An immersive virtual reality interaction method, system and storage medium

Abstract

The application provides an immersive virtual reality interaction method, system and storage medium, and relates to the technical field of virtual reality.The application establishes a personalized behavior baseline within an initial time window after a first user enters a virtual scene; multi-dimensional behavior data of the first user is collected in real time, a normalized deviation degree is calculated based on the personalized behavior baseline, and weighted fusion is performed to determine an emotional trigger state; when the emotional trigger state is determined, a scene unit containing an action sequence, a spatial position and an emotional label is recorded and stored; when a second user enters the virtual scene, a target scene unit is screened based on a spatial matching degree, a time matching degree and a behavior matching degree, a virtual avatar is generated and anchored.The application realizes cross-time transmission of user behavior value, breaks the single-person isolated limitation, realizes asynchronous co-experience, and enables the virtual scene to have dynamic evolution capability.

Description

Technical Field

[0001] This invention relates to the field of virtual reality technology, and in particular to an immersive virtual reality interaction method, system, and storage medium. Background Technology

[0002] With the continuous development of virtual reality technology, virtual reality systems have been widely used in fields such as exhibitions, cultural dissemination, and education and training. After wearing a virtual reality headset and entering a virtual scene, users can obtain an immersive visual experience and a sense of spatial presence.

[0003] However, current virtual reality interactive experiences still have the following problems: the user experience is individual, unable to perceive the behavioral traces of other users, and differs from the atmosphere of a shared space in a real exhibition hall; the system does not actively present the behavioral traces of other users to the user, and the user's interaction behavior loses its value after the experience ends; the system does not record the user's emotional reactions, nor does it have the ability to transform emotional experiences into virtual scene content; the virtual scene content is pre-made and remains unchanged for a long time, lacking the adaptability to dynamically adjust the scene presentation content based on the user's historical behavior data. Summary of the Invention

[0004] This invention provides an immersive virtual reality interaction method, system, and storage medium, aiming to improve the problems of isolated user interaction, inability to transmit emotional experience, and lack of dynamic evolution capability of virtual scenes in existing VR interactive experiences.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an immersive virtual reality interaction method, the method comprising the following steps: Collect behavioral data of the first user in multiple behavioral dimensions in a virtual scene, and establish a personalized behavioral baseline based on the behavioral data within an initial time window; continuously collect the behavioral data after the initial time window, calculate the normalized deviation of each behavioral dimension based on the personalized behavioral baseline, and perform weighted fusion. When the first user is determined to be in an emotionally triggered state based on the preset conditions of the weighted fusion result, a scenario unit containing the action sequence, spatial location, and emotional label assigned according to the emotionally triggered state is recorded and stored. When the second user enters the virtual scene, the target scene unit is filtered based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit. Based on the action sequence and spatial location of the target scenario unit, a virtual avatar is generated to reproduce historical behavioral traces, and the virtual avatar is anchored to the corresponding virtual spatial coordinates.

[0006] The above method, as a preferred embodiment, includes the personalized behavior baseline comprising the mean and standard deviation of each behavior dimension; the normalized deviation is: , Where x is the real-time collected behavioral data, μ is the mean of the corresponding behavioral dimension, and σ is the standard deviation of the corresponding behavioral dimension; the step of determining that the first user is in an emotionally triggered state based on preset conditions of the fusion result includes: The normalized deviation of each behavior dimension is compared with a preset activation threshold. When the normalized deviation of a certain behavior dimension exceeds the preset activation threshold, the behavior dimension is marked as active. When the weighted fusion result exceeds the preset trigger threshold and at least two of the behavioral dimensions are marked as active, the first user is determined to be in the emotional trigger state.

[0007] The above method, as a preferred embodiment, specifically includes calculating the normalized deviation of each behavioral dimension based on the personalized behavioral baseline and performing weighted fusion, which includes: The weight coefficients for each behavioral dimension are configured according to the exhibit type of the current interactive exhibit; When the exhibit type is a visual exhibit, the weighting coefficients for gaze duration and head deflection angle are greater than the weighting coefficients for any other behavioral dimension. When the exhibit type is an interactive exhibit, the weight coefficients for operational behavior and repeated visits are greater than the weight coefficients for any other behavioral dimension. When the exhibit type is narrative exhibit, the weight coefficients of dwell time and return frequency are greater than the weight coefficients of any other behavioral dimension.

[0008] The above method, as a preferred embodiment, includes the recording and storage of scenario units comprising action sequences, spatial locations, and emotion tags, comprising: The action sequence is configured to collect continuous action frame data within a preset time period before and after the trigger time, determine key frames from the action sequence using the inter-frame difference method, smooth the action sequence after extracting key frames using a filtering algorithm, and perform interpolation transition processing on the beginning and end of the smoothed action sequence to make the connection smooth during loop playback. The spatial location is configured as follows: at the trigger time, the first user's three-dimensional spatial coordinates and head orientation angle in the virtual scene. The emotion tag is configured to determine the emotion type and intensity based on the emotion triggering state, and is associated with the scenario unit for storage.

[0009] In the above method, as a preferred embodiment, the spatial matching degree, temporal matching degree, and behavioral matching degree are determined in the following way: The spatial matching degree is determined based on the distance between the current location of the second user and the spatial location of the scenario unit; the smaller the distance, the higher the spatial matching degree. The time matching degree is determined based on the time difference between the current time and the recording time of the scenario unit; the smaller the time difference, the higher the time matching degree. The behavior matching degree is determined based on the similarity between the second user's current behavior pattern and the behavior pattern recorded in the scenario unit. The higher the similarity, the higher the behavior matching degree. The target scenario unit for screening includes: The spatial matching degree, the temporal matching degree, and the behavioral matching degree are weighted and combined to obtain the comprehensive score of each scenario unit; One or more scenario units with the highest comprehensive score are selected as the target scenario units.

[0010] In the above method, as a preferred embodiment, the behavioral dimension includes at least two of the following: position dimension, posture dimension, gaze dimension, and operation dimension. The behavioral data in the location dimension includes location coordinates and / or movement trajectory; The behavioral data in the posture dimension includes head yaw angle and / or posture change data; The behavioral data for the gaze dimension includes gaze direction and / or gaze duration; The behavioral data in the operational dimension includes hand position and / or operational action data.

