Virtual digital human action processing system

Abstract

The invention relates to the technical field of virtual simulation data processing, in particular to a virtual digital human action processing system. The system comprises an image generation module used for generating a virtual digital person whose appearance accords with the expectation of a user; the configuration module is used for determining the importance level of each motion part of the virtual digital human according to the application scene, and configuring a skeleton model of a corresponding precision level for each motion part of the virtual digital human according to the importance level; the higher the importance level of the motion part of the virtual digital human is, the higher the precision of the corresponding skeleton model is, and the higher the precision of the skeleton model represents that the motion of the corresponding motion part of the virtual digital human is closer to the real human; and the action control module is used for controlling each movement part of the virtual digital human to perform actions with corresponding precision according to the instruction information of the user and a preset action rule. By adopting the system provided by the invention, the data processing efficiency can be improved, so that the action performance precision of the virtual digital human and the watching experience of the user are improved.

Description

Technical Field

[0001] The present application relates to the technical field of virtual simulation data processing, and in particular to a virtual digital human action processing system. Background Art

[0002] Driving virtual digital humans has always been a research hotspot in academia and has a wide range of industrial applications. The most common application areas are virtual anchors, virtual customer service, virtual assistants, virtual classrooms, virtual idols, and other interactive games and entertainment. At present, many industries often communicate with users through virtual digital humans, but most online virtual digital humans cannot perform fine and smooth movements. They usually have stiff expressions, jerky movements, and lack of realism, which affects the audience's perception and user experience, and is not conducive to realizing their potential commercial value. In summary, the existing technology has the problem of insufficient accuracy of virtual digital humans' movements due to low data processing efficiency, which needs to be solved urgently. Summary of the invention

[0003] The purpose of the embodiment of the present application is to provide a virtual digital human motion processing system to improve the motion performance accuracy of online virtual digital humans.

[0004] In order to achieve the above-mentioned purpose, the present application provides a virtual digital human motion processing system, the system comprising: an image generation module, used to generate a virtual digital human whose appearance meets the user's expectations; a configuration module, used to determine the importance level of each moving part of the virtual digital human according to the application scenario, and configure a skeletal model of a corresponding accuracy level for each moving part of the virtual digital human according to the importance level; the higher the importance level of the moving part of the virtual digital human, the higher the accuracy of the corresponding skeletal model, and the higher the accuracy of the skeletal model, the closer the movement of the corresponding moving part of the virtual digital human is to that of a real human; an action control module, used to control each moving part of the virtual digital human to perform actions of corresponding accuracy according to the user's instruction information and preset action rules; the instruction information includes action information and voice information.

[0005] Furthermore, the accuracy levels include level I and level II; the motion control module includes: a level I model motion control unit, used to control the movements of type I motion parts of the virtual digital human, and the type I motion parts are correspondingly configured with a skeletal model with level I accuracy; a level II model motion control unit, used to control the movements of type II motion parts of the virtual digital human, and the type II motion parts are correspondingly configured with a skeletal model with level II accuracy.

[0006] Furthermore, the skeletal model of level I accuracy is composed of joints and bones, and the skeletal model of level II accuracy is composed of joints, bones, and muscles attached to the bones.

[0007] Furthermore, the application scenario includes using a virtual digital human to perform sign language translation of voice content; configuring a skeletal model of corresponding accuracy level for each moving part of the virtual digital human includes: configuring a skeletal model of level II accuracy for the arm of the virtual digital human, and configuring a skeletal model of level I accuracy for the remaining parts; the action control module includes: a voice collection module for collecting the user's voice information; controlling each moving part of the virtual digital human to perform actions of corresponding accuracy according to the user's instruction information and preset action rules includes: controlling the arm of the virtual digital human to perform actions of corresponding accuracy according to the user's voice information and sign language execution rules.

