Interaction system and method of 3D virtual digital human

Abstract

The invention discloses an interaction system and method of a 3D virtual digital human, relates to the technical field of man-machine interaction and artificial intelligence, and is used for solving the problem of poor interaction adaptability caused by dependence on a preset template and lack of natural and personalized expression in a traditional method. Preliminarily matching an action model range according to an interaction type; in the interaction stage, the signal intensity of input equipment is detected, an intensity condition is set to extract a data stream, initial response logic is established, dynamic behaviors are collected, and a random forest model is constructed to predict a behavior trend; matching classification is carried out by combining the prediction result and the model, if the matching is low, interaction samples are collected through an auxiliary channel to carry out visualized simulation, and matching samples and models are screened; and matching an interaction scene based on a simulation result, replacing the initial scene and then executing an action, thereby realizing intelligent response and natural interaction of the virtual digital human.

Description

Technical Field

[0001] The present invention relates to the field of human-computer interaction and artificial intelligence technology, and more specifically, to an interaction system and method for a 3D virtual digital human. Background Art

[0002] With the development of virtual reality (VR), augmented reality (AR) and digital twin technologies, 3D virtual digital humans, as an important carrier of human-computer interaction, are widely used in online education, virtual customer service, digital exhibitions, film and television entertainment and other fields; in existing technologies, the interactive behavior of virtual digital humans mostly relies on preset action models and static matching rules, that is, by identifying user input signals and matching corresponding action instructions, the virtual digital humans are driven to perform specific behaviors.

[0003] The existing technology has the following deficiencies: At present, the responses of virtual digital humans in traditional methods mostly rely on preset templates, and cannot generate differentiated actions in real time according to user behavior. They lack natural and personalized performance, resulting in poor interaction adaptability and inability to dynamically optimize action matching strategies based on historical interaction samples, which reduces the intelligence level of the virtual interaction system and the quality of user experience. Therefore, an interaction system and method for 3D virtual digital humans are proposed.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an interactive system and method for 3D virtual digital humans, which solve the problems raised in the above-mentioned background technology by applying a dynamic behavior prediction model, an input signal strength perception mechanism, a sample embodiment simulation method and a scene adaptive control strategy.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a 3D virtual human interaction system and method, comprising the following steps: Step S1: Obtain the current behavior parameters of the virtual digital human, collect user interaction intention information, calculate the basic action logic based on the above information, and preliminarily match the action model range according to the interaction type; Step S2: During the user interaction phase, the input device signal strength is detected, different strength conditions are set to detect the interactive data flow, the initial response logic of the virtual digital human is set through the model range, its dynamic behavior performance is collected, and a random forest model is constructed to predict its dynamic behavior; Step S3: The behavior adaptability is classified by combining the dynamic behavior prediction value with the preliminary matching action model. If the adaptability is low, user interaction samples are collected through the auxiliary channel and simulated in a concrete manner to screen out samples within the standard adaptability range and their corresponding models. Step S4: Based on the matching scenario of the sample interaction data of the embodiment simulation calculation, the scenario matched by the virtual digital human interaction data flow is calculated, and the corresponding action is executed after replacing the initial setting scenario.

[0007] In a preferred embodiment, in step S1, current behavior parameters of the virtual digital human are obtained from the system, including a spatial position vector, an action identifier, an expression state code, and an action intensity, and the current behavior parameters are combined into a current state vector;

[0008] Collect user interaction intention information from the input device, parse the user interaction intention information through intelligent recognition tools to extract interaction semantics, obtain its interaction type, target object, emotional tendency, and combine them into a user intention vector.

[0009] In a preferred embodiment, in step S1, the current state vector and the user intention vector are integrated to construct a basic action logic to obtain a candidate action model: The current state vector and the user intention vector are combined as input conditions, rules are matched in the preset rule library, the action identifiers corresponding to the matching rule items are filtered out, and the candidate action model set is retrieved; Retrieve the action model subset belonging to the current interaction type category in the preset action library, filter the candidate action model set, and retain the candidate action models belonging to the current interaction type subset as the preliminary matching action model range.

