HMI theme generation system and method based on large model driving

Abstract

The invention discloses an HMI theme generation system and method based on large model driving, and relates to the technical field of semantic data processing. The method comprises the following steps: taking pre-read static configuration and dynamic data acquired in real time as context information; on the basis of the context information, the intention and emotion requirements of the user are inferred, and a design token set is obtained; performing quality evaluation on the design token set to obtain scores of each set evaluation index; based on the score of each evaluation index, performing data correction on the design token set to obtain a corrected design token set; analyzing and compiling the revised design token set, and applying the revised design token set to an HMI interface; according to the method, the personalized theme can be generated according to the user semantics, the theme quality is controlled through quantitative indexes, different driving modes, in-vehicle brightness and user emotional states are adapted, and potential safety hazards are reduced.

Description

Technical Field

[0001] This invention relates to the field of semantic data processing technology, and more specifically, to a large model-driven HMI topic generation system and method. Background Technology

[0002] Against the backdrop of the continuous evolution of intelligent cockpit technology, in-vehicle human-machine interfaces have evolved from single-screen static displays to complex systems involving multi-screen collaboration, scene linkage, and remote iterative updates. OEMs need to provide a vast number of interface themes with distinct styles for different vehicle models, configurations, and user groups. The traditional model, which relies entirely on visual designers to manually complete theme color schemes and control style configurations, is no longer suitable for the demands of high-frequency updates and large-scale personalization. To reduce labor costs and accelerate theme deployment cycles, attempts have begun to introduce general image generation models and large language models to convert designers' or users' natural language descriptions into interface theme solutions. Resource building tools then automatically generate theme resource packages that can function on the in-vehicle operating system. However, to ensure the generated results can be processed by subsequent pipelines, the system often only performs format and numerical range checks on the output content. After verification, these design marks are directly mapped to the resource files required for the in-vehicle interface and packaged for release. Some implementations add a simple contrast test before packaging to eliminate extremely low-contrast color combinations, but the detection rules are usually based on general readability standards and are mostly heuristic rules with fixed thresholds, lacking specialized modeling for in-vehicle scenarios and automotive-grade indicators.

[0003] Because the aforementioned technical approaches lack a systematic secondary evaluation mechanism for in-vehicle human-machine interface scenarios during model invocation and result processing, the generated themes often present risks that are difficult to detect in a timely manner in actual vehicle environments. On the one hand, the training objectives of large language models are biased towards semantic rationality and visual aesthetics. This can easily lead to issues such as insufficient contrast between warning text and background colors at instrument panel viewing distances, overly bright decorative elements surrounding important numerical information that attract visual attention, and inconsistent colors used for the same warning level on different interfaces. These problems may not be accurately identified by simple detection rules under a common contrast threshold. On the other hand, existing solutions cannot automatically calculate a unified, quantifiable score based on in-vehicle brightness conditions, screen size and resolution, component semantic roles, and visual specifications after each theme generation. They also cannot use this score to filter out unqualified themes or prompt designers for targeted corrections. The generated results rely more on manual sampling and subjective judgment.

[0004] In this context, while existing technologies have recognized the significant value of using generative AI technologies such as large language models to provide personalized theme design for in-vehicle human-machine interfaces, current practices generally remain focused on the generation capability itself. They lack systematic secondary evaluation and filtering methods for the generated results in the context of in-vehicle scenarios, and lack technical mechanisms to unify automotive safety requirements, driving scenario characteristics, and visual standards into calculable indicators that can be applied to the model output. This results in shortcomings when deploying large-scale model-driven personalized theme design for HMIs.

[0005] In view of this, the present invention proposes an HMI topic generation system and method based on large model-driven approach to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned shortcomings of the prior art and achieve the above objectives, the present invention provides the following technical solution: a method for generating HMI topics based on a large model, comprising: Use pre-read static configurations and real-time collected dynamic data as context information; Based on contextual information, the user's intent and emotional needs are inferred to obtain a set of design tokens; The quality of the design token set is evaluated to obtain scores for each set of evaluation metrics; Based on the scores of each evaluation indicator, the design token set is corrected to obtain the corrected design token set. The modified design token set is parsed and compiled and applied to the HMI interface.

[0007] Furthermore, the static configuration includes visual specifications, a design token map, a theme quality assessment standard, and a mood color map; Dynamic data includes user semantic input, user emotional signals, in-vehicle brightness, and driving mode.

[0008] Furthermore, methods for inferring user intent include: The system performs word segmentation on the user's semantic input and filters out semantic information related to tone, style, and scene. The filtered semantic information is compared with the visual specifications. Semantic information that conforms to the color system and font scheme requirements in the visual specifications is retained, while semantic information that exceeds the scope allowed by the visual specifications is removed. The retained semantic information is adjusted for scene adaptation based on the in-vehicle brightness and driving mode. Integrate semantic information to form a set of user intent elements.

[0009] Furthermore, methods for adapting the retained semantic information to different scenarios include: The specific value of the interior brightness is compared with the preset brightness range threshold in the theme quality assessment standard to determine the current brightness scene level; the theme brightness-related elements are adjusted according to different brightness scene levels, and the theme background brightness and text-background contrast parameters are increased or decreased accordingly. Based on the scene adaptation rules corresponding to different driving modes, adjust the theme complexity-related elements, and accordingly increase or decrease the number of interface decorative elements, simplify or optimize the layout hierarchy, and adjust the intensity of control dynamic effects.

