Three-dimensional digital human interaction system of power system

Abstract

The invention provides a three-dimensional digital human interaction system of a power system. The three-dimensional digital human interaction system is characterized in that an automatic mouth shape matching and facial expression driving module starts and configures a plurality of programs through batch processing files and generates mouth shape animations according to voice data characteristics to drive facial expressions of a three-dimensional digital human model; the hardware optimization and resource management module is used for optimizing the use efficiency of hardware resources, managing the operation cost of the system and supporting graphics card equipment with relatively low performance; the starting process simplification module integrates necessary programs, automatically starts and configures and manages a starting sequence and dependency; the editing module sets role identity backgrounds, character characteristics and language styles through user multi-dimensional parameters to automatically generate and train a virtual role model; the interaction module provides a user interaction interface and three-dimensional digital human visual display; the data storage and management module stores and manages user configuration files and training data.

Description

Technical Field

[0001] This invention relates to the field of information technology for power systems, and more particularly to a three-dimensional digital human interaction system for power systems. Background Technology

[0002] With the widespread application of artificial intelligence technology in the intelligent operation and maintenance of power systems, traditional power (electrical) equipment monitoring and interaction methods suffer from problems such as low visualization and complex operation. Currently, power system dispatching and training scenarios mostly rely on two-dimensional interfaces or simple animation models, making it difficult to achieve a realistic presentation of equipment status and natural interaction. Especially in scenarios such as power (electrical) equipment fault simulation and emergency drills, existing systems lack highly realistic three-dimensional visualization and interaction capabilities, resulting in limited training effectiveness and insufficient immersion for operators.

[0003] Current mainstream power system visualization solutions are typically based on static models or simple animation-driven approaches, with relatively outdated facial expression and lip-syncing technologies. For example, during voice interaction for power dispatch commands, virtual characters cannot automatically generate matching lip movements and expressions based on voice characteristics, requiring manual adjustment tools (such as NVIDIA A2F lip-sync software). This process is cumbersome and relies on high-performance hardware (such as RTX 4080 or higher graphics cards). Furthermore, system startup requires manually launching multiple independent programs (such as speech synthesis, 3D rendering engines, and data interface services), resulting in strict startup order requirements, complex configurations, and high hardware resource consumption. This makes it difficult to run efficiently on medium-configuration equipment deployed in power fields (such as industrial control computers equipped with 3070Ti graphics cards).

[0004] Therefore, there is an urgent need for a 3D digital human system that supports automated lip-syncing and has low hardware dependence, in order to improve the realism and operational efficiency in training, monitoring and interaction scenarios of power and electrical systems. Summary of the Invention

[0005] This invention provides a three-dimensional digital human interaction system for power systems, mainly comprising:

[0006] The system consists of an automated lip-syncing and facial expression driving module, a hardware optimization and resource management module, a simplified startup process module, an editing module, an interaction module, and a data storage and management module. The automated lip-syncing and facial expression driving module starts and configures multiple programs via batch processing files and generates lip-syncing animations based on speech data features to drive the facial expressions of the 3D digital human model. The hardware optimization and resource management module optimizes hardware resource utilization efficiency and manages system operating costs to support lower-performance graphics card devices. The simplified startup process module integrates necessary programs for automatic startup and configures and manages the startup order and dependencies. The editing module automatically generates and trains virtual character models by allowing users to set character identity background, personality traits, and language style using multi-dimensional parameters. The interaction module provides a user interface and a visual display of the 3D digital human. The data storage and management module stores and manages user configuration files and training data. Furthermore, the automated lip-syncing and facial expression driving module includes: starting and configuring multiple programs at once through a batch file, adding configuration parameters for program startup to the batch script to ensure the programs are in the correct state; using audio processing tools to analyze the pitch, volume, and speech rate features of the speech data, generating corresponding lip-syncing animations based on the analyzed audio features; and applying the generated lip-syncing animations to a 3D digital human model using a 3D game engine to drive facial expression changes in real time. Furthermore, the hardware optimization and resource management module includes: calculating hardware resource utilization efficiency E as U / C, where U is the actual GPU utilization rate and C is the total system computing power; calculating operating cost C_run as (P_usedT) / C_eff, where P_used is the actual power used, T is the running time, and C_eff is the optimized resource utilization efficiency; and ensuring that the minimum supported graphics card performance P_min is greater than or equal to the performance value P_3070Ti of the lower-performance graphics card device. Furthermore, the hardware optimization and resource management module includes: optimizing hardware resource utilization efficiency E as follows:

[0007]

[0008] Where AA represents the number of available hardware units, RR represents resource requirements, TT represents system response time, and SS represents the task complexity index. Total computing power.

[0009] Furthermore, the startup process simplification module includes: the total startup time T_total of the system containing N programs is the sum of the startup times T_i of each program; the total configuration time T_config of each program i is the sum of the configuration times T_config,i of each program; the program startup order dependency relationship is that when program i depends on program j in the directed graph, the dependency relationship marker D_i,j is 1; the total time of the reduced manual configuration steps T_reduced is the sum of the times T_step,j of each manual configuration step.

[0010] Furthermore, the editing module includes: a user-defined dimension set S consisting of dimension parameters P_1 to P_N; a role identity background parameter set B consisting of B_1 to B_nB; a personality trait parameter set C consisting of C_1 to C_nC; and a language style parameter set L consisting of L_1 to L_nL. Furthermore, the interaction module includes: an interaction interface function set F consisting of multiple operations O_1 to O_n. Furthermore, the data storage and management module includes: a storage overall data set D consisting of a user profile data set U_user and a training data set T_training; a data management operation O consisting of a data management function M processing the user profile data set U_user and the training data set T_training; and backup data D_backup consisting of a backup function BB processing the storage overall data set D.

[0011] Furthermore, the data storage expression in the data storage and management module is:

[0012]

[0013] in, It is the overall collection of stored data; It is a collection of user profile data; It is the training data set;

[0014] The data management expression is:

[0015] ;in, It is the result of data management operations; It is a function for data management.

[0016] Furthermore, the data storage and management module also includes data backup, the data backup expression being:

[0017]

[0018] in, It is backup data; It is a backup function; It is the overall collection of stored data.

[0019] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0020] 1. This invention discloses a three-dimensional digital human interaction system for power systems. It addresses the interconnected business scenario problems in existing technologies, such as high development threshold, cumbersome startup and configuration, excessive hardware resource requirements, complex lip-sync and facial expression driving, and low efficiency in character editing and data management. The system cleverly integrates multiple core modules through an integrated design to solve these problems.

[0021] 2. This invention uses batch processing files to automatically start and configure multiple programs, achieving real-time and accurate lip-syncing and facial expression control. Simultaneously, hardware resource optimization algorithms significantly improve efficiency and reduce operating costs, ensuring smooth operation even on graphics cards as low as a 3070Ti. The simplified startup process module integrates program dependencies and order, greatly reducing manual configuration time. The editing module provides multi-dimensional and intuitive settings, automatically generating and training personalized virtual characters. The interaction module and data storage management module work together to support efficient user interaction and secure backup.

[0022] 3. This invention reduces the overall difficulty of development and deployment, improves system compatibility, operating efficiency and user experience, and realizes low-threshold, high-performance three-dimensional digital human interactive applications. Attached Figure Description

[0023] Figure 1 This is a flowchart of a three-dimensional digital human interaction system for a power system according to the present invention. Detailed embodiments.

