A cloud-based independent AI service-based personalized massage chair system and a natural language semantic mapping method thereof

Abstract

The application provides a cloud-based independent AI service personalized massage chair system and a natural language semantic mapping method thereof, and the system comprises: a cloud-based independent AI service cluster, which is used for allocating an independent AI service instance indexed by a device unique identifier for each massage chair device; a massage knowledge graph module, which is used for maintaining a symptom-position-technique-parameter four-tuple knowledge base; a user preference learning module, which is used for encoding user historical massage feedback data into a user preference vector and continuously updating by using an incremental learning algorithm; and a device binding management module, which is used for establishing a one-to-one binding relationship between a massage chair device unique identifier and an independent AI service instance. The application solves the problems that the existing personalized service of a massage chair is limited to a static configuration table, cannot dynamically generate personalized massage parameters according to natural language description, and cannot maintain a cross-session user state, and achieves the technical effects of personalized AI massage service and natural language control parameters.

Description

Technical Field

[0001] This invention relates to the field of personalized service technology for intelligent massage equipment, and in particular to a personalized massage chair system based on cloud-based independent AI services and its natural language semantic mapping method. Background Technology

[0002] With the development of artificial intelligence technology, smart massage chairs are gradually evolving towards personalization. However, the personalization mechanisms of existing massage chairs have the following significant limitations: First, the static nature of personalized services. Existing personalization mechanisms are mostly based on static configuration tables or locally embedded user profiles; user profiles are stored in the form of a set of structured parameters, and pre-stored massage programs are invoked through rule matching; this approach cannot dynamically generate personalized massage parameters based on the user's natural language description, nor can it maintain the continuity of the user's state in cross-session scenarios.

[0003] Second, there is a disconnect between natural language and physical parameters. When users express complex semantics such as "I've been working a lot of overtime lately, and my neck is very stiff. I want a deeper massage," which includes symptoms, location, and force preferences, existing systems cannot parse out the specific actuator control parameters (such as motor speed, airbag pressure, heating temperature, and robotic arm trajectory coordinates). Users must memorize fixed instruction keywords or operate through complex menu levels, resulting in a high interaction threshold.

[0004] Third, cross-session state loss. Existing solutions store user profiles on local embedded devices, which have limited storage capacity and cannot maintain the continuity of user state across devices and sessions; preferences established by a user on one massage chair cannot be transferred to another massage chair of the same model; feedback after the previous massage cannot affect the massage plan for the next massage.

[0005] In summary, existing technologies lack a personalized massage system that can combine the intelligence and long-term memory capabilities of independent AI services in the cloud with massage knowledge graph reasoning and continuous learning of user preference vectors, as well as an end-to-end method that directly maps natural language semantics to actuator physical parameters. Summary of the Invention

[0006] The purpose of this invention is to provide a personalized massage chair system based on cloud-based independent AI services and its natural language semantic mapping method. This system can combine the intelligence level and long-term memory capabilities of cloud-based independent AI services with massage knowledge graph reasoning and continuous learning of user preference vectors. It assigns an independent cloud-based AI service instance to each massage chair and each user, thereby enabling continuous learning capabilities of long-term memory, massage knowledge graph, and user preference vectors.

[0007] In a first aspect, the present invention provides a personalized massage chair system based on cloud-based independent AI services, comprising: The massage chair body has a built-in main control MCU, massage actuator and sensor array, and is connected to a cloud server through a communication network; A cloud-based independent AI service cluster, deployed on a cloud server, is used to assign an independent AI service instance indexed by a unique device identifier to each massage chair device; The massage knowledge graph module is used to maintain a knowledge base of symptom-location-technique-parameter quadruples, which include symptom nodes, location nodes, technique nodes, and parameter nodes. The user preference learning module is used to encode users' historical massage feedback data into user preference vectors through an incremental learning algorithm, and continuously update the user preference vectors. The device binding management module is used to establish a one-to-one binding relationship between the unique identifier of the massage chair device and the independent AI service instance, and the lifecycle of the independent AI service instance is bound to the lifecycle of the massage chair device.