[0010] Furthermore, methods for inferring users' emotional needs include: Extracting emotional features from user emotional signals; The emotional characteristics are compared one by one with the characteristic threshold ranges corresponding to each emotional state in the theme quality assessment standard. The number of emotional characteristics falling within the characteristic threshold range of each emotional state is counted, and the emotional state with the highest proportion of emotional characteristics is determined as the user's current emotional state. Query the emotion color mapping table to obtain the recommended color range corresponding to the user's current emotional state; Cross-validate the style-related descriptions in the user's semantic input with the recommended color range to determine the color matching direction; By combining the color system in the visual guidelines, color combinations that match the brand's tone are selected from the color scheme directions to obtain a color scheme style scheme.

[0011] Furthermore, by combining the set of user intent elements, color scheme, visual specifications, in-vehicle brightness, and scene adaptation rules corresponding to driving modes, a set of design tokens is obtained. Obtain a set of design tokens as a structured description of the theme, used to describe all style elements of the theme.

[0012] Furthermore, the established evaluation metrics include contrast score, brand fit score, and contextual fit score. The contrast score, brand fit score, and contextual fit score are combined into a topic quality assessment report.

[0013] Furthermore, the method for calculating the contrast score includes: determining the target brightness scene and the corresponding minimum contrast threshold based on the in-vehicle brightness data and theme quality assessment standards; extracting the RGB color values ​​of the foreground text and background in the design token set and converting them into relative brightness values; calculating the actual contrast value using the contrast calculation standard formula; and then mapping and generating the contrast score according to the pre-design score rules. The method for calculating the brand fit score includes: extracting style features from the design token set, retrieving the corresponding design standards from the visual specifications, comparing the fit between the style features and the design standards item by item, and calculating the brand fit score by weighting the scores according to the preset scoring weights in the theme quality assessment standards. The method for calculating the context matching score includes: retrieving the theme style requirements corresponding to different driving modes in the theme quality assessment criteria and the recommended color scheme and style features corresponding to user emotional signals in the emotion color mapping table; comparing the interface complexity parameters, actual contrast values, decorative element information, and color scheme information in the design token set with the degree of fit between them and the driving mode style requirements, recommended color scheme and style features; and calculating the context matching score by weighting the scores according to the preset driving mode adaptation weight and emotion adaptation weight.

[0014] Furthermore, if the scores of each evaluation indicator in the topic quality assessment report reach the corresponding preset threshold in the topic quality assessment standard, then the verification is passed; if the score of any evaluation indicator does not reach the corresponding preset threshold, then the design token set is corrected. If the absolute value of the difference between the score of the evaluation indicator that does not reach the corresponding preset threshold and the corresponding preset threshold is within the preset slight deviation range in the topic quality evaluation standard, it is judged as slight deviation; otherwise, it is judged as not meeting the requirements. For minor deviations, adjust the relevant parameters of the design token set according to the corresponding evaluation indicators and standards and requirements until the scores of the corresponding evaluation indicators meet the standards. For cases that do not meet the requirements, a feedback message containing the score for failure, the degree of deviation, and the compliance standard requirements is generated. The design token set is then regenerated based on the context information. The quality of the design token set is repeatedly evaluated through multiple rounds of iteration until the scores of each evaluation indicator meet the standards.

[0015] Methods for data correction of the design token set include: When the vehicle is driving at high speed and approaching the tunnel entrance, the sampling frequency of the brightness inside the vehicle is increased to the high-frequency sampling multiple preset in the theme quality assessment standard, and the brightness value inside the vehicle and the current position are continuously collected according to the sampling frequency. The in-vehicle brightness value is input into a pre-built brightness change rate determination model to determine whether the current scene is a high-speed, rapidly changing scene of brightness and darkness. When a scene is determined to be a high-speed scene with rapid changes in brightness, the preset instantaneous dynamic adaptation rules are immediately retrieved, and the instantaneous adjustment values ​​of the relevant parameters in the design token set are calculated based on the instantaneous dynamic adaptation rules. The calculated instantaneous adjustment value is superimposed on the design token set in real time, triggering the instantaneous rendering of the HMI interface, and setting the brightness stability judgment condition. When the brightness change rate collected for a preset number of consecutive times is lower than the set change rate threshold, the instantaneous adjustment value is canceled and the original design token set is restored.

[0016] A large model-driven HMI topic generation system includes: The data acquisition module is used to use pre-read static configuration and real-time acquired dynamic data as context information; The data inference module infers the user's intent and emotional needs based on contextual information to obtain a set of design tokens; The data evaluation module is used to evaluate the quality of the design token set and obtain scores for each set evaluation indicator. The data correction module corrects the design token set based on the scores of each evaluation indicator, resulting in a corrected design token set. The theme application module is used to parse and compile the modified design token set for application in the HMI interface.

[0017] Compared with existing technologies, the technical effects and advantages of the HMI topic generation system and method based on large model-driven approach of the present invention are as follows: This invention pre-reads static configurations such as visual specifications and design token mapping tables, and combines them with dynamic data such as user semantic input and in-vehicle brightness to construct contextual information. A locally deployed LLM (Limited Least Metric) model infers user intent and emotional needs, generating a set of design tokens containing style parameters such as color, font, and layout. After evaluating the quality of the design token set based on three indicators—contrast, brand fit, and contextual relevance—a built-in algorithm automatically adjusts parameters for minor deviations. For cases that do not meet requirements, feedback prompts are generated to drive multiple rounds of iterative optimization of the model, forming a closed-loop correction process. Finally, the validated design token set is parsed and compiled into a theme resource package in Runtime Resource Overlay (RRO) format, which is automatically distributed and applied to all HMI (Hybrid Management Interface) interfaces in the vehicle system.

