A virtual reality interaction analysis method and system

Abstract

The application provides a virtual reality interaction analysis method and system, and relates to the technical field of data processing.The method comprises the following steps: acquiring interaction state record data corresponding to a plurality of experience users respectively, determining a plurality of interaction segments in a target virtual reality scene, constructing a state response feature matrix of each experience user, extracting a segment state change matrix of the experience user, determining segment state interference features of the plurality of interaction segments, constructing a reference state transition link about the plurality of interaction segments, after collecting real-time interaction state data of a target user, constructing a real-time state transformation matrix of the target user, and generating a state transformation correction matrix according to the plurality of segment state interference features, performing interaction state anomaly analysis on the target user according to the reference state transition link and the state transformation correction matrix, and generating an interaction state anomaly analysis result of the target user in the target virtual reality scene.The application realizes accurate identification of potential abnormal interaction states of users.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a virtual reality interactive analysis method and system. Background Technology

[0002] Virtual Reality (VR) technology is increasingly being applied in fields such as education and training, rehabilitation therapy, and human-computer interaction. During the interaction process in a VR system, the ability to perceive and respond to the user's state directly affects the immersion and suitability of the overall experience. Currently, VR systems collect and analyze the user's interaction state in the virtual reality environment, monitoring and analyzing interaction information such as the user's physiological parameters or behavioral trajectories. In this process, some technologies compare this information with preset thresholds to determine whether the user is in an abnormal state and make corresponding feedback or adjustments.

[0003] However, when faced with the complex and ever-changing interactive content in virtual reality, the differences in user states triggered by different interactive segments may be normal content-driven behavior rather than abnormal reactions. If these state fluctuations caused by changes in interactive content cannot be effectively identified, they can easily be misjudged as normal, thus masking potential abnormal user reactions. Furthermore, different users may exhibit individual differences in their state responses within the same interactive segment; failure to delve into the hidden characteristics of segments within historical user data can further affect the accuracy of identifying potential abnormal states. Summary of the Invention

[0004] The purpose of this invention is to provide a virtual reality interaction analysis method and system, which analyzes and mines the interaction data of a group to accurately perceive the virtual reality interaction status of different users and accurately identify potential abnormal states of users in the virtual reality interaction process.

[0005] To achieve the above objectives, the first aspect of the present invention provides a virtual reality interaction analysis method, comprising:

[0006] Acquire interaction state records for multiple users in the target virtual reality scene, identify multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of each user in each interaction segment, and construct the state response feature matrix for each user.

[0007] Extract the fragment state change matrix of the user from the state response feature matrix, extract the fragment state interference features of multiple interaction fragments based on the multiple fragment state change matrices, combine the fragment state interference features and the state response feature matrix to perform state transition analysis on multiple interaction fragments, and construct a reference state transition link for multiple interaction fragments.

[0008] After collecting the real-time interaction state data of the target user in the target virtual reality scene, a real-time state transformation matrix of the target user is constructed. The real-time state transformation matrix is ​​then corrected for interference based on the interference characteristics of multiple fragment states, and a state transformation correction matrix of the target user is generated.

[0009] Based on the reference state transition link and the state transition correction matrix, the interaction state anomaly analysis of the target user is performed, and the analysis results of the interaction state anomaly analysis of the target user in the target virtual reality scene are generated.

[0010] Preferably, the fragment state change matrix of the user is extracted from the state response feature matrix, and fragment state interference features of multiple interaction fragments are extracted based on multiple fragment state change matrices, including:

[0011] Based on the state response feature sequence, generate the fragment state change sequence between two adjacent interaction fragments, and construct a fragment state change matrix containing multiple fragment state change sequences;

[0012] Collective behavior analysis is performed on multiple fragment state change sequences in the fragment state change matrix. Collective change characteristic parameters of multiple interactive state parameters in the fragment state change sequence are extracted, including change direction parameters and change coordination parameters. Based on the collective change characteristic parameters, the behavioral stability parameters of the interactive fragments are calculated.

[0013] Based on multiple behavioral stability parameters of the test users, contextual state stability analysis is performed on multiple interaction segments of the test users to determine the state stability groups of different interaction segments. Based on the state change sequences and stability parameters of multiple segments of the state stability groups, the common state offset parameters of multiple interaction segments with respect to each interaction state parameter are calculated, and the segment state interference characteristics of multiple interaction segments are obtained.