[0008] In one possible implementation, the standalone AI service instance includes: Massage knowledge graph subgraphs are extracted from the global massage knowledge graph and are related to the common symptoms and body parts of users of this massage chair. User preference vectors are used to encode users' preferences for massage intensity, temperature, duration, and type of massage. Dialogue history cache is used to store context information from recent rounds of dialogue; The feedback record is used to store satisfaction ratings and user review texts after each massage.

[0009] In yet another possible implementation, the parameter nodes in the massage knowledge graph module include: The massage head rotation speed parameter has a value range of the preset rotation speed range. The target pressure parameter for the airbag has a value range within a preset pressure range. The target temperature parameter of the heating unit has a range of preset temperature values. The robotic arm trajectory coordinate parameters are used to describe the target position sequence of the massage mechanism in three-dimensional space.

[0010] In other possible implementations, the user preference learning module updates the user preference vector using an incremental learning algorithm, with the update formula as follows: P_new = α · P_old + (1-α) · (P_old + ΔP), Where P_old is the user preference vector before the update, ΔP is the preference increment extracted in this massage session, and α is the learning rate coefficient, with a value range of the preset learning rate range. New users correspond to a smaller α value, and old users correspond to a larger α value.

[0011] In another possible implementation, a cold start guidance module is also included, which is used to collect the user's initial preference information through multiple rounds of dialogue when a new user uses the device for the first time. The initial preference information includes commonly fatigued areas, intensity preference baseline, and single massage duration preference. The initial preference information is used to initialize the user preference vector. The device unique identifier is a globally unique identifier code burned into the device's OTP area when the massage chair is manufactured. The independent AI service instance manager creates a new independent AI service instance when it detects that the device unique identifier is connected to the network for the first time, and loads the existing instance when it detects that a corresponding independent AI service instance already exists.

[0012] Secondly, the present invention also provides a natural language semantic mapping method for a personalized massage chair system based on cloud-based independent AI services, comprising the following steps: S1. Receive natural language input from the user and extract intent elements through the natural language understanding module. The intent elements include at least one of symptom description, location specification, intensity preference, and duration requirement. S2. Query the massage knowledge graph based on the intent element, retrieve the matching symptom-location-technique-parameter quadruple, and obtain the default physical control parameters corresponding to the intent element; S3. Based on the current user preference vector, the default physical control parameters are fine-tuned to generate target physical control parameters. The target physical control parameters include at least one of the following: massage head rotation speed, airbag target pressure, heating unit target temperature, and robotic arm trajectory coordinates. S4. Encapsulate the target physical control parameters into actuator control instructions and send them to the massage chair's main control MCU; S5. During the massage process, the target physical control parameters are dynamically adjusted based on the user's real-time voice feedback.

[0013] In one possible implementation, step S2, querying the massage knowledge graph based on intent elements, includes the following steps: Match symptom nodes based on symptom descriptions to obtain recommended and contraindicated sites associated with the symptom; Based on the specified body part node, obtain the corresponding body part and applicable massage techniques; Based on the force preference, match the technique nodes to determine the movement mode; Obtain the default physical control parameters based on the parameter nodes associated with the part nodes and technique nodes.

[0014] In yet another possible implementation, step S3, the personalized fine-tuning includes the following steps: Based on the value of the intensity preference dimension in the user preference vector, the massage head rotation speed and airbag target pressure are adjusted proportionally. The target temperature of the heating unit is adjusted by offset based on the value of the temperature preference dimension in the user preference vector; The duration of a single massage session is adjusted based on the value of the duration preference dimension in the user preference vector.

[0015] In other possible implementations, step S5, dynamic adjustment includes the following steps: During the massage, the user's voice input is monitored in real time. When an incremental command from the user to adjust the intensity, change the body part, or modify the duration is detected, the incremental value is parsed. In addition, incremental values ​​are added to the currently executing target physical control parameters to generate adjusted physical control parameters, and execution continues.

[0016] In another possible implementation, In step S4, before the step of encapsulating into actuator control instructions, a safety constraint verification step is also included: Verify whether the target physical control parameters are within the preset safety range; Based on health contraindication tags in user profiles, high-risk techniques and body parts are blocked; And check whether the continuous massage duration of the same area exceeds the set limit; It also includes steps for handling exception semantics: When the natural language understanding module detects that the intent elements of the user's input are incomplete or ambiguous, it generates a clarifying question and outputs it to the user. Receive supplementary input from users in response to clarification inquiries; By combining the original input and supplementary input, the complete intent elements are extracted again.