[0008] Furthermore, the Class I model motion control unit includes: a motion parameter calculation model, which is used to calculate the motion parameters of the Class I motion parts, and the motion parameters of the Class I motion parts include joint offset and rotation; a first driving subunit, which drives the corresponding Class I motion parts to move according to the motion parameters; the calculation process of the motion parameter calculation model is as follows: obtaining the target bone length L1, target joint offset o and rotation r of the reference human; the calculation formula of the joint offset o' and rotation r' of the corresponding joint of the virtual digital human is as follows: o'=o(L2 / L1) (1) r'=r(L2 / L1) (2) In the above formulas (1) and (2), L2 represents the corresponding bone length of the virtual digital human.

[0009] Furthermore, the Class II model motion control unit includes: a motion parameter prediction model for predicting motion parameters of Class II motion parts, wherein the motion parameters of Class II motion parts include muscle activation degree; a second driving subunit for driving the corresponding Class II motion parts to move according to the motion parameters; a training method for the motion parameter prediction model includes: S1, obtaining joint posture information of the corresponding motion parts of a reference human; S2, obtaining the expected joint posture of a target virtual digital human based on the joint posture information; S3, calculating the expected torque and expected acceleration applied to each joint based on the expected joint posture; S4, calculating the muscle activation degree that meets the expected acceleration, and calculating the muscle force applied to the joint by the corresponding muscle according to the muscle activation degree; S5, controlling the target virtual digital human to move based on the muscle force and expected torque, and recording the actual joint posture of the virtual digital human; S6, optimizing the method for calculating the muscle activation degree in step S4 based on the deviation between the actual joint posture and the expected joint posture; S7, making training samples using the mapping relationship between the muscle activation degree and the actual joint posture of the virtual digital human; S8, training the motion parameter prediction model using the training samples.

[0010] Furthermore, the level II model motion control unit also includes: A calculation subunit is provided, wherein the calculation subunit calculates the muscle activation degree that meets the expected acceleration based on the following calculation formula: (3) In formula (3), , and They represent the generalized coordinates, velocity and acceleration of the joints in all degrees of freedom in the level II precision skeleton model. represents the mass matrix, represents the Coriolis force and centrifugal force matrices, Indicates external force, Indicates binding force, Indicates the degree of activation of all muscles. represents the coefficient value matrix, represents the constant parameter matrix, Represents the transposed form of the Jacobian matrix that maps constraint forces to generalized coordinates.

[0011] Furthermore, in the training method of the motion parameter prediction model, if there are multiple solutions for the calculated muscle activation degree that meets the expected acceleration, the optimal muscle activation degree is calculated with the goal of minimizing overall muscle energy consumption.

[0012] Further, the desired torque applied to each joint is calculated based on the desired joint posture The calculation expression is as follows: (4) In formula (4), n represents the current time step, represents the joint pose at the next time step, represents the expected joint pose at the next time step, represents the joint velocity at the next time step, represents the proportionality coefficient, represents the differential coefficient.

[0013] Furthermore, the obtaining of joint posture information of corresponding motion parts of referenced humans includes: obtaining a human motion video, and obtaining the joint posture information using a posture recognition technology for the human motion video.

[0014] The virtual digital human motion processing system provided by this application has at least the following beneficial effects: The virtual digital human motion processing system provided by this application configures skeletal models of different precision for different parts of the virtual digital human according to different application scenarios, and can provide customers with more refined services. At the same time, by "ignoring" unimportant parts, the computing power burden of the server can be reduced, the smoothness of the movements of important parts can be ensured, the audience's perception and user experience can be improved, and the scope of application of virtual digital humans can be broadened.

[0015] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings: Figure 1 The structure block diagram of the virtual digital human motion processing system of the embodiment of the present application is schematically shown; Figure 2 The training flowchart of the motion parameter prediction model in the embodiment of the present application is schematically shown.

[0017] Description of Reference Numerals 1-image generation module; 2-configuration module; 3-motion control module; 31-level I model motion control unit; 32-level II model motion control unit; 33-voice acquisition module. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0019] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), such directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0020] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application. Example

[0021] Please refer to Figure 1 This embodiment provides a virtual digital human motion processing system, the system comprising: Image generation module 1, used to generate a virtual digital human whose appearance meets the user's expectations; Configuration module 2 is used to determine the importance level of each moving part of the virtual digital human according to the application scenario, and configure a skeletal model of a corresponding accuracy level for each moving part of the virtual digital human according to the importance level; the higher the importance level of the moving part of the virtual digital human, the higher the accuracy of the corresponding skeletal model, and the higher the accuracy of the skeletal model, the closer the movement of the corresponding moving part of the virtual digital human is to that of a real human; The action control module 3 is used to control the movement parts of the virtual digital human to perform actions with corresponding precision according to the user's instruction information and preset action rules; the instruction information includes action information and voice information.

