semantic encoding; S2. Real-time capturing user pose transformation data and biological feature signals in the virtual exhibition hall topology through an embedded sensor array of a head-mounted display device, and mapping the pose transformation data and biological feature signals to a topology coordinate system of the virtual exhibition hall topology by using a space-time alignment module; S3. According to the topology area coordinates and gaze focus coordinates of the user in the virtual exhibition hall topology, activating the spatial semantic encoding of the corresponding exhibit node cluster, and generating a cognitive intention vector by fusing the biological feature signals through a semantic correlation strength matrix; S4. Based on the cognitive intention vector, the associated attributes in the knowledge graph are retrieved, the physical characteristic parameters of the virtual exhibit are converted into haptic rendering instructions through spatial registration mapping, the distributed haptic array is driven to generate a force field gradient matching the current position in the virtual exhibition hall topology, and spatial audio matching the current position is generated through sound field beamforming technology; S5. Continuously collect the topological behavior trajectory and neurophysiological indicators of the user in the elastic path network, construct a cognitive load and spatial position joint probability model, and dynamically adjust the distribution density and information presentation dimension of the exhibit node cluster.

[0018] The nonlinear manifold learning algorithm of step S1 adopts a hybrid model of Isometric Feature Mapping (Isomap) and Local Linear Embedding (LLE), which includes: the distribution density of the exhibit node cluster is dynamically adjusted by user historical visit data; the adaptive path network is automatically split or merged according to real-time crowd density.

[0019] In constructing the virtual exhibition hall topology (step S1), a nonlinear manifold learning algorithm is used to model the spatial relationship of exhibits. In specific implementation, a hybrid model of Isometric Feature Mapping (Isomap) and Local Linear Embedding (LLE) is used to generate the topology, including: The Isomap algorithm is used to calculate the geodesic distance (non-straight distance) between exhibits based on content theme similarity, ensuring that exhibits with strong theme relevance are located adjacent in the topology space, for example, Renaissance paintings and contemporary sculptures form a natural cluster.

[0020] The LLE algorithm is used to reduce the dimension of the internal structure of the exhibit node cluster, so that the spatial arrangement of the exhibits within the same cluster conforms to their internal logic such as style evolution or time sequence, for example, the early to late works of a painter are arranged in a local linear path according to the creation year.

[0021] The hybrid model supports a dynamic spatial reorganization mechanism, including: Based on user historical visit data (such as average stay time and interaction frequency in a specific exhibition area), the node distribution density of high-frequency attention themes is automatically increased, for example, when it is monitored that the daily access volume of a certain art school exhibition area is continuously higher than the set threshold, the distance between nodes in this area is reduced by 30%, and the information presentation hierarchy is expanded (such as adding a video layer of the artist's life).

[0022] The path network triggers topology deformation based on real-time crowd density data; when the number of users on a path segment exceeds the carrying capacity limit (such as the presence of 8 virtual avatars at the same time), the path is automatically split into two parallel sub-paths (such as dividing the "landscape and portrait" branch according to the theme of the works); when the number of users on adjacent paths is sparse (such as the number of users <2 for 5 consecutive minutes), the paths are merged to simplify the navigation structure.

[0023] This dynamic topology can significantly reduce the cognitive load, and in the peak period of the museum opening day, reduce the flow density of the main hall through path splitting, and shorten the waiting time for interacting with popular exhibits by increasing the node density.

[0024] The spatio-temporal alignment module of step S2 includes a quantum particle swarm optimization (QPSO) algorithm for solving the timestamp drift problem of multi-modal data, and the specific execution method includes: Convert the eye focus coordinates into azimuth / elevation angles in the spherical coordinate system; Map the gesture space vector to the tangent space of the virtual scene through Lie group transformation.