[0017] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a program that can run on the processor, and when the program is executed by the processor, the electronic device performs the method as described in any of the embodiments of the first aspect above.

[0018] Compared with the prior art, the present invention has the following advantages: (1) Truly personalized AI massage service: Through one independent AI service instance per chair and long-term memory mechanism, user state is maintained across sessions and scenarios; the independent AI service instance does not simply call historical configurations, but dynamically generates a customized massage plan based on knowledge graph reasoning and preference vector fine-tuning.

[0019] (2) Natural language, i.e. control parameters: Users can describe their needs in everyday language (e.g., "I have been sitting for a long time and my back hurts. I want a warm, deep massage for half an hour"). The system directly parses and generates specific motor speed, airbag pressure, heating temperature, and trajectory coordinates, without the need to memorize command keywords or operate complex menus.

[0020] (3) Cross-session continuity: User preference vectors are continuously updated, dialogue history cache maintains multi-round context, and effect feedback records save satisfaction from each time, ensuring that users can obtain personalized services based on complete historical data every time they use the service.

[0021] (4) Cold start and rapid convergence: When a new user uses the service for the first time, an initial profile is collected through guided multi-turn dialogue. The first three massages are the "exploration period". Different parameter combinations are actively tried, and preferences are quickly converged based on user feedback.

[0022] In summary, this invention provides a personalized massage chair system that assigns an independent cloud-based AI service instance to each massage chair and each user, possessing continuous learning capabilities based on long-term memory, massage knowledge graph, and user preference vectors. It also provides a mapping method for natural language semantics to actuator physical parameters applied to this massage chair system, enabling users' everyday language descriptions to be directly parsed into specific physical control parameters such as motor speed, airbag pressure, heating temperature, and trajectory coordinates. This solves the technical problems of existing personalized massage chair services being limited to static configuration tables, unable to dynamically generate personalized massage parameters based on natural language descriptions, and unable to maintain user states across sessions, thus achieving a truly personalized AI massage service and the technical effect of natural language as control parameters. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a block diagram of a personalized architecture for a cloud-based independent AI service provided in an embodiment of the present invention. Figure 2 A flowchart illustrating the natural language semantics-to-physical parameter mapping process provided in this embodiment of the invention; Figure 3 This is a block diagram of the independent AI service instantiation and memory architecture provided in the embodiments of the present invention; Figure 4 A flowchart illustrating the massage knowledge graph quadruple structure and semantic mapping process provided in this embodiment of the invention; Figure 5 This is a schematic diagram of an electronic device structure provided in an embodiment of the present invention. Detailed Implementation

[0025] In the description of embodiments of the present invention, the terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be a limitation of the invention. As used in the specification and appended claims of the present invention, the singular expressions “a,” “the,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of the present invention, “at least one” and “one or more” refer to one or more (including two). The term “and / or” is used to describe the relationship between related objects, indicating that three relationships may exist; for example, A and / or B can indicate: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.

[0026] References to "one embodiment" or "some embodiments" as used in this specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the invention. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically emphasized. The term "connection" includes both direct and indirect connections, unless otherwise stated. "First" and "second" 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.

[0027] Reference Figures 1 to 4 As shown, this invention provides a personalized massage chair system based on cloud-based independent AI services, and a method for mapping user natural language semantics to physical control parameters of massage actuators. The steps of this method have clearly defined data inputs, processing procedures, and output results. Specifically: On one hand, the present invention provides a personalized massage chair system based on cloud-based independent AI services, comprising: The massage chair body has a built-in main control MCU, massage actuator and sensor array, and is connected to a cloud server through a communication network; A cloud-based independent AI service cluster, deployed on a cloud server, is used to assign an independent AI service instance indexed by a unique device identifier to each massage chair device; The massage knowledge graph module is used to maintain a knowledge base of symptom-location-technique-parameter quadruples, which include symptom nodes, location nodes, technique nodes, and parameter nodes. The user preference learning module is used to encode users' historical massage feedback data into user preference vectors through an incremental learning algorithm, and continuously update the user preference vectors. The device binding management module is used to establish a one-to-one binding relationship between the unique identifier of the massage chair device and the independent AI service instance, and the lifecycle of the independent AI service instance is bound to the lifecycle of the massage chair device.