[0022] Among them, the image generation module 1 determines the proportion of the overall skeleton model of the virtual digital human. The configuration module 2 fills the skeleton model for each part of the body of the virtual digital human to support the body movement according to the determined appearance of the virtual digital human. The above-mentioned importance level can be set to two levels, and the higher the level, the more important the corresponding movement part is in the current application environment. The basis for judging the importance can be set as: the frequency of use in the application environment or the necessity for realizing its application function. Taking the virtual beauty blogger as an example, when using the virtual digital human to show beauty skills, the parts with the highest appearance rate are the face and arms, while parts such as the legs may not even need to be shown. Therefore, in order to ensure the flexibility of the arm (including fingers) movement, the naturalness of the facial expression and the authenticity of the makeup display when the virtual digital human is performing makeup actions, the face, neck and arm parts can be identified as the second-level importance level, and the remaining parts usually do not require movement, so they are identified as the first-level importance level. Exemplarily, if the accuracy levels include level I and level II, a skeletal model with higher accuracy level II is used for moving parts such as the face, neck, and arms that are identified as level II importance, and a skeletal model with higher accuracy level I is used for the remaining body parts that are identified as level I importance.

[0023] Specifically, the above-mentioned Class I precision skeletal model is composed of joints and bones. Based on this type of model, the existing control method of using joints to drive bones can be used, but since this method is still somewhat different from the actual action principle of the organism, it is easy to cause problems with the mechanical nature of the action. The above-mentioned Class II precision skeletal model is composed of joints, bones, and muscles attached to the bones (the broken line obtained by connecting each attachment point simulates the muscle texture line, and the direction of the broken line is used as the direction of muscle force), that is, a new muscle model is added on the basis of the Class I precision model. The human body muscle completes the movement by pulling the bones through contraction and relaxation, and is the active force-applying party in the movement process. Therefore, the control method of using muscle force to drive joints and bones is used in this embodiment.

[0024] Furthermore, since the two models are different in structure, different control methods are used for the two models with different precisions in this implementation to achieve control of different motion precisions. Specifically, the motion control module 3 includes: The Class I model motion control unit 31 is used to control the Class I motion parts of the virtual digital human, and the Class I motion parts are correspondingly configured with a Class I precision skeleton model; The Class II model motion control unit 32 is used to control the Class II motion parts of the virtual digital human, and the Class II motion parts are correspondingly configured with a Class II precision skeleton model.

[0025] For example, if the application scenario is to use a virtual digital human to perform sign language translation of speech content, in this scenario, the most important action part is the arm (including fingers), and the face does not need to show too many facial expressions, so the arm of the virtual digital human can be configured with a bone model of level II accuracy, and the rest of the body can be configured with a bone model of level I accuracy; since sign language has fixed gestures, a fixed basic action module can be recorded in advance, and then the basic actions can be combined according to the speech content, so the action control module 3 also includes: a voice collection module 33, which is used to collect the user's voice information, and the control of the virtual digital human's various moving parts to perform actions of corresponding accuracy according to the user's instruction information and preset action rules includes: controlling the virtual digital human's arm to perform actions of corresponding accuracy according to the user's voice information and sign language execution rules. However, in order to ensure the fluency of the action and the accuracy of the sign language translation, it is also necessary to repeatedly train the basic actions of the virtual digital human and the mapping relationship between the basic actions and the speech content.