[0011] As a preferred embodiment, the above method further includes: The system detects the spatial distance, orientation angle, and action similarity between the second user and the virtual avatar. When the spatial distance is less than a first preset threshold and the orientation angle is less than a second preset threshold, the virtual avatar is triggered to execute a first-level response, which includes changes in playback speed and / or halo effects. When the action similarity exceeds a third preset threshold, the virtual avatar is triggered to execute a second-level response, which includes changes in action rhythm, orientation shift, and / or visual effects. Record the feedback data of the second user's interaction with the virtual avatar, and update the filtering weight by injecting the data at a preset period.

[0012] In the above method, as a preferred embodiment, the display state of the virtual avatar decays with the retention time according to a negative exponential decay function. When the decay of the display state is lower than a preset elimination threshold, the virtual avatar is automatically removed. When the second user is detected to have paid attention to the virtual avatar, resonance life extension is triggered to restore the display state of the virtual avatar and extend its retention time. The virtual scene is divided into multiple spatial partitions, and each spatial partition has an upper limit on the number of visible avatars. When the number of avatars already presented in a certain spatial partition reaches the upper limit, newly selected avatars will not be presented for the time being.

[0013] In a second aspect, the present invention provides an immersive virtual reality interaction system, comprising: The behavior perception and discrimination module is used to collect behavior data of the first user in multiple behavioral dimensions in a virtual scene, establish a personalized behavior baseline based on the behavior data within an initial time window, and calculate the normalized deviation of each behavior dimension based on the personalized behavior baseline and perform weighted fusion based on the behavior data continuously collected after the initial time window. The scenario recording module is used to record and store a scenario unit containing an action sequence, spatial location, and an emotion label assigned according to the emotion triggering state when the first user is determined to be in an emotion triggering state based on a preset condition judgment of the weighted fusion result. The avatar filtering module is used to filter target scene units based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit when the second user enters the virtual scene. The avatar generation module is used to generate a virtual avatar for reproducing historical behavioral traces based on the action sequence and spatial location of the target scenario unit, and to anchor the virtual avatar to the corresponding virtual space coordinates.

[0014] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the immersive virtual reality interaction method as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) In this invention, a personalized behavioral baseline adapted to the first user's behavioral habits is established within the initial time window, and the normalized deviation of multi-dimensional behavioral data is continuously calculated and weighted and fused to dynamically determine the emotional trigger state. This enables the system to accurately capture high-attention moments based on individual behavioral characteristics, effectively avoiding misjudgments of different user behavior patterns by fixed thresholds. When the emotional trigger state is determined, the action sequence, spatial location, and emotional tag are encapsulated into a scenario unit and persistently stored, transforming instantaneous emotional behavior into a storable spatial digital asset, providing a content foundation for cross-time transmission. When the second user enters the virtual scene, the stored scenario units are filtered in multiple dimensions based on spatial matching degree, temporal matching degree, and behavioral matching degree, and a virtual avatar is generated based on the action sequence and spatial location of the target scenario unit for reproduction. The matching and filtering steps and the avatar generation steps work together to establish a dynamic relationship between the presentation of historical behavioral traces and the second user's current location, current behavior, and current time in three dimensions: space, behavior, and time, realizing the accurate deployment and intelligent reproduction of scenario units. Overall, this invention captures the high-emotional behavior of the first user as a scenario unit and intelligently reproduces it as a virtual avatar when the second user enters, realizing the cross-time transmission of user behavior value, achieving asynchronous co-existence experience, and enabling the virtual scene to have dynamic evolution capabilities.

[0016] (2) In this invention, the deviation is calculated by normalizing the deviation based on the mean and standard deviation of the personalized behavior baseline, eliminating the dimensional differences between different behavior dimensions, so that the deviation of each dimension has a unified and comparable scale; the dual triggering conditions require the comprehensive score to exceed the preset triggering threshold and at least two behavior dimensions to be in an active state. The two work together to ensure that the system will not be falsely triggered due to the random fluctuation of a single dimension, nor will it miss the real high attention moment due to the limitation of the fixed threshold, thus significantly reducing the probability of misjudgment while ensuring triggering sensitivity.

[0017] (3) In this invention, by detecting the spatial distance, orientation angle, and action similarity between the second user and the virtual avatar, a hierarchical first-level response and a second-level response are triggered. This allows the virtual avatar to provide appropriate feedback based on the proximity and behavioral similarity of the second user, triggering the response without interfering with the immersive experience. Simultaneously, feedback data on the interaction between the second user and the virtual avatar is recorded and updated with a preset period of feedback, enabling the matching degree filtering strategy to adaptively adjust based on the actual interaction effect. The synergistic effect of response triggering and feedback feedback forms a closed loop from interaction perception to strategy optimization, continuously improving the filtering accuracy of the scenario units as the system operates.

[0018] (4) In this invention, the virtual avatar decays according to a preset decay function over time, causing avatars with no attention value to fade out naturally; when a second user is detected to have paid attention to an avatar, resonance life extension is triggered, allowing valuable avatars to be retained due to interaction; the setting of the upper limit of the number of visible avatars in the spatial partition ensures that the avatar presentation density in each area is always within a reasonable range. The synergy of decay elimination, resonance life extension, and quantity control enables the virtual scene to both accumulate high-value historical behavior traces and automatically clean up redundant content during long-term operation, maintaining the sustainable evolution capability of the scene. Attached Figure Description

[0019] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts of the immersive virtual reality interaction method provided in the embodiments of the present invention.

[0021] Figure 2 This is one of the flowcharts of the immersive virtual reality interaction method provided in the embodiments of the present invention.

[0022] Figure 3 This is a structural block diagram of an immersive virtual reality interactive system provided in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of an immersive virtual reality interactive scene provided by an embodiment of the present invention.

[0024] Figure labeling: 10, Perception and Judgment Module; 20, Scene Recording Module; 30, Avatar Filtering Module; 40, Avatar Generation Module; 50, Interactive Response Module; 60, Feedback Injection Module. Detailed Implementation

[0025] To make the technical problems solved, the technical solutions, and the beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings of the embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the invention.

[0026] like Figure 1 The image shows an immersive virtual reality interaction method according to an embodiment of the present invention. The method includes the following steps: S1. Collect behavioral data of the first user across multiple behavioral dimensions in a virtual scene, and establish a personalized behavioral baseline based on the behavioral data within an initial time window. Continuously collect the behavioral data after the initial time window, calculate the normalized deviation of each behavioral dimension based on the personalized behavioral baseline, and perform weighted fusion.

[0027] S2. When it is determined that the first user is in an emotionally triggered state based on the preset conditions of the weighted fusion result, a scenario unit containing the action sequence, spatial location, and emotional label assigned according to the emotionally triggered state is recorded and stored.

