A cockpit service recommendation method and device, and electronic equipment

Abstract

The embodiment of the application provides a cockpit service recommendation method and device and electronic equipment, and relates to the technical field of electronic information. The method comprises the following steps: acquiring cockpit multi-source data; the cockpit multi-source data comprises user voice data and environment perception data; determining user information based on the user voice data; the user information comprises user identity, user orientation and user authority; determining a target cockpit service based on the cockpit multi-source data and the user information through a service recommendation model; the service recommendation model is used for determining a plurality of cockpit services matched with user demand intention according to the cockpit multi-source data, and then determining the target cockpit service from the plurality of cockpit services according to the user information; and executing the target cockpit service. The embodiment of the application can be used for an intelligent cockpit system, can realize accurate, safe and personalized service recommendation, and thus improves user satisfaction and safety of cockpit service.

Description

Technical Field

[0001] This application relates to the field of electronic information technology, and in particular to a cabin service recommendation method, apparatus, and electronic device. Background Technology

[0002] Intelligent cockpit systems are the core carriers for enhancing vehicle human-machine interaction and driving experience. Their key lies in accurately identifying users and seamlessly providing personalized services. However, traditional methods typically rely on limited information dimensions for user identification and service matching. This can lead to limitations in the accuracy, timeliness, and personalization of service recommendations in the complex and ever-changing real-world cockpit environment, failing to meet users' actual needs and safety requirements. Summary of the Invention

[0003] The purpose of this application is to provide a cabin service recommendation method, apparatus, and electronic device, which aims to achieve accurate, safe, and personalized service recommendations, thereby improving user satisfaction and safety of cabin services.

[0004] In a first aspect, this application provides a cabin service recommendation method, the method comprising: acquiring multi-source cabin data; the multi-source cabin data including user voice data and environmental perception data; determining user information based on user voice data; the user information including user identity, user location, and user permissions; determining a target cabin service based on the multi-source cabin data and user information through a service recommendation model; the service recommendation model being used to determine multiple cabin services matching the user's needs and intentions based on the multi-source cabin data, and then determining the target cabin service from the multiple cabin services based on the user information; and executing the target cabin service.

[0005] The cabin service recommendation method provided in this application acquires multi-source cabin data, including user voice data and environmental perception data, providing a comprehensive information foundation reflecting user expression and real-time cabin status for subsequent analysis, thus ensuring a comprehensive understanding of user needs and the cabin scenario. Furthermore, based on user voice data, user information including user identity, location, and permissions is extracted, enabling the identification of the current user's personalized characteristics and operational permissions, providing a basis for accurate service adaptation. On this basis, the service recommendation model uses multi-source cabin data to infer user intent and generate multiple possible cabin services. Then, combining user information, it selects the most suitable target cabin service from these services. This process achieves coherent reasoning from raw data to intent recognition and personalized decision-making, ensuring that the recommended service not only meets the user's actual needs but also conforms to the current cabin environment and user permission constraints. Finally, by executing the target cabin service, automated and accurate responses to user needs are directly achieved, thereby improving user satisfaction and security of cabin services.

[0006] In conjunction with the first aspect mentioned above, in one possible implementation, user information is determined based on user voice data, including: extracting voiceprint features from user voice data; determining user identity based on voiceprint features; determining user location based on the time difference of voice data arriving at collection points at different locations within the cockpit; and determining user permissions based on user identity and user location.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, user permissions are determined based on user identity and user location, including: determining user permissions based on the authorization level corresponding to the user identity and the regional attribute corresponding to the user location.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the service recommendation model performs the following operations to determine the target cabin service: Time-aligned and feature-fused multi-source cabin data are performed to obtain fused features; the fused features are used to characterize the user's service demand intent; based on the fused features, a search is conducted in the scene graph to determine multiple cabin services; nodes in the scene graph represent adjustable parameter items of the cabin services, and edges represent the relationships between adjustable parameter items; based on user information, the target cabin service is determined from the multiple cabin services.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, based on fusion features, a search is performed in the scene graph to determine multiple cockpit services, including: determining target parameter items based on the matching degree between the fusion features and the intent tags associated with different adjustable parameter items in the scene graph; and determining multiple feasible parameter value combinations for the target parameter items and other parameter items associated with them based on the edges of the scene graph; wherein each parameter value combination corresponds to a candidate cockpit service.

[0010] In conjunction with the first aspect mentioned above, in one possible implementation, the target cockpit service is determined from multiple cockpit services based on user information, including: filtering multiple cockpit services based on user permissions in the user information to obtain a candidate service set; performing safety arbitration on the adjustment instructions contained in each candidate service based on the current driving status of the vehicle; and determining the target cockpit service from the candidate service set based on the safety arbitration result.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the method further includes: after executing the target cabin service, obtaining user feedback data; the feedback data includes at least the user's cancellation operation, adjustment operation, and continuous usage duration; based on the feedback data, iteratively updating the strategy in the service recommendation model used to determine the user's demand intent or generate cabin service parameter combinations.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the environmental perception data includes at least one of the following: occupant body pressure distribution data acquired through sensors built into the seat; vehicle driving time, speed, and navigation data; light intensity, temperature, and humidity data acquired through cabin environmental sensors; and the user's recent historical operation records within the cabin.

[0013] Secondly, this application provides a cabin service recommendation device, comprising: a data acquisition module, a user verification module, a service recommendation module, and an execution module. The data acquisition module is used to acquire multi-source cabin data, including user voice data and environmental perception data. The user verification module is used to determine user information based on the user voice data, including user identity, user location, and user permissions. The service recommendation module is used to determine a target cabin service based on the multi-source cabin data and user information using a service recommendation model. The service recommendation model is used to determine multiple cabin services matching the user's needs and intentions based on the multi-source cabin data, and then determine the target cabin service from the multiple cabin services based on the user information. The execution module is used to execute the target cabin service.

[0014] In conjunction with the second aspect above, in one possible implementation, the user verification module is specifically used for: extracting voiceprint features from user voice data; determining user identity based on voiceprint features; determining user location based on the time difference of voice data arriving at collection points at different locations within the cockpit; and determining user permissions based on user identity and user location.

[0015] In conjunction with the second aspect above, in one possible implementation, the service recommendation module is specifically used for: performing time alignment and feature fusion on multi-source cabin data to obtain fused features; using the fused features to characterize the user's service demand intent; based on the fused features, searching in the scene graph to determine multiple cabin services; nodes in the scene graph characterize adjustable parameter items of the cabin services, and edges characterize the correlation between adjustable parameter items; and based on user information, determining the target cabin service from the multiple cabin services.

[0016] In conjunction with the second aspect above, in one possible implementation, the device further includes a model update module, specifically used for: obtaining user feedback data after executing the target cabin service; the feedback data includes at least the user's cancellation operation, adjustment operation, and continuous usage duration; and iteratively updating the strategy in the service recommendation model used to determine the user's demand intent or generate cabin service parameter combinations based on the feedback data.

[0017] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory storing processor-executable instructions; when the processor is configured to execute the instructions, the electronic device implements the method of the first aspect described above.