[0026] The virtual digital human motion processing system provided in this embodiment configures skeletal models of different precision for different parts of the virtual digital human according to specific application scenarios. By "emphasizing" important parts and "ignoring" unimportant parts, it can provide customers with more refined services and better perception and experience, while reducing the computing power burden of the server and the memory burden of the device, ensuring the smoothness of the movements of important parts and avoiding redundancy, which helps to improve the development space of virtual digital humans and have better development prospects. Example

[0027] In this embodiment, the motion control of the skeleton model with level I accuracy is discussed, taking the control of the virtual digital human to imitate the motion of a real human as an example. Specifically, the level I model motion control unit 31 includes: The motion parameter calculation model is used to calculate the motion parameters of Class I motion parts, and the motion parameters of Class I motion parts include joint offset and rotation. Since the bone model with Class I accuracy drives the bone movement through the joints, the movement direction and displacement of the motion parts in the world coordinate system can be known after obtaining the joint offset, and the rotation angle of the joint-driven bone rotation can be clarified.

[0028] A first driving subunit drives the corresponding type I motion part to move according to the motion parameters; Since the human parameters of the reference human for reference motion are usually not completely matched with the structural model of the virtual digital human to be driven, even if the topological structures of the skeletons of the two are basically the same, there are still differences in the length, mass and shape of each bone segment. For the virtual digital human, the mass and shape of the bones can be adaptively adjusted without having a significant impact on the body shape, while the adjustment of the bone length will affect the overall body proportions. After adjustment, the target bones of the virtual digital human and the reference human are mainly different in length, so the calculation process of the motion parameter calculation model is as follows: obtain the target bone length L1, target joint offset o and rotation r of the reference human; the calculation formula of the joint offset o' and rotation r' of the corresponding joint of the virtual digital human is as follows: o'=o(L2 / L1) (1) r'=r(L2 / L1) (2) In the above formulas (1) and (2), L2 represents the corresponding bone length of the virtual digital human.

[0029] The above method can achieve simple motion simulation and avoid a strong sense of separation between Class I and Class II moving parts. However, if Class I moving parts do not need to be on camera, there is no need for motion control and they can just remain still, reducing the computing power burden and network burden of the server. Example

[0030] In this embodiment, the motion control of the skeletal model with level II precision is discussed, taking the control of the virtual digital human to imitate the movements of real humans as an example. Since the skeletal model with level II precision is more complex than the skeletal model with level I precision and requires more motion parameters to be adjusted, if the traditional calculation method is used, it is easy to have the problem of difficulty in parameter adjustment. Therefore, in this embodiment, the driving parameters of the skeletal model with level II precision are obtained through a trained neural network model. Exemplarily, the level II model motion control unit 32 includes: a motion parameter prediction model, which is used to predict the motion parameters of the class II motion parts, and the motion parameters of the class II motion parts include the degree of muscle activation; The second driving subunit drives the corresponding type II motion part to move according to the motion parameters; like Figure 2 As shown, the training method of the motion parameter prediction model includes: S1. Acquire joint posture information of corresponding motion parts of reference humans; the acquisition method may be from existing human motion videos, and the joint posture information is acquired by using posture recognition technology for human motion videos.

[0031] S2. Obtaining the expected joint posture of the target virtual digital human based on the joint posture information; the joint posture information from real humans can be used as a reference, but due to the difference in body shape between the virtual digital human and the real human, it is necessary to clarify the mapping relationship between the two in advance.

[0032] S3, based on the expected joint posture, calculate the expected torque and expected acceleration applied to each joint; Exemplarily, based on the expected joint posture, calculate the expected torque applied to each joint The calculation expression is as follows: (4) In formula (4), n represents the current time step, represents the joint pose at the next time step, represents the expected joint pose at the next time step, represents the joint velocity at the next time step, represents the proportionality coefficient, In addition, the expected acceleration can be calculated based on the expected joint velocity time difference.

[0033] S4. Calculate the degree of muscle activation that meets the expected acceleration, and calculate the muscle force applied to the joint by the corresponding muscle based on the degree of muscle activation. When the degree of muscle activation is known, the muscle force and the torque applied to the joint can be derived and calculated based on the muscle model and the dynamic model.