[0025] In the real-time data mapping stage (step S2), the spatio-temporal alignment module solves the time synchronization problem of multi-source sensors through the quantum particle swarm optimization (QPSO) algorithm. Due to the clock drift of the embedded sensors (eye tracker, inertial measurement unit, bioelectric sensor) in the head-mounted device, the pose data and the biological signal timestamp are misaligned. The spatio-temporal alignment module performs the following core operations: The original eye focus coordinates (screen pixel coordinates [X, Y]) are first converted into angle parameters in the spherical coordinate system, including azimuth (φ) and elevation (θ); Azimuth (φ): represents the deflection angle of the line of sight in the horizontal direction (range -180° to 180°); Elevation (θ): represents the inclination angle of the line of sight in the vertical direction (range -90° to 90°); Example: When the user gazes at the virtual sculpture in the upper right corner of the exhibition hall, the system detects φ=65° and θ=22°.

[0026] Align the time drift data using the quantum particle swarm optimization (QPSO) algorithm, including: Treat each sensor data stream as a "particle swarm", and define the timestamp offset as the particle position. Through the quantum behavior model, the optimal time offset parameter is efficiently searched in the solution space. For example, when the gesture data is delayed by 120ms compared to the eye movement data, the QPSO finds the compensation value within 3 iterations, reducing the time difference between the two to <5ms.

[0027] Map the space vector of the gesture action (composed of hand joint rotation matrices) to the local tangent space of the virtual scene through Lie group transformation, including: Construct the Lie algebra representation of the rotation group and the Euclidean group; project the gesture motion trajectory onto the tangent plane of the virtual object surface using exponential mapping. For example, when the user stretches his hand to touch the virtual vase, the motion trajectory of his index finger is mapped as a tangent vector field on the surface of the vase.

[0028] The spatio-temporal alignment module can significantly improve the interaction authenticity. In the scenario where the user simultaneously performs the gaze and grabbing actions, the spatio-temporal alignment error can be effectively reduced, thereby improving the synchronization rate of the tactile feedback and the visual focus.

[0029] The semantic correlation strength matrix of step S3 is generated by pre-training of a graph convolution network (GCN), and the knowledge dynamic visualization method comprises the following steps: When it is detected that the gaze duration of the user on a certain exhibit node exceeds the gaze duration threshold, the k-hop correlation path of the node in the knowledge graph is activated; and the correlation path is rendered in transparency gradient according to the attention weight distribution.

[0030] In the cognitive intention analysis stage (step S3), the semantic correlation strength matrix pre-trained by the graph convolution network (GCN) is used to realize knowledge dynamic visualization, and the specific execution process is as follows: In the pre-training stage, the exhibit attributes (such as the creation year, art school, and material type) are constructed as knowledge graph nodes; the neighbor node information is aggregated by a multi-layer GCN to generate a weight matrix representing the semantic correlation strength between the exhibits; for example, in the art exhibition scenario, Monet's “Water Lilies” has strong correlation edges (weight > 0.85) with nodes such as the Impressionist movement and outdoor sketching techniques.

[0031] The gaze duration of the user on the virtual exhibit node is monitored in real time, and when the gaze duration exceeds the dynamic gaze duration threshold (for example, the basic gaze duration threshold is 2 seconds, which is adaptively adjusted according to the user's cognitive load): The k-hop correlation path is activated, and the k-layer associated entities are retrieved and diffused along the knowledge graph edges from the current exhibit node; The transparency gradient control is implemented on the correlation path according to the attention weight distribution output by the GCN; The high-weight correlation edges (such as direct mentorship) are visualized with opacity 100%; The low-weight correlation edges (such as indirect cultural influence) are semi-transparently faded with opacity 30%; Implementation scenario example: When the user continuously gazes at the Renaissance painting “Academy in Athens” for more than the threshold value: The system activates the 3-hop correlation path (k=3) and presents: Direct correlation layer (1-hop): author Raphael, contemporary painter Michelangelo (opacity 100%); Indirect correlation layer (2-hop): ancient Greek philosophical thought, perspective drawing innovation (opacity 60%); Cultural influence layer (3-hop): Neoclassical movement (opacity 30%); The associated entities are suspended around the virtual exhibit according to the weight gradient, forming a spatio-temporal knowledge network.