[0028] S3. When the second user enters the virtual scene, the target scene unit is selected based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit.

[0029] S4. Generate a virtual avatar for reproducing historical behavioral traces based on the action sequence and spatial location of the target scenario unit, and anchor the virtual avatar to the corresponding virtual space coordinates.

[0030] It is worth noting that in this embodiment, firstly, during the initial user data collection phase, a personalized behavioral baseline is established using behavioral data collected within an initial time window, determining a unique behavioral reference standard for each individual user. Subsequently, behavioral data is continuously collected, and the baseline is used to calculate a normalized deviation, transforming behavioral data of different dimensions into a deviation index with a unified scale. Finally, the deviation indexes are weighted and fused, assigning differentiated weights to different behavioral dimensions based on the actual application scenario. The weighted fusion result comprehensively reflects the user's true level of attention in a specific scenario.

[0031] Then, based on whether the user is in an emotionally triggered state according to the weighted fusion result, only when the user is in an emotionally triggered state, the user's action sequence and spatial location are recorded and stored as a scenario unit, and the scenario unit is labeled with the user's emotional label, which is converted into a structured data format to provide a content source for subsequent cross-time reproduction.

[0032] Next, in the second user virtual reenactment stage, the spatial matching degree between the user's current location and the stored scenario units is filtered to ensure that the reenacted historical behavioral traces are spatially relevant to the second user. The matching degree between the user's current behavior and the behavioral characteristics of the scenario units is filtered to associate the reenacted behavioral traces with the second user's current behavioral patterns. Finally, the matching degree between the user's current time and the time of the scenario units is filtered to associate the reenacted behavioral traces with the second user's access period in the time dimension. This comprehensive filtering of the three matching degrees intelligently selects the target scenario unit most relevant to the user from a large number of stored scenario units, achieving precise deployment of scenario units.

[0033] Finally, a virtual avatar is generated based on the action sequence of the target scenario unit, and the historical behavioral traces are reproduced in a visualized virtual image, so that the current user can intuitively perceive the behavior of the historical user in the emotional trigger state.

[0034] Therefore, according to the embodiments of the present invention, it is possible to capture the high emotional behavior of the first user as a scenario unit and intelligently reproduce it as a virtual avatar when the second user enters, thereby realizing the cross-time transmission of user behavior value, realizing asynchronous co-existence experience, and enabling the virtual scene to have dynamic evolution capabilities.

[0035] In this embodiment, the specific implementation of each of the above steps is as follows.

[0036] To achieve the objective of this embodiment, step S1 specifically includes four sub-steps, S11 to S14.

[0037] S11. Establish a personalized behavior baseline within the initial time window after the first user enters the virtual scene.

[0038] In this embodiment, taking a visual exhibit scenario as an example, the behavioral dimensions include gaze dimension, posture dimension, position dimension, and operation dimension. Specifically, the gaze dimension includes gaze direction and gaze duration acquired through an eye-tracking sensor; the posture dimension includes head yaw angle and posture change data acquired through an inertial measurement unit; the position dimension includes position coordinates and movement trajectory acquired through a spatial positioning system; and the operation dimension includes hand position and operation action data acquired through a gesture tracking device.

[0039] The initial time window is set to the first Ms (e.g., 60 seconds) after entering the scene. Within this window, behavioral data sequences of the first user across the aforementioned behavioral dimensions are continuously collected. The mean and standard deviation of each behavioral dimension are calculated to form the first user's personalized behavioral baseline. This personalized behavioral baseline includes the mean (e.g., the mean gaze duration is 2.5 seconds, and the mean movement speed is 1.0 m / s) and standard deviation (e.g., the standard deviation of gaze duration is 1.2 seconds, and the standard deviation is 0.3 m / s) for each behavioral dimension. This personalized behavioral baseline serves as a reference for subsequent normalized deviation calculations. Illustratively, the behavioral data sequences of the first user across the aforementioned behavioral dimensions, after calculation, yield the following baseline data: mean gaze duration of 2.5 seconds, standard deviation of 1.2 seconds; mean head yaw angle of 20 degrees, standard deviation of 10 degrees; mean movement speed of 1.0 m / s, standard deviation of 0.3 m / s; and mean hand operation frequency of 0.2 Hz, standard deviation of 0.1 Hz.

[0040] S12. Calculate the normalized deviation of each behavioral dimension.

[0041] In this embodiment, the behavioral data of the first user across all behavioral dimensions, continuously collected after the initial time window, are compared with the personalized behavioral baseline to calculate the normalized deviation. The formula for calculating the normalized deviation is: , Where x represents the real-time collected behavioral data, μ is the mean of the corresponding behavioral dimension, and σ is the standard deviation of the corresponding behavioral dimension. Taking gaze duration as an example, let the mean of gaze duration in the personalized behavioral baseline be μ, the standard deviation be σ, and the real-time collected gaze duration be x. If σ is greater than a preset small threshold (e.g., 1e-6), then... If σ is less than or equal to this tiny threshold, it indicates that the data in this dimension is essentially constant within the initial time window, and in this case, d = 0 is directly set. It is understandable that this process avoids the problem of division operations failing when the standard deviation is zero.

[0042] The normalized deviation of each behavioral dimension is compared with a preset activation threshold. When the normalized deviation of a certain behavioral dimension exceeds the preset activation threshold (e.g., the activation threshold is 1.5), that behavioral dimension is marked as active. Through normalization, the deviations of different behavioral dimensions have a comparable scale, providing a basis for subsequent weighted fusion. Taking the gaze duration dimension as an example, let μ = 2.5s and σ = 1.2s (σ > 1e-6) be the gaze duration of the first user on the current exhibit, and let x = 5.5s be the gaze duration. Then the normalized deviation d is 2.5, which exceeds the preset activation threshold of 1.5, and this gaze duration dimension is marked as active.

[0043] For example, real-time behavioral data of the first user at a certain moment after the initial time window regarding the current visual exhibit is obtained: gaze duration is 7.5s, head yaw angle is 15°, movement speed is 0.4 m / s, and hand operation frequency is 0.1 Hz. Based on the above baseline data and real-time behavioral data, the normalized deviation of each behavioral dimension is calculated. The normalized deviation of gaze duration d1≈4.17, exceeding the preset activation threshold of 1.5, and this dimension is marked as active. The normalized deviation of head yaw angle d2=0.5, not exceeding the preset activation threshold of 1.5. The normalized deviation of movement speed d3=2.0, exceeding the preset activation threshold of 1.5, and this dimension is marked as active. The normalized deviation of hand operation frequency d4=1.0, not exceeding the preset activation threshold of 1.5.