[0014] Preferably, state transition analysis is performed on multiple interaction segments by combining the segment state interference features and state response feature matrices to construct a reference state transition link for the multiple interaction segments, including:

[0015] Based on the state interference characteristics of the fragments, state interference correction is performed on each state response feature matrix to generate a state response correction matrix for each user. Based on the multiple state response correction matrices, an initial state transition link for multiple interaction fragments is constructed.

[0016] Based on the initial state transition link, state deviation analysis is performed on each state response correction matrix to construct a state residual matrix for multiple experience users;

[0017] Based on the behavioral stability parameters of the user in different interaction segments, the residual aggregation weight of the user in each interaction segment in the state residual matrix is ​​determined. Based on the multiple residual aggregation weights corresponding to each user, the state residual matrix is ​​weighted to generate a state aggregation matrix. The state aggregation matrix is ​​then fused with the initial state transition link to obtain a reference state transition link for multiple interaction segments.

[0018] Preferably, the interaction state anomaly analysis of the target user is performed based on the reference state transition link and the state transition correction matrix, including:

[0019] Based on the reference state transition link and the state transition correction matrix, a real-time deviation feature sequence of the target user in each interaction segment is constructed, and the real-time deviation parameter of each real-time deviation feature sequence is calculated.

[0020] The reference deviation index for each interaction segment is calculated by referencing the state transition link and multiple state response correction matrices under different interaction segments.

[0021] Based on multiple reference deviation indices, segment anomaly analysis is performed on the real-time deviation parameters of the target user under different interaction segments. The local state anomaly parameters of the target user under each interaction segment are calculated. Based on multiple local state anomaly parameters, state evolution anomaly analysis is performed on the target user, and the interaction state anomaly analysis results of the target user in the target virtual reality scene are generated.

[0022] Preferably, the population change feature parameters of multiple interaction state parameters in the fragment state change sequence are extracted, including:

[0023] Based on the sequence of state changes in the segment, determine the direction of change of different interactive state parameters in the interactive segment, and count the percentage of users in the interactive segment whose direction of change of interactive state parameters is the same as that of interactive state parameters to obtain the direction of change of interactive state parameters.

[0024] The interaction segment is segmented into windows. The reference window of the interaction segment is determined based on the eigenvalues ​​corresponding to the eigenvalues ​​of the multiple interaction state parameters in the interaction segment. The window difference value of the interaction state parameters is calculated based on the positional distance between the window to which the eigenvalue of the interaction state parameter belongs and the reference window of the interaction segment. The change coordination parameter of the interaction state parameter is determined based on the window difference value.

[0025] Preferably, for the real-time deviation parameter and the reference deviation index, it further includes:

[0026] The norm features of the real-time deviation feature sequence are extracted to obtain the real-time deviation parameters of the real-time deviation feature sequence. Multiple norm features of the state response correction matrix are extracted to obtain multiple individual deviation parameters of the state response correction matrix. Multiple individual deviation parameters under each interaction segment are fused to obtain the reference deviation index under each interaction segment.

[0027] A second aspect of the present invention provides a virtual reality interaction analysis system for implementing the above-described virtual reality interaction analysis method, comprising:

[0028] The state feature extraction module is used to acquire the interaction state record data corresponding to multiple experience users of the target virtual reality scene, determine multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of the experience user in each interaction segment, and construct the state response feature matrix of each experience user.

[0029] The reference state link construction module is used to extract the fragment state change matrix of the user from the state response feature matrix, extract the fragment state interference features of multiple interactive fragments based on the multiple fragment state change matrices, and perform state transition analysis on multiple interactive fragments by combining the fragment state interference features and the state response feature matrix to construct a reference state transition link for multiple interactive fragments.

[0030] The state change correction module is used to construct the real-time state change matrix of the target user after collecting the real-time interaction state data of the target user in the target virtual reality scene, and to perform interference compensation correction on the real-time state change matrix according to the interference characteristics of multiple fragment states, so as to generate the state change correction matrix of the target user.

[0031] The interaction state anomaly analysis module is used to perform interaction state anomaly analysis on the target user based on the reference state transition link and the state transformation correction matrix, and generate the interaction state anomaly analysis results of the target user in the target virtual reality scene.