[0018] Fourthly, this application provides a computer-readable storage medium comprising: computer software instructions; which, when executed in an electronic device, cause the electronic device to implement the method described in the first aspect.

[0019] Fifthly, this application provides a computer program product that, when run on a computer, causes the computer to perform the steps of the relevant method described in the first aspect above, so as to implement the method of the first aspect above.

[0020] The beneficial effects of the second to fifth aspects mentioned above can be referred to the corresponding description of the first aspect, and will not be repeated here. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a cabin service recommendation method provided in an embodiment of this application; Figure 2 A flowchart illustrating a method for determining a target cabin service, provided as an embodiment of this application; Figure 3 A detailed flowchart illustrating a method for determining a target cabin service, provided for an embodiment of this application; Figure 4 A detailed flowchart illustrating a cabin service recommendation method provided in this application embodiment; Figure 5 A schematic diagram illustrating the composition of a cabin service recommendation device provided in this application embodiment; Figure 6 This is a schematic diagram of a cabin service recommendation device provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that in the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0025] In the embodiments of this application, the terms "first," "second," "third," "fourth," "fifth," and "sixth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," "fourth," "fifth," and "sixth" may explicitly or implicitly include one or more of that feature.

[0026] In embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0027] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0028] As described in the background section, current intelligent cockpit systems suffer from two fundamental limitations in personalized services. The first limitation is the static identification and recall problem: existing systems cannot accurately distinguish between different users in the same seat, and their services heavily rely on fixed scenarios manually set by the user beforehand, failing to address complex, undefined dynamic needs. The second limitation is the insufficient intelligence of the system: existing solutions are essentially mechanical replays based on "IF-THEN" rules; the system itself lacks the ability to continuously perceive and comprehensively understand the user's real-time state, behavioral intentions, and the vehicle environment, and is even less capable of autonomously creating or optimizing service scenarios based on this understanding.

[0029] Based on this, this application provides a cabin service recommendation method. By acquiring multi-source cabin data, including user voice data and environmental perception data, it provides a comprehensive information foundation reflecting user expression and real-time cabin status for subsequent analysis, thereby ensuring a comprehensive understanding of user needs and cabin scenarios. Furthermore, based on user voice data, user information including user identity, location, and permissions is extracted, enabling the identification of the current user's personalized characteristics and operational permissions, providing a basis for accurate service adaptation. On this basis, the service recommendation model uses multi-source cabin data to infer user intent and generate multiple possible cabin services. Then, combining user information, it selects the most suitable target cabin service from these services. This process achieves coherent reasoning from raw data to intent recognition and personalized decision-making, ensuring that the recommended service conforms to both the user's actual needs and the current cabin environment and user permission constraints. Finally, by executing the target cabin service, it directly achieves automated and accurate response to user needs, thereby improving user satisfaction and security of cabin services.

[0030] The cockpit service recommendation method provided in this application can be widely applied to various intelligent cockpits, intelligent driving cockpits, or enclosed space scenarios that require personalized human-computer interaction. This application does not impose any restrictions on the specific implementation form of the cockpit, cockpit architecture, or interactive terminal.

[0031] For example, embodiments of this application can be applied to family-shared vehicle scenarios that support multi-user identification and personalized configuration. When the same vehicle is frequently driven or ridden by multiple family members (such as parents and children), the method provided in this application can accurately identify the current user's identity and location based on voice data, and, combined with personal preset preferences (such as seat position and frequently used entertainment options), automatically recommend and execute personalized cabin environment settings and content services, achieving a seamless "one person, one cabin" switching experience.

[0032] For example, the embodiments of this application can be applied to multi-passenger service scenarios such as business receptions or premium ride-hailing services. When there are passengers with different identities and permissions (such as VIP customers and ordinary passengers) in the vehicle at the same time, the method provided by this application can distinguish users through voice and location information, and filter out matching target services from the recommended service set according to their permissions (such as VIP passengers having access to all entertainment functions), so as to meet personalized needs while ensuring the safety and order of service management.

[0033] For example, the embodiments of this application can be applied to intelligent cockpit scenarios that integrate proactive interaction and safety warnings. Faced with situations where drivers may be distracted or fail to perceive risks in a timely manner in complex traffic environments, the method provided in this application can integrate environmental perception data (such as detecting sudden road conditions ahead) with the driver's voice intent, proactively recommend and execute integrated services such as adjusting the air conditioning fan speed to refresh the driver, lowering the media volume, or highlighting key navigation information, thereby enhancing contextual awareness and driving safety through natural interaction.

[0034] The following detailed description of a cabin service recommendation method provided by this application, in conjunction with specific embodiments and accompanying drawings, provides an example of such a method.

[0035] The cockpit service recommendation method provided in this application can be implemented by a cockpit service recommendation system deployed in the in-vehicle computing unit or an edge / cloud server communicating with it. This system can be modularly designed, either deployed independently as a core intelligent module in the cockpit domain controller, or integrated as a software service into existing vehicle operating systems or cockpit middleware platforms. The system possesses closed-loop processing capabilities throughout the entire process, from multi-source heterogeneous data access and fusion, real-time user status perception, intent understanding and decision-making to service scheduling and execution.

[0036] Figure 1 This is a flowchart illustrating a cabin service recommendation method provided in an embodiment of this application. For example, as shown... Figure 1 As shown, it includes the following: S101. Acquire multi-source data from the cockpit.

[0037] The cockpit multi-source data includes user voice data and environmental perception data.

[0038] In this embodiment, cockpit multi-source data refers to the heterogeneous data set collected by the cockpit service recommendation system to achieve personalized service recommendations, covering various aspects such as user status, cockpit environment, and vehicle operation. This data serves as the foundational data support for subsequent user identification, intent understanding, and service generation. User voice data refers to various voice signals emitted by the user within the cockpit, including audio data from active commands, vague complaints, and everyday conversations from which voiceprint features and semantic information can be extracted. Environmental perception data refers to relevant data collected through internal and external cockpit sensors, the vehicle control system, and the user operation recording module, reflecting the cockpit environment status, vehicle operation, and user behavior habits.

[0039] In some embodiments, user voice data may include: explicit control commands issued by the user for cabin functions (such as "turn on seat heating" or "lower the air conditioning temperature"), vague expressions of needs or complaints made by the user while driving (such as "a little tired" or "too noisy"), everyday conversations between the user and other occupants in the cabin, and unconscious voice fragments that have voiceprint recognition value. This voice data includes both acoustic information that can be used for identity verification and voiceprint feature extraction, and semantic or emotion-related information that can indirectly reflect the user's needs and intentions.

[0040] In some embodiments, environmental perception data may include: occupant body pressure distribution data, muscle tension patterns, and the movement trajectory of the center of pressure in the seat, obtained through a high-density pressure distribution sensor built into the seat; continuous driving time, current average vehicle speed, navigation route planning, and real-time traffic information obtained through the vehicle control system; ambient light intensity, cabin temperature, and humidity data obtained through light sensors, temperature sensors, and humidity sensors arranged in the cabin; records of recent active adjustment operations by the user in the cabin (such as repeatedly adjusting the air conditioning vents, quickly switching radio stations, changing seat positions, etc.) recorded by the vehicle system; and external weather data obtained through a surround-view camera.