[0044] S13. Weighted fusion is performed by configuring weight coefficients according to the exhibit type. In this embodiment, the weight coefficients of each behavioral dimension are configured according to the exhibit type of the current interactive exhibit.

[0045] For example, when the exhibit type is a visual exhibit, the weight coefficients for gaze duration and head yaw angle are greater than the weight coefficients for any other behavioral dimension. The weight coefficient for gaze duration can range from 0.25 to 0.40, the weight coefficient for head yaw angle can range from 0.25 to 0.40, and the combined weight coefficients for the other dimensions can range from 0.20 to 0.50.

[0046] In other embodiments, for example, when the exhibit type is interactive, the weight coefficients for operational behavior and repeated visits are greater than the weight coefficients for any other behavioral dimension. For example, when the exhibit type is narrative, the weight coefficients for dwell time and return frequency are greater than the weight coefficients for any other behavioral dimension.

[0047] The formula for weighted fusion is: , Where s is the weighted fusion result, w i Let d be the weight coefficient for the i-th behavioral dimension. i Let be the normalized deviation of the i-th behavioral dimension, and n be the number of behavioral dimensions. Taking visual exhibits as an example, the normalized deviations for gaze duration, head yaw angle, movement speed, and head yaw angle are d1 = 4.17, d2 = 1.5, d3 = 2.0, and d4 = 1.5, respectively. Weight coefficients are assigned to each behavioral dimension according to the exhibit type. For example, the weight coefficient for gaze duration is w1 = 0.35, the weight coefficient for head yaw angle is w2 = 0.35, the weight coefficient for movement speed is w3 = 0.15, and the weight coefficient for hand operation frequency is w4 = 0.15. The sum of these weight coefficients is 1. The calculation process for the weighted fusion result is as follows: .

[0048] S14. Determine the emotional trigger state based on dual conditions and a cooling-off mechanism.

[0049] In this embodiment, when the weighted fusion result exceeds a preset trigger threshold (e.g., a value of 2.0), and at least two behavioral dimensions are marked as active, the first user is determined to be in an emotion-triggered state and a trigger signal is sent. For example, in the aforementioned example data, the weighted fusion result of 2.185 exceeds the preset trigger threshold of 2.0, and both gaze duration and movement speed are active, satisfying the dual condition that at least two behavioral dimensions are active.

[0050] After sending the trigger signal, a preset cooldown period (e.g., 10 seconds) is entered, during which the triggering judgment of the first user's emotional state is stopped. It is understandable that the introduction of the cooldown mechanism can prevent the first user from repeatedly triggering recordings in a short period of time due to a sustained high emotional state in front of the same exhibit.

[0051] Therefore, by establishing a personalized behavioral baseline and coordinating dual triggering conditions in step S1, a reference benchmark can be determined based on the first user's own behavioral habits within the initial time window, and the emotional triggering state can be accurately captured in subsequent continuous judgment. At the same time, false triggers caused by sensor noise or accidental actions are filtered out, ensuring that each scenario unit stored in the database corresponds to the first user's real high attention moment, providing a reliable source of content for presenting valuable historical behavioral traces to the second user in the future.

[0052] To achieve the objective of this embodiment, step S2 specifically includes three sub-steps, S21 to S23.

[0053] S21. Construct the data structure for the scenario unit.

[0054] In this embodiment, the data structure of the scenario unit includes an action sequence field, a spatial location field, an emotion tag field, and a timestamp field. The action sequence field stores continuous action frame data collected from the first user within a preset time period before and after the triggering of the emotion triggering state. Specifically, this includes continuous action frame data such as head micro-movements, body leaning forward, and hand pointing. For example, storing the action sequence from 2 seconds before the triggering moment to 3 seconds after the triggering moment includes action frames such as the head tilt angle slowly changing from 15° to 10°, the torso leaning forward angle increasing from 5° to 15°, and the right hand raising from a naturally drooping position to point towards the exhibit. The spatial location field stores the first user's three-dimensional spatial coordinates in the virtual scene at the triggering moment and the head orientation angle. The emotion tag field stores the emotion type identifier and emotion intensity identifier determined according to the emotion triggering state, and stores them in association with the scenario unit. The emotion type identifier includes types such as high attention, deep thinking, and repeated exploration. Based on the numerical range of the weighted fusion result and the combination of activated behavioral dimensions, it is mapped to a preset emotion type library (e.g., a high weighted fusion result and activated gaze dimension correspond to "high attention," while repeated visits to the same area correspond to "repeated exploration"). The timestamp field stores the time recorded by the scenario unit and is used for subsequent time matching calculations and time decay management.

[0055] S22. Process the action sequence.

[0056] In this embodiment, the processing of the motion sequence includes: determining keyframes from the motion sequence using inter-frame differencing (e.g., a Euclidean distance threshold of 0.05 for pixel changes), i.e., performing inter-frame differencing calculations on consecutive motion frames; when the difference between adjacent frames exceeds a preset difference threshold, the frame is marked as a keyframe and retained; frames with differences below the threshold are merged or discarded, thereby reducing the data storage volume of the motion sequence. A filtering algorithm is used to smooth the motion sequence after keyframe extraction (e.g., using Kalman filtering) to eliminate trajectory jitter caused by sensor noise, making the subsequent virtual avatar's recreated actions more natural and smooth. Interpolation transition processing is performed on the beginning and end of the smoothed motion sequence, i.e., inserting transition frames (e.g., generating transition frames through linear interpolation) before the first frame and after the last frame of the motion sequence to ensure smooth transitions during loop playback.

[0057] S23. Assign emotion tags and store scenario units.

[0058] In this embodiment, emotion tags are directly assigned based on the judgment result of the emotion triggering state, including an emotion type identifier and an emotion intensity identifier, and are stored in association with the scenario unit (e.g., the emotion type identifier "high attention" and the emotion intensity identifier "2.185" are written into the emotion tag field of the scenario unit). After the scenario unit is generated, it is stored in the system database, completing the transformation of instantaneous emotional behavior reactions into persistent spatial digital assets.

[0059] Thus, by using step S2, the behavior of the first user in the emotionally triggered state is encapsulated into a structured scenario unit containing action sequence, spatial location and emotional label and stored persistently. This allows the user behavior traces that would otherwise disappear when the experience ends to be preserved, providing a standardized and searchable source of content for the second user to retrieve and reproduce these behavior traces when entering the virtual scene.