[0032] The dynamic visualization method improves the information density controllable test data display user's understanding efficiency of core associated information, and reduces the cognitive load index.

[0033] The distributed tactile array of step S4 includes a piezoelectric ceramic micro-brake matrix, and the generation method of the force field gradient includes: The surface texture feature Fourier descriptor of the virtual exhibit is extracted; The descriptor is converted into a tactile excitation signal sequence through a waveform superposition algorithm.

[0034] In the tactile feedback generation stage (step S4), the distributed tactile array realizes force field gradient regulation through the piezoelectric ceramic micro-brake matrix, and the specific execution process is as follows: The Fourier descriptor of the virtual exhibit surface geometry is extracted, the three-dimensional model surface is discretized into a point cloud grid, and the spatial coordinate sequence is converted into a frequency domain feature vector through fast Fourier transform (FFT); for example, the concave-convex texture of the baroque style relief is characterized as a feature harmonic component in the frequency band of 0.5-120Hz.

[0035] The frequency domain features are converted into tactile excitation signals using a waveform superposition algorithm, including: The amplitude spectrum of the Fourier descriptor is mapped to the micro-brake amplitude; the phase spectrum is mapped to the excitation timing of the piezoelectric ceramic unit; and a spatial gradient force field is generated through wave field interference principle.

[0036] For example: the silk material virtual exhibit generates a main vibration mode with an amplitude of 0.2N and a frequency of 80Hz, while the stone carving surface generates an overlay waveform with an amplitude of 1.5N and a frequency of 25Hz.

[0037] The excitation signal distribution is dynamically adjusted according to the user gesture contact position, including: The high-frequency component (>50Hz) is activated in the direct contact area of the fingertip to transfer texture details; The low-frequency component (<10Hz) is enhanced in the palm covering area to simulate the weight feeling of the material; For example, when the user touches the Ming Dynasty blue and white porcelain virtual exhibit: The fingertip contact pattern triggers 200Hz micro-vibration to simulate the concave-convex of the glaze; The palm center generates a 5Hz slowly varying wave field to simulate the thermal conductivity of the ceramic; The bottle mouth and the bottle bottom form an amplitude gradient difference (top 0.8N→bottom 1.6N) to strengthen the center of gravity perception.

[0038] The method improves the accuracy of tactile recognition while controlling the force feedback delay within the tactile perception threshold (<10ms).

[0039] The sound field beamforming adopts a deformable microphone array, and the implementation method includes: The beam main lobe direction is dynamically adjusted according to the user's head azimuth angle; when it is detected that the number of users gathered is greater than 1, non-overlapping independent sound zones are automatically generated.

[0040] In the sound field regulation link (biological feature signal in step S3), the deformable microphone array realizes sound field spatial orientation through beamforming technology, and the specific method includes: Real-time analysis of user head azimuth angle (accuracy ± 3°), including: The infrared depth sensor tracks the temporal bone position; The head deflection angle is calculated in combination with IMU data; The sound wave main lobe direction is dynamically adjusted to keep the beam center axis always aligned with the user's ear canal: Example: When the user turns left by 30° to watch the mural, the beam main lobe is synchronously deflected by 28°-32° (phase modulation delay <5ms).

[0041] Multi-user sound zone isolation, including: When it is detected that the distance between users is less than the critical distance of sound interference (such as 1.2m): Calculate the Voronoi partition of the spatial coordinates of each user's head; Generate non-overlapping independent sound zones for each user (beam sidelobe suppression >20dB); Implementation scenario example: Two users (user A and user B) stand side by side (distance about 0.8 meters, less than the preset 1.2 meter critical distance of sound interference), and jointly observe a fine bronze ding in the showcase at close range.

[0042] Head tracking and main lobe dynamic alignment: The infrared depth sensor of the deformable microphone array deployed above the showcase captures the temporal bone positions of the two users in real time; In combination with the head posture data provided by the lightweight IMU (such as integrated in smart glasses or guide earphones) worn by the user, the system accurately calculates that user A's head is currently slightly tilted about 15° to the left to observe the details of the ding body decoration, and user B's head remains straight but his gaze is focused on the connection between the ding feet; According to the calculated azimuth angle (accuracy ± 3°), the phase delay of the array is dynamically adjusted (delay <5ms), so that the two main lobes generated by beamforming have the following directions: The main lobe center axis of user A accurately points to his left ear canal (since the head is tilted to the left, the main lobe needs to be synchronously tilted to the left by about 13°-17°).