[0041] In some embodiments, the cabin service recommendation system can continuously collect user voice data after the vehicle is started via a multi-microphone array deployed in the cabin. The collection process is not limited by user-initiated triggering and can capture all valid voice signals in the cabin in real time. When the vehicle is powered on, the cabin service recommendation system can synchronously acquire environmental perception data through seat-built-in pressure sensors, cabin environment sensors, the vehicle controller area network (CAN) bus interface, the navigation system, and the vehicle's storage module. Among these, data such as body pressure distribution and light intensity are collected in real time, while data such as driving duration and historical operation records are dynamically updated to ensure data timeliness.

[0042] In one possible implementation, the cabin service recommendation system can employ a combined trigger-based and continuous data acquisition strategy. After the vehicle is unlocked and powered on, the system automatically activates a multi-microphone array to continuously collect cabin voice data at a sampling frequency of 16kHz. Simultaneously, it triggers the seat pressure sensor, environmental sensor, and vehicle status monitoring module to collect environmental perception data in parallel at preset frequencies. Specifically, body pressure distribution data is sampled at 10Hz, temperature, humidity, and light intensity data at 1Hz, and vehicle status data such as driving duration and speed are synchronized in real-time with the vehicle's CAN bus signal. All collected data carries a unified format timestamp and data source identifier, is transmitted via Ethernet, and stored in a temporary data buffer for subsequent processing.

[0043] As can be seen from step S101, by acquiring multi-source data from the cockpit, reliable basic data can be provided for accurate identification of user identity and accurate determination of user location in subsequent steps. At the same time, by integrating potential demand information in user voice and contextual information in environmental perception data, rich data source support is provided for a deep understanding of user needs and intentions, avoiding problems such as user identification bias and misjudgment of intent caused by single or missing data, ensuring the relevance and accuracy of subsequent service recommendations, and laying a solid data foundation for the generation of personalized cockpit services.

[0044] S102. Determine user information based on user voice data.

[0045] User information includes user identity, user location, and user permissions.

[0046] In this embodiment, user information refers to the core information set generated by the cabin service recommendation system based on user voice data recognition, used to support personalized service recommendations and safety management. Its core components include user identity, user location, and user permissions. User identity refers to individual characteristic information that uniquely identifies a passenger in the cabin, serving as the core basis for distinguishing different users. User location refers to the specific location area of ​​the user within the cabin, including but not limited to the space corresponding to the driver, front passenger, and rear seats. User permissions refer to the scope and level restrictions on cabin functions that the user can operate, determined based on user identity and location, and are a key basis for balancing service convenience and operational safety.

[0047] In some embodiments, user identity may include: a unique identifier ID registered by the user in the system, the user name bound to the voiceprint feature, the user's preset personalized tags (such as "driver user", "family user 1"), and the user's corresponding authorization level (such as administrator level, ordinary user level, visitor level). This information is associated with the user's voiceprint feature template to ensure the uniqueness and accuracy of identity recognition.

[0048] In some embodiments, user location may include: seating area classification within the cabin (driver's seat, front passenger seat, rear left seat, rear right seat, rear middle seat), and specific user location data calculated based on spatial coordinates (accurate to the centimeter level), which not only clarifies the functional area where the user is located, but also provides a location benchmark for targeted service execution.

[0049] In some embodiments, user permissions may include: sensitive operation permissions (such as accessing personal privacy data, modifying core vehicle settings, and adjusting other users' personalized configurations), general operation permissions (such as adjusting the local air conditioning temperature, switching personal music playlists, and controlling local window raising and lowering), and restricted operation permissions (such as only being able to receive service push notifications and not being able to modify any system settings). Different permissions correspond to different operation scopes and verification requirements.

[0050] In some embodiments, the cabin service recommendation system can collect user voice data through a multi-microphone array deployed in the cabin. The voice data is first pre-processed for noise reduction, and then voiceprint features are extracted from the processed voice data and compared with a locally encrypted voiceprint template library to obtain the user's identity. During voice data collection, the user's location is obtained by calculating the time difference between the voice signal reaching different microphone collection points and combining this with preset cabin space layout parameters. Based on the authorization level corresponding to the determined user identity and the regional functional attributes corresponding to the user's location (e.g., driver's seat associated with core driving operation permissions, rear seats associated with comfort adjustment permissions), user permissions are obtained through the system's built-in permission mapping rules. The entire process is completed within one second after the user emits a voice signal, achieving seamless recognition.

[0051] In some embodiments, the cabin service recommendation system may also: during the user identification process, set up a dual verification mechanism, in addition to voiceprint feature comparison, for identity verification corresponding to sensitive operations, the user may be additionally required to input a short voice command (such as a preset verification phrase) for secondary confirmation. When the user's location changes (such as moving from the passenger seat to the driver's seat), the user's permissions are dynamically adjusted, and the system can automatically recalculate and update the corresponding operation permissions.

[0052] In some embodiments, the cabin service recommendation system can also store the identified user identity, location, and permission information in a temporary cache using encryption. This cache is only accessed temporarily during service execution and is automatically cleared after the service is completed to prevent information leakage. Simultaneously, it records the user identification history, including identification time, identification result, and permission access details, providing data support for system optimization.

[0053] In one possible implementation, the cabin service recommendation system can use a digital signal processor to process the collected user voice data. First, an ambient noise is filtered using a built-in adaptive filter. Then, 39-dimensional voiceprint feature parameters (including Mel-frequency cepstral coefficients, linear predictive cepstral coefficients, etc.) are extracted from the voice data to form a voiceprint feature vector. Next, a Gaussian mixture model-general background model algorithm is used to compare this voiceprint feature vector with a locally encrypted voiceprint template library. When the similarity exceeds a preset threshold (e.g., 95%), the user's identity and corresponding authorization level are determined. Subsequently, a time delay estimation algorithm based on the generalized cross-correlation function is used to calculate the time difference of the voice signal arriving at each microphone. Combined with the coordinates of the cabin microphone array, the user's location is calculated (positioning accuracy error less than 3 cm), thus determining the user's corresponding seating area. Finally, through the system's built-in permission mapping table, combined with the user's authorization level and seating area attributes (e.g., core settings permissions for the driver's seat, basic comfort adjustment permissions for the guest seat), the scope of operations the user can perform is determined, forming complete user permission information. If authentication fails three times in a row, the system will automatically lock the voiceprint recognition function for 5 minutes to prevent malicious attempts.