[0032] The present invention has the following beneficial effects:

[0033] This invention performs fragmented modeling of user interaction processes in virtual reality scenes, extracts the state response features and evolution trends of users in continuous interaction segments, and combines them with the structural behavioral characteristics of group users to construct a dynamic reference state transition link, enabling contextual consistency comparison and deviation identification of new user state changes. Furthermore, through mechanisms such as interference feature compensation and behavioral stability weighted residual fusion, it eliminates analysis errors caused by differences in scene content and heterogeneity of user constitutions. Finally, based on the reference path and deviation index, it completes segment-level local anomaly identification and overall interaction state evaluation, achieving dynamic and accurate identification and analysis of potential abnormal user interaction states. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a virtual reality interactive analysis method according to an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the structure of a virtual reality interactive analysis system according to an embodiment of the present invention. Detailed Implementation

[0036] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0037] This invention provides a virtual reality interaction analysis method, please refer to [link / reference]. Figure 1 The method includes:

[0038] Step S1: Obtain the interaction state record data corresponding to multiple experience users of the target virtual reality scene, determine multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of the experience users in each interaction segment, and construct the state response feature matrix of each experience user.

[0039] In this embodiment, the interactive state records of different users during the interaction process include, but are not limited to, data on users' physiological and behavioral parameters, such as heart rate, skin conductance, eye movement data, and other physiological data, as well as data representing head, hand, and body behaviors. To further analyze the state changes of different users during the experience of the target virtual reality scene, multiple interactive segments within the target virtual reality scene can be further identified. This process can be based on the scene's specific content structure, dividing the target virtual reality scene into multiple semantically independent interactive segments, such as specific task steps, key interaction nodes, or scene turning points, or segmenting by time period, with each segment lasting a fixed duration. This embodiment does not specifically limit these methods.

[0040] For each user, extract their state response feature sequence in each interaction segment. This sequence can be represented by multi-dimensional parameters, specifically including feature values ​​corresponding to different interaction state parameters, such as statistical values ​​of different physiological parameters in this time period, and statistical values ​​corresponding to different behavioral parameters such as movement speed, rotation speed of different limbs, click frequency, etc. Combine the state response features extracted from the same user in each interaction segment to form a state response feature matrix, which is used to express the state evolution trajectory of the user in the entire virtual scene as the segment changes.

[0041] Step S2: Extract the fragment state change matrix of the user from the state response feature matrix, extract the fragment state interference features of multiple interactive fragments based on the multiple fragment state change matrices, and perform state transition analysis on multiple interactive fragments by combining the fragment state interference features and the state response feature matrix to construct a reference state transition link for multiple interactive fragments.

[0042] In this embodiment, after constructing the state response feature matrix, the changing trends of the user's state in each interactive segment are extracted to obtain the segment state change matrix, which is used to characterize the relevant dynamic features of each user's state fluctuation or stability in different segments. Furthermore, based on the state changes of multiple users in the same segment, the state stability and interference trends under each segment are evaluated, and segment state interference features are extracted to identify the content impact of each interactive segment on the user's state in the virtual reality scene.

[0043] As an optional implementation process, the fragment state change matrix of the experiencing user is extracted from the state response feature matrix, and fragment state interference features of multiple interaction fragments are extracted based on multiple fragment state change matrices, specifically including:

[0044] Based on the state response feature sequence, generate the fragment state change sequence between two adjacent interactive segments. For example, calculate the difference vector between two state response feature sequences as the fragment state change sequence between interactive segments, thereby constructing a fragment state change matrix containing multiple fragment state change sequences.

[0045] Then, a group behavior analysis was performed on multiple fragment state change sequences in the fragment state change matrix to extract the group change feature parameters of multiple interactive state parameters in the fragment state change sequence, including change direction parameters and change coordination parameters. The behavioral stability parameters of the interactive fragments were calculated based on the group change feature parameters.

[0046] Specifically, for different interaction segments, an analysis is conducted from the perspective of group differences. This involves combining segment state change sequences with group behavior analysis to extract group change characteristic parameters for different interaction states. The segment state change sequence represents the difference in state changes between two adjacent interaction segments, or the state change trend of one of the interaction segments. In the extraction of group change characteristic parameters, relevant group parameters are extracted from the perspectives of consistency in the direction of change and synergy of change.