[0043] The main lobe center axis of user B is accurately directed to his right ear canal (because the head is straight, the main lobe direction basically remains the original setting).

[0044] Multi-user sound zone isolation generation: The system detects that the distance between users is less than 1.2 meters, and automatically triggers multi-user sound zone management.

[0045] Based on the real-time head space coordinates (3D position) of user A and user B, the system performs Voronoi space partitioning.

[0046] The deformable array generates a highly directional, spatially non-overlapping independent sound zone for each user according to the partitioning results, and the array form may be slightly adjusted to optimize beam directivity.

[0047] Through beamforming algorithm, the sidelobe energy directed to the main lobe of user A (especially towards user B) is suppressed by more than 20dB, while the sidelobe energy directed to the main lobe of user B (towards user A) is also suppressed by more than 20dB, effectively avoiding mutual interference between the two sound zones.

[0048] In the independent sound zone of user A, clear and stable reception of detailed explanation of the symbolic meaning of the bronze ding surface decoration is obtained, for example, "the body of the ding symbolizes communication between heaven and earth, the cloud and thunder pattern represents eternal life... ", the sound energy is mainly concentrated in the left ear area, and the sound pressure level is stable at about 65dB.

[0049] In the independent sound zone of user B, synchronous reception of in-depth explanation of the bronze ding casting process is obtained, for example, "this ding is cast by lost wax method, with visible mold line traces, and the foot is cast separately and then welded... ", the sound energy is mainly concentrated in the right ear area, and the sound pressure level is stable at about 63dB.

[0050] This method improves the signal-to-noise ratio of the target sound zone and controls the sound field reconstruction delay within 10ms (below the perceptual threshold of human ear).

[0051] The neurophysiological indicators of step S5 include the θ / β frequency band power ratio of brain waves (EEG), and the state space definition of the cognitive load and spatial position joint probability model includes: State: User cognitive load index = f(pupil diameter change rate, interaction action frequency); Action: Discrete level set of exhibit information granularity; Reward function: product of user stay time and knowledge absorption efficiency.

[0052] In the adaptive adjustment phase (step S5), the system uses the neurophysiological indicators for real-time monitoring and dynamically optimizes the user experience using the cognitive load and spatial position joint probability model. The specific implementation mechanism is as follows: Continuous acquisition of the power ratio of the theta / beta band of the user's brain waves (EEG) as the core physiological indicator: The increase in theta wave (4-8 Hz) power represents cognitive resource tension; The increase in beta wave (13-30 Hz) power reflects focused information processing; Example: When the user tries to understand an abstractist painting, the theta / beta ratio increases from the baseline value of 1.2 to 2.5, indicating cognitive overload.

[0053] Fusion of behavioral indicators to construct the cognitive load index (CLI): Among them, the change rate of pupil diameter reflects the degree of cognitive effort, the frequency of interactive actions indicates the operation burden, and is the coefficient.

[0054] Reinforcement learning control, including: State space: Take CLI as the environmental state representation (e.g. CLI=0.7 represents moderate load); Action space: Select adjustment strategies from a set of discrete levels of exhibit information granularity, including: Text summary intensity (3 levels: 1 detailed, 2 moderate, and 3 concise); Associated path expansion depth (k=1 to 4 layers); Tactile feedback intensity (5 adjustable levels); Reward function including: ; In the formula, is the user's stay time, is the knowledge absorption efficiency (evaluation score increment rate).

[0055] Implementation scenario example: When the user's CLI abnormally increases (theta / beta=2.1) is detected in the bronze ware exhibition: Model action: Reduce the inscriptions interpretation granularity from detailed to concise (reduce 45% of the text amount); Synchronize the associated path to k=1 layer (only keep the direct historical background); After adjustment, the theta / beta ratio falls to 1.3 within 8 seconds, and the reward value increases by 120%.

