Sub-health conditioning cabin self-adaptive regulation system based on big data cloud platform

Abstract

The application discloses a sub-health conditioning cabin self-adaptive regulation and control system based on a big data cloud platform and relates to the technical field of health management equipment.The application constructs a multi-source data set through multi-dimensional data integration preprocessing, so that a regulation and control model can accurately extract characteristic data to diagnose conditioning effects, divides scenes in combination with multiple factors, and associates historical data to generate initial regulation and control parameters, provides a matching benchmark for a self-adaptive regulation and control module, enables the self-adaptive regulation and control module to accurately output parameter adjustment instructions to control an execution mechanism, extracts characteristic data to input a trained model to diagnose and generate a scheme, realizes accurate conditioning, matches parameter instructions to control multiple execution mechanisms, dynamically adapts energy fields and sub-health states, and a conditioning effect feedback module collects feedback data and evaluation reports, provides adjustment basis for the self-adaptive regulation and control module, timely intervenes in abnormal conditioning, avoids excessive early warning or insufficient early warning through hierarchical response, guarantees user safety, reduces unnecessary conditioning interruption, and improves user experience.

Description

Technical Field

[0001] This invention relates to the field of health management equipment technology, and in particular to an adaptive control system for a sub-health conditioning cabin based on a big data cloud platform. Background Technology

[0002] Currently, most sub-health conditioning cabins on the market adopt fixed parameter control modes. For example, Chinese patent application CN120037048A discloses a multi-sensory sub-health sleep conditioning cabin, which constructs a device system including modules such as an intelligent control system, a phototherapy system, a sound therapy system, an aromatherapy system, a physical massage system, and an intelligent safety monitoring system. This system achieves multi-factor synergistic treatment and personalized treatment, while enhancing safety and comfort. It not only has significant sleep conditioning effects but also improves treatment safety, emphasizes data privacy protection, and improves system stability and compatibility. It can comprehensively improve sleep disorders in a comfortable state. Moreover, thanks to the integrated application of multi-sensory stimulation therapy and the intelligent control system, it can achieve comprehensive and personalized treatment plans, possessing outstanding technological advantages and good prospects for promotion.

[0003] Although the above patents exist, the following problems still exist:

[0004] In existing technologies, although conditioning chambers have basic parameter adjustment functions, they mostly rely on users to manually set or preset fixed programs. They cannot dynamically generate suitable conditioning programs based on users' age, gender, health status and real-time sleep data, making it difficult to achieve accurate matching of parameters with individual user differences and real-time status, which affects the pertinence and effectiveness of conditioning effects. Furthermore, they lack real-time monitoring of the chamber environment and user status, or can only issue basic alarms after monitoring, making it difficult to fully avoid potential risks during the conditioning process. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive regulation system for a sub-health conditioning cabin based on a big data cloud platform. By integrating and preprocessing multi-dimensional data to construct a multi-source dataset, combining multiple factors to divide scenarios and associate historical data to generate initial regulation parameters, the adaptive regulation module controls the actuator to dynamically adapt to the energy field, achieving precise conditioning, ensuring that the system continuously adapts to the user's state, effectively improving the user experience, and solving the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The sub-health conditioning cabin adaptive regulation system based on a big data cloud platform includes a sub-health data collection module, a cloud platform analysis module, an adaptive regulation module, and a conditioning effect feedback module. Each module constructs a real-time data interaction network through the big data cloud platform, forming the adaptive regulation link of the sub-health conditioning cabin.

[0008] The sub-health data acquisition module is used to collect real-time physiological parameters and operational status data of users in the sub-health conditioning cabin. It preprocesses the collected physiological parameters and operational status data to generate multi-source datasets.

[0009] The conditioning scenario adaptation module is used to divide the usage scenarios based on the user's conditioning needs, activity status and cabin environment conditions, and to construct the initial control parameters of the sub-health conditioning cabin by combining the historical conditioning data sample library in the big data cloud platform.

[0010] The cloud platform analysis module is used to extract the operational and physiological characteristics data of multi-source datasets and input them into the constructed sub-health conditioning adaptive regulation model for conditioning effect diagnosis. Based on the diagnostic results of the sub-health conditioning adaptive regulation model, the corresponding conditioning plan is generated.