[0028] In one possible implementation, the standalone AI service instance includes: Massage knowledge graph subgraphs are extracted from the global massage knowledge graph and are related to the common symptoms and body parts of users of this massage chair. User preference vectors are used to encode users' preferences for massage intensity, temperature, duration, and type of massage. Dialogue history cache is used to store context information from recent rounds of dialogue; The feedback record is used to store satisfaction ratings and user review texts after each massage.

[0029] In yet another possible implementation, the parameter nodes in the massage knowledge graph module include: The massage head rotation speed parameter has a value range of the preset rotation speed range. The target pressure parameter for the airbag has a value range within a preset pressure range. The target temperature parameter of the heating unit has a range of preset temperature values. The robotic arm trajectory coordinate parameters are used to describe the target position sequence of the massage mechanism in three-dimensional space.

[0030] In other possible implementations, the user preference learning module updates the user preference vector using an incremental learning algorithm, with the update formula as follows: P_new = α · P_old + (1-α) · (P_old + ΔP), Where P_old is the user preference vector before the update, ΔP is the preference increment extracted in this massage session, and α is the learning rate coefficient, with a value range of the preset learning rate range. New users correspond to a smaller α value, and old users correspond to a larger α value.

[0031] In another possible implementation, a cold start guidance module is also included, which is used to collect the user's initial preference information through multiple rounds of dialogue when a new user uses the device for the first time. The initial preference information includes commonly fatigued areas, intensity preference baseline, and single massage duration preference. The initial preference information is used to initialize the user preference vector. The device unique identifier is a globally unique identifier code burned into the device's OTP area when the massage chair is manufactured. The independent AI service instance manager creates a new independent AI service instance when it detects that the device unique identifier is connected to the network for the first time, and loads the existing instance when it detects that a corresponding independent AI service instance already exists.

[0032] On the other hand, the present invention also provides a natural language semantic mapping method for a personalized massage chair system based on cloud-based independent AI services, comprising the following steps: S1. Receive natural language input from the user and extract intent elements through the natural language understanding module. The intent elements include at least one of symptom description, location specification, intensity preference, and duration requirement. S2. Query the massage knowledge graph based on the intent element, retrieve the matching symptom-location-technique-parameter quadruple, and obtain the default physical control parameters corresponding to the intent element; S3. Based on the current user preference vector, the default physical control parameters are fine-tuned to generate target physical control parameters. The target physical control parameters include at least one of the following: massage head rotation speed, airbag target pressure, heating unit target temperature, and robotic arm trajectory coordinates. S4. Encapsulate the target physical control parameters into actuator control instructions and send them to the massage chair's main control MCU; S5. During the massage process, the target physical control parameters are dynamically adjusted based on the user's real-time voice feedback.

[0033] In one possible implementation, step S2, querying the massage knowledge graph based on intent elements, includes the following steps: Match symptom nodes based on symptom descriptions to obtain recommended and contraindicated sites associated with the symptom; Based on the specified body part node, obtain the corresponding body part and applicable massage techniques; Based on the force preference, match the technique nodes to determine the movement mode; Obtain the default physical control parameters based on the parameter nodes associated with the part nodes and technique nodes.

[0034] In yet another possible implementation, step S3, the personalized fine-tuning includes the following steps: Based on the value of the intensity preference dimension in the user preference vector, the massage head rotation speed and airbag target pressure are adjusted proportionally. The target temperature of the heating unit is adjusted by offset based on the value of the temperature preference dimension in the user preference vector; The duration of a single massage session is adjusted based on the value of the duration preference dimension in the user preference vector.

[0035] In other possible implementations, step S5, dynamic adjustment includes the following steps: During the massage, the user's voice input is monitored in real time. When an incremental command from the user to adjust the intensity, change the body part, or modify the duration is detected, the incremental value is parsed. In addition, incremental values ​​are added to the currently executing target physical control parameters to generate adjusted physical control parameters, and execution continues.