[0024] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0025] like Figure 1 This embodiment of a three-dimensional digital human interaction system for power systems may specifically include:

[0026] S1, a three-dimensional digital human interaction system for power systems, characterized in that the system includes an automated lip-syncing and facial expression driving module, a hardware optimization and resource management module, a startup process simplification module, an editing module, an interaction module, and a data storage and management module. The automated lip-syncing and facial expression driving module is used to start and configure multiple programs through batch files to achieve automatic matching and driving of lip shapes and facial expressions.

[0027] The batch file integrates the startup commands of multiple programs and embeds preset configuration parameters, sequentially launching the audio processing tool and the 3D animation environment, ensuring all programs run in the background and are in standby mode. The audio processing tool acquires audio data generated by speech synthesis in real time, extracting pitch, volume, and speech rate features to form corresponding lip-sync animation sequences. The 3D animation environment directly receives the lip-sync animation sequences, mapping them to a digital human facial skeleton driver to drive facial expression changes and synchronously output a visualized digital human. The hardware resource manager monitors the program's memory usage during runtime, dynamically adjusting the rendering precision of the 3D animation environment based on the usage, maintaining smooth operation even on low-end graphics cards.

[0028] The generation steps are as follows:

[0029] 1. Batch files automate processes by integrating startup commands and preset configuration parameters.

[0030] In one embodiment, the batch file first loads the path of the audio processing tool and sets its sampling rate parameters, and then starts the 3D animation environment in sequence to ensure background operation. This reduces manual intervention and improves system stability. The beneficial effect is that the program directly enters standby mode, avoiding startup delay.

[0031] 2. Specific methods by which audio processing tools extract features from speech-synthesized audio data.

[0032] Understandably, Fourier transform analysis is used to analyze pitch features, decompose the audio waveform into frequency components, and then calculate the average pitch value. At the same time, volume features are extracted by calculating the amplitude peak, and speech rate features are based on the syllable interval time statistics to form a mouth shape animation sequence. This can accurately match mouth shape changes, and the beneficial effect is to generate animation sequences in real time, improving the naturalness of digital human expression.

[0033] 3. The process of the 3D animation environment receiving the mouth animation sequence and mapping it to the facial bone driver.

[0034] Specifically, keyframe data from the animation sequence is applied to skeletal nodes. For example, the rotation angle of the chin and lips is adjusted in each frame to drive changes in facial expressions and output a visual digital human. This enables synchronous driving and has the beneficial effect of enhancing the immersiveness of the interaction, allowing users to feel a more realistic response from the virtual character.

[0035] 4. Methods for monitoring video memory usage using Hardware Resource Manager.

[0036] In one possible implementation, the current utilization rate is obtained by periodically querying the graphics card API. If the utilization rate increases, the rendering resolution is reduced or the texture details are simplified to dynamically adjust the rendering accuracy of the 3D animation environment and maintain smooth operation under low-end graphics cards. This can optimize resource allocation and has the beneficial effect of lowering the hardware threshold, making the system more accessible to ordinary devices.

[0037] 5. After the batch file is started, the features extracted by the audio processing tool directly affect the mapping accuracy of the 3D animation environment.

[0038] For example, audio sequences with higher pitch characteristics trigger more exaggerated lip-syncing animations, which has the beneficial effect of enhancing the diversity of facial expressions and ensuring that digital humans adapt to different speech styles.

[0039] 6. When driving facial expressions, the 3D animation environment works in conjunction with the hardware resource manager to adjust the rendering.

[0040] For example, when video memory usage is close to the threshold, non-critical facial animations are simplified first, which helps maintain overall smoothness and avoid system lag.

[0041] 7. The mouth animation sequence formed by extracting audio data can be fine-tuned by combining it with preset expression templates during the mapping process.

[0042] For example, adding blinking animations when the speech rate is fast enhances the vividness, which has the beneficial effect of making digital human interaction more human-like and reducing the sense of stiffness.

[0043] 8. The dynamic adjustment of the hardware resource manager not only monitors video memory, but can also be extended to CPU load.

[0044] For example, reducing the number of parallel threads in audio processing can have the beneficial effects of comprehensively optimizing system performance, supporting a wider range of hardware configurations, and enabling low-cost operation of automated lip-syncing and facial expression-driven processes.

[0045] S2, the hardware optimization and resource management module is used to optimize the efficiency of hardware resource utilization and manage the system's operating costs to ensure that the system supports graphics card devices as low as 3070Ti. The hardware resource utilization efficiency E is calculated as U / C, where U is the actual GPU utilization rate and C is the total computing power of the system.

[0046] The actual GPU utilization and the total computing power of the 3070Ti graphics card are obtained. Hardware resource utilization efficiency is determined by dividing the actual GPU utilization by the total computing power. Based on resource requirements, response time, task complexity index, and the number of available hardware, optimized hardware resource utilization efficiency is determined from the actual GPU utilization and total computing power. Optimized hardware resource utilization efficiency decreases when resource requirements increase and increases when the number of available hardware increases. The actual power consumption and runtime are obtained. System operating costs are determined by multiplying the actual power consumption and runtime, then dividing by the optimized hardware resource utilization efficiency. Maintaining a high level of optimized hardware resource utilization efficiency reduces system operating costs, thus ensuring that hardware resource utilization optimization and operating cost management are implemented on the 3070Ti graphics card.

[0047] For example, when the graphics processor utilization is low, the actual utilization of the graphics processor and the total computing power of the 3070Ti graphics card can be obtained, for example, the utilization rate is 50% and the total computing power is a specific floating-point operation capability. The hardware resource utilization efficiency can be determined by dividing the actual utilization of the graphics processor by the total computing power. This can bring more accurate resource evaluation results and help the accuracy of the subsequent optimization process.

[0048] In one possible implementation, for situations with high resource demands, the optimized hardware resource utilization efficiency is determined based on the actual utilization rate of the graphics processor and the total computing power. For example, increased resource demands may lead to decreased efficiency, while increased available hardware may improve efficiency. This approach offers the benefit of dynamic adjustment, ensuring that the system maintains high efficiency under different loads.

[0049] For example, the actual power used and the running time are obtained, such as a medium power level and a running time of several hours. The operating cost of the management system is determined by multiplying the actual power used and the running time and then dividing by the optimized hardware resource utilization efficiency. This can bring about cost control and support the stable operation of low-end devices such as 3070Ti graphics cards.

[0050] In one possible implementation, while maintaining a high level of optimized hardware resource utilization efficiency, the management system's operating costs are reduced, for example by reducing unnecessary power consumption through efficiency improvements. This ensures that hardware resource utilization efficiency optimization and operating cost management are achieved on the 3070Ti graphics card device, which can bring about overall system performance improvements and energy savings.

[0051] For example, the acquisition of graphics processor utilization can be considered from multiple perspectives. For instance, the utilization fluctuates greatly in video rendering tasks. The efficiency value can be obtained by dividing by the total computing power, which supports the accuracy of cost calculation. Optimizations related to response time support each other and improve the system response speed.

[0052] In one possible implementation, the handling of task complexity index, such as increasing response time for complex tasks, thereby affecting optimization efficiency, can bring about targeted adjustments. This is supported by an increase in the number of available hardware, ensuring that efficiency does not drop significantly when resource demands are high.

[0053] For example, the product of power and runtime divided by efficiency determines the cost. In continuous computing scenarios, this can provide the benefit of real-time monitoring, which, in conjunction with the aforementioned efficiency optimization, reduces the overall cost.

[0054] In one possible implementation, cost reduction at a higher level of efficiency, such as in a multitasking environment, can lead to sustainable operation, which, together with the support of the 3070Ti graphics card, achieves the goal of hardware optimization.