[0060] The specific implementation method of step S3 in this embodiment is as follows.

[0061] In this embodiment, when the second user enters the virtual scene, stored scene units are retrieved. Target scene units are filtered based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit. Spatial matching degree is determined by the distance between the second user's current location and the spatial location of the scene unit; the smaller the distance, the higher the spatial matching degree. Temporal matching degree is determined by the time difference between the current time and the recorded time of the scene unit; the smaller the time difference, the higher the temporal matching degree (e.g., the time difference between the second user's current entry time and the recorded time of the scene unit is 3 days, indicating a moderate temporal matching degree). Behavioral matching degree is determined by the similarity between the second user's current behavioral pattern and the behavioral pattern recorded in the scene unit; the higher the similarity, the higher the behavioral matching degree (e.g., the second user's current moving speed is 0.5 m / s, and the gaze duration is 6.0 s, similar to the first user's slow movement and long gaze behavior pattern recorded in the scene unit, indicating a high behavioral matching degree). The spatial matching degree, temporal matching degree, and behavioral matching degree are weighted and combined (for example, spatial matching degree weight 0.4, temporal matching degree weight 0.3, and behavioral matching degree weight 0.3 to calculate the comprehensive score of each scenario unit). The comprehensive score of each scenario unit is obtained, and one or more scenario units with the highest comprehensive score are selected as target scenario units.

[0062] To achieve the objective of this embodiment, step S4 specifically includes three sub-steps, S41 to S43.

[0063] S41. Generate a virtual avatar.

[0064] In this embodiment, the virtual avatar is generated based on the action sequence of the target scenario unit, rendered using a standard digital human model, and appears as a realistic digital character. The action sequence drives the avatar to perform actions and plays in a loop. The virtual avatar does not present any of the real user's personal facial features or physical appearance information to protect the visual privacy of the previous user. Virtual avatars from different retention periods are distinguished by different identifiers: avatars generated from scenario units recorded within the last day use a specific identifier, those recorded within 1 to 7 days use another identifier, and those recorded more than 7 days ago use yet another identifier, so that the second user can intuitively perceive the time hierarchy of the historical behavioral traces represented by the virtual avatar.

[0065] S42, Anchoring Virtual Avatar.

[0066] In this embodiment, the generated virtual avatar is anchored to the virtual spatial coordinates corresponding to the spatial position of the target scene unit (e.g., anchoring the virtual avatar to coordinates (12.5, 0, 8.2), consistent with the three-dimensional spatial coordinates recorded by the scene unit). The virtual scene is divided into multiple spatial partitions (e.g., each partition is a square area with a side length of 10m), and each spatial partition has an upper limit on the number of visible avatars (e.g., the upper limit for each partition is 3). If there are other virtual avatars or virtual scene objects within a preset radius around the anchor point (e.g., the preset radius is 1m), the virtual avatar is offset to a nearby empty space (e.g., offset to coordinates (12.8, 0, 8.2)), and a directional halo is superimposed to indicate the original anchor position. When the number of avatars already presented in a certain spatial partition reaches the upper limit, newly added virtual avatars are not presented temporarily, waiting for existing avatars to be eliminated before being deployed.

[0067] It can be seen that by combining the three-dimensional matching degree filtering mechanism and the avatar generation anchoring mechanism in steps S3 and S4, the historical scenario units most relevant to the second user's current location, current behavior, and current time can be selected as target scenario units, and a virtual avatar can be generated in the corresponding spatial location. This enables the presentation of historical behavior traces to establish a dynamic relationship with the second user in three dimensions: space, behavior, and time, thereby achieving precise deployment and intelligent reproduction of scenario units.

[0068] like Figure 2 As shown, the above method further includes step S5, which detects the spatial distance, orientation angle, and action similarity between the second user and the virtual avatar, triggers a hierarchical interactive response based on the detection results, and records the interactive feedback data to update the filtering weights at a preset period. This step specifically includes three sub-steps, S51 to S53.

[0069] S51. Detect the spatial and behavioral relationships between the second user and the virtual avatar.

[0070] In this embodiment, spatial relationships include the spatial distance between the second user and the virtual avatar (e.g., the second user's current position coordinates are (13.0, 0, 8.5), the virtual avatar's anchor coordinates are (12.5, 0, 8.2), and the Euclidean distance is approximately 0.58m), and the angle between the second user's head orientation and the virtual avatar's direction (e.g., the second user's head orientation deflection angle is 10°, the virtual avatar's direction deflection angle is 15°, and the angle is 5°). Behavioral relationships include the similarity between the second user's current action sequence and the historical action sequence looped by the virtual avatar (e.g., the second user also makes a forward lean and slight head movements, achieving a 75% similarity to the looped action sequence played by the virtual avatar). Detection continues in the background and is triggered simply by the second user maintaining natural browsing behavior; no active operation is required.

[0071] S52. Trigger a progressive two-way interactive response based on preset hierarchical conditions.

[0072] In this embodiment, the progressive two-way interactive response refers to an interactive feedback mechanism that is triggered in layers based on the spatial and behavioral relationships between the second user and the virtual avatar. The response includes two levels, which are triggered progressively from shallow to deep.

[0073] Specifically, the first-level response is triggered when the spatial distance is less than a first preset threshold (e.g., 1.5 m) and the facing angle is less than a second preset threshold (e.g., 30 deg) (e.g., the spatial distance is 0.58 m < 1.5 m and the facing angle is 5 deg < 30 deg, which satisfies the first-level triggering condition). The response content includes changes in the playback speed of the virtual avatar and halo effects. This level response conveys the signal that the virtual avatar has perceived the presence of the second user.

[0074] Second-level response: Triggered when the action similarity exceeds the third preset threshold (e.g., 70%) (e.g., the above action similarity of 75% > 70%, which meets the second-level triggering condition). The response content includes changes in the virtual avatar's action rhythm, orientation deviation, and visual effects. This level of response is triggered when the second user performs a similar behavior to the first user, forming a cross-time response at the behavioral level.

[0075] It is worth noting that when the second user does not trigger the above-mentioned hierarchical conditions, the virtual avatar continues to loop the basic action sequence without interfering with the second user's immersive experience.

[0076] S53. Record feedback data and update the filtering weights.