[0047] In this process, for any user's interaction state parameters in an interactive segment, the group change characteristic parameters are extracted. Based on the segment's state change sequence, the direction of change for different interaction state parameters within the segment is determined (e.g., increasing or decreasing). The proportion of users with this change direction relative to the total number of users is then determined as the direction parameter of the interaction state parameters for a specific user in that segment. Furthermore, the interactive segment is divided into multiple equally spaced windows. The window in which the extreme value of the state characteristic is reached is determined (e.g., maximum heart rate) or the extreme value of the state change rate (e.g., maximum interaction frequency). Then, based on the windows corresponding to multiple users, the window with the largest proportion of users is used as the reference window. The window difference value between the user and the reference window is calculated; that is, the positional distance between the window corresponding to the extreme value of the interaction state parameter in a user's interaction segment and the reference window. Based on the window difference value, the change coordination parameter of the interaction state parameters is determined; the smaller the window difference value, the larger the change coordination parameter. The change direction parameter and change coordination parameter are important indicators that measure the consistency between the changes in different interaction state parameters of different users in a certain interaction segment and the change behavior of the group. The larger the change direction parameter and change coordination parameter are, the smaller the difference between them and the group behavior. Finally, the two are fused, for example, by weighting the fusion with preset weight parameters, to obtain the behavior stability parameters of the experience user in a certain interaction segment with respect to different change coordination parameters.

[0048] Finally, based on multiple behavioral stability parameters of the test users, contextual state stability analysis is performed on multiple interaction segments of the test users to determine the state stability groups of different interaction segments. Based on the state change sequences and stability parameters of multiple segments of the state stability groups, the common state offset parameters of multiple interaction segments with respect to each interaction state parameter are calculated, and the segment state interference characteristics of multiple interaction segments are obtained.

[0049] Specifically, for selecting the stable state group, we can first determine the behavioral stability feature vector of each user under different interaction segments, which is composed of multiple behavioral stability parameters. Then, we determine the two adjacent interaction segments for each interaction segment and calculate the state difference parameter between the user and the two adjacent interaction segments, which is the distance between the stability feature vectors. If both calculated state difference parameters are less than a preset state difference threshold, it indicates that the user is in a locally stable state in that interaction segment, and they are classified into the stable state group for that interaction segment. In this way, we can determine the stable state group corresponding to each interaction segment. Then, for the state change sequences of multiple segments in the stable state group, we calculate the common state offset parameter corresponding to each interaction state parameter by weighting according to the stability parameters. The larger the stability parameter, the less the user's state parameter is affected by local content changes, and the closer it is to normal state changes, such as the natural increase in heart rate after stimulation. By determining the representative group in each interaction segment in the above way, we finally determine the common state offset parameter of each interaction state parameter, which is a quantitative parameter that represents the state offset related to the influence of interaction content changes, and uses it as the segment state interference feature of the interaction segment with respect to multiple interaction state parameters.

[0050] Based on the aforementioned fragment state interference characteristics, and combined with the state response feature matrix, state transition structure modeling is performed on multiple interaction fragments to construct a reference state transition link exhibited by the group of users throughout the entire interaction process. As an optional implementation process, the construction process of the reference state transition link for multiple interaction fragments specifically includes:

[0051] Based on the state interference characteristics of the fragments, state interference correction is performed on each state response feature matrix to generate a state response correction matrix for each user. Based on multiple state response correction matrices, an initial state transition link for multiple interaction fragments is constructed.

[0052] Specifically, for the state response feature sequence of each interaction segment in the state response correction matrix, the common state offset parameter corresponding to that segment is used as the state offset brought about by the interaction content. The offset is subtracted according to the common state offset parameter. Specifically, if the parameter is too high due to scene influence, the actual data collected in that interaction segment is subtracted according to the common state offset parameter; if the parameter is too low, it is supplemented according to the common state offset parameter. This results in a state response correction matrix after removing the influence of actual content changes in the interaction scene, reflecting a more realistic user behavior trajectory after eliminating the interference of segment content. These corrected multi-segment state data are integrated to construct an initial state transition link representing the entire user group in the scene. This link can be formed by statistically aggregating the corrected response feature sequences of all users in each interaction segment, for example, by calculating the mean, representing the reference state nodes of the group in each segment, and forming a coherent state change path. It is worth noting that the above analysis involves analyzing adjacent interactive segments and the contextual content of the interactive segments. For interactive segments at the edges, due to the lack of contextual content, those skilled in the art can conceive of discarding the interactive segments at the edges during the calculation process, or using interpolation to supplement the data preprocessing, so as to successfully complete the above data analysis and calculation process.