[0036] In another possible implementation, In step S4, before the step of encapsulating into actuator control instructions, a safety constraint verification step is also included: Verify whether the target physical control parameters are within the preset safety range; Based on health contraindication tags in user profiles, high-risk techniques and body parts are blocked; And check whether the continuous massage duration of the same area exceeds the set limit; In another possible implementation, the invention further includes the step of anomaly semantic processing: When the natural language understanding module detects that the intent elements of the user's input are incomplete or ambiguous, it generates a clarifying question and outputs it to the user. Receive supplementary input from users in response to clarification inquiries; By combining the original input and supplementary input, the complete intent elements are extracted again.

[0037] Figure 1 This is a block diagram of the cloud-based independent AI service personalized architecture of the present invention. It shows the binding relationship of the independent AI service instance manager to create independent AI service instances with device UUID as index, the extraction logic of massage knowledge graph subgraph from global knowledge graph, the incremental learning and update process of user preference vector, the context storage structure of dialogue history cache, the satisfaction closed-loop mechanism of effect feedback record, and the multi-round dialogue collection process of cold start guidance module.

[0038] Depend on Figure 1As can be seen, the personalized massage chair system based on cloud-based independent AI services described in this invention includes: a cloud-based independent AI service cluster, a device binding management module, a massage knowledge graph module, and a user preference learning module. Wherein: The cloud-based independent AI service cluster assigns an independent AI service instance to each massage chair device. The independent AI service instance is indexed by the massage chair device's unique identifier (UUID). The lifecycle of the independent AI service instance is bound to the lifecycle of the massage chair device.

[0039] The massage knowledge graph module maintains a knowledge base of symptom-location-technique-parameter quadruples; symptoms include subjective descriptions from users (such as "soreness", "stiffness", "fatigue"); locations include massage areas on the human body; techniques include the action patterns of the massage mechanism (such as kneading, tapping, acupressure, and tuina); parameters include the physical control values ​​of the massage actuator.

[0040] The user preference learning module uses an incremental learning algorithm to encode users' historical massage feedback data (such as manual adjustment records, voice evaluations, and post-massage satisfaction ratings) into user preference vectors and continuously updates them.

[0041] When an independent AI service instance receives user interaction data, it combines a massage knowledge graph and user preference vectors to generate a personalized massage control strategy.

[0042] Figure 2 This is a flowchart illustrating the natural language semantics-to-physical parameter mapping process of the present invention. It shows the end-to-end mapping process from user voice input to intent element extraction, symptom node matching, location node retrieval, technique node determination, parameter node acquisition, personalized fine-tuning of user preference vector, safety constraint verification, and actuator control instruction encapsulation, as well as the dynamic adjustment closed loop of real-time voice feedback during massage execution.

[0043] Depend on Figure 2 As can be seen, the mapping method from natural language semantics to actuator physical parameters described in this invention, applied to massage chair control, includes the following steps: Step A1: Receive natural language input from the user and extract intent elements through the Natural Language Understanding (NLU) module. Intent elements include at least one of symptom description, location specification, intensity preference, and duration requirement. Step A2: Query the massage knowledge graph based on intent elements to retrieve matching symptom-location-technique-parameter quadruples; Step A3: Based on the current user preference vector, fine-tune the parameter values ​​in the quadruple to generate target physical control parameters; the target physical control parameters include at least one of the following: massage head rotation speed (RPM), airbag target pressure (kPa), heating unit target temperature (°C), and robotic arm trajectory coordinates (X, Y, Z); Step A4: Encapsulate the target physical control parameters into actuator control instructions and send them to the massage chair's main control MCU; Step A5: During the massage, dynamically adjust the target physical control parameters based on the user's real-time voice feedback.

[0044] The above method will be illustrated with specific examples below.

[0045] Example 1: Cloud-based AI Independent Service Instantiation and Personalization like Figure 3 The diagram shown is a block diagram of the independent AI service instantiation and memory architecture. It illustrates the device UUID generation and OTP burning process, the creation / loading / destruction logic of the independent AI service instance manager, the voiceprint registration and user identity binding process, the dynamic extraction mechanism of the massage knowledge graph subgraph, the incremental learning and update formula of the user preference vector, and the exploration period parameter convergence strategy of the cold start guidance module. Specifically, the independent AI service instantiation process of this invention is as follows: Step S301: When the massage chair device leaves the factory, a globally unique UUID (such as "MC-2026A-8F3E12D4") is generated and written into the device's OTP area.