[0010] In a preferred embodiment, in step S2, the signals of the user input devices are detected in real time, and the strengths of different input devices are weighted by preset weights to obtain the device signal strength value; The first strength threshold and the second strength threshold are preset to divide the device signal strength into different user intent strength levels: If the device signal strength is greater than or equal to the first strength threshold, it is determined to be a high intention interaction strength; If the device signal strength is between the first strength threshold and the second strength threshold, it is determined to be a medium intention interaction strength; If the device signal strength is less than or equal to the first strength threshold, it is determined to be low intended interaction strength.

[0011] In a preferred embodiment, in step S2, the interaction data streams at different user intention strength levels are detected, including the current state vector, the user intention vector, and the device signal strength; Build the initial response logic based on the user's intention intensity level and the range of candidate action models, and match the response action features of the current intention intensity level as the initial response action model; Construct a multi-level condition matching mechanism based on the user intention intensity level and candidate action model: Condition 1: The user intention strength matches the action strength to screen candidate action models; Condition 2: Filter the candidate action model with the largest number of coordinate points in the action path parameters; Condition 3: Expression adjustment parameters match emotional tendencies to screen candidate action models; The candidate action model finally screened out is used as the initial response action model, and the virtual digital human performs the corresponding action while recording the current state vector during the response execution process.

[0012] In a preferred embodiment, a random forest model is constructed based on the interactive data stream and the current state vector to predict the movement trend of the virtual digital human; The interaction data stream and the current state vector are combined as samples. Non-numerical features are converted into numerical types through feature encoding, and a time series sample set is generated using a time sliding window. After random forest integration and training, the predicted behavior vector is obtained, including: Action identifier: Select the action identifier with the highest frequency among the action identifiers predicted by the sub-decision tree as the action identifier in the predicted behavior vector;

[0013] Action intensity: The arithmetic average of the action intensity values ​​predicted by the corresponding sub-decision tree is used as the action intensity in the predicted behavior vector.

[0014] In a preferred embodiment, in step S3, the initial response action model is classified for suitability based on the predicted behavior vector: If the action identifiers are consistent, it is judged that the adaptability is high, otherwise it is judged that the adaptability is low; When the adaptability is judged to be high, the initial response action model is used as the standard adaptability action model; When the adaptability is judged to be low, interaction samples are collected through auxiliary channels, including the user intention vector, current state vector, initial response action model, and environmental context information recorded during the interaction process; By receiving interaction samples through concrete simulation, a virtual interaction scene matching the interaction sample information is constructed, and the virtual digital human is driven to perform corresponding actions based on the candidate action model.

[0015] In a preferred embodiment, in step S3, dynamic behavior response data parameters during the embodiment simulation are collected, and the parameters are normalized and summed to obtain a comprehensive response score; Determine whether the interaction sample is a standard adaptability interaction sample based on the comprehensive response score: If the comprehensive response score is greater than the preset score threshold, it is judged as a standard adaptive action model; If the response comprehensive score is less than the preset score threshold, auxiliary channel sampling and materialization simulation are performed.

[0016] In a preferred embodiment, in step S4, a matching scenario of sample interaction data is simulated and calculated based on the standard adaptability sample, and the dynamic behavior response data is normalized and merged into an interaction feature vector; Each scene label in the preset scene label library corresponds to a different scene feature vector, and the matching degree between the interaction feature vector and the scene feature vector is calculated by Euclidean distance; The current scene label of the virtual digital human is replaced with the scene label with the greatest similarity, and the specific response of the action execution is completed according to the standard adaptive action model.

[0017] The interactive system of 3D virtual digital human includes interaction intention analysis module, dynamic behavior prediction module, adaptability judgment and simulation module, and scene reasoning and action execution module; The interaction intention analysis module is used to collect virtual digital human behavior parameters and user interaction intentions, build basic action logic, and match the action model range; The dynamic behavior prediction module sets the initial response logic based on signal strength, collects dynamic behavior data, uses random forest to predict behavior, and passes it to subsequent modules for behavior adaptability assessment; The adaptability determination and simulation module is used to evaluate behavioral adaptability, assist in channel sample collection and simulation, and screen out samples within the standard adaptability range and their corresponding models;

[0018] The scene reasoning and action execution module is used for scene reasoning and replacement, realizing action scheduling and virtual digital human execution.