[0053] Based on the initial state transition link, state deviation analysis is performed on each state response correction matrix. Specifically, based on the initial state transition link, the state residual of each user in each interaction segment is calculated to assess the degree of deviation between their behavior and the ideal evolution path of the group, and state residual matrices corresponding to different experience users are constructed.

[0054] Based on the behavioral stability parameters of the user in different interaction segments, the residual aggregation weight of the user in each interaction segment in the state residual matrix is ​​determined. Based on the multiple residual aggregation weights corresponding to each user, the state residual matrix is ​​weighted to generate a state aggregation matrix. The state aggregation matrix is ​​then fused with the initial state transition link to obtain a reference state transition link for multiple interaction segments.

[0055] Specifically, to reasonably measure the contribution of each user's residual information to the overall reference construction, a behavioral stability parameter is introduced for differentiated weighting. Specifically, the behavioral stability parameters of each user in each interaction segment are standardized and used as residual aggregation weights. Based on multiple residual aggregation weights, the state residual matrix is ​​weighted and aggregated to generate a state aggregation matrix, resulting in aggregated state sequences for different interaction segments. For example, the residual vectors of users in different interaction segments are weighted using corresponding residual aggregation weights to obtain the reliable offset correction amount of different users regarding the reference path in each interaction segment. This process controls the reliability of residual features in different interaction segments through behavioral stability parameters, which can identify collective abnormal offsets caused by the overall scene, thus preventing the overall path from being judged as abnormal due to severe local interference. The analysis process does not directly judge the feasibility of individual user behavior based on residual features, but rather analyzes the consistency between offsets and the group, which can better identify complex behavioral structures such as offset but reasonable, offset and unreasonable, and no offset but pseudo-consistent. Finally, the state aggregation matrix is ​​fused with the initial state transition path. The state aggregation matrix contains the aggregated state features of the group under different interaction segments. The constructed reference state transition path represents the optimized state evolution path after eliminating scene interference and identifying state deviations, and is used as a benchmark for judging the state of new users.

[0056] Step S3: After collecting the real-time interaction state data of the target user in the target virtual reality scene, construct the real-time state transformation matrix of the target user, and perform interference compensation correction on the real-time state transformation matrix according to the interference characteristics of multiple fragment states to generate the state transformation correction matrix of the target user.

[0057] In this embodiment, for the actual application process of the reference state transition link, after collecting the real-time interaction state data of the target user in the target virtual reality scene, the real-time data is first divided into multiple interaction segments according to the time information, and the real-time state feature sequence of each segment is extracted. Based on this, the real-time state transition matrix of the target user is constructed. Considering the normal state offset caused by the scene content itself, the state transition matrix of the target user is corrected for interference based on the segment state interference characteristics of the aforementioned multiple segments. By eliminating the offset of the state feature sequence, the common influence brought by the scene itself is removed, resulting in a state transition correction matrix that better reflects the user's real interaction response.

[0058] Step S4: Perform anomaly analysis on the target user's interaction state based on the reference state transition link and state transformation correction matrix, and generate the anomaly analysis results of the target user's interaction state in the target virtual reality scene.

[0059] In this embodiment, by combining the aforementioned constructed reference state transition link with the target user's state transition correction matrix, a dynamic evolution analysis of the target user's state evolution trend during the interaction process is performed. This identifies whether the target user exhibits behavior that significantly deviates from the group reference path, while also considering the dynamic changes in the deviated behavior. This generates anomaly analysis results of the user's interaction state in the current virtual reality scene, which can be used for subsequent content adaptation, risk warning, interaction adjustment, or behavior feedback, thereby improving the intelligent response capability and personalized experience level of the virtual reality system.

[0060] As an optional implementation process, anomaly analysis of the target user's interaction state is performed based on the reference state transition link and the state transition correction matrix, specifically including:

[0061] Based on the reference state transition link and the state transformation correction matrix, a real-time deviation feature sequence of the target user is constructed for each interaction segment. This sequence represents the target user's deviation state relative to the group reference path in a specific interaction scenario across different interaction segments. Then, real-time deviation parameters for the corresponding interaction segments are calculated based on the real-time deviation feature sequence, for example, using the L2 norm to represent the intensity of the target user's state deviation in that interaction segment. Simultaneously, for multiple state response correction matrices under different interaction segments, individual deviation parameters corresponding to each state response correction matrix are calculated using the same method. These individual deviation parameters are then fused, for example, by calculating the mean, as a reference deviation index for the interaction segment. This index represents a reference indicator that the specific interaction segment is in a normal deviation state within the group.