[0046] Step S302: When the device connects to the network for the first time, the Independent AI Service Instance Manager checks whether an independent AI service instance indexed by this UUID exists. If it does not exist, a new instance is created; if it exists (e.g., when the device is reconnected after repair), the existing instance is loaded.

[0047] Step S303: Upon first use, a user's identity is established through voiceprint registration. The voiceprint feature vector is bound to the user ID and stored in the user preference learning module. Multiple users can register their voiceprints on the same massage chair, with each user corresponding to an independent user preference vector.

[0048] Step S304: A standalone AI service instance contains the following core data structures: Massage Knowledge Graph Subgraphs: Extract subgraphs related to users’ common symptoms and body parts from the global knowledge graph to reduce the search space during reasoning; User preference vector: A floating-point vector of dimension D that encodes a user's preference for intensity, temperature, duration, and technique. Dialogue history cache: The context of the most recent N rounds of dialogue, used for the coherence of multi-round dialogues; Results feedback records: Satisfaction ratings and user reviews after each massage.

[0049] Step S305: The user preference vector is updated using an incremental learning algorithm. Let the current preference vector be P_old. After this massage, the user manually adjusted the intensity (from level 3 to level 5) and gave a voice evaluation of "This intensity is just right." The system then extracts the preference increment ΔP from this session, and the update formula is: P_new = α · P_old + (1-α) · (P_old + ΔP) Where α is the learning rate (0.5 ≤ α ≤ 0.9), new users have a smaller α (learn quickly), while old users have a larger α (remain stable).

[0050] Step S306: Cold start strategy for independent AI service instances. When a new user uses the service for the first time, an initial profile is collected through guided dialogue: "Which part of your body gets tired most easily?" "Do you prefer a gentle and soothing massage or a deep and powerful massage?" "How long do you usually have a massage session?" The collected data is used to initialize the user preference vector. The first 3 massages are the "exploration period," where different parameter combinations are actively tried, and preferences are quickly converged based on user feedback.

[0051] Example 2: Mapping Natural Language Semantics to Physical Parameters like Figure 4 The diagram shown illustrates the four-tuple structure and semantic mapping flowchart of the massage knowledge graph. It demonstrates the relationships between symptom nodes, location nodes, technique nodes, and parameter nodes, showing the complete mapping path from user natural language input to NLU intent element extraction, knowledge graph query, personalized fine-tuning, control command encapsulation, and CAN bus transmission, as well as the incremental command parsing process during dynamic adjustments. Specifically, the massage knowledge graph adopts a four-tuple structure: Symptom nodes: User's subjective description of feelings, such as "aches and pains," "stiffness," "fatigue," "throbbing pain," and "numbness." Each symptom node is associated with recommended and contraindicated areas.

[0052] Location nodes: Massage areas of the body, such as "neck," "shoulder," "waist," and "legs." Each location node is associated with applicable massage techniques and default parameters.

[0053] Technique nodes: movement patterns of the movement, such as "kneading" (simulating the kneading action of a human hand, with the movement performing periodic compression), "tapping" (rapid striking), "acupressure" (deep pressure at a fixed point), and "massage" (long-stroke pushing and pressing). Each technique corresponds to a set of basic physical parameters.

[0054] Parameter nodes: physical control values, including: massage head rotation speed 20-120 RPM, airbag pressure 10-80 kPa, heating temperature 35-55℃, trajectory coordinates (target position sequence of the mechanism in XYZ three-dimensional space), and single session duration 5-60 minutes.

[0055] Mapping process example: User voice input: "I've been working overtime every day lately, and my neck is really stiff. I'd like a warm, deep massage for half an hour." Step S401: NLU analysis extracts intent elements: Symptom "Neck stiffness" → Match symptom node "Neck stiffness"; Location "Neck" → Match location nodes "Neck" and "Shoulder"; Strength preference "Deep" → Technique preference "Acupressure" + "Kneading", high strength level; Temperature preference "Warm" → Turn on heating, target temperature 42℃; Duration "Half an hour" → 30 minutes.