[0054] As shown in step S102, by synchronously determining the user's identity, location, and permissions through user voice data, different users in the same seat can be accurately distinguished. This effectively solves the identity confusion problem caused by relying on physical identifiers in existing technologies and avoids the security risks of unauthorized users operating sensitive functions. By clearly defining the user's location, a location basis is provided for subsequent targeted service execution, ensuring that services only apply to the area where the target user is located. Through hierarchical permission control, both the user's personal privacy and the security of the vehicle's core settings are protected, while providing authorized users with a convenient operating experience. At the same time, the seamless recognition feature of this step, which requires no manual user intervention, further enhances the intelligence and humanization of the cabin service, laying the foundation for the accurate recommendation and safe execution of subsequent target cabin services.

[0055] S103. Based on multi-source cockpit data and user information, determine the target cockpit service through a service recommendation model.

[0056] Among them, the service recommendation model is used to determine multiple cabin services that match the user's needs and intentions based on multi-source cabin data, and then determine the target cabin service from multiple cabin services based on user information.

[0057] In this embodiment, the service recommendation model refers to the core algorithm model set used in the cabin service recommendation system to realize user demand intent recognition, multi-dimensional cabin service generation, and target service selection. It is the core connecting data input and service output. User demand intent refers to the service demand tendency of users that is not explicitly expressed or has been initially reflected, derived from user status, environmental conditions, and behavioral characteristics. It is the core guide for service generation. Multiple cabin services refer to multiple candidate service schemes with different parameter combinations generated by the service recommendation model based on user demand intent and combined with adjustable cabin resources. Target cabin service refers to the personalized service scheme that meets user needs, is safe and compliant, and is adapted to user permissions, finally determined after filtering and optimization from multiple cabin services in combination with user information (identity, location, permissions).

[0058] In some embodiments, the service recommendation model may include: a multimodal data fusion unit for time alignment and feature fusion of multi-source cockpit data, an intent recognition unit for identifying user needs and intents based on fused features, a scene generation unit for generating multiple candidate services based on scene graphs, a permission filtering unit for filtering candidate services according to user permissions, and a conflict resolution unit for security arbitration of candidate services. The units work together to complete the entire process from data processing to target service output.

[0059] In some embodiments, user intent may include: intent related to driving status (such as "relieving fatigue" after long-term driving, "improving focus" under complex road conditions), intent related to environmental adaptation (such as "rapid cooling" in high-temperature environments, "adjusting ambient lighting and sunshades" in bright light environments), intent related to leisure (such as "relaxing and relaxing" while waiting in a car, "adapting to children's mode" when traveling with children), and intent related to personalized preferences (such as "matching commonly used air conditioning temperature and music style").

[0060] In some embodiments, multiple cabin services may include: seat adjustment related services (combinations of different massage intensities, backrest angles, and lumbar support), environment adjustment related services (combinations of temperature, airflow, airflow mode, and steering wheel heating / seat ventilation), infotainment related services (combinations of different music playlists, navigation modes, and interface layouts), and multi-dimensional collaborative services (linked combinations of seat, environment, and infotainment functions, such as the combination of seat massage, constant air conditioning temperature, and soft music playback corresponding to "relieve fatigue").

[0061] In some embodiments, the target cabin service may include: personalized parameter combinations adapted to user identity and permissions (such as the driver's usual seat position + exclusive playlist), regional services adapted to the user's location (such as air conditioning adjustment only for the passenger area corresponding to the needs of the passenger), safety and compliance services adapted to the current driving state (such as soothing services that do not involve significant seat adjustments when driving at high speeds), and dynamically optimized precision service solutions (such as the combination of air conditioning temperature and massage mode finely adjusted based on the user's recent preferences).

[0062] In some embodiments, the cabin service recommendation system can acquire multi-source cabin data (including user voice data, body pressure distribution, driving status, environmental parameters, historical operation records, etc.) in real time through the multi-microphone array, seat sensors, environmental sensors, vehicle CAN bus, and vehicle storage module deployed in step S101. Through voiceprint recognition, sound source localization, and permission mapping logic in step S102, user information such as user identity, location, and permissions is obtained. When the vehicle is in motion or idling, and the system detects user voice commands, changes in user status (such as frequent posture adjustments), or sudden changes in environmental parameters (such as a sudden temperature rise), the service recommendation model is automatically invoked. Based on the acquired multi-source cabin data and user information, the process of determining the target cabin service is initiated.

[0063] In one possible implementation, the specific implementation of step S103 is described below. Figure 2 The corresponding implementation examples are described in detail here.

[0064] As shown in step S103, the target cabin service is determined by the service recommendation model based on multi-source cabin data and user information. This fully utilizes the complementary value of multi-dimensional data, combines the user's unique identity, specific location, and hierarchical permissions, and accurately captures the user's explicit and implicit needs and intentions. This avoids misjudgment of needs caused by a single data source, and realizes a leap from "passively responding to instructions" to "actively understanding needs and generating services". This improves the intelligence level of cabin services and user experience, and lays a key foundation for the accurate execution of subsequent services.

[0065] S104, Perform target cabin service.

[0066] In some embodiments, the cabin service recommendation system can establish a data transmission channel with the output port of the service recommendation model. After the target cabin service is determined in step S103, the corresponding service execution instruction set is acquired in real time. This instruction set contains the specific adjustment parameters, execution order, and priority information of each execution module. The acquisition process uses the Ethernet transmission protocol to ensure the stability and real-time performance of data transmission. At the same time, the instruction set is checked for integrity to ensure that no data is lost or tampered with, providing a reliable basis for the accurate execution of the service.

[0067] In some embodiments, the cabin service recommendation system can also: monitor the operating status of each actuator in real time during service execution, determine whether functions such as seat adjustment, environmental control, and infotainment are executed normally according to instructions through sensor feedback data, and automatically trigger a fault warning mechanism if an execution failure occurs (such as seat adjustment malfunction or air conditioning not operating at the preset temperature), issuing a gentle prompt to the user and attempting to restart the execution operation. It supports real-time user intervention during service execution; if the user manually adjusts service parameters or cancels the current service, the system will immediately respond and terminate the original execution instruction, while recording the user's intervention behavior as feedback data. After service execution is completed, the system will notify the user through voice prompts in the cabin or a slight pop-up on the instrument panel to avoid disturbing the user. Furthermore, the cabin service recommendation system can also encrypt and store key service execution information (such as execution time, adjustment parameters, and execution results) in association with user information, providing data support for subsequent service optimization.

[0068] In one possible implementation, the cabin service recommendation system can break down the acquired set of instructions for the target cabin service into functional modules and send them separately to the seat control unit, environmental adjustment unit, and infotainment control unit. Upon receiving the instructions, the seat control unit uses a 12-way electric adjustment motor to drive the seat to adjust its position, backrest angle, lumbar support, and massage mode, simultaneously recording the user's personalized parameters and updating them to the user's dedicated storage area. The environmental adjustment unit adjusts the air conditioning temperature, fan speed, and airflow mode according to the instructions, triggering the steering wheel heating or seat ventilation functions, and adjusting the ambient lighting color and brightness to match the scene. The infotainment control unit switches to the user's associated music playlists and navigation favorites, adapting the interface layout to the user's preferred settings, ensuring zero-delay audio and video synchronization. All execution modules work collaboratively according to the priority order in the instruction set, with the entire execution process strictly controlled within 3 seconds, achieving a seamless personalized service experience for the user. Simultaneously, if a driving safety warning is received during execution (such as sudden acceleration or braking), unnecessary adjustments are suspended to prioritize driving safety.