[0056] When the cognitive load index exceeds the adaptive threshold, the method includes: Split the current exhibit node cluster into a sub-topology structure; insert buffer nodes in the path network and reduce the information transmission rate.

[0057] When the reinforcement learning model detects that the user cognitive load index (CLI) exceeds the adaptive threshold (e.g., CLI>0.8), trigger the topology reconstruction mechanism to reduce information complexity, including: Disassemble the current associated exhibit aggregation cluster into a sub-topological structure, including: The closely connected exhibit group in the original knowledge graph (such as the "Renaissance Trio" cluster); Split into independent sub-groups according to historical periods or art genres (Da Vinci / Michelangelo / Raphael sub-clusters); Example: When the user CLI rises to 0.92, the Baroque art cluster is split into "Rubens School", "Caravaggio School", and "Bernini Sculpture" three sub-topologies.

[0058] Insert buffer nodes dynamically in the associated path network, including: Add historical background explanation nodes (such as "The Impact of the Reformation on Art"); Embed interactive question and answer modules (such as "Please select the direction you want to delve into: technique / history / genre"); Synchronize the information transmission rate to 30%-60% of the base value, the text push speed from 5 lines / second to 2 lines / second, and the tactile signal update frequency from 100Hz to 45Hz.

[0059] Example of implementation scenario: When the user is exploring the "Impressionist Development Context" and the CLI continues to be higher than the adaptive threshold (detection θ / β=2.3): Split the 12-node cluster of Monet, Renoir, etc. into: Sub-cluster A: Early Impressionism (1874-1886) (4 nodes); Sub-cluster B: Post-1886 Neo-Impressionism (5 nodes); Sub-cluster C: Influences of Japanese Ukiyo-e (3 nodes); Insert the "Pointillism Technique Evolution" buffer node between sub-clusters, reduce information flow rate by 50% (video commentary frame rate from 30fps to 15fps), and after reconstruction, the user CLI falls to 0.65 within 10 seconds.

[0060] This mechanism can improve information absorption rate in high-load scenarios, while reducing topology entropy (complexity index).

[0061] Example Two: As Figure 2As shown, the embodiment is further improved on the basis of embodiment one, except that in the actual operation of embodiment one, it is found that there is a cumulative error between the virtual pose and the physical space in the fast moving scene (such as IMU drift up to 0.15m when continuously turning in the arc-shaped exhibition hall), and the spatial misplacement leads to distorted cognitive intention analysis (such as the system misactivating the semantic coding of the calligraphy and painting area when the user actually observes the bronze ware), resulting in spatial mismatch between the tactile feedback and the visual focus (such as the user touching the virtual showcase but receiving force feedback from the empty area), and failing to maintain the consistency of the virtual and real space experience. Based on this, the digital interactive exhibition method based on VR technology further includes a virtual and real space error compensation mechanism: A Kalman filter of the user kinematics model is constructed through data fusion of the inertial measurement unit (IMU) and the external optical tracking system. When the residual error between the physical displacement and the virtual pose exceeds the tolerance, the elastic deformation compensation of the topological structure is triggered.

[0062] In the virtual and real space alignment link, the error compensation mechanism ensures spatial consistency through multi-source sensor fusion: The Kalman filter dynamically optimizes the six-degree-of-freedom pose (position + quaternion attitude) of the user by fusing the inertial measurement unit (IMU) angular velocity data (±0.5° / s accuracy) and the external optical tracking system position information (±2mm resolution), including: ; Among them, is the optimal pose estimation at time k, is the state transition matrix, is the Kalman gain matrix, is the observation matrix, is the predicted state, is the optical observation value.