[0056] Step S402: Knowledge Graph Query: Query the parameter quadruple set of "Neck Stiffness" → "Neck / Shoulder" → "Acupressure / Kneading" and obtain the default values: rotation speed 60RPM, airbag pressure 45kPa, temperature 40℃, duration 20 minutes.

[0057] Step S403: Personalized fine-tuning: Based on the current user preference vector, the user's historical preference is "high intensity" (the third dimension of the preference vector = 0.7), then the parameter fine-tuning is as follows: speed 80 RPM (+33%), airbag pressure 55 kPa (+22%), temperature 42℃ (+2℃), duration 30 minutes (as per user requirements).

[0058] Step S404: Generate control commands: Encapsulate the above parameters into a CAN bus command frame and send it to the main control MCU. After parsing, the MCU drives the mechanism motor to run the acupressure program at 80 RPM, adjusts the airbag solenoid valve to 55 kPa, and raises the heating unit temperature to 42°C. The total program duration is 30 minutes.

[0059] Step S405: Dynamic Adjustment: After 10 minutes of massage, the user says, "The intensity can be increased a little more." The system interprets the incremental instruction "Intensity +1 level," and based on the original parameters, the rotation speed is increased to 90 RPM and the airbag pressure is increased to 60 kPa. The massage program continues for the remaining 20 minutes.

[0060] Example 3: Security Constraint Verification Before the step of encapsulating into actuator control instructions, a safety constraint verification step is also included: Parameter range verification: Rotation speed ∈ [20, 120] RPM, pressure ∈ [10, 80] kPa, temperature ∈ [35, 55] ℃, duration ∈ [5, 60] minutes; Contraindication verification: Based on health contraindication tags in the user profile (such as "osteoporosis", "pregnant woman", "hypertension"), high-risk techniques and areas are automatically blocked; Cumulative load check: The duration of continuous massage on the same area does not exceed the set upper limit to prevent over-massage.

[0061] In other embodiments of the present invention, an electronic device 500 is disclosed, which can integrate the above-described system, such as... Figure 5 As shown, the hardware components include: one or more processors 501; memory 502; display 503; one or more application programs (not shown); and one or more computer programs 504. These components can be connected via one or more communication buses 505. The one or more computer programs 504 are stored in the memory 502 and configured to be executed by the one or more processors 501. The one or more computer programs 504 include instructions.

[0062] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer, implements the method described in the above-described method embodiments. Specific beneficial effects can be found in the above-described method embodiments.

[0063] The present invention also provides a computer program product that, when executed by a computer, implements the method described in the above-described method embodiments. Specific beneficial effects can be found in the above-described method embodiments.

[0064] Through the above description of the embodiments, those skilled in the art will clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical 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. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0065] In the various embodiments of the present invention, the functional units 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.

[0066] 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 computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as flash memory, portable hard disk, read-only memory, random access memory, magnetic disk, or optical disk.

[0067] The above description is merely a specific implementation of the embodiments of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of the present invention should be covered within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be determined by the protection scope of the claims.

Claims

1. A personalized massage chair system based on cloud-based independent AI services, characterized in that, include: The massage chair body has a built-in main control MCU, massage actuator and sensor array, and is connected to a cloud server through a communication network; A cloud-based independent AI service cluster, deployed on a cloud server, is used to assign an independent AI service instance indexed by a unique device identifier to each massage chair device; The massage knowledge graph module is used to maintain a knowledge base of symptom-location-technique-parameter quadruples, which include symptom nodes, location nodes, technique nodes, and parameter nodes. The user preference learning module is used to encode users' historical massage feedback data into user preference vectors through an incremental learning algorithm, and continuously update the user preference vectors. The device binding management module is used to establish a one-to-one binding relationship between the unique identifier of the massage chair device and the independent AI service instance, and the lifecycle of the independent AI service instance is bound to the lifecycle of the massage chair device.

2. The system as described in claim 1, characterized in that, The standalone AI service instances include: Massage knowledge graph subgraphs are extracted from the global massage knowledge graph and are related to the common symptoms and body parts of users of this massage chair. User preference vectors are used to encode users' preferences for massage intensity, temperature, duration, and type of massage. Dialogue history cache is used to store context information from recent rounds of dialogue; The feedback record is used to store satisfaction ratings and user review texts after each massage.