[0019] Technical effects and advantages of the present invention: 1. The present invention obtains the current behavior parameters of the virtual digital human, collects user interaction intention information, calculates the basic action logic based on the above information, and preliminarily matches the action model range according to the interaction type. During the user interaction stage, the input device signal strength is detected, and different strength conditions are set to detect the interaction data flow. The initial response logic of the virtual digital human is set according to the model range, and its dynamic behavior performance is collected. A random forest model is constructed to predict its dynamic behavior. The behavior adaptability is classified by combining the dynamic behavior prediction value with the preliminarily matched action model. If the adaptability is low, user interaction samples are collected through auxiliary channels, and a concrete simulation is performed to screen out samples within the standard adaptability range and their corresponding models. Based on the concrete simulation, the matching scenarios of the sample interaction data are calculated, and the scenarios matched by the virtual digital human interaction data flow are inferred. The corresponding actions are executed after replacing the initially set scenarios, thereby achieving the virtual digital human's deep understanding of user input and natural and smooth action behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the method for interacting with a 3D virtual human according to the present invention.

[0021] Figure 2 Schematic diagram of the modules of the interactive system of 3D virtual digital human of the present invention. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] Example 1

[0024] See also Figure 1 , the interaction method of 3D virtual digital human, the specific operation process is as follows: Step S1: Obtain the current behavior parameters of the virtual digital human, collect user interaction intention information, calculate the basic action logic based on the above information, and preliminarily match the action model range according to the interaction type; Step S2: During the user interaction phase, the input device signal strength is detected, different strength conditions are set to detect the interactive data flow, the initial response logic of the virtual digital human is set through the model range, its dynamic behavior performance is collected, and a random forest model is constructed to predict its dynamic behavior; Step S3: The behavior adaptability is classified by combining the dynamic behavior prediction value with the preliminary matching action model. If the adaptability is low, user interaction samples are collected through the auxiliary channel and simulated in a concrete manner to screen out samples within the standard adaptability range and their corresponding models. Step S4: Based on the matching scenario of the sample interaction data of the embodiment simulation calculation, the scenario matched by the virtual digital human interaction data flow is calculated, and the corresponding action is executed after replacing the initial setting scenario.

[0025] The specific implementation is as follows: In step S1, the current behavior parameters of the virtual digital human are obtained from the system, including the spatial position vector, action identifier, expression state code and action intensity, and the current behavior parameters are combined into a current state vector; The spatial position vector is the current position coordinate of the virtual digital human in three-dimensional space, the action identifier is the action model number performed by the current virtual digital human, the expression state code represents the set of code values ​​of the current state of the facial expression, and the action intensity is the dynamic amplitude of the current action execution.

[0026] Collect user interaction intention information from input devices, including voice input information, image input information, touch input information, motion input information, and text input information; Intelligent recognition tools are used to parse user interaction intention information and extract interaction semantics to unify user interaction intentions in different modalities, obtain their interaction type, target object, and emotional tendency, and combine them into a user intention vector. The interaction type is the user's interactive behavior category, the target object is the object identifier or location information of the intended action, and the emotional tendency is the user's emotional orientation during interaction.