[0069] As shown in step S104, by accurately executing the target cabin service determined by the service recommendation model, the user's needs and intentions are transformed into cabin function adjustments, allowing the user to obtain a personalized experience tailored to their own needs and scenarios. The multi-module collaboration and priority control during the execution process ensure the efficiency and smoothness of service delivery, and the feature of completing the adjustment within 3 seconds improves the user experience.

[0070] In this embodiment of the application, in step S103 above, the cabin service recommendation system can perform reasoning through a service recommendation model to determine the target cabin service. For example, as shown... Figure 2 As shown, the above step S103 can be specifically implemented as S201-S203: S201. Perform time alignment and feature fusion on the multi-source data of the cockpit to obtain fused features.

[0071] Among them, the fusion feature is used to characterize the user's service request intent.

[0072] In this embodiment, time alignment refers to the process of calibrating and synchronizing multi-source data from different sources and with different acquisition frequencies within the cockpit according to a unified time reference, ensuring consistency of each data point in the time dimension. Feature fusion refers to the operation of integrating the feature information of preprocessed multimodal data (such as voice features, sensor data, vehicle status data, etc.) to form a comprehensive feature set that can fully reflect the user's state and environmental conditions. Fusion features refer to feature vectors with multi-dimensional complementary information obtained after time alignment and feature fusion processing. These vectors can accurately capture the correlation between user state, environmental changes, and behavioral tendencies, thereby effectively representing the user's service demand intentions.

[0073] In some embodiments, time alignment may include: determining a uniform timestamp precision (e.g., millisecond level), performing interpolation or downsampling on data collected at different frequencies, accurately matching the time segments of user voice data with the time segments of body pressure distribution data, vehicle driving status data, and environmental perception data from the same period, correcting time deviations caused by sensor response delays, and ensuring that all data form a corresponding relationship at the same time node.

[0074] In some embodiments, feature fusion may include: concatenating features after normalizing numerical data (such as temperature, vehicle speed, and body pressure), weighted fusion of voiceprint features and semantic features extracted from speech data, cross-modal fusion of visual features extracted from image data (such as ambient lighting images) and physical features from sensor data, and a fusion method that dynamically allocates feature weights based on the confidence level of each modality of data.

[0075] In some embodiments, the fused features may include: features related to the user's physiological state (such as muscle tension derived from body pressure distribution and fatigue based on voiceprint feature analysis), features related to the environmental state (such as temperature deviation, light intensity level, and road condition complexity), features related to the user's behavior (such as historical operation preference features and voice emotion features), and features related to the vehicle's operating state (such as driving time and vehicle speed stability). The various feature dimensions are interconnected to form a unified feature vector.

[0076] In some embodiments, the cabin service recommendation system can acquire multi-source cabin data through the multi-microphone array, seat pressure sensor, environmental sensor, vehicle CAN bus interface, and vehicle storage module deployed in step S101. This data already carries the original acquisition timestamp. Simultaneously, it retrieves device response parameters and latency records from the system data cache during data acquisition to provide a basis for time alignment correction. All data is transmitted to the data processing unit in a standardized format to ensure the integrity and availability of the acquired data.

[0077] In some embodiments, the cabin service recommendation system can also: perform secondary cleaning of multi-source cabin data before time alignment, automatically identifying and removing abnormal data caused by sensor malfunctions, noise data generated by environmental interference, and redundant data collected repeatedly, thereby improving data quality. It can also evaluate the effectiveness of each modal feature during the fusion process; if the confidence level of a certain modal data is lower than a preset threshold (e.g., 60%), its fusion weight is automatically reduced or it is temporarily excluded from the current fusion. Simultaneously, the cabin service recommendation system can also record processing logs for time alignment and feature fusion, including processing time, data volume, fusion strategy, feature dimensions, etc., providing a reference for system optimization.

[0078] In one possible implementation, the cabin service recommendation system can use the Transformer model as the core processing model. First, a millisecond-level unified timestamp benchmark is established, and low-frequency ambient temperature and humidity data are interpolated to complete the data. High-frequency body pressure distribution data is downsampled, and the start and end times of user voice data are marked. The timestamps of each data point are corrected by combining sensor response delay records, achieving precise time alignment of all multi-source data. Subsequently, features are extracted from each modality. Numerical data is normalized to feature values ​​in the [0,1] interval using min-max normalization. 39-dimensional voiceprint features and 20-dimensional semantic features are extracted from the voice data, and 10-dimensional state features are extracted from the vehicle driving data. Using the Transformer model's self-attention mechanism, the correlation weights between modal features are calculated, and weighted fusion of different modal features is performed. The weights for physiological state-related features are set to 0.4, environmental state-related features to 0.2, user behavior-related features to 0.3, and vehicle operating state-related features to 0.1, ultimately generating a 128-dimensional fused feature vector.

[0079] As shown in step S201, by performing time alignment and feature fusion on multi-source cockpit data to obtain fused features, the inconsistency and information fragmentation of multi-source heterogeneous data in the time dimension are effectively solved, allowing effective information scattered in different modal data to be integrated and complemented. The generated fused features can comprehensively and three-dimensionally reflect the correlation between user status, environmental conditions and vehicle operating status, avoiding misjudgment of demand intent caused by the limitations of single modal data.

[0080] S202. Based on the fusion features, search in the scene map to determine multiple cockpit services.

[0081] In the scene graph, nodes represent adjustable parameters of the cockpit service, and edges represent the relationships between adjustable parameters.

[0082] In this embodiment, the scenario graph refers to a structured knowledge graph used to depict the relationships between adjustable resources in the cabin service, serving as the core data support for dynamically combining service parameters. Adjustable parameter items refer to all customizable functional parameters within the cabin, forming the basic units of the service plan. Relationships refer to the logical relationships such as constraints, collaboration, and mutual exclusion between adjustable parameter items, used to regulate the rationality of parameter combinations. Multiple cabin services refer to a parameter set with complete service logic, formed by retrieving and combining multiple adjustable parameter items from the scenario graph based on user needs and intentions; this is the prototype of the target cabin service.

[0083] In some embodiments, the scene graph may include: a topology structure constructed with adjustable parameter items as core nodes and association relationships as edges, a built-in parameter item attribute library (including parameter value range, function priority, and adapted scene tags) and an association rule library (including collaborative activation conditions, mutual exclusion logic, and priority conflict resolution strategies), and a reserved dynamic update interface to support the addition of new parameter items and the expansion of association relationships.

[0084] In some embodiments, adjustable parameters may include: seat system position adjustment parameters, backrest angle parameters, lumbar support strength parameters, massage mode and intensity parameters, ambient system air conditioning temperature, air volume, air outlet mode parameters, ambient light color and brightness parameters, fragrance type and concentration parameters, infotainment system music style, volume, navigation broadcast frequency parameters, interface display layout parameters, and comfort auxiliary parameters such as steering wheel heating level and seat ventilation intensity.