[0018] This invention effectively solves the problems of traditional manual design methods being unable to adapt to high-frequency updates and large-scale personalization needs, as well as existing AI generation solutions lacking a specialized evaluation mechanism for in-vehicle scenarios, posing safety hazards, exhibiting poor theme consistency, relying on manual review, and facing difficulties in engineering implementation. The advantages of this invention are that it can quickly generate personalized themes without extensive manual intervention. Theme quality is strictly controlled through quantitative indicators and closed-loop optimization, meeting both readability and safety requirements in driving scenarios while maintaining brand visual consistency. It also adapts to different driving modes, in-vehicle brightness, and user emotional states, reducing safety hazards. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a large model-driven HMI topic generation system disclosed in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of a large model-driven HMI topic generation system disclosed in Embodiment 2 of the present invention; Figure 3 This is a flowchart of a large model-driven HMI topic generation method disclosed in an embodiment of the present invention; Figure 4 This is a flowchart of a method for inferring user intent disclosed in an embodiment of the present invention; Figure 5 This is a flowchart of a method for inferring users' emotional needs as disclosed in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0021] Example 1: Please see Figure 1 As shown, this embodiment discloses an HMI topic generation system based on a large model, including a data acquisition module, a data inference module, a data evaluation module, a data correction module, and a topic application module. Each module is connected by wired and / or wireless means to realize data transmission.

[0022] The data acquisition module is used to use pre-read static configurations and real-time acquired dynamic data as context information.

[0023] When preparing to generate a theme, the system first performs data acquisition initialization. It reads a static configuration file and loads visual specifications, a design token mapping table, theme quality assessment criteria, and an emotional color mapping table into memory. The visual specifications are the brand's visual design standards, including the specified color scheme, font scheme, logo, and its usage guidelines. This type of data is typically provided by the design department and stored in the vehicle's infotainment system configuration file for reference during theme generation. The design token mapping table defines the correspondence between design tokens and specific interface resources. For example, color tokens correspond to system color resource names, and font size tokens correspond to the dimens items in style files. This mapping table is statically stored in the system as a configuration file and used by the resource generation module. The theme quality assessment criteria are the quantitative evaluation rules and threshold configurations for theme quality. Examples include the minimum contrast standard between text and background, the tolerance threshold for theme colors deviating from the visual specifications, and the acceptable brightness range under different driving scenarios. These standards are determined by design specifications and safety regulations and are built into the system as a configuration file. The emotional color mapping table is a mapping table between emotions / scenes and color schemes. For example, it maps soothing emotions to cool blue-green tones and passionate scenes to highly saturated warm colors. This table, developed by the design team based on user research, is statically stored in the system configuration for reference during theme generation and evaluation. Simultaneously, real-time environmental data, including in-vehicle brightness and driving mode, is acquired through sensors and system interfaces. Furthermore, current user emotion signals are also acquired. These signals are detected through in-vehicle cameras, voice emotion analysis, and other means, indicating the user's current emotional state, such as excitement, calmness, or fatigue. The system's multimodal sensors or AI algorithms collect this signal in real time, providing additional contextual basis for theme generation. In-vehicle brightness is the current cabin light intensity measured by the in-vehicle ambient light sensor, used to determine whether the environment is bright or dim. The vehicle's operating system periodically acquires this value through sensor interfaces. Driving mode information includes the vehicle's current driving mode, such as Eco, Sport, and Autopilot modes. This is obtained through the vehicle's CAN bus or system API, triggering notifications when modes switch to adjust the theme style to suit different driving scenarios. The above data covers the user preferences, environmental perception, and design specifications required for theme generation. Static data is provided through configuration files or built-in resources, while dynamic data is collected in real time via sensors and system interfaces.

[0024] The data inference module infers the user's intent and emotional needs based on contextual information, and obtains a set of design tokens.

[0025] The locally deployed LLM model receives user semantic input, static configuration, and dynamic data. Visual specifications, a design token map, topic quality assessment criteria, an emotion color map, in-vehicle brightness, driving mode, and user emotional signals collectively constitute the contextual information required for topic generation, providing static constraints and dynamic scene basis for topic generation. User semantic input consists of a description of the topic request issued by the user via voice or text, including natural language descriptions of topic style and scene intent selection, such as "I want a refreshing summer theme." The vehicle's infotainment system acquires user semantic input through voice recognition or a touch interface.