[0077] In this embodiment, feedback data of the interaction between the second user and the virtual avatar is recorded, and the algorithm weight parameters for filtering target scenario units in step S3 are updated back at a preset period. The feedback data includes the second user's behavior sequence, the frequency and duration of responses to each interaction response level (e.g., 8 seconds of viewing time, 3 responses per day), and implicit feedback data such as the second user's behavior patterns after approaching the avatar (e.g., staying for 15 seconds after triggering a second-level response before leaving). Target scenario units that have triggered interactions are statistically analyzed at a preset period. The weighted composite index of the average dwell time and response frequency for each scenario unit is calculated and compared with a preset benchmark value. If it is higher than the benchmark value, the weight coefficient corresponding to the matching degree dimension that contributes the most to the filtering of that scenario unit is increased (e.g., the behavior matching degree weight coefficient is increased from 0.30 to 0.35); if it is lower than the benchmark value, it is decreased accordingly. The comprehensive matching degree threshold can also be adjusted according to the ratio of the first-level to the second-level responses. Through the above closed-loop feedback, the filtering algorithm continuously iterates with the interaction data, improving the accuracy of scenario unit filtering and the effectiveness of interaction responses.

[0078] Therefore, through the detection of spatial and behavioral relationships and the hierarchical response mechanism in step S5, the virtual avatar transforms from a passive, static display into an interactive object capable of providing hierarchical feedback based on the spatial proximity and behavioral similarity of the second user. This achieves asynchronous, bidirectional interaction between the previous and current users without increasing the operational burden on the second user. The feedback and injection mechanism allows the filtering strategy to adaptively optimize based on actual interaction effects, continuously improving the interactive experience of the virtual scene as the system runs.

[0079] In this embodiment, the method further includes step S6, which performs lifecycle management on the virtual avatar. The display state of the virtual avatar decays according to a preset decay function over time. When the decay of the display state is lower than a preset elimination threshold, the virtual avatar is automatically removed. When the second user's attention behavior towards the virtual avatar is detected to meet the resonance condition, resonance extension is triggered, restoring the display state of the virtual avatar and extending its retention time. The virtual scene is divided into multiple spatial partitions, and each spatial partition has an upper limit on the number of visible avatars. When the number of avatars already presented in a spatial partition reaches the upper limit, newly selected avatars are not presented temporarily. This step specifically includes three sub-steps, S61 to S63.

[0080] S61, Manage avatar display status decay and automatic removal.

[0081] In this embodiment, the display status of the virtual avatar decays naturally with a negative exponential function over time. When the decay rate of the display status falls below a preset elimination threshold, the virtual avatar is automatically removed and the corresponding scene unit data is cleared. This mechanism ensures that the virtual scene does not become cluttered due to the infinite accumulation of avatars.

[0082] S62, trigger resonance to extend life.

[0083] In this embodiment, when the second user's attention behavior is detected to meet the following resonance conditions, resonance is triggered and a life-extending operation is executed: the distance between the second user's position and the avatar anchor point is less than a first preset threshold (e.g., 1.5m); the angle between the virtual reality headset orientation and the avatar anchor point direction is less than a second preset threshold (e.g., 30°); and the viewing duration exceeds a third preset threshold (e.g., 5s). The life-extending operation includes restoring the avatar's visibility to a preset value (e.g., restoring it to 50%) and extending the retention time. The life-extending mechanism allows valuable virtual avatars to naturally emerge due to the second user's attention, while worthless ones automatically disappear, achieving self-sustaining of the virtual scene content.

[0084] S63, Control of the number of spatial partitions.

[0085] In this embodiment, the virtual scene is divided into multiple spatial partitions, each with a maximum limit on the number of visible avatars (e.g., a maximum of 3 visible avatars per spatial partition). When the number of avatars already displayed in a spatial partition reaches the limit, newly added virtual avatars are not displayed temporarily, waiting for existing avatars to be deprecated before being deployed. This mechanism ensures that the avatar display density in each area is always within a reasonable range.

[0086] This demonstrates that through the coordinated efforts of the decay and elimination mechanism, the resonance lifespan mechanism, and the spatial partition quantity control in step S6, virtual avatars achieve adaptive management throughout their entire lifecycle: avatars without attention value naturally fade out, while valuable avatars are retained due to interaction, and the density of avatars in each area remains under control. The synergy of decay and elimination, resonance lifespan, and quantity control enables virtual scenes to both accumulate high-value historical behavioral traces and automatically clean up redundant content during long-term operation, maintaining the scene's sustainable evolutionary capability.

[0087] In summary, the method provided in this embodiment establishes a personalized behavioral baseline and calculates the normalized deviation in step S1, adapting the judgment of emotional trigger states to the first user's own behavioral habits and avoiding misjudgments based on a uniform threshold. Step S2 transforms instantaneous emotional behaviors into persistently storable scenario units. Steps S3 and S4 filter target scenario units based on three-dimensional matching degree and generate virtual avatars anchored to corresponding spatial locations, establishing a dynamic association between historical behavioral traces and the second user across three dimensions: space, behavior, and time. These steps collaboratively store the first user's behavior under emotional trigger states as scenario units, which are then reproduced as virtual avatars when the second user enters. Matching degree filtering enables precise deployment and intelligent reproduction of scenario units, thereby achieving cross-temporal transmission of user behavioral value, creating an asynchronous coexistence experience, and enabling the virtual scene to dynamically evolve.

[0088] like Figure 3 As shown, based on the same inventive concept, embodiments of the present invention also relate to an immersive virtual reality interaction system for implementing the immersive virtual reality interaction method described above. The system includes a behavior perception and discrimination module 10, a scene recording module 20, an avatar screening module 30, an avatar generation module 40, an interaction response module 50, and a feedback injection module 60.

[0089] Specifically, the behavior perception and discrimination module 10 collects behavioral data from multiple dimensions of the first user, establishes a personalized behavior baseline based on the behavioral data within an initial time window, and calculates and weights the normalized deviation based on the behavioral data continuously collected after the initial time window. The scenario recording module 20 records and stores scenario units containing action sequences, spatial locations, and emotion tags when the first user is determined to be in an emotionally triggered state. The avatar filtering module 30 filters target scenario units based on spatial matching degree, behavioral matching degree, and temporal matching degree when the second user enters the virtual scene. The avatar generation module 40 generates a virtual avatar based on the action sequence and spatial location of the target scenario unit and anchors it to the corresponding virtual spatial coordinates. The interaction response module 50 detects the spatial distance, orientation angle, and action similarity between the second user and the virtual avatar, triggers a hierarchical interaction response, and records feedback data to update the filtering weights at a preset period. The feedback injection module 60 is used to record the behavioral sequence, acceptance data and implicit feedback data of the second user's interaction with the virtual avatar, perform batch statistical summary at a preset time period, update the algorithm weight parameters of the avatar screening module 30, and perform lifecycle management of the virtual avatar.