[0055] For example, from the perspective of power consumption, such as extended operating time in low-power mode and better cost calculation, this forms a logical chain with the application of efficiency formulas, supporting efficient management of the system on low-end devices.

[0056] In one possible implementation, an example of efficiency reduction when resource demand increases, such as when demand exceeds capacity, is addressed by adjusting the formula. This provides an early warning effect and is linked to subsequent cost management to ensure overall optimization.

[0057] Examples of ways to improve efficiency include increasing the number of available hardware components, such as adding auxiliary hardware, which provides scalability benefits and, in conjunction with reduced operating costs, enables system support down to a 3070Ti graphics card.

[0058] S3, the startup process simplification module is used to integrate necessary programs and automatically start configuration management startup order and dependencies to reduce manual configuration steps, wherein the system contains N programs and the total startup time T_total is the sum of the startup times T_i of each program.

[0059] For a startup process involving N programs, the startup time Ti of each program i is obtained. The total startup time Ttotal is obtained by summing the startup times Ti of all programs, where Ttotal represents the total time required for all programs to start, and Ti represents the time required for a single program to start. Based on the total startup time Ttotal, the dependencies and startup order between programs are determined. The startup order of dependent programs is adjusted so that subsequent programs are started after the preceding program has finished. For each program with the adjusted startup order, the configuration time Tconfig,i corresponding to the configuration parameter Pi required by each program i is obtained. The total configuration time Tconfig is obtained by summing the configuration times Tconfig,i of all programs, where Tconfig represents the total time required for all programs to configure, and Tconfig,i represents the time required for a single program to configure. Based on the total configuration time Tconfig, configuration items that originally required manual completion are identified and converted to automatic execution. After automatic execution, the total configuration time Tconfig is directly used in the startup process, thereby reducing the number of manual configuration items and completing the automatic startup and configuration of all programs.

[0060] For example, in the startup process involving multiple programs, the startup time of each program is first obtained. For instance, a system has three programs: a database service, an application server, and a front-end interface. The startup time of the database service is 10 seconds, the application server is 15 seconds, and the front-end interface is 5 seconds. By summing these, the total startup time is obtained as 30 seconds. This clearly quantifies the overall startup time and is beneficial for subsequent optimization of dependencies to shorten the waiting time.

[0061] In one possible implementation, the dependencies between programs are determined based on the total startup time. For example, the database service must start before the application server because the application server needs to connect to the database. If the order is not adjusted, it may lead to startup failure or retry. After adjustment, the subsequent program is triggered as soon as the preceding program is completed, thereby ensuring smooth startup and reducing the total time wasted. The purpose of this method is to improve system stability and avoid errors caused by manual intervention.

[0062] It should be noted that for the programs after the order has been adjusted, the configuration time of each program is obtained. For example, configuring database service parameters such as connection strings takes 2 seconds, configuring application server ports takes 3 seconds, and configuring the front-end interface path takes 1 second. The total configuration time is 6 seconds. This is beneficial for integrating configuration steps, making automatic management more efficient, and reducing delays caused by human configuration.

[0063] In one possible implementation, based on the total configuration time, manual items such as manually entered parameters are identified and converted into automatic script execution. For example, parameters can be preloaded using configuration files. After automatic execution, the total configuration time is directly integrated into the startup process. This not only reduces the number of manual steps but also enables the automatic startup and configuration of all programs, resulting in a simplified overall process. For example, in enterprise-level systems, this approach can shorten the startup time from minutes to seconds, improving deployment efficiency.

[0064] For example, extending from the perspective of dependency adjustment, if a new monitoring program depends on the application server, it can be further sorted based on the total startup time to ensure that monitoring is triggered after the server starts, which is beneficial for real-time monitoring rather than post-event remediation.

[0065] In one possible implementation, the summation of configuration time can also be combined with parallel processing, such as configuring independent programs simultaneously, reducing the total configuration time and further supporting the benefits of automated execution, such as reducing downtime and improving availability in cloud environments.

[0066] It should be noted that these examples support each other from multiple aspects, including time quantification, dependency management, and configuration automation, to achieve the integration and simplification of the startup process. For example, time summation provides basic data, dependency adjustment uses this data to optimize the order, configuration summation is built on the optimized order, and automatic startup directly applies these time metrics, ultimately reducing manual steps and managing startup order and dependencies, resulting in an overall technical effect of efficient startup.

[0067] S4, the editing module is used to provide an intuitive interface for editing the virtual human character of the large artificial intelligence model through multi-dimensional settings. It supports setting the character's identity, background, personality traits and language style. It automatically generates and trains the virtual character model according to the user's settings, where the user-defined dimension set S is the parameter P_1 to P_N of each dimension.

[0068] The system acquires multiple dimensional parameters set by the user for the virtual character in the editing interface. Each dimensional parameter corresponds to a specific description of the character's identity, background, personality traits, or language style, forming an initial dimensional set containing the original descriptions of all dimensional parameters. Each dimensional parameter in the initial dimensional set undergoes structured processing, decomposing each dimensional parameter into quantifiable feature terms, and combining all feature terms into a unified parameter feature set, which preserves the correspondence between dimensions. The parameter feature set is input into a pre-trained large-scale language model, which directly generates a complete prompt description corresponding to the virtual character based on the parameter feature set. This complete prompt description integrates the settings of all dimensional parameters. The large-scale language model is then fine-tuned based on the complete prompt description, resulting in a virtual character model specific to this dimensional set. This virtual character model reflects the user-defined character identity, background, personality traits, and language style during interaction.

[0069] Specifically, the implementation process of virtual character design can be broken down and illustrated in detail from multiple stages, from user input to final model training, to demonstrate its logic and practicality.

[0070] First, when obtaining the dimension parameters set by the user, the user may input information such as the character's identity background, personality traits, and language style through an intuitive interface.

[0071] For example, a user might want to create a virtual character as a "power system safety interpreter," with a personality described as "gentle and wise" and a language style as "clear, concise, and organized." This information, once collected, forms an initial set of dimensions containing all the raw descriptive content input by the user. This method of collection ensures the integrity of the user's intent, laying the foundation for subsequent processing.

[0072] Next, during the structuring of the initial set of dimensions, the aforementioned background of a "power system safety interpreter" can be decomposed into specific characteristics, such as "possessing solid professional knowledge of power systems" and "focusing on safety procedures and risk prevention"; the personality trait of "gentle and wise" can be decomposed into "calm tone, soothing the audience's anxiety about complex risks" and "able to simplify complex concepts and explain principles logically"; the language style of "clear, concise, and organized" can be decomposed into "popularizing terminology" and "following the logical chain of 'phenomenon-principle-measure' in the narrative structure." These characteristics are integrated into a unified set of parameter features, preserving the correspondence between the dimensions. The purpose of this decomposition and integration is to transform the user's abstract description into operable data units, facilitating subsequent generation and training.

[0073] Furthermore, after the parameter feature set is input into a pre-trained large language model, the model generates a complete prompt description based on the feature terms.

[0074] For example, based on the aforementioned features, the model can generate an integrated description: "A professional focused on power system safety, possessing solid knowledge of power engineering and safety regulations, with a gentle and calm personality and clear logic, adept at using plain, accurate, and well-organized language to explain technical principles and risk prevention measures." This description organically integrates the structured features of identity, personality, and language style, ensuring both the professionalism and consistency of the role setting and providing clear and actionable textual guidance for subsequent generation or training.

[0075] Finally, during fine-tuning training based on complete prompt descriptions, large language models adjust their output style and content bias according to the generated descriptions.