[0011] The adaptive control module is used to match the corresponding parameter adjustment instructions based on the conditioning plan and the initial control parameters of the sub-health conditioning cabin. The sub-health conditioning cabin receives the parameter adjustment instructions and controls the corresponding actuators to perform the control operation.

[0012] The treatment effect feedback module is used to collect users' physiological feedback data during and after the treatment, compare it with the baseline value before treatment, issue abnormal warnings, generate an effect evaluation report, and transmit the evaluation data and feedback information to the big data cloud platform.

[0013] Furthermore, the sub-health data acquisition module integrates an operational status acquisition sensor and a user physiological parameter acquisition sensor; wherein, the operational status acquisition sensor is used to acquire data on cabin temperature, humidity, oxygen concentration, far-infrared radiation intensity, magnetic flux density, negative oxygen ion concentration, and cabin temperature uniformity, and the user physiological parameter acquisition sensor is used to acquire data on the user's heart rate, heart rate variability, blood oxygen saturation, skin resistance, body surface temperature distribution, microcirculation velocity, and electroencephalogram rhythm.

[0014] Furthermore, the conditioning scenario adaptation module includes:

[0015] The scenario classification unit is used to acquire cabin environment data and user input of target conditioning needs and target activity status, divide the usage scenarios into static conditioning scenarios, dynamic intervention scenarios and emergency relief scenarios, and output the initial scenario classification results.

[0016] The scene label construction unit is used to extract scene feature data corresponding to the scene classification result based on the initial scene classification result and the historical conditioning data sample library, and generate the corresponding scene label;

[0017] The data calibration unit is used to retrieve the standard physiological parameter range that is the same as the scene label in the historical conditioning data sample library, calculate the deviation between the user physiological parameters in the multi-source dataset and the standard physiological parameter range, perform calibration processing on the user physiological parameters in the multi-source dataset, and output the calibrated physiological parameter data.

[0018] Furthermore, the specific process of the scene tag construction unit generating the corresponding scene tags includes:

[0019] The current scene corresponding to the initial scene classification result is analyzed to determine the scene data, which includes cabin environment data, user basic data, conditioning demand data and activity status data.

[0020] The scene data is divided into multiple scene blocks according to scene category, including environmental parameter blocks, user attribute blocks, conditioning demand blocks, and activity status blocks;

[0021] Obtain the type of each scene block and perform numerical processing to obtain the type value of each scene block;

[0022] Calculate the correlation degree between each scene block based on the type value, retain the related scene blocks with a correlation degree less than the preset correlation degree threshold, and remove the abnormal blocks with no correlation.

[0023] A scene tree is constructed based on the correlation between each related scene block and each scene block, where each scene block corresponds to a scene node in the scene tree.

[0024] Key features are extracted from the scene nodes in the scene tree from top to bottom, and the extraction method of object data included in the scene nodes is determined according to the type of scene node;

[0025] The feature range that meets the extraction rules in each level is determined as the target feature region, and the feature data in the target feature region is validated. After the validation is passed, an initial scene label description is generated based on the extracted node feature data.

[0026] Retrieve historical processing data sample library from the big data cloud platform, perform similarity matching between the initial scene label description and the historical scene label, correct the deviation data in the initial label description, and form a standardized scene feature description;

[0027] Based on the effective feature data of all scene nodes in the scene tree, several associated scene label descriptions are obtained;

[0028] Based on the feature extraction order of the several associated scene tags, a queue is established, the matching degree with historical valid scene tags is calculated, and the standardized scene feature description with the highest matching degree is selected as the final scene tag of the current scene and stored in the big data cloud platform.

[0029] Furthermore, the specific process of calculating the correlation degree between each scene block based on the type value includes:

[0030] Retrieve the type value for each scene block;

[0031] Generate a type value vector corresponding to each scene block using the type value of each scene block;

[0032] Retrieve the type value vector corresponding to the scene block for each data collection;

[0033] Obtain the L2 norm of the type value vector corresponding to each data collection based on the type value vector corresponding to each data collection in each scene segment;

[0034] The type fluctuation coefficient corresponding to each scene block is obtained based on the L2 norm of the type value vector determined by each data collection for each scene block.