[0026] Please see Figure 4 As shown, the method for inferring user intent includes: LLM performing word segmentation on user semantic input, extracting core words related to the topic, and filtering semantic information related to tone, style, and scene; comparing the filtered semantic information with visual specifications, retaining semantic information that conforms to the color system and font scheme requirements in the visual specifications, and removing semantic information that exceeds the allowable range of the visual specifications; and combining in-vehicle brightness and driving mode to perform scene adaptation adjustments on the retained semantic information. The specific adaptation process is as follows: first, obtain the specific value of in-vehicle brightness, compare the specific value of in-vehicle brightness with the preset brightness range threshold in the topic quality assessment standard, and determine the appropriate brightness range threshold. The system first determines the current brightness level of the scene and adjusts the theme's brightness-related elements accordingly, increasing or decreasing the background brightness and contrast parameters between text and background to ensure that text information is clearly legible under the current brightness conditions. Then, it acquires the driving mode and adjusts the theme's complexity-related elements based on the scene adaptation rules corresponding to different driving modes. This includes increasing or decreasing the number of interface decorative elements, simplifying or optimizing the layout hierarchy, and adjusting the intensity of control dynamic effects to ensure that the theme's complexity matches the attention requirements of the driving scene, achieving precise adaptation of theme elements to the real-time driving scene. Finally, it integrates the filtered, compared, and adjusted semantic information to form a clear set of user intent elements.

[0027] Please see Figure 5As shown, the method for inferring users' emotional needs includes: The emotional features extracted by LLM from user emotional signals refer to facial expression features, eye state features, and facial muscle tension features collected by the in-vehicle camera, as well as user speech rate features, tone features, and speech energy features obtained through voice emotion analysis; the process of determining the emotional state is as follows: LLM compares each extracted emotional feature with the preset feature threshold range corresponding to each emotional state in the topic quality assessment standard, counts the number of emotional features falling within each emotional state feature threshold range, and determines the emotional state with the highest proportion of emotional features as the user's current emotional state; it queries the emotional color mapping table to obtain the recommended color scheme range corresponding to the user's current emotional state; and it cross-validates the style-related descriptions in the user's semantic input with the recommended color scheme range. Specifically, the process involves LLM extracting the color tendency corresponding to the style-related descriptions in the user's semantic input, judging the degree of overlap between the color tendency and the recommended color scheme range, and if the overlap is high... If the degree of overlap reaches the preset matching threshold in the theme quality assessment standard, the recommended color scheme range is directly determined as the color matching direction. If the degree of overlap does not reach the preset matching threshold, the core semantics of the style description in the user's semantic input are extracted, and the color matching logic corresponding to the emotional state in the emotional color mapping table is combined to adjust the recommended color scheme range to form a color matching direction consistent with the core semantics of the style description. Combining the color system in the visual specifications, specific color combinations that conform to the brand tone are selected from the color matching directions. The specific selection process is as follows: First, colors belonging to the color system of the visual specifications are selected from the color matching directions, and colors explicitly prohibited by the visual specifications are eliminated. Then, according to the definition and usage ratio requirements of the main color, auxiliary color, and accent color in the visual specifications, the candidate range of theme main color, auxiliary color, and accent color is determined from the selected colors. Finally, according to the color matching harmony requirements in the visual specifications, the colors in the candidate range are combined and verified, and the combination that meets the color matching ratio and the visual effect is consistent with the brand tone is selected to form the final color scheme style scheme.

[0028] LLM integrates a set of user intent elements, color schemes, rigid requirements of visual guidelines, and scene adaptation rules corresponding to in-vehicle brightness and driving modes to generate a corresponding set of design tokens, which is a structured description of the theme. The set of design tokens includes style parameters such as color, font, and layout, as well as metadata such as usage tags and hierarchical relationships.

[0029] The data evaluation module is used to evaluate the quality of the design token set and obtain scores for each set evaluation metric.

[0030] The established evaluation metrics include contrast score, brand fit score, and contextual fit score. The method for evaluating and scoring the design token set, calculating each metric according to the theme quality evaluation criteria, includes: The method for calculating the contrast score is as follows: The in-vehicle brightness data is compared with the preset brightness scene division thresholds in the theme quality assessment standard to determine the current target brightness scene. The brightness scene division threshold is the lowest contrast threshold corresponding to each brightness scene. The RGB values ​​of the foreground text color and the background color are extracted from the design token set. According to a preset relative brightness conversion formula, the RGB values ​​of the foreground text color and the background color are converted into corresponding relative brightness values. The relative brightness conversion formula is a standardized calculation formula set according to human visual characteristics and color perception standards. Based on the two converted relative brightness values, the actual contrast value is calculated using the standard contrast calculation formula. The standard formula for calculating contrast is: (relative brightness value of the darker area + 0.05) divided by (relative brightness value of the darker area + 0.05). The contrast score is calculated based on the comparison between the actual contrast value and the minimum contrast threshold corresponding to the current target brightness scene: if the actual contrast value is greater than or equal to 1.5 times the minimum contrast threshold, the score is the full contrast score set in the theme quality assessment standard; if the actual contrast value is greater than or equal to the minimum contrast threshold but less than 1.5 times the minimum contrast threshold, the score is calculated by dividing (actual contrast value - minimum contrast threshold) by (0.5 times the minimum contrast threshold) and then multiplying by the full contrast score; if the actual contrast value is less than the minimum contrast threshold, the score is zero. This calculation method ensures that the score quantifies the readability of foreground text and background under current lighting conditions, thereby guaranteeing readability and safety in driving scenarios.