[0027] The current state vector and user intention vector are combined to build the basic action logic to obtain the candidate action model: The current state vector and the user intention vector are combined as the input conditions for rule matching. Each rule item in the preset rule library is matched in sequence. When the matching conditions of the rule item are consistent with the input conditions, the corresponding rule item is judged as a matching rule item. The action identifier in the matching rule item is filtered out, and the corresponding candidate action model is retrieved and merged into a candidate action model set; The candidate action model includes action identifier, action start and end time, action intensity, action path parameters, and expression adjustment parameters; Obtain the current interaction type and further filter the candidate action model set. Classify the virtual digital human action models according to different interaction types in the preset action library and divide them into different action model subsets corresponding to the semantics of different interaction types, for example, into action model subsets such as "greeting", "request", "feedback", and "command"; Filter the action model subset corresponding to the interaction type, and select the candidate action models belonging to the action model subset from the candidate action model set as the preliminary matching action model range; It should be noted that the input device refers to a hardware device used to collect user input and obtain multimodal input signals of the user during the interaction process, including voice commands, facial expressions, gestures, keyboard input, etc.; intelligent recognition tools include but are not limited to: automatic speech recognition engines, image or video analysis frameworks, facial expression recognition tools and device signal monitoring modules to realize the analysis of user interaction intention information. For example, the voice input information is transcribed by the automatic speech recognition engine and input into the natural language understanding model to extract the interaction type label in the sentence. For the image or video stream captured by the camera device, the coordinates of the user's skeletal key points are extracted and calculated by the image or video analysis framework to determine the user's action type, etc.; the preset rule library is a set of behavior decision rules pre-defined and stored in the system, which is used to determine the adaptive action logic output based on the current behavior state of the virtual digital human and the user's interaction intention; the preset action library is a set of predefined and callable structured action templates according to different interaction types. For example, if the interaction type is a greeting, the matching actions include greeting, waving, smiling, etc. In step S2, the signals of the user input device are detected in real time, including the volume intensity in the voice channel, the gesture recognition confidence in the video channel, and the hand movement amplitude in the touch channel; Calculate the device signal strength based on the integrated input device signals: ,in, 、 、 , 、 、 They are volume intensity, gesture recognition confidence, and hand movement amplitude, is the device signal strength value; A first strength threshold and a second strength threshold are preset, where the first strength threshold is greater than the second strength threshold, and the device signal strength is divided into different user intention strength levels to achieve hierarchical recognition of user interaction intentions: If the device signal strength is greater than or equal to the first strength threshold, it is determined to be a high intention interaction strength; If the device signal strength is between the first strength threshold and the second strength threshold, it is determined to be a medium intention interaction strength; If the device signal strength is less than or equal to the first strength threshold, it is determined to be low intended interaction strength.

[0028] Detect interaction data streams at different levels of user intent strength, including user intent vectors and device signal strength, and use the interaction data streams as input data for the random forest model; Build the initial response logic based on the user's intention intensity level and the range of candidate action models, and match the response action features of the current intention intensity level as the initial response action model; Construct a multi-level condition matching mechanism based on the user intention intensity level and candidate action model: Condition 1: Match the user intention strength with the action strength to screen candidate action models; Condition 2: Filter the candidate action model with the largest number of coordinate points in the action path parameters; Condition 3: Expression adjustment parameters match emotional tendencies to screen candidate action models; The candidate action model finally screened is used as the initial response action model, and the virtual digital human performs the corresponding action while recording the current state vector during the response execution process; A random forest model is built based on the interaction data stream and the current state vector to predict the subsequent movement trend of the virtual digital human: Data collection: Combine the interactive data stream and the current state vector as samples, map the non-numerical features in the samples into numerical form through the feature encoding conversion mechanism, collect samples at different times through a time sliding window, and merge them into a sample set in time series; The process of building a single subject decision tree: Randomly extract m sub-features from all the features of the sample, where m is less than the total number of features, to form the candidate feature set of the current tree; Traverse each feature in the candidate feature set and calculate the split point based on the mean square error minimization principle: For each candidate feature, traverse all split points, calculate the sum of the variances of the left and right subset target variables after splitting, and select the feature with the smallest total variance and its split point as the basis for splitting the current node; Repeat the above steps for the generated left and right child nodes, and recursively build the subtree until the number of node samples is less than the preset sample threshold; Random forest ensemble and training: K subsamples are randomly sampled from the sample set with replacement to ensure that the training data of each tree is diverse. The process of building a single-subject decision tree is repeated for each decision tree, and a subset of candidate features is randomly selected when splitting nodes at each layer. Prediction result output: The output of a single decision tree is a multidimensional prediction vector, including action identifiers and action strengths. The frequency of occurrence of the action identifiers predicted by all decision trees is counted, and the one with the highest frequency is selected as the action identifier. The arithmetic mean of the action strengths predicted by the corresponding single decision tree is calculated as the action strength and combined with the action identifier to form a predicted behavior vector.