[0085] In some embodiments, the association may include: collaborative association (such as enhancing the relaxation effect when "seat massage" and "air conditioning constant temperature" are turned on at the same time), constraint association (such as "seat heating" cannot be enabled at the same time under "air conditioning cooling mode"), dependency association (such as "seat ventilation" needs to be effective based on the "seat power on" state), and priority association (such as "driving safety related parameters" having higher priority than "leisure and entertainment related parameters"). Each association is accompanied by corresponding effective conditions.

[0086] In some embodiments, multiple cabin services may include: a service corresponding to "gentle massage + 115° backrest + medium lumbar support + 25° constant temperature" generated with "seat massage mode" as the target parameter; or a service corresponding to "24°C + 3-level fan speed + upward airflow + warm yellow ambient lighting" generated with "air conditioning temperature" as the target parameter; or a service corresponding to "light music + low volume + simplified navigation announcements" generated with "music style" as the target parameter. Each cabin service includes a complete combination of parameter values ​​and function execution logic, and is accompanied by matching information with the user's needs and intentions.

[0087] In some embodiments, the cabin service recommendation system can obtain a fused feature vector after time alignment and feature fusion through step S201. This vector has been standardized and carries information related to user needs and intentions. Simultaneously, a pre-built scene graph is retrieved from a locally encrypted knowledge graph repository.

[0088] In some embodiments, the cabin service recommendation system can also: dynamically update the scene graph, for example, automatically add or optimize associations and adjust association weight coefficients based on high-frequency and effective parameter combinations in users' historical operation records; validate the validity of multiple cabin service parameter combinations generated by the retrieval, and eliminate invalid combinations that violate mutual exclusion rules, exceed the parameter value range, or do not conform to the current cabin hardware configuration; assign priority scores to different parameter combinations according to the confidence level of the demand intent corresponding to the fusion features, with higher confidence levels corresponding to higher priority parameter combinations. Simultaneously, the cabin service recommendation system can also record node matching logs and association call information during the retrieval process, providing data support for graph optimization and retrieval algorithm upgrades.

[0089] In one possible implementation, the cabin service recommendation system can first obtain the 128-dimensional fusion feature vector generated in step S201, and then transform it into a demand intent label with confidence (e.g., "Relieve Fatigue" with a confidence of 0.92) through a fully connected neural network. Subsequently, using this intent label as the search keyword, the system traverses the parameter attribute library of the scene graph, filtering out adjustable parameter items with the "Relieve Fatigue" adaptation label (e.g., seat massage parameters, air conditioning constant temperature parameters, gentle music parameters, etc.). Then, based on the association rule library of the scene graph, it retrieves the collaborative relationships between the filtered parameter items, excluding parameter combinations with mutually exclusive relationships, and combines parameters such as seat massage mode (gentle), backrest angle (115°), air conditioning temperature (25℃), airflow (level 2), and music style (light music) according to the principle of "core parameters first, collaborative parameters complete". Simultaneously, combining the value range and effective conditions of the parameter items, multiple logically consistent cabin service parameter combinations are generated. Each combination is accompanied by a compliance score for the association relationship and a confidence score for the adaptation intent, forming a candidate cabin service set.

[0090] As shown in step S202, all adjustable cockpit resources are integrated in the form of a structured graph, solving the problems of rigid parameter combinations and limited coverage in traditional preset scenarios. The constraint of the correlation ensures the rationality and safety of parameter combinations, avoiding invalid or conflicting parameter configurations. The precise matching of fusion features and scenario graphs allows the generated parameter combinations to closely match the user's potential needs and intentions, covering more long-tail scenarios that have not been preset. The generation of multiple candidate cockpit service sets also provides ample selection space for subsequent selection of target services based on user information, ensuring the personalization and adaptability of the final service solution, and laying the foundation for achieving "scenario creation" rather than "scenario playback".

[0091] S203. Based on user information, determine the target cabin service from multiple cabin services.

[0092] In this embodiment of the application, the target cockpit service refers to the complete service solution that best matches the user's personalized needs, operation permissions and scenario after filtering and optimization from multiple candidate cockpit services obtained based on fusion features and scene graph retrieval, combined with user identity, location and permissions. Its unique combination of parameter values ​​is the basis for the final cockpit service execution.

[0093] In some embodiments, the cabin service recommendation system can obtain multiple candidate cabin services generated based on the target parameter items matched by the fusion features through step S202. These services are accompanied by compliance scores and intent-fit reliability. At the same time, the system retrieves the user information determined in step S102 from the system's temporary encrypted cache, including the user's unique identifier, precise location coordinates, hierarchical permission range, and historical operation preference records.

[0094] In some embodiments, the cabin service recommendation system can also: pre-rank multiple cabin services using the user's historical service preferences as matching weights, assigning higher weights to service types that the user has frequently selected in the past; automatically trigger a brief voice secondary verification (such as "Confirm adjustment of seat memory settings?") for cabin services involving sensitive operation permissions; label the reasons for services excluded during the filtering process (such as "insufficient permissions" or "incompatible orientation") and associate them with corresponding parameter items and intent tags, providing a reference for optimizing the intent tag weights in subsequent scene graphs. Simultaneously, the cabin service recommendation system also supports users manually marking frequently used services, which the system synchronizes to the user's exclusive configuration library, improving the efficiency of subsequent filtering processes.

[0095] In one possible implementation, the cabin service recommendation system can first retrieve the hierarchical permissions from the user's information to perform an initial screening of multiple candidate cabin services, eliminating services that the user has no operational permissions for (such as "modify vehicle core settings" services for visitor users). Then, based on the user's cabin location (e.g., the right rear seat), services suitable for that area are selected (e.g., rear seat adjustment, regional air conditioning control services, excluding the driver-only steering wheel heating service). Next, the system retrieves the user's historical service preference records and calculates the matching degree between each remaining service and the user's past selections (e.g., if the user frequently selects "gentle massage" services, the matching degree weight of the corresponding service increases by 0.3). Finally, considering the current vehicle driving status (e.g., high-speed driving), a safety arbitration is performed, excluding services that may affect driving safety (e.g., services that significantly adjust the seat backwards), and selecting the service with the highest comprehensive score in terms of intent matching, permission compliance, location adaptability, and preference matching from the remaining services, thus determining it as the target cabin service.

[0096] As shown in step S203, determining the target cockpit service from multiple cockpit services using user information filters service options that do not conform to the user's permissions and location, avoiding unauthorized operations and invalid services across regions, thus ensuring the compliance and relevance of service execution. Combining the filtering logic with the user's historical preferences makes the target cockpit service more aligned with the user's personalized usage habits, improving user satisfaction. The multi-dimensional filtering and safety arbitration mechanism ensures that the target service meets the user's core needs while also considering safety in driving scenarios.