[0031] The method for calculating the brand fit score is as follows: Extract the specific color value of the main theme color, font family and font size parameters, and style features of key UI elements from the design token set. Key UI elements include the outline styles and fill rules of buttons, title bars, and alert components. Retrieve the color system, font scheme, and key UI element design standards specified in the visual specifications. The color system includes standard color values ​​and allowable deviation ranges for primary, secondary, and accent colors; the font scheme includes specified font families, standard font sizes for each level of text, and font size adjustment restrictions; and the key UI element design standards include style specifications for each core component. Compare the extracted main theme color value with the standard color value of the primary color in the visual specifications' color system, and calculate the color value in the RGB color space. The Euclidean distance is used to compare the main color scheme with the preset color tolerance threshold in the visual specifications to determine the degree of color matching. The extracted font family and font size parameters are compared with the specified content in the font scheme of the visual specifications to determine whether the font family meets the requirements and whether the font size is within the standard font size and adjustment limits. The extracted key UI element style features are compared with the corresponding design standards in the visual specifications item by item to check whether the outline style and filling rules meet the specifications. Based on the preset brand matching score weights in the theme quality assessment standards, the degree of color matching, font compliance, and key UI element style compliance are scored separately, and the weighted calculation is used to obtain the total brand matching score to evaluate whether the use of color and font maintains brand consistency.

[0032] The method for calculating the context matching score is as follows: Retrieve the theme style requirements corresponding to different driving modes from the preset theme quality assessment criteria, including interface complexity level, contrast standard, and restrictions on the number of decorative elements. Simultaneously, query the emotional color mapping table to obtain the recommended color range and style characteristics corresponding to the user's emotional signals. Extract the theme's interface complexity parameters, actual contrast value, number and type of decorative elements, and color scheme information from the design token set. Based on the current driving mode, compare the extracted interface complexity parameters with the interface complexity level corresponding to that driving mode, and compare the actual contrast value with the corresponding... The system compares the number of decorative elements with the limit for the number of decorative elements corresponding to the driving mode to determine whether each parameter meets the style requirements of the driving mode. It also compares the extracted color scheme information with the recommended color range corresponding to the user's emotional signal to determine whether the color scheme matches the style characteristics corresponding to the emotion. Based on the preset driving mode adaptation weight and emotion adaptation weight in the theme quality assessment standard, the system scores the degree of conformity of each parameter of the driving mode and the degree of fit between the color scheme and the recommended emotional style. The total context matching score is calculated by weighting the scores to complete the judgment of the degree of fit between the theme style and the context.

[0033] The contrast score, brand fit score, and context matching score are summarized into a topic quality assessment report, which includes the specific indicator values ​​of the contrast score, brand fit score, and context matching score, as well as the judgment results of whether each score passes the preset threshold in the topic quality assessment standard.

[0034] The data correction module corrects the design token set based on the scores of each evaluation indicator, resulting in a corrected design token set.

[0035] The theme results are screened and optimized based on the theme quality assessment report. If the contrast score, brand fit score, and contextual matching score in the theme quality assessment report all reach the corresponding preset thresholds in the theme quality assessment criteria, the theme passes the verification; if the score of any of the above assessment indicators does not reach the corresponding preset threshold, the feedback optimization process is triggered.

[0036] The specific method for judging minor deviations is as follows: First, identify the type of indicator for the non-compliant item. Contrast score is a key safety indicator, while brand fit score and context matching score are non-key compliance indicators. Second, calculate the absolute value of the difference between the actual score of the non-compliant item and the corresponding preset threshold. Finally, determine whether this absolute value of the difference is within the preset minor deviation range in the theme quality assessment standard. If the non-compliant item is only a non-key compliance indicator, or the absolute value of the difference in the key safety indicator is within the minor deviation range and the recognition time of the core information does not exceed the preset maximum recognition time threshold for core information in the theme quality assessment standard, and the visual recognition parameter of the core information is not lower than the preset minimum visual recognition threshold for core information in the theme quality assessment standard, then it is judged as a minor deviation. The core information includes vehicle speed, RPM, warning prompts, and navigation key guidance information. The visual recognition parameter is the quantified value of the distinction between the core information and surrounding elements.

[0037] For cases deemed minor deviations, the system uses a built-in algorithm to automatically adjust the parameters of the design token set. The specific correction method is as follows: If the non-compliance item is a contrast score, the built-in algorithm extracts the RGB color values ​​of the foreground text color and background color from the design token set. It adjusts the color value parameters by increasing the brightness value of the foreground text color or decreasing the brightness value of the background color. After each adjustment, the contrast under the current lighting conditions is calculated until the contrast score reaches a preset threshold. If the non-compliance item is a brand fit score, for deviations in the main color tone, the algorithm adjusts the RGB color values ​​of the main color tone in the design token set according to the standard color value of the main color in the visual specifications, reducing its Euclidean distance from the standard color value in the RGB color space until it meets the color tolerance threshold requirements. For deviations in font compliance, the algorithm directly adjusts the design token set... The font families and font size parameters in the set are replaced with the font families and corresponding standard font sizes specified in the visual specifications. For style deviations of key UI elements, the algorithm adjusts the outline styles and fill rule parameters of key UI elements in the design token set according to the corresponding design standards in the visual specifications. If the non-compliant item is the context matching score, for driving mode adaptation deviation, the algorithm adjusts the number parameters and layout hierarchy parameters of decorative elements in the design token set according to the interface complexity level and decorative element quantity limit corresponding to the current driving mode, or adjusts the color contrast parameters according to the corresponding contrast standard. For emotion adaptation deviation, the algorithm queries the recommended color range corresponding to the user's emotion signal in the emotion color mapping table and adjusts the color parameters of the color scheme in the design token set until the context matching score reaches the preset threshold.