[0035] The correlation between different scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block.

[0036] Furthermore, the correlation between different scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block, including:

[0037] Retrieve the type value vector for each scene block;

[0038] The cosine similarity between any two scene blocks is obtained using the type value vector of each scene block.

[0039] Retrieve the type fluctuation coefficient corresponding to each scene block;

[0040] The type fluctuation coefficients corresponding to each pair of scene blocks are coupled to obtain the type fluctuation coupling coefficients of each pair of scene blocks.

[0041] The correlation between each pair of scene blocks is obtained by using the cosine similarity and type fluctuation coupling coefficient corresponding to each pair of scene blocks.

[0042] Furthermore, the specific process of constructing initial control parameters for the conditioning scenario adaptation module includes:

[0043] Based on the current scenario type and the final scenario label, retrieve the historical treatment data sample library for the corresponding scenario. The historical treatment data sample library contains the baseline range of user physiological parameters, treatment parameter configurations, and treatment effect data for users with different basic attributes in the same scenario.

[0044] The baseline values ​​of the current user's physiological parameters are determined by matching the mean of the calibrated physiological parameter data with the baseline range of the user's physiological parameters.

[0045] Set the data collection cycle for user physiological parameters and cabin operation status data according to the treatment need level corresponding to the scene label;

[0046] By integrating the baseline values ​​of physiological parameters with the data acquisition cycle and combining the segmented features of environmental parameters in the scene tags, the initial control parameters of the conditioning chamber are generated, and the association mapping relationship between the initial control parameters and scene type and scene tag is established.

[0047] Furthermore, the specific process by which the feature processing module generates operational feature data and physiological feature data includes:

[0048] Based on the set data acquisition cycle, each acquisition cycle is evenly divided into several consecutive time nodes, and each time node is synchronized with the acquisition cycle.

[0049] Two adjacent time points are selected to form a monitoring period, and the rate of change of the operational status data and the rate of change of the user's physiological parameters are calculated within each monitoring period.

[0050] The rate of change of operational status data and the rate of change of user physiological parameters for all monitoring periods within the same data collection cycle are summarized to generate operational characteristic data and physiological characteristic data.

[0051] Furthermore, the parameter adjustment instruction generation process of the adaptive control module includes:

[0052] The difference between the target parameters and the initial control parameters in the conditioning plan is calculated to obtain the parameter adjustment amount for each actuator, and the adjustment level is divided according to the absolute value of the parameter adjustment amount.

[0053] For different adjustment levels, a corresponding adjustment strategy is matched, and the adjustment strategy is converted into a drive signal for the actuator to generate an executable parameter adjustment instruction.

[0054] Furthermore, the big data cloud platform is also used to learn from the received evaluation data and feedback information, specifically including:

[0055] The quantitative indicators in the effect evaluation report are associated and stored with the corresponding treatment plan, scene tag, and scene type. The historical treatment data sample library is updated, and treatment effect cases under the same scene are added.

[0056] Time-series analysis was performed on multiple treatment data of the same user to extract the trend characteristics of the user's physiological indicators, and the treatment plan for the user was optimized based on the trend characteristics.

[0057] Migrate conditioning data from new scenarios that are not included in the historical conditioning data sample library to train the sub-health conditioning adaptive regulation model in similar scenarios, thus supplementing the model training samples.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] By integrating and preprocessing multi-dimensional data to construct a multi-source dataset, the regulatory model can accurately extract feature data to diagnose the conditioning effect. Simultaneously, by combining multiple factors to segment scenarios and linking historical data, initial regulatory parameters are generated, providing a matching benchmark for the adaptive regulatory module. This allows the module to accurately output parameter adjustment commands to control the execution mechanism, avoiding the problem of poor adaptability of general parameters. Feature data is extracted and input into the trained model for diagnosis and solution generation, achieving precise conditioning. Matching parameter commands controls multiple execution mechanisms, dynamically adapting to the energy field and sub-health state. Feedback data and evaluation reports collected by the conditioning effect feedback module provide adjustment basis for the adaptive regulatory module, preventing the conditioning effect from becoming fixed and ensuring the system continuously adapts to the user's state. This not only allows for timely intervention in conditioning anomalies but also avoids over-warning or under-warning through tiered responses, ensuring user safety, reducing unnecessary conditioning interruptions, and improving the user experience. Attached Figure Description