[0076] For example, when interacting with users, the virtual character answers questions in a refined tone and demonstrates profound insights into historical topics. This training method allows the virtual character to authentically reflect the user's established identity, personality traits, and language style, enhancing the personalized experience of the interaction. At the same time, this method can adapt to the diverse needs of different users, ensuring consistency and depth in the character's performance.

[0077] It's important to note that each step is closely interconnected, forming a complete closed loop from user input to structured processing, prompt generation, and model training. Each step directly impacts the subsequent results; for example, the quality of structured processing determines the accuracy of the prompt description, which in turn affects the training outcome. This interlocking design not only improves the efficiency of virtual character generation but also ensures that the final character's performance closely matches user expectations, providing strong support for personalized virtual interaction.

[0078] S5, the interactive module is used to provide a user interactive interface and a 3D digital human visualization display.

[0079] The system extracts interactive intents from user-input voice or text data. These intents are determined by parsing the voice content or text semantics to identify the user's interaction needs. The parsed intents are then transformed into structured instruction data for subsequent processing and response. For this structured instruction data, a pre-built 3D digital human facial animation library is invoked to extract facial expression and lip-sync parameters corresponding to the instruction data. These parameters are then mapped onto the 3D digital human model to form preliminary visual feedback. Based on this preliminary visual feedback, the 3D digital human model is dynamically adjusted using real-time rendering tools. These adjustments include optimizing facial details and lighting effects to ensure consistency of the visual feedback across different viewpoints, generating the final display. This final display is presented to the user through an interactive interface. This interface supports real-time updates to the 3D digital human's visualization and dynamically adjusts the display based on new user input, achieving continuous interaction between the user interface and the 3D digital human's visualization.

[0080] The generation steps are as follows:

[0081] 1. The process of extracting interactive intent from user-input voice or text data.

[0082] In one possible implementation, when a user speaks a question about the weather, the system first uses a speech recognition tool to convert the speech into text, and then analyzes the keywords in the text, such as weather and location, to determine the interaction intent as querying weather information. This ensures that subsequent responses are more accurate and helps improve the consistency of the user experience.

[0083] For example, in another scenario, if a user inputs text describing a problem, the semantics are parsed, taking into account contextual relationships, thus transforming it into structured instruction data. This helps avoid ambiguity and supports the continuity of multi-turn dialogue. In this way, the acquisition of interactive intent not only improves processing efficiency but also provides reliable basic data for the expression-driven processing of 3D digital humans.

[0084] 2. The specific implementation of calling a pre-established 3D digital human facial animation library for structured instruction data.

[0085] It should be noted that this animation library is a database that stores various facial expressions and lip-sync parameters.

[0086] For example, when the instruction data represents a happy emotion, the system extracts the corresponding smile parameters, such as the numerical changes in the upturn of the corners of the mouth and the curvature of the eyes, and maps them onto the model to form preliminary visual feedback. The purpose of this is to make the digital human perform more naturally and to enhance the immersive experience of the interaction.

[0087] In one embodiment, if the instruction involves the theme of anger, the extracted parameters will include adjustments to furrowed eyebrows and closed mouth, which complement the happiness parameter to ensure the diversity of expressions. Through the mapping process, the initial visual feedback can reflect the user's intent in real time, thereby providing continuous input for subsequent rendering.

[0088] 3. Based on the initial visual feedback, details are dynamically adjusted using real-time rendering tools.

[0089] Understandably, real-time rendering tools, such as graphics processing unit-based software, will optimize facial details such as skin texture and lighting effects.

[0090] For example, in low-light environments, the tool adjusts the shadow distribution to keep expressions clear, which helps ensure consistency from multiple perspectives and provides users with a stable visual experience when rotating to view.

[0091] For example, in dynamic scenes, if the digital human's head turns, the optimization process will update the lighting and shadows to match the new perspective. This adjustment is closely linked to the mapping parameters of the previous step, and together they generate the final display image, thereby improving the overall realism and interactive smoothness of the visualization.

[0092] 4. The mechanism for presenting the final display screen to the user through an interactive interface.

[0093] It should be noted that this interface is a graphical user interface that supports real-time updates.

[0094] For example, when a user enters a new question, the interface will dynamically adjust the digital human's mouth shape and facial expression according to the new input, such as changing from a smile to surprise. This can achieve continuous interaction and help maintain the natural flow of the conversation.

[0095] In one possible implementation, if the user asks questions continuously, the interface will accumulate previous feedback and smoothly transition to the new screen. This linkage and the aforementioned adjustment process support each other, ensuring a close integration between the user interaction interface and the 3D digital human visualization, thereby providing efficient response and immersive experience in various interactive scenarios.

[0096] S6, the data storage and management module is used to store and manage user configuration files and training data, wherein the overall data set D includes the user configuration file data set U_user and the training data set T_training, and the data management operation O is the function M that processes U_user and T_training.

[0097] In the data storage and management module, the overall dataset is defined as the union of the user profile dataset and the training dataset. The overall dataset contains both user profile datasets and training datasets. The user profile dataset stores user-related configuration information, while the training dataset stores sample data used for model training. Data management operations are performed on the overall dataset. These operations use pre-defined functions to merge and organize the user profile dataset and training dataset, resulting in a unified and managed data set. A data backup is then performed on this unified data set. The backup function completely copies the overall dataset, generating a backup dataset that retains all the contents of the overall dataset for recovery. The backup dataset maintains a consistent relationship with the unified data set, ensuring that the processing results of the data management operations on the overall dataset are fully reflected in the backup dataset. This achieves the storage and management of the user profile dataset and training dataset.

[0098] For example, when defining the overall dataset in the data storage and management module, the user profile dataset can be regarded as a structured information that stores user preferences and settings. For example, user preferences include personalized training parameters, while the training dataset contains historical sample data such as images or text sequences. Defining the overall dataset as the union of the two can ensure data integrity, thereby avoiding information loss in subsequent management and improving the reliability of data processing.

[0099] In one possible implementation, when performing data management operations on the overall dataset, a pre-established function can first sort the user profile dataset, for example, by user ID, and then merge it with the training dataset to form a unified data result. This merging and organization process helps to quickly retrieve data, such as directly accessing relevant configurations during model training, thereby reducing query time and improving system efficiency, which is beneficial to optimizing overall data access performance.

[0100] For example, when performing data backup based on the data results after unified management, the backup function generates a backup data set by completely copying the overall data set, such as mirroring the original data to cloud storage. The backup data set retains all contents for recovery use. This enables rapid restoration in the event of data loss, which is beneficial for ensuring data security and continuity. For example, if the original data is damaged, the backup can seamlessly take over.

[0101] In one possible implementation, when the backup dataset maintains a consistent association with the unified managed data results, a periodic synchronization mechanism can be used to ensure that the processing results are fully reflected in the backup. For example, the backup can be updated immediately after the data management operation updates the overall dataset, thereby enabling the storage and management of the user configuration dataset and the training dataset. This consistent association helps prevent data inconsistency issues and is beneficial for supporting reliable model training and user configuration applications, such as maintaining data synchronization in a multi-device environment.

[0102] For example, the above process can be examined from multiple perspectives. In practical applications, user profile data sets may involve privacy information. By defining the overall dataset through union and performing management operations, access control can be strengthened, while backup provides redundancy protection. These aspects support each other to ensure data integrity and security. For example, merging processes support efficient training, while backups prevent risks and together improve the robustness of the system, which is beneficial for long-term data management.