[0038] The specific method for determining non-compliance is as follows: if the non-compliant item includes a key safety indicator, namely the contrast score, and the absolute value of the difference between its actual score and the preset threshold exceeds the slight deviation range preset in the theme quality assessment standard; or if the non-compliant item is a non-key compliance indicator, but the absolute value of the difference between its actual score and the preset threshold exceeds the slight deviation range; or if the number of non-compliant items reaches the multiple deviation thresholds preset in the theme quality assessment standard, then it is determined to be non-compliant.

[0039] For cases deemed non-compliant, the system generates a feedback message. This message includes the name of the non-compliant evaluation indicator, the degree of deviation between the actual score and the preset threshold, and the corresponding compliance standard requirements. This feedback message is input into the locally deployed LLM (Local Management Module). The LLM, combined with visual specifications, a design token mapping table, theme quality assessment standards, an emotional color mapping table, in-vehicle brightness, driving mode, user emotional signals, and user semantic input, regenerates the design token set. This optimization process can be iterated multiple times. After each regeneration of the design token set, the quality assessment is repeated until all scores in the theme quality assessment report reach the corresponding preset threshold. A corrected design token set that passes quality verification is obtained, satisfying all evaluation requirements of the theme quality assessment standards and ready to enter the resource building phase. The entire optimization loop is based on the locally deployed LLM and built-in algorithms to achieve a closed-loop optimization process, independent of external network transmission, ensuring the stability of the generation process and the controllability of the results.

[0040] The theme application module is used to parse and compile the modified design token set for application in the HMI interface.

[0041] The automated conversion and packaging of theme resources is performed as follows: The resource generation module loads the design token mapping table and parses the preset correspondence between design tokens and the resource files required by the vehicle system. This correspondence clarifies the association between color design tokens and Android theme resources (colors.xml), font design tokens and Android theme resources (styles.xml), spacing design tokens and Android theme resources (dimens.xml), and layout design tokens and corresponding layout resource files. For each style parameter in the correction design token set, the resource generation module automatically parses and assigns values ​​according to the above correspondence, accurately assigning the color value parameters contained in the color design tokens to the corresponding fields in the colors.xml file, and configuring the font family and font size parameters contained in the font design tokens. The system configures the spacing values ​​contained in the spacing design tokens to the corresponding dimension items in the dimens.xml file, and configures the layout hierarchy and control position parameters contained in the layout design tokens to the corresponding layout resource files, forming a complete resource overlay file set covering all interface style requirements. The resource generation module calls the Android system's native compilation toolchain and compiles the complete resource overlay file set according to the RRO format packaging specification, completing the format conversion and integration of resource files. The entire process of design token parsing, resource file generation, and theme package compilation is executed through a preset CI / CD automated pipeline without manual intervention, ensuring the standardization and efficiency of the process. If the compilation process terminates without exception, the system confirms that the theme package file meets the installation requirements of the vehicle system and completes the generation of the theme resource package.

[0042] The final result is a compiled theme resource package, i.e., a theme package file in RRO format, along with corresponding metadata descriptions. The metadata descriptions include information such as theme version number, applicable vehicle models, and theme function identifiers.

[0043] The system initiates an automated theme distribution and application process, which is as follows: The theme management module of the vehicle infotainment system receives the theme package file in RRO format and its corresponding metadata description. Simultaneously, it synchronizes the theme package file and metadata description to the theme store backend of the vehicle infotainment system, completing the distribution and registration of theme resources. The theme management module performs compatibility checks on the received RRO format theme package file, verifying whether the applicable vehicle model range in the metadata description matches the current vehicle infotainment system model and whether the theme version number meets the version compatibility requirements of the vehicle infotainment system. Upon successful verification, the local installation process is triggered. The theme management module deploys the RRO format theme package file to the system's preset theme resource directory according to the application installation specifications of the vehicle infotainment system's Android operating system. When the vehicle infotainment system's Android operating system is running, the RRO mechanism automatically triggers the resource allocation process. The source overlay logic, through preset resource priority rules, replaces the original UI resources in the vehicle system with color resources, font resources, layout resources, and control style resources from the new theme resource package, achieving a global theme resource replacement. The theme management module sends interface refresh commands to all HMI applications, including instrument panel applications, central control screen applications, and HUD applications, through the vehicle system's internal broadcast mechanism or preset interface call methods. After receiving the interface refresh command, each HMI application immediately loads the replaced new theme resources and re-renders its own interface components to ensure that the visual style of each application interface is consistent with the new theme. The entire process of theme distribution, installation, resource overlay, and interface refresh is completed within a preset time threshold to ensure the immediacy of theme switching after user operation and to achieve rapid effect of theme application.

[0044] After the new theme is fully applied to all HMI application interfaces of the vehicle system, users can intuitively obtain the updated personalized theme visual effects. At the same time, information such as the storage path of the theme package file, metadata description, installation time, and application status is recorded in the theme management database of the vehicle system, providing data support for subsequent theme query, switching, uninstallation and other management operations.

[0045] Through the above steps, this embodiment implements multi-round quality control of the generated theme locally in the vehicle's infotainment system. It utilizes quantitative indicators to evaluate and automatically optimize the generated results, ensuring a stable theme style and controllable quality, meeting driving safety and brand consistency requirements. The final generated theme is automatically built and distributed, instantly applied to the entire vehicle's HMI system. All application interfaces maintain a unified style, adapting to day and night environments and different driving scenarios, providing users with a reliable real-time theme customization experience.