[0062] Then, based on multiple reference deviation indices, segment anomaly analysis is performed on the real-time deviation parameters of the target user under different interaction segments. The difference between the real-time deviation parameters and the reference deviation indices is calculated to obtain the local state anomaly parameters of the target user under each interaction segment. Based on multiple local state anomaly parameters, state evolution anomaly analysis is performed on the target user. This includes identifying multiple state anomaly segments of the target user based on preset local state anomaly thresholds, and then performing state evolution anomaly analysis on multiple interaction segments using a sliding window approach. If continuous state anomalies are detected, such as multiple interaction segments within a feature window all belonging to state anomaly segments, it indicates that the user has an abnormal trend. The above analysis process specifically considers the instability of different interaction segments themselves. For the same degree of state deviation, anomalies in high-stability scenarios are more suspicious, while for scenarios with lower inherent stability, the stimuli brought by the scenario may lead to local anomalies. Furthermore, through state evolution anomaly analysis, the persistence of local anomalies is monitored. If they only appear briefly, it indicates a normal state; if they persist, the user's state needs to be closely monitored to accurately identify potential abnormal states of the user during virtual reality interaction and improve the user's interactive experience.

[0063] Based on the virtual reality interaction analysis method described above, this invention also provides a virtual reality interaction analysis system. Please refer to [link to relevant documentation]. Figure 2 The system includes:

[0064] The state feature extraction module 01 is used to acquire the interaction state record data corresponding to multiple experience users of the target virtual reality scene, determine multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of the experience user in each interaction segment, and construct the state response feature matrix of each experience user.

[0065] The reference state link construction module 02 is used to extract the fragment state change matrix of the user from the state response feature matrix, extract the fragment state interference features of multiple interactive fragments based on the multiple fragment state change matrices, and perform state transition analysis on multiple interactive fragments by combining the fragment state interference features and the state response feature matrix to construct a reference state transition link for multiple interactive fragments.

[0066] The state change correction module 03 is used to construct the real-time state change matrix of the target user after collecting the real-time interaction state data of the target user in the target virtual reality scene, and to perform interference compensation correction on the real-time state change matrix according to the interference characteristics of multiple fragment states, so as to generate the state change correction matrix of the target user.

[0067] The interaction state anomaly analysis module 04 is used to perform interaction state anomaly analysis on the target user based on the reference state transition link and the state transformation correction matrix, and generate the interaction state anomaly analysis results of the target user in the target virtual reality scene.

[0068] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A virtual reality interaction analysis method, characterized in that, include: Acquire interaction state records for multiple users in the target virtual reality scene, identify multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of each user in each interaction segment, and construct the state response feature matrix for each user. The fragment state change matrix of the user is extracted from the state response feature matrix. Based on these fragment state change matrices, fragment state interference features of multiple interaction fragments are extracted. Combining these fragment state interference features and the state response feature matrix, state transition analysis is performed on multiple interaction fragments to construct a reference state transition chain for these fragments. Specifically, this includes: Based on the state response feature sequence, generate the fragment state change sequence between two adjacent interaction fragments, and construct a fragment state change matrix containing multiple fragment state change sequences; Collective behavior analysis is performed on multiple fragment state change sequences in the fragment state change matrix. Collective change characteristic parameters of multiple interactive state parameters in the fragment state change sequence are extracted, including change direction parameters and change coordination parameters. Based on the collective change characteristic parameters, the behavioral stability parameters of the interactive fragments are calculated. Based on multiple behavioral stability parameters of the test users, context state stability analysis is performed on multiple interaction segments of the test users to determine the state stability groups of different interaction segments. Based on the state change sequences and stability parameters of multiple segments of the state stability groups, the common state offset parameters of multiple interaction segments with respect to each interaction state parameter are calculated, and the segment state interference characteristics of multiple interaction segments are obtained. Based on the state interference characteristics of the fragments, state interference correction is performed on each state response feature matrix to generate a state response correction matrix for each user. Based on the multiple state response correction matrices, an initial state transition link for multiple interaction fragments is constructed. Based on the initial state transition link, state deviation analysis is performed on each state response correction matrix to construct a state residual matrix for multiple experience users; Based on the behavioral stability parameters of the user in different interaction segments, the residual aggregation weight of the user in each interaction segment in the state residual matrix is ​​determined. Based on the multiple residual aggregation weights corresponding to each user, the state residual matrix is ​​weighted to generate a state aggregation matrix. The state aggregation matrix is ​​then fused with the initial state transition link to obtain a reference state transition link for multiple interaction segments. After collecting the real-time interaction state data of the target user in the target virtual reality scene, a real-time state transformation matrix of the target user is constructed. The real-time state transformation matrix is ​​then corrected for interference based on the interference characteristics of multiple fragment states, and a state transformation correction matrix of the target user is generated. Based on the reference state transition link and the state transition correction matrix, the interaction state anomaly analysis of the target user is performed, and the analysis results of the interaction state anomaly analysis of the target user in the target virtual reality scene are generated.