[0034] Among them, acceleration and muscle activation The mapping relationship between them is as follows: (3) In formula (3), , and They represent the generalized coordinates, velocity and acceleration of the joints in all degrees of freedom in the level II precision skeleton model. represents the mass matrix (derived from the skeleton model parameters, in this embodiment, it refers to the weight of each bone segment), represents the Coriolis force and centrifugal force matrix (which can be calculated based on known parameters such as (bone) mass and (expected) angular velocity of joint rotation), Represents external force (customizable and can be used as an adjustment parameter for optimizing calculation formulas), represents the constraint force (derived from the bone, related to its mass, and a known parameter), represents the degree of activation of all muscles required to achieve the desired joint posture, represents the coefficient value matrix, represents the constant parameter matrix, represents the transposed form of the Jacobian matrix that maps constraint forces to generalized coordinates; Every posture generated during human movement is the result of coordinated cooperation of muscle groups. In the above formula (3): (3-1); (3-2); In formula (3-1), m represents the number of bones. Represents the muscle force matrix of m muscles (there is a mapping relationship between muscle force and acceleration and activation degree respectively), Indicates that The transposed form of the Jacobian matrix mapped to generalized coordinates; represents the activation degree matrix of m muscles, and In formula (3-2), Indicates maximum muscle force.

[0035] Furthermore, in this step, if there are multiple solutions for the calculated muscle activation degree that meets the expected acceleration, the optimal muscle activation degree is calculated with the goal of minimizing the overall muscle energy consumption. That is, there may be multiple muscle combinations that can control joints and bones to complete the desired action, but the human brain usually calculates the global optimum when controlling human movement. For example, when two muscles can produce the same joint control effect, the brain usually gives priority to muscles with larger force arms and less effort to complete the control. Therefore, in this embodiment, the thinking mode of the human brain is simulated to determine the optimal muscle control strategy with the goal of minimizing muscle energy consumption.

[0036] S5, controlling the target virtual digital human to move based on the muscle force and the expected torque, and recording the actual joint posture of the virtual digital human; S6: Optimize the method for calculating the muscle activation degree in step S4 based on the deviation between the actual joint posture and the expected joint posture, such as adjusting the above parameters Or modify the skeleton model parameters , wait.

[0037] S7, using the mapping relationship between the muscle activation degree and the actual joint posture of the virtual digital human to create training samples; S8. Using the training samples to train a motion parameter prediction model.

[0038] Different from the joint skeleton, it is difficult to collect the real muscle activity data when training the motion parameter prediction model based on muscle control. In this embodiment, the experience replay mechanism in deep reinforcement learning is referred to to create training samples. At the same time, the method of calculating the degree of muscle activation is optimized according to the deviation between the actual joint posture and the expected joint posture, which helps to speed up the training progress of the motion parameter prediction model.

[0039] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0040] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0041] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0042] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0043] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0044] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0045] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0046] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0047] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A virtual digital human motion processing system, characterized in that: The system comprises: An image generation module (1) is used to generate a virtual digital human whose appearance meets the user's expectations; A configuration module (2) is used to determine the importance level of each moving part of the virtual digital human according to the application scenario, and configure a skeletal model of a corresponding accuracy level for each moving part of the virtual digital human according to the importance level; the higher the importance level of the moving part of the virtual digital human, the higher the accuracy of the corresponding skeletal model; the higher the accuracy of the skeletal model, the closer the movement of the corresponding moving part of the virtual digital human is to that of a real human; The action control module (3) is used to control the movement parts of the virtual digital human to perform actions with corresponding precision according to the user's instruction information and preset action rules; the instruction information includes action information and voice information.

2. The virtual digital human motion processing system according to claim 1, characterized in that: The accuracy levels include level I and level II; the motion control module (3) includes: A Class I model motion control unit (31) is used to control the movements of Class I motion parts of a virtual digital human, wherein the Class I motion parts are correspondingly configured with a Class I precision skeleton model; The class II model motion control unit (32) is used to control the movements of the class II motion parts of the virtual digital human, wherein the class II motion parts are correspondingly configured with a skeletal model with class II accuracy.

3. The virtual digital human motion processing system according to claim 2, characterized in that: The skeletal model with level I accuracy is composed of joints and bones, and the skeletal model with level II accuracy is composed of joints, bones and muscles attached to the bones.