[0090] The system described above is used to implement the corresponding immersive virtual reality interaction method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0091] The following three application scenario examples, combined with the collaboration process of various modules in the system, further illustrate the technical solution of this invention.

[0092] Application Example 1 In this application example, refer to Figure 4 As shown, the virtual exhibition hall contains multiple rectangular display stands connected by passageways. User A, wearing a virtual reality headset, re-enters the porcelain section of the virtual museum. When User A approaches exhibit 1, they see a digital avatar standing in front of it, leaning slightly forward with its head close to the exhibit, occasionally turning its head slightly to observe details from different angles. This digital avatar possesses a complete human appearance and natural movements, consistent with a real person, and blends seamlessly into the virtual scene.

[0093] The presentation process of the digital avatar is as follows: The avatar selection module 30 acquires the stored scenario units and calculates the spatial matching degree, temporal matching degree, and behavioral matching degree of each scenario unit based on user A's current location, current behavior, and entry time. Since user A is currently located in the vicinity of exhibit 1, and the entry time period has a periodic correspondence with the recording time of a certain scenario unit, the overall matching degree of this scenario unit is high, and it is selected as the target scenario unit. The avatar generation module 40 generates a digital human virtual avatar based on the action sequence of the target scenario unit, uses a preset digital human model to give it a realistic appearance, and anchors the virtual avatar to the virtual spatial coordinates corresponding to the spatial location recorded by the scenario unit, completing the avatar deployment. User A is attracted by the digital human avatar during the browsing process and approaches to observe it.

[0094] As User A, wearing a virtual reality headset, enters the exhibition area, the behavior perception and discrimination module 10 collects User A's behavioral data in real time in the background. This data includes location coordinates and movement trajectory, gaze direction and duration, head tilt angle and posture changes, and hand position and operation data. Within an initial time window, a personalized behavioral baseline for User A is established. When User A browses to exhibit 1 and is attracted by its intricate details, pausing to gaze at it for an extended period, the behavior perception and discrimination module 10 calculates the normalized deviation of each behavioral dimension and performs weighted fusion, determining that User A is in an emotionally triggered state. The scenario recording module 20 then records the sequence of actions before and after the trigger moment, including User A's head movements and forward leaning. After keyframe extraction via inter-frame difference, filtering, smoothing, and first-to-last interpolation, the action sequence, along with spatial location coordinates, head orientation angle, and emotion tags, is encapsulated into a scenario unit and stored in the system database.

[0095] When User A browses to exhibit 1, they are attracted by its intricate patterns and gaze at it for approximately 40 seconds. The behavior perception and discrimination module 1010 calculates the normalized deviation of each behavior dimension in real time. The deviation of the gaze duration dimension exceeds 1.5 standard deviations, and the deviation of the concentration dimension exceeds 1.8 standard deviations, both indicating an active state. Based on the exhibit type being a visual exhibit, the behavior perception and discrimination module 1010 assigns greater weighting coefficients to gaze duration and concentration than to the other behavior dimensions. The calculated weighted fusion result exceeds a preset trigger threshold, simultaneously satisfying the dual condition that at least two behavior dimensions are active. Therefore, User A is determined to be in an emotionally triggered state, and a trigger signal is sent, followed by a preset cooling-off period.

[0096] Through the above process, User A encounters and interacts with historical behavioral traces in the virtual scene, achieving an asynchronous coexistence experience across time dimensions. User A's behavioral traces under emotional triggering states are automatically captured and stored by the system as scenario units, which can be reproduced as interactive virtual avatars. As User A and other visitors continuously generate new emotionally triggered behaviors in the exhibition hall, scenario units recorded at different times gradually accumulate at each booth location, making the virtual scene content continuously rich and evolving.

[0097] In other application examples, such as the space exploration-themed exhibition area in a science museum, the behavior perception and discrimination module 10 and the scenario recording module 20 have completed the recording and storage of scenario units where historical users crouch down to look at the ground in a certain interactive area. After a new user enters the area, the avatar filtering module 30 filters the scenario unit as the target scenario unit based on the three-dimensional matching degree, and the avatar generation module 40 generates a digital human virtual avatar and anchors it to the corresponding spatial position. The digital human avatar plays the action sequence of crouching down to look at the ground in a loop. After seeing the digital human avatar, the new user, out of curiosity, also crouches down to look at the same position. The interaction response module 50, through action similarity calculation, detects that the new user's current action sequence has a high matching degree with the digital human avatar's historical action sequence, determines that the behavior mirroring trigger condition is met, and drives the digital human avatar to execute the second-level response, including the digital human avatar turning its head towards the new user, slightly slowing down the action rhythm, and generating faint visual effects around it, forming a cross-time response at the behavioral level.

[0098] In other application examples, such as a virtual process production experience zone in a manufacturing enterprise, the behavior perception and discrimination module 10 and the scenario recording module 20 have completed the recording and storage of a scenario unit in which a senior process engineer repeatedly operates and observes before a precision machining step. After a potential customer enters the experience zone, the avatar screening module 30 filters the scenario unit as the target scenario unit based on the three-dimensional matching degree, and the avatar generation module 40 generates a digital human virtual avatar, which repeatedly demonstrates the action sequence of operation, pause, and fine adjustment and anchors it to the corresponding spatial position. The customer notices the details of pausing and observing in the action sequence of the digital human avatar, realizes that there are process points that need attention in this step, and slows down the operation speed to observe carefully. The interaction response module 50 detects the customer's attention behavior and triggers resonance extension. During long-term operation, the scenario unit is continuously perceived and resonated by new visitors, triggering multiple extensions. The feedback injection module 60 continuously records the interaction data and periodically updates the screening algorithm weight parameters of the avatar screening module 30, so that the matching accuracy and interaction response effectiveness of the scenario unit continuously improve with the use time.

[0099] Based on the same inventive concept, corresponding to any of the above embodiments, this invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, can implement the steps of the above-described immersive virtual reality interaction method.