[0060] Figure 1 This is a block diagram of the adaptive control system of the sub-health conditioning cabin based on a big data cloud platform according to the present invention;

[0061] Figure 2 This is a flowchart of the adaptive control system for the sub-health conditioning cabin based on a big data cloud platform according to the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] To address the technical issues of existing conditioning chambers that rely on manual settings and fixed plans, failing to dynamically generate tailored conditioning plans based on individual user differences and real-time sleep data, thus affecting the targeted effectiveness of conditioning, and lacking a comprehensive real-time monitoring and early warning mechanism for the chamber environment and user condition, making it difficult to fully avoid potential risks during conditioning, please refer to [link to relevant documentation]. Figures 1-2This embodiment provides the following technical solution:

[0064] The sub-health conditioning cabin adaptive control system based on a big data cloud platform includes a sub-health data acquisition module, a cloud platform analysis module, an adaptive control module, and a conditioning effect feedback module. Each module constructs a real-time data interaction network through the big data cloud platform, forming an adaptive control link for the sub-health conditioning cabin. The sub-health data acquisition module integrates operating status acquisition sensors and user physiological parameter acquisition sensors. The operating status acquisition sensors acquire data on cabin temperature, humidity, oxygen concentration, far-infrared radiation intensity, magnetic flux density, negative oxygen ion concentration, and cabin temperature uniformity. The user physiological parameter acquisition sensors acquire data on the user's heart rate, heart rate variability, blood oxygen saturation, skin resistance, body surface temperature distribution, microcirculation velocity, and electroencephalogram rhythm.

[0065] The sub-health data acquisition module is used to collect real-time physiological parameters of users in the sub-health conditioning chamber, such as heart rate, blood oxygen saturation, skin resistance, body temperature fluctuations, and initial brain wave rhythm, as well as chamber operation status data, such as chamber temperature, humidity, oxygen concentration, and initial operating parameters of each energy component. The module preprocesses the collected user physiological parameters and chamber operation status data, including using filtering algorithms to remove environmental interference signals and normalizing the noise-reduced data to generate a multi-source dataset.

[0066] The conditioning scenario adaptation module is used to divide the usage scenarios based on the user's conditioning needs, activity status and cabin environment conditions, and to construct the initial control parameters of the sub-health conditioning cabin by combining the historical conditioning data sample library in the big data cloud platform.

[0067] The cloud platform analysis module is used to extract the operational and physiological characteristics data of multi-source datasets and input them into the constructed sub-health conditioning adaptive regulation model for conditioning effect diagnosis. Based on the diagnostic results of the sub-health conditioning adaptive regulation model, the corresponding conditioning plan is generated.

[0068] In this embodiment, the sub-health conditioning adaptive regulation model takes historical operational characteristic data and historical physiological characteristic data as input, and takes the conditioning effect score composed of the user's subjective rating and the improvement rate of physiological indicators as output. The model parameters are optimized through iterative training. When obtaining the diagnostic results, the conditioning effect is divided into four levels: excellent, good, moderate, and poor.

[0069] Excellent rating: Physiological indicator improvement rate > 30% and subjective score ≥ 8 points;

[0070] Good grade: Physiological indicators improve by 15%-30% and subjective score is 6-7;

[0071] Medium level: Physiological indicators improved by 5%-15% and subjective scores were 4-5.

[0072] Poor grade: Improvement rate of physiological indicators <5% and subjective score <4 points;

[0073] The adaptive control module is used to match the corresponding parameter adjustment instructions based on the conditioning plan and the initial control parameters of the sub-health conditioning cabin. The sub-health conditioning cabin receives the parameter adjustment instructions and controls the corresponding actuators to perform the control operations. The actuators include an infrared radiation module, an electromagnetic resonance device, a photon energy component and an acoustic resonance system, so as to realize the dynamic adaptation of the energy field to the user's sub-health state.