[0046] Example 2: In existing solutions, in-vehicle brightness is collected periodically and the adjustment strategy is based on stable brightness scenarios. However, when driving at high speeds, vehicles frequently enter and exit tunnels, tree-lined roads, and other similar scenarios, causing rapid changes in in-vehicle brightness within a short period. Due to the visual adaptation delay of the human eye, the contrast and brightness parameters of regular subjects cannot be quickly adjusted, potentially leading to momentary unreadable information and affecting driving safety. Therefore, this embodiment further improves upon Embodiment 1 to address the aforementioned problems.

[0047] Please see Figure 2 As shown, this embodiment discloses an HMI topic generation system based on a large model, which also includes a brightness adjustment module to solve the problem of instantaneous readability under rapid switching between high brightness and darkness.

[0048] The brightness adjustment module establishes real-time linkage with the existing data acquisition module. When the vehicle is driving at high speed, the acquisition frequency of the in-vehicle brightness is increased to the high-frequency acquisition multiple preset in the theme quality assessment standard. The current driving speed and position information of the vehicle are collected simultaneously to ensure accurate capture of the instantaneous state of rapid brightness changes. When the vehicle is driving at high speed and approaching the tunnel entrance, the in-vehicle brightness value and current position are continuously collected according to the acquisition frequency.

[0049] The collected in-vehicle brightness values ​​are input into a pre-built brightness change rate determination model. The brightness change rate determination model uses the preset brightness change safety threshold in the theme quality assessment standard as a benchmark to calculate the ratio of the brightness difference between two adjacent collections to the collection time interval. By comparing the ratio with the brightness change safety threshold, it is determined whether it belongs to a high-speed, rapid switching scene between light and dark. If the ratio of the brightness difference between two adjacent collections to the time interval exceeds the preset threshold, it is determined to be a rapid switching scene.

[0050] When a scene is determined to be changing abruptly, the preset instantaneous dynamic adaptation rules are immediately retrieved. The instantaneous dynamic adaptation rules are based on the visual priority settings of the core information, which are the vehicle speed, RPM, warning prompts, and key navigation guidance information defined above. The instantaneous dynamic adaptation rules specify the style adjustment logic of the core information corresponding to different brightness change directions. When the brightness drops suddenly, the contrast and text brightness of the core information are temporarily increased, and when the brightness rises suddenly, the reflectivity of the background of the core information is temporarily reduced.

[0051] Extract style parameters corresponding to core information from the currently active set of design tokens, including text color value, background color value, font size and border style parameters of core information, to ensure that the adjustment only applies to core information and does not change the theme style of non-core information. Extract and display the color value of text, background color value and current font size and other core information style parameters.

[0052] Based on the instantaneous dynamic adaptation rules and the extracted core information style parameters, instantaneous adjustment values ​​are calculated. By increasing the brightness value of the text color, decreasing the brightness value of the background color, or increasing the font size, the distinction between the core information and the surrounding environment is improved. Based on the adjustment logic of sudden brightness drop in the instantaneous dynamic adaptation rules, combined with the extracted core information text and background color values, relevant parameters are calculated and adjusted to ensure that the human eye can still quickly recognize the information during the adaptation delay period.

[0053] The calculated instantaneous adjustment value is superimposed on the original design token set in real time. The instantaneous refresh interface of the vehicle system triggers the real-time rendering of the core information area of ​​the HMI interface. At the same time, a brightness stability judgment condition is set. When the brightness change rate collected for a preset number of consecutive times is lower than the change rate threshold, the instantaneous adjustment value is automatically canceled and the parameter configuration of the original design token set is restored. After the vehicle leaves the tunnel, the brightness change rate continues to be lower than the threshold, and the system automatically restores the original style of the core information. This ensures instantaneous readability during high-speed brightness switching without affecting the overall style consistency of the theme.

[0054] Example 3: Please see Figure 3 As shown, this embodiment provides a method for generating HMI topics based on a large model, including: Use pre-read static configurations and real-time collected dynamic data as context information; Based on contextual information, the user's intent and emotional needs are inferred to obtain a set of design tokens; The quality of the design token set is evaluated to obtain scores for each set of evaluation metrics; Based on the scores of each evaluation indicator, the design token set is corrected to obtain the corrected design token set. The modified design token set is parsed and compiled and applied to the HMI interface.