2. The virtual reality interaction analysis method according to claim 1, characterized in that, Anomaly analysis of the target user's interaction state is performed based on the reference state transition link and the state transition correction matrix, including: Based on the reference state transition link and the state transition correction matrix, a real-time deviation feature sequence of the target user in each interaction segment is constructed, and the real-time deviation parameter of each real-time deviation feature sequence is calculated. The reference deviation index for each interaction segment is calculated by referencing the state transition link and multiple state response correction matrices under different interaction segments. Based on multiple reference deviation indices, segment anomaly analysis is performed on the real-time deviation parameters of the target user under different interaction segments. The local state anomaly parameters of the target user under each interaction segment are calculated. Based on multiple local state anomaly parameters, state evolution anomaly analysis is performed on the target user, and the interaction state anomaly analysis results of the target user in the target virtual reality scene are generated.

3. The virtual reality interaction analysis method according to claim 1, characterized in that, Extract the population change feature parameters of multiple interaction state parameters in the fragment state change sequence, including: Based on the sequence of state changes in the segment, determine the direction of change of different interactive state parameters in the interactive segment, and count the percentage of users in the interactive segment whose direction of change of interactive state parameters is the same as that of interactive state parameters to obtain the direction of change of interactive state parameters. The interaction segment is segmented into windows. The reference window of the interaction segment is determined based on the eigenvalues ​​corresponding to the eigenvalues ​​of the multiple interaction state parameters in the interaction segment. The window difference value of the interaction state parameters is calculated based on the positional distance between the window to which the eigenvalue of the interaction state parameter belongs and the reference window of the interaction segment. The change coordination parameter of the interaction state parameter is determined based on the window difference value.

4. The virtual reality interaction analysis method according to claim 2, characterized in that, For real-time deviation parameters and reference deviation indices, the following are also included: The norm features of the real-time deviation feature sequence are extracted to obtain the real-time deviation parameters of the real-time deviation feature sequence. Multiple norm features of the state response correction matrix are extracted to obtain multiple individual deviation parameters of the state response correction matrix. Multiple individual deviation parameters under each interaction segment are fused to obtain the reference deviation index under each interaction segment.

5. A virtual reality interactive analysis system, characterized in that, The system is used to implement the virtual reality interactive analysis method according to any one of claims 1-4, comprising: The state feature extraction module is used to acquire the interaction state record data corresponding to multiple experience users of the target virtual reality scene, determine multiple interaction segments in the target virtual reality scene, extract the state response feature sequence of the experience user in each interaction segment, and construct the state response feature matrix of each experience user. The reference state link construction module is used to extract the fragment state change matrix of the user from the state response feature matrix, extract the fragment state interference features of multiple interactive fragments based on the multiple fragment state change matrices, and perform state transition analysis on multiple interactive fragments by combining the fragment state interference features and the state response feature matrix to construct a reference state transition link for multiple interactive fragments. The state change correction module is used to construct the real-time state change matrix of the target user after collecting the real-time interaction state data of the target user in the target virtual reality scene, and to perform interference compensation correction on the real-time state change matrix according to the interference characteristics of multiple fragment states, so as to generate the state change correction matrix of the target user. The interaction state anomaly analysis module is used to perform interaction state anomaly analysis on the target user based on the reference state transition link and the state transformation correction matrix, and generate the interaction state anomaly analysis results of the target user in the target virtual reality scene.

Images