[0097] In this embodiment of the application, the detailed process of steps S201-S203 is as follows: Figure 3As shown. For example, the cabin service recommendation system first acquires four types of input data: first, voiceprint / location features extracted from user voice data; second, cabin pressure data collected by sensors built into the seat; third, the user's historical operations and vague intention information; and fourth, vehicle and environmental status feedback from the vehicle's CAN bus and environmental sensors. Then, multi-source data alignment is performed. The cabin service recommendation system calibrates and synchronizes these four types of multi-source data with different acquisition frequencies and formats using millisecond-level timestamps. After time alignment, the features of each data point are weighted and fused to form multimodal context-aware information (i.e., the fused features obtained in S201) that comprehensively reflects the user's state, behavior, and environmental conditions. Next, dynamic intent understanding is performed based on this multimodal context-aware information. By capturing the correlations between data, a user intent label with confidence level is derived. Then, a scene graph is retrieved. The cabin service recommendation system uses the intent label as a keyword, traverses the pre-built scene graph, selects adjustable parameter items that match the intent, and combines them according to the association rules within the graph to form a candidate parameter combination service set. Finally, a safety arbitration is conducted. The cabin service recommendation system retrieves the confirmed user information (identity, location, permissions) and, in conjunction with the current vehicle driving status, filters and prioritizes the candidate service set, eliminating combinations with incompatible permissions, unsuitable locations, or security risks, and ultimately determining the target cabin service with the highest overall suitability.

[0098] In this embodiment, after executing the target cabin service, the cabin service recommendation system can also obtain user feedback data. Then, based on the feedback data, the strategies used in the service recommendation model to determine user intent or generate cabin service parameter combinations are iteratively updated.

[0099] The feedback data includes at least the user's undo actions, adjustment actions, and duration of continuous use.

[0100] In this embodiment, feedback data refers to the user's behavioral response and status change information collected by the cabin service recommendation system after the target cabin service is executed, which is the direct basis for evaluating service suitability. Strategy iteration update refers to the process by which the cabin service recommendation system optimizes and adjusts core strategies such as intent recognition logic and parameter combination generation rules in the service recommendation model based on feedback data, in order to improve the accuracy of subsequent recommendations.

[0101] In some embodiments, feedback data may include: user-initiated cancellation operations during service execution (such as turning off the activated seat massage function), manual adjustment operations of service parameters (such as adjusting the air conditioning temperature from 25°C to 23°C), continuous usage duration of the service from start to stop (such as music playback lasting 30 minutes), and may also include user voice evaluations (such as "This fragrance is too strong"), operation frequency in a short period of time (such as switching music styles 3 times in a row), and changes in the user's physiological state collected by seat pressure sensors and heart rate sensors (such as whether the tension of body pressure distribution has decreased). This information reflects the user's satisfaction with the current service and actual adaptation needs from different dimensions.

[0102] In some embodiments, the cabin service recommendation system can obtain the specific content and time points of the user's cancellation and adjustment operations through the vehicle's system operation recording module during service execution and within 10 minutes after the service ends. The system automatically obtains the duration of service usage through the vehicle's timing unit. It also collects the user's voice feedback in real time through a multi-microphone array within the cabin. Furthermore, it acquires physiological data such as changes in body pressure distribution and heart rate fluctuations through pressure sensors built into the seat and wrist heart rate sensors (if the user is wearing an in-vehicle smart device). All data acquisition is performed silently in the background, without requiring active user intervention.

[0103] In some embodiments, the cabin service recommendation system can also: classify the effectiveness of feedback data, for example, marking user-initiated manual adjustment and cancellation operations as high-weight data, user voice evaluations as medium-weight data, and continuous usage duration and physiological state changes as low-weight data. The system also associates feedback data with scene tags, binding and storing them with corresponding user intent tags and service parameter combinations for accurate identification of strategy modules to be optimized later. When negative feedback data for a certain type of service continuously exceeds a preset threshold, the cabin service recommendation system can automatically trigger a re-verification of the association rules for the corresponding parameter items in the scene graph. Simultaneously, the cabin service recommendation system also supports users manually submitting detailed feedback (such as "I don't like this massage intensity") via voice or the vehicle's infotainment system interface, and sets the weight of manual feedback to the highest level.

[0104] In one possible implementation, the cabin service recommendation system can initiate a feedback data collection process after the target cabin service (such as "relieving fatigue" corresponding to gentle seat massage + 24°C air conditioning) is completed. First, it captures the cancellation operation "the seat massage was turned off by the user 5 minutes after it started" and the adjustment operation "the air conditioning temperature was adjusted to 22°C" through the vehicle's operation log. The timing unit obtains the service's continuous usage duration as 12 minutes. The seat pressure sensor detects that the user's body pressure distribution tension did not show a significant reduction. These data are then labeled: "cancelling the seat massage" is marked as negative feedback, "adjusting the air conditioning temperature" is marked as suitability adjustment feedback, and "short duration + no reduction in body pressure tension" is marked as low suitability feedback. Next, based on the labeled feedback data, the feature weights corresponding to the intent of "relieving fatigue" in the service recommendation model are adjusted (such as increasing the weight of the feature of "body pressure relaxation"). At the same time, the parameter combination rules under this intent in the scene graph are optimized (the default air conditioning temperature is adjusted to 22℃ and associated with the collaborative parameter of "reducing the intensity of seat massage by 1 level"), thus completing the iterative update of the service recommendation model. The whole process is executed silently in the background without interfering with the user's current operation.

[0105] In this embodiment, by acquiring feedback data and iteratively updating the service recommendation model based on it, the limitation of traditional cabin service systems being unable to evolve autonomously is overcome. Feedback data assesses service suitability from multiple dimensions, including user behavior and physiological state, allowing the model optimization to better align with users' actual needs. By dynamically adjusting the combination rules of intent recognition weights and scenario parameters, the accuracy of the service recommendation model continuously improves with increased user usage, gradually shifting cabin service from general adaptation to personalized customization, ultimately achieving a more intelligent and user-friendly personalized service experience.

[0106] In summary, Figure 4 A detailed flowchart illustrating a cabin service recommendation method provided in this application embodiment. For example, as shown... Figure 4As shown, the cabin service recommendation system first acquires the user's voice, extracts its features, and extracts feature parameters containing voiceprint information and semantic tendencies. Then, it completes user authentication based on these feature parameters. Simultaneously, the system collects multimodal environmental perception data (such as seat pressure distribution, cabin temperature and humidity, and real-time vehicle driving status) and performs dynamic intent understanding based on the authentication results and multimodal environmental perception data. Based on the demand tendencies derived from dynamic intent understanding, the system generates personalized scene commands adapted to the user's identity and scenario status. Subsequently, based on identity authentication, the system performs security arbitration and command orchestration on these commands, verifies the command's operational permissions and driving safety, and plans the execution order of various function adjustments. Then, the system triggers cabin collaborative adjustments, simultaneously adapting and adjusting functions such as seat angle, air conditioning parameters, and infotainment configuration. After the service execution process is complete, the system can initiate a user feedback collection process to obtain feedback data such as user manual adjustments, function cancellations, and continuous usage time. Finally, based on this feedback data, the strategy is optimized, and the feature weights for dynamic intent understanding and the generation rules for personalized scenario instructions are continuously updated to form a closed-loop iteration of the service process.