[0074] The treatment effect feedback module is used to collect users' physiological feedback data during and after the treatment, compare it with the baseline value before the treatment and provide abnormal warnings, generate an effect evaluation report including microcirculation improvement rate, cell activity enhancement, and brain wave rhythm recovery, and transmit the evaluation data and feedback information to the big data cloud platform.

[0075] In this embodiment, an early warning is triggered when any of the following conditions are detected: the microcirculation improvement rate decreases for three consecutive collection cycles and the cumulative decrease is >15%; the cell activity improvement rate becomes negative during the treatment process and the absolute value is >8%; the brain wave rhythm smoothness is <0 and the duration is >5 minutes; the early warning information is simultaneously pushed to the cloud platform analysis module, triggering the treatment plan optimization process;

[0076] The system includes preset multiple levels of abnormal thresholds. When operational status data or user physiological parameters exceed the threshold range of the corresponding level, an alarm of that level is triggered.

[0077] Level 1 Alarm (Minor Abnormality): Activates local green light indicator and low-frequency buzzer;

[0078] Level 2 Alarm (Moderate Abnormality): Initiates flashing yellow light and a medium-frequency buzzer, and pushes an alert to the user's terminal;

[0079] Level 3 alarm (serious abnormality): Red light flashes and high-frequency buzzer sounds, suspends operation and turns on emergency ventilation.

[0080] In this embodiment, a multi-source dataset is constructed through multi-dimensional data integration and preprocessing, enabling the regulation model to accurately extract feature data to diagnose the regulation effect. Simultaneously, by combining multiple factors to classify scenarios and associating them with historical data, initial regulation parameters are generated, providing a matching benchmark for the adaptive regulation module. This allows the module to accurately output parameter adjustment commands to control the execution mechanism, avoiding the problem of poor adaptability of general parameters. Feature data is extracted and input into the trained model for diagnosis and solution generation, achieving precise regulation. Matching parameter commands controls multiple execution mechanisms, dynamically adapting to the energy field and sub-health state. Feedback data and evaluation reports collected by the regulation effect feedback module provide adjustment basis for the adaptive regulation module, preventing the regulation effect from becoming fixed and ensuring the system continuously adapts to the user's state. This not only allows for timely intervention in regulating abnormalities but also avoids excessive or insufficient warnings through tiered responses, ensuring user safety, reducing unnecessary regulation interruptions, and improving the user experience.

[0081] In this embodiment, the conditioning scene adaptation module includes:

[0082] The scenario classification unit is used to acquire in-cabin environmental data and user-inputted target conditioning needs (fatigue relief, sleep improvement, metabolic regulation) and target activity states (sitting, light activity, semi-reclining). It uses a multi-dimensional feature matching algorithm to classify the usage scenarios into static conditioning scenarios, such as in a normal in-cabin environment, in a sitting or semi-reclining state and matching any target conditioning need; dynamic intervention scenarios, such as in a normal in-cabin environment, in a light activity state and matching metabolic regulation needs; and emergency relief scenarios, such as in an abnormal in-cabin environment or sudden discomfort, in any target activity state and matching any target conditioning need. The initial scenario classification results are then output.

[0083] The scene label construction unit is used to extract scene feature data corresponding to the scene classification result based on the initial scene classification result and the historical conditioning data sample library, and generate the corresponding scene label;

[0084] The data calibration unit is used to retrieve the standard physiological parameter range that is the same as the scene label in the historical conditioning data sample library, calculate the deviation between the user physiological parameters in the multi-source dataset and the standard physiological parameter range, calibrate the user physiological parameters in the multi-source dataset, and output the calibrated physiological parameter data to eliminate the acquisition error caused by individual physiological differences.

[0085] In this embodiment, the specific process by which the scene tag construction unit generates the corresponding scene tags includes:

[0086] The current scene corresponding to the initial scene classification result is analyzed to determine the scene data, which includes cabin environment data, user basic data, conditioning demand data and activity status data.

[0087] The scene data is divided into multiple scene blocks according to scene category, including environmental parameter blocks, user attribute blocks, conditioning demand blocks, and activity status blocks;

[0088] Obtain the type of each scene block and perform numerical processing to obtain the type value of each scene block;

[0089] Calculate the correlation degree between each scene block based on the type value, retain the related scene blocks with a correlation degree less than the preset correlation degree threshold, and remove the abnormal blocks with no correlation.