[0103] In one possible implementation, another example is in an enterprise environment where the training dataset contains a large number of samples. After being processed by functions to form a unified result and then backed up, it can support team collaboration, such as multiple people accessing it simultaneously without conflict. These examples support the original process from the perspectives of efficiency and security, which is beneficial for realizing scalable data storage.

[0104] For example, from a recovery perspective, if the unified data results are lost due to a failure, the complete retention of the backup data set allows for the immediate recovery of user profiles and training data, thereby maintaining business continuity. These multi-faceted examples collectively demonstrate the effectiveness of the process and contribute to enhancing the practical value of data management.

[0105] S21, the specific steps of the automated lip-syncing and facial expression driving module are as follows: multiple programs are started and configured at once through a batch file to ensure that the programs are in the correct state; audio processing tools are used to analyze the pitch, volume and speech rate characteristics of the speech data to generate corresponding lip-syncing animations; and the lip-syncing animations are applied to the three-dimensional digital human model through the 3D game engine Unreal Engine to drive facial expression changes in real time.

[0106] The batch file launches the speech synthesis program and audio processing tool at once, configuring parameters so that the speech synthesis program can directly transfer the generated audio file to the audio processing tool for processing. The audio processing tool extracts pitch, volume, and speech rate features from the received audio file, forming a feature sequence for each frame. This feature sequence is mapped to a pre-established set of lip-shape keyframes, generating continuous lip-shape animation data. This lip-shape animation data is then applied in real-time to the facial skeleton of the 3D digital human model via Unreal Engine, achieving synchronized driving of facial expressions and lip movements.

[0107] A batch file is a scripting tool used to automate the execution of a series of commands. In this scenario, it first starts a speech synthesis program that converts text input into an audio file, such as converting the user's query into natural speech output. Then, it configures parameters to ensure that the audio file format is standardized, such as a sampling rate of 44.1kHz to match subsequent processing requirements. Next, it starts an audio processing tool and sets its input path to directly point to the generated audio file. This avoids the manual transmission step and helps improve the overall workflow efficiency because it reduces human error and speeds up the conversion from text to audio.

[0108] In one possible implementation, assuming a user inputs a greeting, the batch file first calls a speech synthesis program to generate the corresponding audio file, and then automatically passes it to the audio processing tool through parameter settings such as file naming rules. This not only simplifies the operation but also ensures data continuity, thus providing a stable input foundation for subsequent feature extraction. The audio processing tool is software specifically designed for sound signal processing. After receiving the audio file, it extracts pitch features using methods such as Fourier transform. The fundamental frequency of the audio represents the pitch of the sound; for example, identifying the vowel pronunciation corresponding to the high-pitched part in a dialogue audio. Volume features measure amplitude to reflect sound intensity; for example, the amplitude increases when shouting. Speech rate features calculate the number of syllables per second to represent speaking speed; for example, the syllables are denser when narrating quickly. These features form a sequence corresponding to each frame, such as a sequence array with one frame every 20 milliseconds. This helps capture the dynamic changes in the audio, and its beneficial effect is the accurate quantification of speech elements, providing data support for lip-sync generation.

[0109] In one possible implementation, for a news broadcast audio clip, the tool extracts pitch sequences to display intonation fluctuations, volume sequences to highlight emphasized words, and speech rate sequences to mark pauses. This resulting feature sequence comprehensively describes the audio characteristics from a temporal perspective, supporting more realistic animation mapping. A pre-built set of lip-shape keyframes is a standard library based on facial animation, containing various lip-shape poses such as open, closed, or specific consonant shapes. These keyframes are preset 3D model poses, such as a wide-open mouth shape for the vowel 'a'. Through a mapping process, feature sequences are matched with keyframes; for example, higher pitch corresponds to a wider mouth opening, and faster speech rate corresponds to rapid keyframe switching, thereby generating continuous lip-shape animation data. This is similar to interpolation animation techniques, ensuring smooth animation transitions and contributing to natural facial agglomeration because it bridges audio data with visual expression.

[0110] In one possible implementation, if the feature sequence shows a sudden increase in volume, the mapping selects keyframes emphasizing the mouth shape and inserts transition frames to form a smooth animation data sequence, which supports real-time performance from a dynamic matching perspective. Unreal Engine, a 3D game engine used for rendering and animation control, applies mouth-shape animation data to the facial skeletons of a 3D digital human model. These skeletons are the model's joints, such as the chin and lip nodes, and are driven in real-time by the engine's skeletal animation system. For example, it adjusts the rotation and position of the bones based on the animation data to achieve synchronization between facial expressions and mouth shapes. For instance, when speaking, the lips open and close synchronously with the slight raising of the eyebrows. This is beneficial for creating immersive interactions because it integrates audio-driven and visual rendering, ensuring that the digital human responds instantly and naturally.

[0111] In one possible implementation, for real-time dialogue scenarios, the engine updates the model skeleton every frame based on the incoming animation data, synchronously driving facial expression changes, thereby improving the user experience from the perspective of visual feedback.

[0112] S22, the operating cost C_run of the hardware optimization and resource management module is calculated as (P_usedT) / C_eff, where P_used is the actual power used, T is the running time, and C_eff is the resource utilization efficiency after hardware optimization, and ensures that the minimum supported graphics card performance P_min is greater than or equal to the 3070Ti graphics card performance P_3070Ti.

[0113] The actual power consumption and runtime are obtained from real-time monitoring data of the graphics card, and the resource utilization efficiency after hardware optimization is determined by the ratio of hardware resource utilization to total computing power. The operating cost is obtained by multiplying the actual power consumption and runtime by the efficiency after hardware optimization. During the operating cost calculation, it is determined whether the minimum supported graphics card performance is greater than or equal to that of a 3070Ti graphics card. If so, the current operating cost calculation result continues. The operating cost that meets the minimum supported graphics card performance condition is taken as the final operating cost value of the hardware optimization and resource management module.

[0114] The generation steps are as follows:

[0115] For example, the process of obtaining actual power consumption and runtime from real-time monitoring data of a graphics card involves using monitoring tools to capture the instantaneous power consumption and duration of the graphics card during task execution. This provides a useful technical effect of accurately reflecting the actual hardware load because real-time data avoids estimation bias. The process of determining the optimized resource utilization efficiency of the hardware from the ratio of hardware resource utilization to total computing power is to derive the optimized efficiency value by comparing the current utilization ratio with the total capacity of the graphics card, which helps to improve the accuracy of cost calculations.

[0116] In one possible implementation, the specific method for calculating the operating cost by multiplying the actual power used and the running time after hardware optimization by the efficiency includes first multiplying the power and time to form the total energy consumption, and then dividing by the efficiency value to generate a cost index. This approach can bring about useful technical effects in optimizing resource allocation because it incorporates efficiency into the calculation, thereby reducing unnecessary overhead, and it is connected with the aforementioned real-time data acquisition to ensure that the cost is based on the actual optimization results.

[0117] For example, a specific method for determining whether the minimum supported graphics card performance is greater than or equal to that of a 3070Ti graphics card during the calculation of operating costs is to compare the performance metrics of the current graphics card with the standard performance benchmark of the 3070Ti. If the performance meets the benchmark, the current operating cost calculation result is used. This implementation involves setting a performance threshold for verification. This has the useful technical effect of ensuring system compatibility because it prevents the use of substandard hardware from causing calculation failures and continues the aforementioned cost calculation logic, ensuring that only results with acceptable performance are adopted.