[0107] The cabin service recommendation method provided in this application acquires multi-source cabin data, including user voice data and environmental perception data, providing a comprehensive information foundation reflecting user expression and real-time cabin status for subsequent analysis. This ensures a comprehensive understanding of user needs and the cabin scenario. Furthermore, based on user voice data, user information including user identity, location, and permissions is extracted, enabling the identification of the current user's personalized characteristics and operational permissions, providing a basis for accurate service adaptation. On this basis, the service recommendation model uses multi-source cabin data to infer user intent and generate multiple possible cabin services. Then, combining user information, it selects the most suitable target cabin service from these services. This process achieves coherent reasoning from raw data to intent recognition and personalized decision-making, ensuring that the recommended service conforms to both the user's actual needs and the current cabin environment and user permission constraints. Finally, by executing the target cabin service, automated and accurate responses to user needs are directly achieved, thereby improving user satisfaction and security of cabin services.

[0108] In an exemplary embodiment, Figure 5 This is a schematic diagram illustrating the composition of a cabin service recommendation device provided in an embodiment of this application. Figure 5As shown, the cabin service recommendation device includes: a data acquisition module 501, a user verification module 502, a service recommendation module 503, and an execution module 504. The data acquisition module 501 acquires multi-source cabin data, including user voice data and environmental perception data. The user verification module 502 determines user information based on the user voice data, including user identity, user location, and user permissions. The service recommendation module 503 determines the target cabin service based on the multi-source cabin data and user information using a service recommendation model. The service recommendation model determines multiple cabin services matching the user's needs based on the multi-source cabin data, and then selects the target cabin service from these services based on the user information. The execution module 504 executes the target cabin service.

[0109] In this embodiment of the application, the user verification module 502 is specifically used for: extracting voiceprint features from voice data; determining user identity based on voiceprint features; determining user location based on the time difference of voice data arriving at collection points at different locations in the cockpit; and determining user permissions based on user identity and user location.

[0110] In this embodiment, the service recommendation module 503 is specifically used for: performing time alignment and feature fusion on multi-source cockpit data to obtain fused features; using the fused features to characterize the user's service demand intent; searching in the scene graph based on the fused features to determine multiple cockpit services; nodes in the scene graph characterize adjustable parameter items of the cockpit services, and edges characterize the association between adjustable parameter items; and determining the target cockpit service from the multiple cockpit services based on user information.

[0111] In this embodiment of the application, the device further includes a model update module, specifically used for: obtaining user feedback data after executing the target cabin service; the feedback data includes at least the user's cancellation operation, adjustment operation, and continuous usage duration; and iteratively updating the strategy in the service recommendation model used to determine the user's demand intent or generate cabin service parameter combinations based on the feedback data.

[0112] In an exemplary embodiment, this application also provides an electronic device, which may be the front-end code generation device in the above method embodiments. Figure 6 This is a schematic diagram of a cabin service recommendation device provided in an embodiment of this application. Figure 6 As shown, the front-end code generation device may include: a processor 601 and a memory 602; the memory 602 stores instructions executable by the processor 601; when the processor 601 is configured to execute instructions, it causes an electronic device, network device, or manager to implement the system functions described in the foregoing method embodiments.

[0113] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0114] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0117] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0118] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for recommending cabin services, characterized in that, The method includes: Acquire multi-source cockpit data; the multi-source cockpit data includes user voice data and environmental perception data; Based on the user's voice data, user information is determined; the user information includes user identity, user location, and user permissions. Based on the multi-source cabin data and the user information, a target cabin service is determined through a service recommendation model. The service recommendation model is used to determine multiple cabin services that match the user's needs and intentions based on the multi-source cabin data, and then determine the target cabin service from the multiple cabin services based on the user information. Perform the target cabin service.

2. The method according to claim 1, characterized in that, The process of determining user information based on the user's voice data includes: Extract voiceprint features from the user's voice data; User identity is determined based on the aforementioned voiceprint features; The user's location is determined based on the time difference between the arrival of the voice data at different collection points inside the cockpit. User permissions are determined based on the user's identity and location.

3. The method according to claim 2, characterized in that, The process of determining user permissions based on the user's identity and location includes: The user's permissions are determined based on the authorization level corresponding to the user's identity and the regional attribute corresponding to the user's location.

4. The method according to claim 1, characterized in that, The service recommendation model performs the following operations to determine the target cabin service: The cockpit multi-source data is time-aligned and feature-fused to obtain fused features; the fused features are used to characterize the user's service request intent. Based on the fusion features, a search is performed in the scene graph to determine multiple cabin services; The nodes in the scene graph represent adjustable parameters of the cockpit service, and the edges represent the relationships between the adjustable parameters. Based on the user information, the target cabin service is determined from the multiple cabin services.

5. The method according to claim 4, characterized in that, Based on the fused features, a search is performed in the scene graph to determine multiple cabin services, including: The target parameter is determined based on the matching degree between the fusion features and the intent tags associated with different adjustable parameter items in the scene graph; Based on the edges of the scene graph, multiple feasible parameter value combinations are determined for the target parameter item and other associated parameter items; wherein each parameter value combination corresponds to a candidate cockpit service.

6. The method according to claim 4, characterized in that, The step of determining the target cabin service from the plurality of cabin services based on the user information includes: Based on the user permissions in the user information, the multiple cockpit services are filtered to obtain a candidate service set; Based on the vehicle's current driving status, the adjustment instructions contained in each candidate service are subject to safe arbitration. Based on the results of the safety arbitration, the target cabin service is determined from the pool of candidate services.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: After the target cabin service is executed, user feedback data is obtained; the feedback data includes at least the user's cancellation operation, adjustment operation, and continuous usage duration; Based on the feedback data, the strategies used in the service recommendation model to determine user demand intentions or generate cabin service parameter combinations are iteratively updated.

8. The method according to claim 1, characterized in that, The environmental sensing data includes at least one of the following: Occupant body pressure distribution data obtained through sensors built into the seat; The vehicle's own driving time, speed, and navigation data; Light intensity, temperature, and humidity data are obtained through in-cabin environmental sensors; The user's recent operation history in the cockpit.

9. A cabin service recommendation device, characterized in that, The device includes: The data acquisition module is used to acquire multi-source data from the cockpit; the multi-source data from the cockpit includes user voice data and environmental perception data. The user verification module is used to determine user information based on the user's voice data; the user information includes user identity, user location, and user permissions. The service recommendation module is used to determine the target cabin service based on the multi-source cabin data and the user information through a service recommendation model; the service recommendation model is used to determine the user's demand intent and multiple cabin services based on the multi-source cabin data, and then determine the target cabin service from the multiple cabin services based on the user information. An execution module is used to perform the target cockpit service.

10. An electronic device, characterized in that, The electronic device includes: a processor and a memory; The memory stores instructions that the processor can execute; When the processor is configured to execute the instructions, the electronic device performs the method as described in any one of claims 1-8.

Images