[0029] It should be noted that the random forest model is an integrated learning algorithm that makes predictions by constructing multiple decision trees; the encoding conversion mechanism refers to the conversion of non-numeric category features into numerical representations through encoding tools; action identifiers are discrete labels and do not require numerical conversion. The system directly performs frequency statistics on the action identifiers output by all decision trees; the preset sample threshold is used to determine whether to continue splitting the current node and is set by professionals.

[0030] In step S3, the initial response action model is classified for suitability based on the predicted behavior vector: If the action identifier in the predicted behavior vector is consistent with the action identifier in the initial response action model, it is judged that the adaptability is high; otherwise, it is judged that the adaptability is low.

[0031] When the adaptability is judged to be high, the initial response action model is used as the standard adaptability action model; When the adaptability is judged to be low, user interaction samples are collected through the auxiliary channel, including the user intention vector, current state vector and initial response action model, as well as the environmental context information recorded during the interaction process. The environmental context information includes the interaction scene identifier, interaction time, user status, and the behavior sequence and results of the previous interaction; By receiving interaction samples collected by the auxiliary sampling channel through concrete simulation, a virtual interaction scene matching the interaction sample information is constructed. Based on the initial response action model, the virtual digital human is driven to perform corresponding actions. The action posture, expression changes and action strength of the virtual digital human are rendered in real time to generate a continuous and natural 3D animation. In the constructed virtual interactive scene, the embodied simulation engine initially responds to the action model and drives the virtual digital human to perform the corresponding action. The system renders the virtual digital human's action posture, expression changes, and action strength in real time, generating a continuous and natural 3D animation process. During the action simulation process, the embodiment simulation engine continuously collects dynamic behavior response data, including action trajectory similarity, expression and emotion matching score, spatial interaction target hit rate, behavior continuity score, and action response delay time. The dynamic behavior response data is normalized and summed to obtain a comprehensive response score. Determine whether the interaction sample is a standard adaptability interaction sample based on the comprehensive response score: If the response comprehensive score is greater than the preset score threshold, it is judged as a standard adaptive interaction sample and associated with its corresponding candidate action model; If the response comprehensive score is less than the preset score threshold, auxiliary channel sampling and materialization simulation are performed; It should be noted that the auxiliary channel refers to the additional data collection and sample generation path that is initiated when the system detects that the current action model library cannot meet the high adaptability requirements during the interaction between the virtual digital human and the user; the embodiment simulation refers to the use of three-dimensional simulation technology to truly reproduce the interaction process between the user and the virtual digital human in a virtual environment based on the collected user interaction samples and the current behavior parameters of the virtual digital human, so as to realize the visualization and dynamic verification of the interaction scenario; the preset scoring threshold is set by professionals. For example, the sum of the mean and standard deviation of the adaptability score is used as the preset scoring threshold. The method is not unique and will not be elaborated here.

[0032] In step S4, based on the standard adaptability sample and the standard adaptability action model, a matching scenario of sample interaction data is simulated and calculated, dynamic behavior response data of the simulated is continuously collected, and normalized and merged into an interaction feature vector; Each set of scene labels in the preset scene label library corresponds to a different scene feature vector, where the dimension and meaning of each parameter are consistent with the interaction feature vector. The matching degree between the interaction feature vector and the scene feature vector is calculated using the Euclidean distance: ,in, is the i-th parameter in the interaction feature vector, is the parameter corresponding to the i-th scene in the scene feature vector, m is the total number of parameters, is similarity; The current scene label of the virtual digital human is replaced with the scene label with the greatest similarity. In the updated matching scene, the system completes the specific driving of action execution according to the standard adaptive action model, including action trajectory control, expression state encoding and action intensity adjustment.