[0118] In one possible implementation, the specific process of using the operating cost that meets the minimum supported graphics card performance conditions as the final operating cost value of the hardware optimization and resource management module includes directly assigning this cost as the module output after the judgment is passed. This can bring about the effective technical effect of unified management of hardware resources because it integrates the chain of performance verification and cost calculation and connects with the previous steps to form a complete optimization process, thereby achieving precise control of operating costs and maximization of resource efficiency.

[0119] For example, one aspect of the above process in practical applications is that when a graphics card is processing a graphics rendering task, it acquires a moderate power consumption and a runtime of several hours. It then calculates the cost based on the efficiency ratio from the utilization rate and determines the final value after judging whether the performance meets the standard. This can bring about the useful technical effect of reducing energy waste because it supports efficient management.

[0120] In one possible implementation, and exemplified from another perspective, when the system is running machine learning training, real-time data capture is performed at high power and for long durations to calculate efficiency and cost. Then, after verifying that the performance is greater than or equal to that of the 3070Ti, the final cost is determined. This approach can bring about a useful technical effect of improving computational reliability because it ensures mutual support between cost optimization and performance assurance under high loads.

[0121] For example, taking an extended perspective, in video processing scenarios, acquiring low-power, short-duration data, determining the efficiency calculation cost, and using it as the module value after performance evaluation can bring about a useful technical effect of flexibly adapting to different tasks. This is because it integrates real-time monitoring, efficiency calculation, performance verification, and final assignment into a tight technical chain, supporting diverse hardware optimization needs.

[0122] S23, the optimized hardware resource utilization efficiency E of the hardware optimization and resource management module is further calculated as follows:

[0123]

[0124] Where AA represents the number of available hardware units, RR represents resource requirements, TT represents system response time, and SS represents the task complexity index. Total computing power.

[0125] Since actual usage efficiency is affected by the amount of resources used, the number of available hardware units is introduced. ;

[0126] To consider the ratio of resource requirements to computing power when calculating the total computing capacity of the computing system, the following considerations are added: if resource requirements... Exceeding total capacity This will increase computational load, thereby reducing efficiency;

[0127] This indicates that the ratio of task response time to complexity is taken into account, which increases the system's adaptability to tasks of different complexities. Higher response times will reduce efficiency with more complex tasks.

[0128] The hardware optimization and resource management module expression is also subject to hardware compatibility metrics. The impact, calculated using the following formula:

[0129]

[0130] in: This is a hardware compatibility metric, with a value ranging from 0 to 1. 1 indicates full compatibility, and 0 indicates incompatibility. The value reflects the degree of compatibility between the system and the graphics card. The higher the compatibility, the closer the actual usable performance is to that of the 3070Ti.

[0131] In the hardware resource optimization process, the first step is to obtain the number of currently available hardware units from the operating environment and record it as the first resource value. Simultaneously, resource demand data is collected and recorded as the first demand value, serving as the baseline data for subsequent resource allocation calculations. Based on the first resource value and the first demand value, and combined with a pre-established baseline value for total computing power, the ratio of resource demand to total computing power is calculated to form the first ratio data, used to measure the pressure level of resource allocation. After obtaining the first ratio data, system response time data is collected and recorded as the first response value. Combined with task complexity index data, recorded as the first complexity value, the ratio of response time to task complexity is calculated to generate the second ratio data, used as the basis for further adjusting resource allocation. By combining the first and second ratio data, the final calculated value of hardware resource utilization efficiency is determined. This calculated value characterizes the hardware optimization effect under the current resource demand, response time, and task complexity, completing the evaluation of hardware resource utilization efficiency.

[0132] In one possible implementation, the hardware resource optimization process obtains data on the number of currently available hardware units from the operating environment in real time and records it as the first resource value. At the same time, it simultaneously collects data on the resource requirements of the current task and records it as the first requirement value. This provides an accurate initial basis for subsequent resource pressure assessment.

[0133] Understandably, the first resource value directly reflects the number of available computing units, while the first demand value reflects the specific amount of computing, storage, or bandwidth used by the task. Together, they constitute the basic data for resource matching.

[0134] Specifically, the first resource value and the first demand value are used to calculate a ratio based on a pre-established benchmark value for total computing power. This ratio is calculated by dividing the first demand value by the benchmark value for total computing power to obtain the first ratio data. The first ratio data is used to quantify the load of resource demand relative to the overall computing power. A larger first ratio data indicates higher resource demand, which may lead to increased computing pressure, thus providing a quantitative reference for subsequent efficiency adjustments.

[0135] It should be noted that the total computing power baseline is predetermined by the hardware configuration and covers the peak processing power of all available units.

[0136] In one embodiment, after obtaining the first ratio data, the system continues to collect response time data and record it as the first response value. The task complexity index data is also collected and recorded as the first complexity value. Subsequently, the ratio of the first response value to the first complexity value is calculated to generate the second ratio data. The second ratio data reflects the relative level of response latency with respect to task processing difficulty. When the response time is long and the task complexity is low, this ratio data will increase, indicating that there is room for optimization. This data, together with the first ratio data, contributes to the efficiency calculation.

[0137] In one possible implementation, after combining the first and second proportional data, the first resource value is multiplied by a preset utilization coefficient to obtain the numerator. Then, the total computing power baseline value is multiplied by 1, added to the first proportional data, and then multiplied by 1 again and added to the second proportional data to obtain the denominator. Finally, the optimized hardware resource utilization efficiency value is obtained by dividing the numerator by the denominator. This efficiency value comprehensively considers the impact of available hardware quantity, resource demand pressure, and response time on task complexity, achieving a complete representation of the hardware resource utilization status, thereby supporting the efficiency evaluation objectives of the hardware optimization and resource management modules.

[0138] S31, in the startup process simplification module, the total configuration time T_config of each program i is the sum of the configuration times T_config,i of each program, and the program startup order dependency relationship is set to D_i,j is 1 when program i depends on program j in the directed graph.

[0139] In the program startup process, for the configuration process of each program, individual configuration time data for each program is obtained. This time data is determined by recording the time period required for each program from initialization to configuration completion. The configuration times of all programs are summed to obtain the overall configuration time, which is used as a basis for subsequent process optimization. For the overall configuration time, the distribution characteristics of the configuration time of each program are analyzed, and programs with longer configuration times are extracted as key optimization targets. The dependency data of these programs during startup is recorded. This dependency data is determined by the calling order and condition constraints between programs. After obtaining the dependency data, the dependencies between programs are constructed into a directed graph structure. Each node in the directed graph represents a program. If a program depends on another program, a directed edge is established between the corresponding nodes, and the dependency value is marked as 1 for subsequent startup order planning. For the constructed directed graph structure, the dependency paths between programs are analyzed to determine the priority of the startup order, ensuring that programs with a dependency value of 1 only begin configuration after the dependent program has started. This optimizes the overall configuration time and achieves the goal of simplifying the program configuration process.

[0140] In one possible implementation, the program startup process involves obtaining the time consumed by the configuration process. For example, for a database connection program, its individual configuration time is determined by recording the time period from loading initialization parameters to completing connection verification. This record can reveal configuration bottlenecks, thus providing a basis for overall optimization. When the time consumed by multiple programs, such as interface rendering programs and data processing programs, is accumulated to obtain the total overall configuration time, it is helpful to identify the main sources of startup delay and achieve process acceleration.

[0141] In one possible implementation, a distribution characteristic analysis is performed on the total configuration time. For example, programs with long configuration times, such as complex algorithm loaders, are extracted as key optimization targets. This extraction is based on time-consuming sorting, which can highlight the role of high-time-consuming links. Then, the dependency data of these programs is recorded. The order of calls between programs, such as the condition constraint that a data processing program needs to wait for a database connection program, is used to determine the order. This helps to avoid errors caused by disordered startup and improve system stability.