3. The system as described in claim 1, characterized in that, The parameter nodes in the massage knowledge graph module include: The massage head rotation speed parameter has a value range of the preset rotation speed range. The target pressure parameter for the airbag has a value range within a preset pressure range. The target temperature parameter of the heating unit has a range of preset temperature values. The robotic arm trajectory coordinate parameters are used to describe the target position sequence of the massage mechanism in three-dimensional space.

4. The system as described in claim 1, characterized in that, The user preference learning module uses an incremental learning algorithm to update the user preference vector, and the update formula is as follows: P_new = α · P_old + (1-α) · (P_old + ΔP), Where P_old is the user preference vector before the update, ΔP is the preference increment extracted in this massage session, and α is the learning rate coefficient, with a value range of the preset learning rate range. New users correspond to a smaller α value, and old users correspond to a larger α value.

5. The system as described in claim 1, characterized in that, It also includes a cold start guidance module, which is used to collect the user's initial preference information through multiple rounds of dialogue when a new user uses it for the first time. The initial preference information includes commonly fatigued areas, intensity preference baseline, and single massage duration preference. The initial preference information is used to initialize the user preference vector. The device unique identifier is a globally unique identifier code burned into the device's OTP area when the massage chair is manufactured. The independent AI service instance manager creates a new independent AI service instance when it detects that the device unique identifier is connected to the network for the first time, and loads the existing instance when it detects that a corresponding independent AI service instance already exists.

6. A natural language semantic mapping method for a personalized massage chair system based on cloud-based independent AI services as described in any one of claims 1 to 5, characterized in that, Includes the following steps: S1. Receive natural language input from the user and extract intent elements through the natural language understanding module. The intent elements include at least one of symptom description, location specification, intensity preference, and duration requirement. S2. Query the massage knowledge graph based on the intent element, retrieve the matching symptom-location-technique-parameter quadruple, and obtain the default physical control parameters corresponding to the intent element; S3. Based on the current user preference vector, the default physical control parameters are fine-tuned to generate target physical control parameters. The target physical control parameters include at least one of the following: massage head rotation speed, airbag target pressure, heating unit target temperature, and robotic arm trajectory coordinates. S4. Encapsulate the target physical control parameters into actuator control instructions and send them to the massage chair's main control MCU; S5. During the massage process, the target physical control parameters are dynamically adjusted based on the user's real-time voice feedback.

7. The method as described in claim 6, characterized in that, In step S2, querying the massage knowledge graph based on intent elements includes the following steps: Match symptom nodes based on symptom descriptions to obtain recommended and contraindicated sites associated with the symptom; Based on the specified body part node, obtain the corresponding body part and applicable massage techniques; Based on the force preference, match the technique nodes to determine the movement mode; Obtain the default physical control parameters based on the parameter nodes associated with the part nodes and technique nodes.

8. The method as described in claim 6, characterized in that, In step S3, the personalized fine-tuning includes the following steps: Based on the value of the intensity preference dimension in the user preference vector, the massage head rotation speed and airbag target pressure are adjusted proportionally. The target temperature of the heating unit is adjusted by offset based on the value of the temperature preference dimension in the user preference vector; The duration of a single massage session is adjusted based on the value of the duration preference dimension in the user preference vector.

9. The method as described in claim 6, characterized in that, In step S5, the dynamic adjustment includes the following steps: During the massage, the user's voice input is monitored in real time. When an incremental command from the user to adjust the intensity, change the body part, or modify the duration is detected, the incremental value is parsed. In addition, incremental values ​​are added to the currently executing target physical control parameters to generate adjusted physical control parameters, and execution continues.

10. The method as described in claim 6, characterized in that: In step S4, before the step of encapsulating into actuator control instructions, a safety constraint verification step is also included: Verify whether the target physical control parameters are within the preset safety range; Based on health contraindication tags in user profiles, high-risk techniques and body parts are blocked; And check whether the continuous massage duration of the same area exceeds the set limit; It also includes steps for handling exception semantics: When the natural language understanding module detects that the intent elements of the user's input are incomplete or ambiguous, it generates a clarifying question and outputs it to the user. Receive supplementary input from users in response to clarification inquiries; By combining the original input and supplementary input, the complete intent elements are extracted again.

Images