[0033] The system realizes precise scene matching and dynamic scene replacement based on concrete simulation samples, ensuring the scene adaptability and execution effect of virtual digital human's interactive actions, and realizing efficient and natural multimodal user interaction experience.

[0034] It should be noted that the preset scene label library is used to establish a label system for various interactive scenes. The scene label corresponds to the scene feature vector, which is used to numerically describe the interactive properties of the corresponding scene.

[0035] Example 2

[0036] Interactive systems based on 3D virtual digital humans, such as Figure 2 As shown, it is used to implement an interaction method based on 3D virtual digital human, including an interaction intention analysis module, a dynamic behavior prediction module, an adaptability judgment and simulation module, and a scene reasoning and action execution module; The interaction intention analysis module is used to collect virtual digital human behavior parameters and user interaction intentions, build basic action logic, and match the action model range; The dynamic behavior prediction module sets the initial response logic based on signal strength, collects dynamic behavior data, uses random forest to predict behavior, and passes it to subsequent modules for behavior adaptability assessment; The adaptability determination and simulation module is used to evaluate behavioral adaptability, assist in channel sample collection and simulation, and screen out samples within the standard adaptability range and their corresponding models; The scene reasoning and action execution module is used for scene reasoning and replacement, realizing action scheduling and virtual digital human execution.

[0037] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0038] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0039] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0040] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0041] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0042] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0043] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0044] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0045] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0046] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0047] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0048] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A 3D virtual human interaction method, characterized by: The following steps are involved: Step S1: Obtain the current behavior parameters of the virtual digital human, collect user interaction intention information, calculate the basic action logic based on the above information, and preliminarily match the action model range according to the interaction type; Step S2: During the user interaction phase, the input device signal strength is detected, different strength conditions are set to detect the interactive data flow, the initial response logic of the virtual digital human is set through the model range, its dynamic behavior performance is collected, and a random forest model is constructed to predict its dynamic behavior; Step S3: The behavior adaptability is classified by combining the dynamic behavior prediction value with the preliminary matching action model. If the adaptability is low, user interaction samples are collected through the auxiliary channel and simulated in a concrete manner to screen out samples within the standard adaptability range and their corresponding models. Step S4: Based on the matching scenario of the sample interaction data of the embodiment simulation calculation, the scenario matched by the virtual digital human interaction data flow is calculated, and the corresponding action is executed after replacing the initial setting scenario.

2. The 3D virtual human interaction method according to claim 1, characterized in that: In step S1, the current behavior parameters of the virtual digital human are obtained from the system, including the spatial position vector, action identifier, expression state code and action intensity, and the current behavior parameters are combined into a current state vector; Collect user interaction intention information from the input device, parse the user interaction intention information through intelligent recognition tools to extract interaction semantics, obtain its interaction type, target object, emotional tendency, and combine them into a user intention vector.

3. The 3D virtual human interaction method according to claim 2, characterized in that: In step S1, the current state vector and the user intention vector are combined to construct the basic action logic to obtain the candidate action model: The current state vector and the user intention vector are combined as input conditions, rules are matched in the preset rule library, the action identifiers corresponding to the matching rule items are filtered out, and the candidate action model set is retrieved; Retrieve the action model subset belonging to the current interaction type category in the preset action library, filter the candidate action model set, and retain the candidate action models belonging to the current interaction type subset as the preliminary matching action model range.

4. The 3D virtual human interaction method according to claim 3, characterized in that: In step S2, the signals of the user input devices are detected in real time, and the strengths of different input devices are weighted by preset weights to obtain the device signal strength value; The first strength threshold and the second strength threshold are preset to divide the device signal strength into different user intent strength levels: If the device signal strength is greater than or equal to the first strength threshold, it is determined to be a high intention interaction strength; If the device signal strength is between the first strength threshold and the second strength threshold, it is determined to be a medium intention interaction strength; If the device signal strength is less than or equal to the first strength threshold, it is determined to be low intended interaction strength.