[0142] In one possible implementation, a directed graph structure is constructed after obtaining the dependency data. For example, nodes represent programs such as security authentication programs and user interface programs. If the user interface program depends on the security authentication program, a directed edge from authentication to the interface is established and marked with a value of 1. This construction process includes listing all program nodes first, and then adding edges according to the dependency data to ensure the integrity of the graph structure. This is used to plan the startup order and is beneficial for orderly execution and reducing waiting time.

[0143] In one possible implementation, dependency paths are analyzed for the directed graph structure. For example, the startup priority is determined to ensure that programs with a dependency value of 1, such as background service programs, are configured only after the network connection program they depend on has started. This analysis achieves priority sorting by traversing the graph paths, optimizes the overall configuration time, and helps to simplify the program configuration process and improve system response speed.

[0144] For example, from another perspective, in a multi-threaded environment, such as for a parallel graphics rendering program, obtaining configuration time-consuming data can capture the latency caused by thread conflicts. The sum of these times helps to balance the load, achieve more efficient resource allocation, and reduce startup bottlenecks.

[0145] For example, when analyzing distribution characteristics, such as extracting time-consuming remote callers in a distributed system, recording dependencies such as call order depending on authentication constraints can support optimized object selection and facilitate cross-node coordinated startup.

[0146] For example, when constructing a directed graph, such as adding an edge with a value of 1 in a microservice architecture, ensures that dependencies such as service discoverers depend on the registry center, which is beneficial to the accuracy of path analysis.

[0147] For example, analyzing dependency paths to determine priorities, such as ensuring configuration order in real-time applications, can optimize total time consumption and benefit overall process simplification.

[0148] In one possible implementation, these examples support each other from time consumption to graph analysis. For example, the distribution characteristics of the total time consumption are directly input into the dependent records to form a chain that supports the final optimization goal.

[0149] For example, in embedded systems, the cumulative configuration time reveals hardware limitations, while dependency graph construction resolves startup conflicts and improves the efficiency of embedded environments. These aspects together enhance the technical effect of process simplification.

[0150] S32, the total time T_reduced for the reduced manual configuration steps in the startup process simplification module is the sum of the times T_step,j for each manual configuration step.

[0151] S41, in the editing module, the set of character identity background parameters B is B_1 to B_nB, the set of personality traits parameters C is C_1 to C_nC, and the set of language style parameters L is L_1 to L_nL.

[0152] During character editing, the editing module first obtains a set of parameters for the character's identity background, defining it as the first parameter group, encompassing a range from the first identity parameter to the next multiple identity parameters, and records the specific description of each parameter. Based on the description of the first parameter group, the editing module further obtains a set of parameters for the character's personality traits, defining it as the second parameter group, encompassing a range from the first personality parameter to the next multiple personality parameters, and maps these parameters to the identity background description of the first parameter group. After completing the mapping of the second parameter group, the editing module continues to obtain a set of parameters for the character's language style, defining it as the third parameter group, encompassing a range from the first language parameter to the next multiple language parameters, and performs matching and adjustment based on the personality trait description of the second parameter group. Based on the matching and adjustment results of the third parameter group, the editing module integrates the first, second, and third parameter groups into a complete character parameter configuration, ensuring that the parameter sets for the character's identity background, personality traits, and language style each correspond to a complete range from the first to the next multiple parameters, forming a unified character feature description.

[0153] In one possible implementation, the character editing process begins by obtaining a set of parameters representing the character's identity background, which is defined as the first parameter group. This parameter group includes a range from the first identity parameter to the subsequent identity parameters. Each parameter records specific descriptive content. This ensures the diversity and completeness of the character's background. For example, when creating a virtual assistant character, the first parameter group may include professions such as teacher, age such as middle-aged, and background such as urban resident. These descriptive contents are recorded for subsequent association. The beneficial effect is that it provides a solid background foundation, avoids character settings from being detached from reality, and thus enhances the credibility and immersion of the virtual character.

[0154] In one possible implementation, based on the description of the first parameter group, a set of parameters for the character's personality traits is further obtained and defined as the second parameter group. This second parameter group includes a range from the first personality parameter to a multiple personality parameter, and is associated and mapped with the identity background description of the first parameter group. In detail, this mapping process matches personality parameters such as extroversion or introversion with background such as the teaching profession. For example, the teacher role may be mapped to the personality parameters of patience and meticulousness. Through this association, the beneficial effect is that the personality traits are more in line with the background, avoiding contradictory settings, improving the logical consistency of the character, and laying the foundation for the subsequent language style.

[0155] In one possible implementation, after completing the association mapping of the second parameter group, the parameter set of the character's language style is obtained and defined as the third parameter group, which includes the range from the first language parameter to the multiple language parameters. This third parameter group is then matched and adjusted in conjunction with the personality trait description of the second parameter group. Specifically, this adjustment process involves matching language parameters such as formal or colloquial with personality traits such as rigorous. For example, a rigorous personality is adjusted to a language style that uses formal vocabulary. The beneficial effect is to enhance the authenticity of the character's expression, ensure the coordination between language and personality, and thus improve the coherence of the user interaction experience.

[0156] In one possible implementation, based on the matching adjustment results of the third parameter group, the first, second, and third parameter groups are integrated into a complete role parameter configuration. This ensures that each parameter set corresponds to the complete range from the first to the next set of parameters, forming a unified role characteristic description. This integration process includes cross-validating the consistency of each group of parameters, such as verifying the overall harmony between teacher background, rigorous personality, and formal language. The beneficial effect is to generate a comprehensive role model that supports multi-dimensional applications such as games or chatbots, improving the flexibility and practical value of custom roles.

[0157] S51, the interactive interface function set F in the interactive module includes multiple operations O_1 to O_n.

[0158] An operation set is pre-established in the interactive interface. This set contains multiple independent operation items, each corresponding to a specific interactive function. At least one operation item is retrieved from this set to form an initial function list. The initial function list is then expanded by decomposing each operation item into multiple sub-operations. These sub-operations are sequentially linked to form an operation chain, thereby increasing the number of operation items in the initial function list. Multiple operation items are selected from the expanded operation chain and combined according to a preset interface layout rule. The combined operation items form a complete interactive interface function set, which includes multiple operations O_1 to O_n. The interactive interface function set is verified by traversing each operation item in the function set to confirm its executableness in the interface, thus ensuring that the function set completely contains multiple operations O_1 to O_n.

[0159] The generation steps are as follows:

[0160] In one possible implementation, the process of pre-establishing the operation set can be understood as extracting core interactive functions from user needs in the early stages of interface design. For example, chat input, message sending, and history viewing can be treated as independent operation items. Each of these operation items corresponds to a specific interactive function, and at least one operation item is obtained from them to form an initial function list. This ensures that the basic functions of the interface cover the user's basic operational needs, which is beneficial to improving the usability and responsiveness of the interface. At the same time, in this way, the operation set becomes the starting point for subsequent expansion, avoiding arbitrariness in function design.

[0161] 2. The specific implementation of expanding the initial function list can be to decompose each operation item, such as sending a message, into sub-operations, including text editing, attachment uploading, and sending confirmation. These sub-operations are linked sequentially to form an operation chain. For example, text editing can automatically trigger attachment uploading, thereby expanding the number of operation items. This expansion is beneficial to enriching the depth of the interactive interface and making the user experience smoother.