[0100] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An immersive virtual reality interaction method, characterized in that, The method includes the following steps: Collect behavioral data of the first user in multiple behavioral dimensions in a virtual scene, and establish a personalized behavioral baseline based on the behavioral data within an initial time window; continuously collect the behavioral data after the initial time window, calculate the normalized deviation of each behavioral dimension based on the personalized behavioral baseline, and perform weighted fusion. When the first user is determined to be in an emotionally triggered state based on the preset conditions of the weighted fusion result, a scenario unit containing the action sequence, spatial location, and emotional label assigned according to the emotionally triggered state is recorded and stored. When the second user enters the virtual scene, the target scene unit is filtered based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit. Based on the action sequence and spatial location of the target scenario unit, a virtual avatar is generated to reproduce historical behavioral traces, and the virtual avatar is anchored to the corresponding virtual spatial coordinates.

2. The immersive virtual reality interaction method according to claim 1, characterized in that, The personalized behavior baseline includes the mean and standard deviation of each behavior dimension; the normalized deviation is: , Where x is the real-time collected behavioral data, μ is the mean of the corresponding behavioral dimension, and σ is the standard deviation of the corresponding behavioral dimension; the step of determining that the first user is in an emotionally triggered state based on preset conditions of the fusion result includes: The normalized deviation of each behavior dimension is compared with a preset activation threshold. When the normalized deviation of a certain behavior dimension exceeds the preset activation threshold, the behavior dimension is marked as active. When the weighted fusion result exceeds the preset trigger threshold and at least two of the behavioral dimensions are marked as active, the first user is determined to be in the emotional trigger state.

3. The immersive virtual reality interaction method according to claim 2, characterized in that, The calculation of normalized deviations for each behavioral dimension based on the personalized behavioral baseline and the subsequent weighted fusion specifically includes: The weight coefficients for each behavioral dimension are configured according to the exhibit type of the current interactive exhibit; When the exhibit type is a visual exhibit, the weighting coefficients for gaze duration and head deflection angle are greater than the weighting coefficients for any other behavioral dimension. When the exhibit type is an interactive exhibit, the weight coefficients for operational behavior and repeated visits are greater than the weight coefficients for any other behavioral dimension. When the exhibit type is narrative exhibit, the weight coefficients of dwell time and return frequency are greater than the weight coefficients of any other behavioral dimension.

4. The immersive virtual reality interaction method according to claim 1, characterized in that, The recording and storage of the aforementioned contextual unit includes action sequences, spatial locations, and emotion tags, comprising: The action sequence includes: continuous action frame data collected within a preset time period before and after the triggering time; key frames are determined from the action sequence using the inter-frame difference method; the action sequence after key frame extraction is smoothed using a filtering algorithm; and the beginning and end of the smoothed action sequence are interpolated for smooth transition during loop playback. The spatial location includes: the three-dimensional spatial coordinates and head orientation angle of the first user in the virtual scene at the time of triggering; The emotion tag includes: the emotion type and emotion intensity determined according to the emotion triggering state, and is stored in association with the scenario unit.

5. The immersive virtual reality interaction method according to claim 1, characterized in that, The spatial matching degree, temporal matching degree, and behavioral matching degree are determined in the following ways: The spatial matching degree is determined based on the distance between the current location of the second user and the spatial location of the scenario unit; the smaller the distance, the higher the spatial matching degree. The time matching degree is determined based on the time difference between the current time and the recording time of the scenario unit; the smaller the time difference, the higher the time matching degree. The behavior matching degree is determined based on the similarity between the second user's current behavior pattern and the behavior pattern recorded in the scenario unit. The higher the similarity, the higher the behavior matching degree. The target scenario unit for screening includes: The spatial matching degree, the temporal matching degree, and the behavioral matching degree are weighted and combined to obtain the comprehensive score of each scenario unit; One or more scenario units with the highest comprehensive score are selected as the target scenario units.

6. The immersive virtual reality interaction method according to claim 1, characterized in that, The behavioral dimension includes at least two of the following: position dimension, posture dimension, gaze dimension, and operation dimension; The behavioral data in the location dimension includes location coordinates and / or movement trajectory; The behavioral data in the posture dimension includes head yaw angle and / or posture change data; The behavioral data for the gaze dimension includes gaze direction and / or gaze duration; The behavioral data in the operational dimension includes hand position and / or operational action data.

7. The immersive virtual reality interaction method according to claim 1, characterized in that, The method also includes: The system detects the spatial distance, orientation angle, and action similarity between the second user and the virtual avatar. When the spatial distance is less than a first preset threshold and the orientation angle is less than a second preset threshold, the virtual avatar is triggered to execute a first-level response, which includes changes in playback speed and / or halo effects. When the action similarity exceeds a third preset threshold, the virtual avatar is triggered to execute a second-level response, which includes changes in action rhythm, orientation shift, and / or visual effects. Record the feedback data of the second user's interaction with the virtual avatar, and update the filtering weight by injecting the data at a preset period.

8. The immersive virtual reality interaction method according to claim 1, characterized in that, The display status of the virtual avatar decays with the retention time according to a negative exponential decay function. When the decay of the display status is lower than a preset elimination threshold, the virtual avatar is automatically removed. When the second user is detected to have paid attention to the virtual avatar, resonance life extension is triggered to restore the display state of the virtual avatar and extend its retention time. The virtual scene is divided into multiple spatial partitions, and each spatial partition has an upper limit on the number of visible avatars. When the number of avatars already presented in a certain spatial partition reaches the upper limit, newly selected avatars will not be presented for the time being.

9. An immersive virtual reality interactive system, characterized in that, include: The behavior perception and discrimination module is used to collect behavior data of the first user in multiple behavioral dimensions in the virtual scene, and to establish a personalized behavior baseline based on the behavior data within the initial time window. It is used to calculate the normalized deviation of each behavioral dimension based on the personalized behavioral baseline and perform weighted fusion based on the behavioral data continuously collected after the initial time window; The scenario recording module is used to record and store a scenario unit containing an action sequence, spatial location, and an emotion label assigned according to the emotion triggering state when the first user is determined to be in an emotion triggering state based on a preset condition judgment of the weighted fusion result. The avatar filtering module is used to filter target scene units based on the spatial matching degree between the second user's current location and the scene unit, the behavioral matching degree between the second user's current behavior and the scene unit, and the temporal matching degree between the second user's current time and the scene unit when the second user enters the virtual scene. The avatar generation module is used to generate a virtual avatar for reproducing historical behavioral traces based on the action sequence and spatial location of the target scenario unit, and to anchor the virtual avatar to the corresponding virtual space coordinates.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the immersive virtual reality interaction method as described in any one of claims 1 to 8.

Images