5. The 3D virtual human interaction method according to claim 4, characterized in that: In step S2, the interaction data streams at different user intention strength levels are detected, including the current state vector, the user intention vector, and the device signal strength; Build the initial response logic based on the user's intention intensity level and the range of candidate action models, and match the response action features of the current intention intensity level as the initial response action model; Construct a multi-level condition matching mechanism based on the user intention intensity level and candidate action model: Condition 1: The user intention strength matches the action strength to screen candidate action models; Condition 2: Filter the candidate action model with the largest number of coordinate points in the action path parameters; Condition 3: Expression adjustment parameters match emotional tendencies to screen candidate action models; The candidate action model finally screened out is used as the initial response action model, and the virtual digital human performs the corresponding action while recording the current state vector during the response execution process.

6. The 3D virtual human interaction method according to claim 5, characterized in that: Build a random forest model based on the interactive data stream and the current state vector to predict the movement trend of the virtual digital human; The interaction data stream and the current state vector are combined as samples. Non-numerical features are converted into numerical types through feature encoding, and a time series sample set is generated using a time sliding window. After random forest integration and training, the predicted behavior vector is obtained, including: Action identifier: Select the action identifier with the highest frequency among the action identifiers predicted by the sub-decision tree as the action identifier in the predicted behavior vector; Action intensity: The arithmetic average of the action intensity values ​​predicted by the corresponding sub-decision tree is used as the action intensity in the predicted behavior vector.

7. The 3D virtual human interaction method according to claim 6, characterized in that: In step S3, the initial response action model is classified for suitability based on the predicted behavior vector: If the action identifiers are consistent, it is judged that the adaptability is high, otherwise it is judged that the adaptability is low; When the adaptability is judged to be high, the initial response action model is used as the standard adaptability action model; When the adaptability is judged to be low, interaction samples are collected through auxiliary channels, including the user intention vector, current state vector, initial response action model, and environmental context information recorded during the interaction process; By receiving interaction samples through concrete simulation, a virtual interaction scene matching the interaction sample information is constructed, and the virtual digital human is driven to perform corresponding actions based on the candidate action model.

8. The 3D virtual human interaction method according to claim 7, characterized in that: In step S3, dynamic behavior response data parameters during the embodiment simulation are collected, and the parameters are normalized and summed to obtain a comprehensive response score; Determine whether the interaction sample is a standard adaptability interaction sample based on the comprehensive response score: If the comprehensive response score is greater than the preset score threshold, it is judged as a standard adaptive action model; If the response comprehensive score is less than the preset score threshold, auxiliary channel sampling and materialization simulation are performed.

9. The 3D virtual human interaction method according to claim 8, characterized in that: In step S4, a matching scenario of sample interaction data is simulated and calculated based on the standard adaptability sample, and the dynamic behavior response data is normalized and merged into an interaction feature vector; Each scene label in the preset scene label library corresponds to a different scene feature vector, and the matching degree between the interaction feature vector and the scene feature vector is calculated by Euclidean distance; The current scene label of the virtual digital human is replaced with the scene label with the greatest similarity, and the specific response of the action execution is completed according to the standard adaptive action model.

10. A 3D virtual human interaction system, configured to implement the 3D virtual human interaction method according to any one of claims 1 to 9, characterized in that: It includes interaction intention analysis module, dynamic behavior prediction module, adaptability judgment and simulation module, and scene reasoning and action execution module; The interaction intention analysis module is used to collect virtual digital human behavior parameters and user interaction intentions, build basic action logic, and match the action model range; The dynamic behavior prediction module sets the initial response logic based on signal strength, collects dynamic behavior data, uses random forest to predict behavior, and passes it to subsequent modules for behavior adaptability assessment; The adaptability determination and simulation module is used to evaluate behavioral adaptability, assist in channel sample collection and simulation, and screen out samples within the standard adaptability range and their corresponding models; The scene reasoning and action execution module is used for scene reasoning and replacement, realizing action scheduling and virtual digital human execution.

Images