4. The virtual digital human motion processing system according to claim 3, characterized in that: The application scenario includes using a virtual digital human to perform sign language translation of speech content; configuring a skeletal model of a corresponding accuracy level for each moving part of the virtual digital human includes: configuring a skeletal model of level II accuracy for the arm of the virtual digital human, and configuring a skeletal model of level I accuracy for the rest of the parts; The action control module (3) comprises: A voice collection module (33), used to collect voice information of the user; The controlling of each moving part of the virtual digital human to perform actions of corresponding precision according to the user's instruction information and preset action rules includes: According to the user's voice information and sign language execution rules, the virtual digital human's arms are controlled to perform movements with corresponding precision.

5. The virtual digital human motion processing system according to claim 2, characterized in that: The level I model motion control unit (31) comprises: A motion parameter calculation model, used to calculate the motion parameters of the Class I motion parts, wherein the motion parameters of the Class I motion parts include joint offset and rotation; A first driving subunit drives the corresponding type I motion part to move according to the motion parameters; The calculation process of the motion parameter calculation model is as follows: Obtain the target bone length L1, target joint offset o and rotation r of the reference human; The calculation formulas for the joint offset o' and rotation r' of the corresponding joints of the virtual digital human are as follows: o'=o(L2 / L1) (1) r'=r(L2 / L1) (2) In the above formulas (1) and (2), L2 represents the corresponding bone length of the virtual digital human.

6. The virtual digital human motion processing system according to claim 2, characterized in that: The class II model motion control unit (32) comprises: A motion parameter prediction model, used to predict the motion parameters of the Class II motion parts, wherein the motion parameters of the Class II motion parts include the degree of muscle activation; The second driving subunit drives the corresponding type II motion part to move according to the motion parameters; The training method of the motion parameter prediction model includes: S1, obtaining the joint posture information of the corresponding motion parts of the reference human; S2, acquiring the expected joint posture of the target virtual digital human based on the joint posture information; S3, calculating the expected torque and expected acceleration applied to each joint based on the expected joint posture; S4, calculating the degree of muscle activation that meets the expected acceleration, and calculating the muscle force applied to the joint by the corresponding muscle according to the degree of muscle activation; S5, controlling the target virtual digital human to move based on the muscle force and the expected torque, and recording the actual joint posture of the virtual digital human; S6, optimizing the method for calculating the muscle activation degree in step S4 based on the deviation between the actual joint posture and the expected joint posture; S7, using the mapping relationship between the muscle activation degree and the actual joint posture of the virtual digital human to create training samples; S8. Using the training samples to train a motion parameter prediction model.

7. The virtual digital human motion processing system according to claim 6, characterized in that: The class II model motion control unit (32) further includes: A calculation subunit is provided, wherein the calculation subunit calculates the muscle activation degree that meets the expected acceleration based on the following calculation formula: （3） In formula (3), , and They represent the generalized coordinates, velocity and acceleration of the joints in all degrees of freedom in the level II precision skeleton model. represents the mass matrix, represents the Coriolis force and centrifugal force matrices, Indicates external force, Indicates binding force, Indicates the degree of activation of all muscles. represents the coefficient value matrix, represents the constant parameter matrix, Represents the transposed form of the Jacobian matrix that maps constraint forces to generalized coordinates.

8. The virtual digital human motion processing system according to claim 6, characterized in that: In the training method of the motion parameter prediction model, if there are multiple solutions for the calculated muscle activation degree that meets the expected acceleration, the optimal muscle activation degree is calculated with the goal of minimizing overall muscle energy consumption.

9. The virtual digital human motion processing system according to claim 6, characterized in that: Calculate the desired torque applied to each joint based on the desired joint posture The calculation expression is as follows: （4） In formula (4), n represents the current time step, represents the joint pose at the next time step, represents the expected joint pose at the next time step, represents the joint velocity at the next time step, represents the proportionality coefficient, represents the differential coefficient.

10. The virtual digital human motion processing system according to claim 6, characterized in that: The obtaining of joint posture information of corresponding motion parts of a reference human includes: A human body motion video is obtained, and the joint posture information is obtained by using posture recognition technology on the human body motion video.

Images