[0063] When the residual error between the physical displacement and the virtual pose exceeds the tolerance (such as >0.1m), including: Calculate the deformation gradient field of the exhibit topological network; implement elastic deformation compensation along the user motion direction; For example: when the user actually moves 0.15m to the right but the virtual pose lags behind by 0.12m: Shift the nodes of the three bronze wares in front of the field of view by 8° of visual angle towards the center; Bend the associated knowledge path to adapt to the new spatial coordinates; The residual error after compensation is reduced to 0.02m (below the minimum resolution threshold of the human eye).

[0064] Implementation scenario example: when walking quickly in the circular exhibition hall: IMU detects 2.3m / s 2 Acceleration mutation; The optical system feedback displacement hysteresis is 0.18m (exceeding the 0.1m tolerance); The topology is stretched by 12% in the movement direction in the virtual distance; The virtual projection of the bronze ware decoration is real-time attached to the physical exhibition cabinet position; Embodiment three: As Figure 3 shown, based on the same inventive concept as the digital interactive exhibition method based on VR technology in the foregoing embodiments, the present application provides a digital interactive exhibition system based on VR technology, and the system and method embodiments in the present application are based on the same inventive concept. Among them, the system comprises: A topology generation module, which generates a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, the virtual exhibition hall topology comprising a dynamically reconfigurable exhibit node cluster and an elastic path network connecting the exhibit node cluster, the exhibit node cluster being preloaded with spatial semantic encoding; A sensing alignment module, which captures user pose transformation data and biological feature signals in the virtual exhibition hall topology in real time through an embedded sensor array of a head-mounted display device, and maps the pose transformation data and biological feature signals to a topology coordinate system of the virtual exhibition hall topology using a space-time alignment module; An intention analysis module, which activates the spatial semantic encoding of the corresponding exhibit node cluster according to the user's current topology area coordinates and gaze focus coordinates in the virtual exhibition hall topology, and generates a cognitive intention vector by fusing the biological feature signals through a semantic association strength matrix; A multi-sense feedback module, which retrieves associated attributes in a knowledge graph based on the cognitive intention vector, converts the physical characteristic parameters of virtual exhibits into haptic rendering instructions through spatial registration mapping, drives a distributed haptic array to generate a force field gradient matching the current position in the virtual exhibition hall topology, and generates spatial audio matching the current position through sound field beamforming technology.

[0065] A dynamic tuning module, which continuously collects user topology behavior trajectories and neurophysiological indicators in the elastic path network, constructs a cognitive load and spatial position joint probability model, and dynamically adjusts the distribution density and information presentation dimension of the exhibit node cluster.

[0066] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

[0067] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A digital interactive exhibition method based on VR technology, characterized in that, The method comprises: Generating a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, the virtual exhibition hall topology comprising dynamically reconfigurable exhibit node clusters and an elastic path network connecting the exhibit node clusters, the exhibit node clusters being preloaded with spatial semantic encoding; Real-time capturing of user pose transformation data and biometric signals in the virtual exhibition hall topology through an embedded sensor array of a head-mounted display device, and mapping the pose transformation data and biometric signals to a topological coordinate system of the virtual exhibition hall topology using a spatio-temporal alignment module; According to the user's current topological region coordinates and gaze focus coordinates in the virtual exhibition hall topology, activating the spatial semantic encoding of the corresponding exhibit node cluster, and generating a cognitive intention vector by fusing the biometric signals through a semantic association strength matrix; Retrieving associated attributes in the knowledge graph based on the cognitive intention vector, converting the physical characteristic parameters of the virtual exhibits into haptic rendering instructions through spatial registration mapping, driving a distributed haptic array to generate a force field gradient matching the current position in the virtual exhibition hall topology, and generating spatial audio matching the current position through sound field beamforming technology; Continuously collecting topological behavior trajectories and neurophysiological indicators of the user in the elastic path network, constructing a cognitive load and spatial position joint probability model, and dynamically adjusting the distribution density and information presentation dimension of the exhibit node clusters. 2.The digital interactive exhibition method based on VR technology of claim 1, wherein, The nonlinear manifold learning algorithm uses a hybrid model of isometric feature mapping and local linear embedding, which includes: The distribution density of the exhibit node cluster is dynamically adjusted by user historical visit data; The adaptive path network automatically splits or merges according to real-time crowd density. 3.The digital interactive exhibition method based on VR technology of claim 1, wherein, The spatio-temporal alignment module includes a quantum particle swarm optimization algorithm to solve the timestamp drift problem of multi-modal data, and the execution method includes: Converting the eye focus coordinates to azimuth and elevation angles in a spherical coordinate system; Mapping the gesture space vector to the tangent space of the virtual scene through Lie group transformation. 4.The digital interactive exhibition method based on VR technology of claim 1, wherein, The semantic association strength matrix is generated by pre-training with a graph convolution network, and the knowledge dynamic visualization method includes: When the user's gaze duration on a certain exhibit node is detected to exceed a threshold, the k-hop associated path of the node in the knowledge graph is activated; According to the attention weight distribution, the associated path is rendered with a transparency gradient.