[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0056] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A large model driven HMI theme generation method based on, characterized in that, The application relates to a design token generation method based on user context information. The method comprises the following steps: reading pre-read static configuration and real-time collected dynamic data as context information; based on the context information, inferring the user's intention and emotional needs to obtain a design token set; quality evaluation of the design token set to obtain the scores of the set evaluation indexes; based on the scores of the evaluation indexes, data correction is performed on the design token set to obtain a corrected design token set; 2. The HMI theme generation method based on a large model drive according to claim 1, characterized in that, analysis and compilation of the corrected design token set for application to an HMI interface. The static configuration comprises visual specifications, a design token mapping table, theme quality evaluation standards and an emotional color mapping table; 3. The HMI theme generation method based on a large model drive according to claim 2, characterized in that, The dynamic data comprises user semantic input, user emotional signals, in-vehicle brightness and driving modes. The method for inferring the user's intention comprises the following steps: word segmentation processing is performed on the user semantic input, and semantic information related to color, style and scene is screened out; the screened semantic information is compared with the visual specifications, and the semantic information meeting the color system and font scheme requirements in the visual specifications is retained, and the semantic information exceeding the allowed range of the visual specifications is removed; scene adaptation adjustment is performed on the retained semantic information in combination with the in-vehicle brightness and the driving mode; 4. The HMI theme generation method based on a large model drive according to claim 3, characterized in that, the semantic information is integrated to form a user intention element set. The method for inferring the user's emotional needs comprises the following steps: emotional features in the user emotional signals are extracted; the emotional features are compared with feature threshold ranges corresponding to preset emotional states in the theme quality evaluation standards one by one, the number of the emotional features falling into the feature threshold ranges of the emotional states is counted, and the emotional state with the highest emotional feature number ratio is determined as the current emotional state of the user; the emotional color mapping table is queried to obtain a recommended color range corresponding to the current emotional state of the user; cross verification is performed on the style-related description in the user semantic input and the recommended color range to determine a color matching direction; 5. The HMI theme generation method based on a large model drive according to claim 4, characterized in that, in combination with the color system in the visual specifications, a color combination meeting the brand keynote is screened out from the color matching direction to obtain a color matching style scheme. The user intention element set, the color matching style scheme, the visual specifications, the in-vehicle brightness and the scene adaptation rules corresponding to the driving mode are combined to obtain a design token set; 6. The HMI theme generation method based on a large model drive according to claim 2, characterized in that, a structured description of the design token set is obtained to describe all style elements of the theme. The set evaluation indexes comprise contrast score, brand fit score and context matching score; 7. The HMI theme generation method based on a large model drive according to claim 6, characterized in that, the contrast score, the brand fit score and the context matching score are summarized as a theme quality evaluation report. The method for calculating the contrast score comprises the following steps: determining a target brightness scene and a corresponding contrast minimum threshold value according to in-vehicle brightness data and theme quality evaluation standards, extracting RGB color values of foreground text and background in the design token set and converting the RGB color values into relative brightness values, calculating actual contrast values according to a contrast calculation standard formula, and generating a contrast score according to a pre-design rule; The method for calculating the brand fit score comprises the following steps: extracting style features in the design token set, calling corresponding design standards in the visual specifications, comparing the style features with the design standards item by item in terms of fit degree, and calculating the brand fit score according to preset score weights in the theme quality evaluation standards; The method for calculating the context matching degree score comprises: calling the theme style requirements corresponding to different driving modes in the theme quality evaluation standard and the recommended color system and style features corresponding to the user emotional signals in the emotional color mapping table, respectively comparing the interface complexity parameters, the contrast actual values, the decoration element information, and the color scheme color system information in the design token set with the matching degrees of the driving mode style requirements, the emotional recommended color system, and the style features, calculating the context matching degree score by weighting according to the preset driving mode adaptation weight and the emotional adaptation weight.

8. The HMI theme generation method based on a large model drive according to claim 6, characterized in that, If the scores of each evaluation index in the theme quality evaluation report all reach the corresponding preset threshold in the theme quality evaluation standard, it is determined to pass the verification; if the score of any evaluation index does not reach the corresponding preset threshold, the design token set is subjected to data correction. If the absolute value of the difference between the score of the evaluation index that does not reach the corresponding preset threshold and the corresponding preset threshold is within the preset slight deviation interval in the theme quality evaluation standard, it is determined to be a slight deviation, otherwise, it is determined to be not in compliance with the requirements. For the case of slight deviation, the relevant parameters of the design token set are adjusted according to the corresponding standard and requirements for the corresponding evaluation index until the score of the corresponding evaluation index meets the standard. For the case of not in compliance with the requirements, a feedback prompt containing the unqualified score, the deviation degree, and the compliance standard requirements is generated, the design token set is regenerated in combination with the context information, and the quality evaluation of the design token set is repeatedly performed through multiple rounds of iteration until the scores of each evaluation index meet the standard.

9. The HMI theme generation method based on a large model drive according to claim 8, characterized in that, The method for correcting the data of the design token set comprises: When the vehicle is close to the entrance of a tunnel at high speed, the collection frequency of the indoor brightness is increased to a high-frequency collection multiple preset in the theme quality evaluation standard, and the indoor brightness value and the current position are continuously collected at the collection frequency; The indoor brightness value is input into a pre-constructed brightness change rate determination model to determine whether the current situation belongs to a high-speed bright-dark sharp switching scenario; When it is determined to be a high-speed bright-dark sharp switching scenario, the preset instantaneous dynamic adaptation rule is immediately called, and the instantaneous adjustment value of the relevant parameters in the design token set is calculated based on the instantaneous dynamic adaptation rule; The calculated instantaneous adjustment value is superimposed in real time into the design token set, triggering the immediate rendering of the HMI interface, and the brightness stability determination condition is set, when the brightness change rate collected for a continuous preset number of times is lower than the set change rate threshold, the instantaneous adjustment value is cancelled, and the original design token set is restored.

10. A large model driven based HMI theme generation system for implementing the large model driven based HMI theme generation method of any one of claims 1-9, characterized in that, It comprises: A data collection module for collecting pre-read static configurations and real-time collected dynamic data as context information; A data inference module for inferring the user's intention and emotional needs based on the context information to obtain the design token set; A data evaluation module for quality evaluating the design token set to obtain the scores of each evaluation index; A data correction module for correcting the data of the design token set based on the scores of each evaluation index to obtain the corrected design token set; A theme application module for parsing and compiling the corrected design token set and applying it to the HMI interface.

Images