[0162] In one possible implementation, the formation of this operation chain can also support multimodal interaction, such as combining voice input to further enhance the flexibility of the function, and closely connect with the aforementioned initial list to ensure that the expansion does not deviate from the original foundation.

[0163] 3. The process of selecting multiple operation items from the expanded operation chain and combining them according to the preset interface layout rules. For example, after selecting text editing and sending confirmation, they are placed in the bottom bar of the interface to form a combination. The combined operation items finally constitute a complete set of interactive interface functions, including multiple operations O_1 to O_n. This combination is beneficial to optimize the user interface layout and improve operation efficiency.

[0164] In one possible implementation, the layout rules can prioritize frequently used operations based on user habits, thereby making the set of functions more practical and directly derived from the extension of the aforementioned operation chain.

[0165] 4. The specific method for verifying the set of interactive interface functions is to traverse the text editing of each operation item, such as O_1, to confirm its executableness in the interface. For example, after simulating a user click, check the response time and execute without errors. This ensures that the set of functions completely includes multiple operations O_1 to O_n. This verification is helpful in discovering potential problems and improving overall stability.

[0166] In one possible implementation, traversal can also integrate a feedback mechanism to further enhance the reliability of the set and connect it with the composition step to ensure that the final set of functions meets the requirements of the interactive module.

[0167] S61, the data storage and management module includes data backup, wherein the backup data D_backup is the processing of the overall data set D by the backup function BB.

[0168] The data storage expression in the data storage and management module is:

[0169]

[0170] in, It is the overall collection of stored data; It is a collection of user profile data; It is the training data set;

[0171] The data management expression is:

[0172] ;in, It is the result of data management operations; It is a function for data management.

[0173] The data storage and management module also includes data backup, and the data backup expression is:

[0174]

[0175] in, It is backup data; It is a backup function; It is the overall collection of stored data.

[0176] In the data storage and management module, the overall stored data set is first organized. This overall data set consists of user configuration file data sets and training data sets. These two sets of data are categorized and integrated to form a unified stored data set for subsequent processing. A backup operation is then performed on the organized stored data set. This backup operation processes the stored data set using a preset backup function, saving the processed data as a backup data set to ensure data recoverability during storage. After the backup data set is generated, it is verified by comparing its consistency with the original stored data set to determine if the backup is complete. If the consistency meets preset standards, the backup data set is confirmed as usable. The verified backup data set is then stored in a designated storage location, a pre-established independent storage area, to ensure the security of the backup data set, thereby achieving the data backup goal of the data storage and management module.

[0177] The generation steps are as follows:

[0178] For example, in the data storage and management module, when organizing the overall stored data set, one can start with the user profile data set and process it separately from the training data set. For instance, one can first classify the personal information and preference settings in the user profile data set, and then integrate the historical records in the training data set. This approach can bring about the beneficial effect of a clear data structure, which facilitates the smooth execution of subsequent backup operations.

[0179] In one possible implementation, a backup operation is performed on the organized storage data set. When processing through a preset backup function, the data can be copied to a temporary buffer and then compressed and saved as a backup data set. This can bring the beneficial effect of data recoverability, because the backup function ensures the rapid restoration of data in the event of accidental loss, thereby supporting the reliability of the overall module.

[0180] For example, when verifying the backup data set after it is generated, the backup data set can be compared byte by byte with the original stored data set. If the consistency meets the preset standard, such as a complete match, it can be confirmed as usable. This can bring the beneficial effect of backup integrity, avoid data corruption problems, and connect with the aforementioned backup operation to ensure the continuity from preparation to verification.

[0181] In one possible implementation, when storing the verified backup data set to a designated storage location, an independent cloud storage area can be selected for placement. For example, the backup data set can be encrypted before uploading. This can bring beneficial effects to data security because the pre-established independent storage area prevents unauthorized access. It is connected to the verification process to form a complete chain from backup generation to secure storage, thereby achieving the data backup goal in the data storage and management module.

[0182] Based on the embodiments of the present invention described above, and through the above description, those skilled in the art can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A three-dimensional digital human interaction system for an electrical power system, characterized by, The system comprises an automated lip-sync matching and facial expression driving module, a hardware optimization and resource management module, a startup process simplification module, an editing module, an interaction module, and a data storage and management module. The automated lip-sync matching and facial expression driving module starts and configures multiple programs through batch files and generates mouth animation to drive the facial expressions of a three-dimensional digital human model based on voice data characteristics. The hardware optimization and resource management module optimizes hardware resource usage efficiency and manages system running costs to support lower-performance graphics card devices. The startup process simplification module integrates necessary programs for automatic startup and configures management startup order and dependencies. The editing module automatically generates and trains virtual character models by setting user multi-dimensional parameters for character identity background, personality characteristics, and language style. The interaction module provides a user interaction interface and a three-dimensional digital human visual display. The data storage and management module stores and manages user configuration files and training data. The automated lip-sync matching and facial expression driving module comprises:

2. The system of claim 1, wherein, Multiple programs are started and configured at once through batch files, and configuration parameters during program startup are added to the batch script to ensure that the programs are in the correct state. Audio processing tools are used to analyze the pitch, volume, and speech rate characteristics of the voice data, and corresponding mouth animations are generated based on the analyzed audio characteristics. The generated mouth animations are applied to the three-dimensional digital human model through a 3D game engine to drive real-time facial expression changes. The hardware optimization and resource management module comprises:

3. The system of claim 1, wherein, The hardware resource usage efficiency E is calculated as U / C, where U is the actual used GPU occupancy rate and C is the total computing power of the system. The running cost C_run is calculated as (P_usedT) / C_eff, where P_used is the actual used power, T is the running time, and C_eff is the hardware-optimized resource usage efficiency. The minimum supported graphics card performance P_min is greater than or equal to the performance value P_3070Ti of the lower-performance graphics card device. The hardware optimization and resource management module comprises:

4. The system of claim 3, wherein, The optimized hardware resource usage efficiency E is calculated as: The startup process simplification module comprises: ； where AA is the number of available hardware, RR is the resource requirement, TT is the system response time, SS is the task complexity index, is the total computing power.

5. The system of claim 1, wherein, The total system startup time T_total is the sum of the startup times T_i of each program.

6. The system of claim 1, wherein, The total configuration time T_config of each program is the sum of the configuration times T_config,i of each program.

7. The system of claim 1, wherein, The program startup order dependency relationship is represented as a directed graph, where the dependency relationship marker D_i,j is 1 when program i depends on program j. The reduced manual configuration step total time T_reduced is the sum of the manual configuration step times T_step,j. The editing module comprises: The user-set dimension set S is composed of dimension parameters P_1 to P_N. The character identity background parameter set B is composed of B_1 to B_nB. The personality characteristic parameter set C is composed of C_1 to C_nC. The language style parameter set L is composed of L_1 to L_nL. The interaction module comprises: The interaction interface function set F includes multiple operations O_1 to O_n.

8. The system of claim 1, wherein, The data storage and management module comprises: a storage overall data set D comprising a user profile data set U_user and a training data set T_training; a data management operation O being a processing of the user profile data set U_user and the training data set T_training by a data management function M; and a backup data D_backup being a processing of the storage overall data set D by a backup function BB.

9. The system of claim 8, wherein, The data storage expression in the data storage and management module is: ； wherein, is a stored overall data set; is a user profile data set; is a training data set; The data management expression is: ; wherein, is the result of a data management operation; is a function of data management.

10. The system of claim 9, wherein, The data storage and management module further comprises a data backup, and the data backup expression is: ； wherein, is the data backed up; is the function backed up; is the total data set stored.

Images