5. The digital interactive exhibition method based on VR technology according to claim 1, characterized in that, The generation method of the force field gradient includes: Extracting the surface texture feature Fourier descriptor of the virtual exhibit; Converting the descriptor into a haptic excitation signal sequence through a waveform superposition algorithm. 6.The digital interactive exhibition method based on VR technology of claim 1, wherein, The sound field beamforming technology uses a deformable microphone array, and the implementation method includes: Dynamically adjusting the beam lobe direction according to the user's head azimuth angle; when the number of users detected is greater than 1, automatically generating non-overlapping independent sound zones. 7.The digital interactive exhibition method based on VR technology of claim 1, wherein, The neurophysiological indicators include the θ / β band power ratio of the brain waves, and the state space definition of the cognitive load and spatial position joint probability model includes: State: user cognitive load index = f(pupil diameter change rate, interaction action frequency); Action: a set of discrete levels of exhibit information granularity; Reward function: the product of user dwell time and knowledge absorption efficiency. 8.The digital interactive exhibition method based on VR technology of claim 7, wherein, When the cognitive load index exceeds the adaptive threshold, the execution method includes: Splitting the current exhibit node cluster into a sub-topology structure; Inserting a buffer node in the path network and reducing the information transmission rate. 9.The digital interactive exhibition method based on VR technology of claim 1, wherein, The method further includes a virtual-real space error compensation mechanism: A Kalman filter of the user kinematics model is constructed through data fusion of the inertial measurement unit and the external optical tracking system; When the residual error between the physical displacement and the virtual pose exceeds the tolerance, the elastic deformation compensation of the topology structure is triggered.

10. A digital interactive exhibition system based on VR technology, characterized in that, The system comprises: A topology generation module that generates a virtual exhibition hall topology with curvature attributes based on a nonlinear manifold learning algorithm, the virtual exhibition hall topology including dynamically reconfigurable exhibit node clusters and an elastic path network connecting the exhibit node clusters, the exhibit node clusters being preloaded with spatial semantic encoding; A sensory alignment module that captures pose transformation data and biological feature signals of a user in the virtual exhibition hall topology in real time through an embedded sensor array of a head-mounted display device, and maps the pose transformation data and biological feature signals to a topology coordinate system of the virtual exhibition hall topology using a space-time alignment module; An intention analysis module that activates the spatial semantic encoding of the corresponding exhibit node cluster according to the topology region coordinates and the gaze focus coordinates of the user in the virtual exhibition hall topology, and generates a cognitive intention vector by fusing the biological feature signals through a semantic correlation strength matrix; A multi-sense feedback module that retrieves associated attributes in a knowledge graph based on the cognitive intention vector, converts the physical characteristic parameters of virtual exhibits into haptic rendering instructions through spatial registration mapping, drives a distributed haptic array to generate a force field gradient matching the current position in the virtual exhibition hall topology, and generates spatial audio matching the current position through sound field beamforming technology; A dynamic tuning module that continuously collects topology behavior trajectories and neurophysiological indicators of a user in the elastic path network, constructs a cognitive load and spatial position joint probability model, and dynamically adjusts the distribution density and information presentation dimension of the exhibit node clusters.

Images