[0090] A scene tree is constructed based on the correlation between various related scene blocks and each scene block. Each scene block corresponds to a scene node in the scene tree, including the root node, first-level child nodes, second-level child nodes, and third-level child nodes.

[0091] Key features are extracted from the scene nodes in the scene tree from top to bottom (root node to third-level child nodes). Environmental parameter features are extracted from the root node, such as temperature 22-26℃, humidity 40%-60%, and oxygen concentration 21%-23%. User attribute features are extracted from the first-level child nodes, such as age 55 and sleep disorders. Conditioning needs features are extracted from the second-level child nodes, such as sleep improvement. Activity status features are extracted from the third-level child nodes, such as semi-reclining. The method of extracting object data included in the scene node is determined according to the type of scene node.

[0092] Among them, the environmental parameter segment node extracts features according to the preset normal / abnormal parameter thresholds to ensure that the parameter values ​​correspond to the labels; the user attribute segment node extracts features according to the adaptive standard of whether there are chronic underlying symptoms to ensure that the physical condition label corresponds to the previous symptom records; the conditioning demand segment node extracts features according to the urgency level of the frequency of the demand and the duration of discomfort to ensure that the demand label corresponds to the actual demand intensity; and the activity status segment node extracts features according to the dynamic range of limb movement amplitude and body displacement range to ensure that the activity label corresponds to the actual action status.

[0093] The feature ranges that conform to the extraction rules in each level are identified as target feature regions. For example, in the environmental parameter block node, the parameter range of 22-26℃ temperature and 40%-60% humidity is the target feature region, and in the user attribute block node, the keyword regions of 55 years old and sleep disorders are the target regions. The feature data in the target feature regions are then validated to remove abnormal data that exceeds the reasonable range. For example, environmental parameters such as temperature greater than 30℃ and humidity of 70%, or users with hypertension records in the basic layer but labeled as having normal physical condition, are all judged as abnormal. After the validation is passed, an initial scene label description is generated based on the extracted feature data of each node.

[0094] Retrieve historical conditioning data sample library from big data cloud platform, match the initial scene label description with the historical scene label, correct the deviation data in the initial label description, such as correcting the environmental 22-26℃ to the normal cabin temperature 22-26℃, and form a standardized scene feature description;

[0095] Based on the effective feature data of all scene nodes in the scene tree, several associated scene label descriptions are obtained;

[0096] Based on the feature extraction order of the several associated scene labels, a queue is established, the matching degree with historical valid scene labels is calculated, and the standardized scene feature description with the highest matching degree is selected as the final scene label of the current scene and stored in the big data cloud platform for subsequent initial adjustment parameter construction.

[0097] Specifically, the process of calculating the correlation between each scene block based on the type value includes:

[0098] Retrieve the type value for each scene block;

[0099] Generate a type value vector corresponding to each scene block using the type value of each scene block;

[0100] Retrieve the type value vector corresponding to the scene block for each data collection;

[0101] Obtain the L2 norm of the type value vector corresponding to each data collection based on the type value vector corresponding to each data collection in each scene segment;

[0102] The type fluctuation coefficient corresponding to each scene block is obtained based on the L2 norm of the type value vector determined by each data collection for each scene block.

[0103] The type fluctuation coefficient corresponding to each scene block is obtained by the following formula:

[0104]

[0105] Where B represents the type fluctuation coefficient corresponding to each scene block; n represents the total number of data collections already performed; L i L represents the L2 norm of the type value vector corresponding to the i-th data acquisition; c This represents the preset L2 norm reference value; L z L represents the median L2 norm of the type value vector corresponding to n data collections; b The L2 norm standard deviation of the type value vector corresponding to n data collections;

[0106] The correlation between different scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block.

[0107] Traditional methods (such as variance and standard deviation) can only roughly describe the overall dispersion of data and cannot accurately depict the "deviation relationship between a single data point and a reference value" or the "dynamic scaling law of fluctuations." In this embodiment, the molecule... Accurately capture the direction and magnitude of the deviation between the type value vector (L2 norm) of a single data acquisition and the preset reference value (positive deviation indicates exceeding the reference value, negative deviation indicates insufficient value);

[0108] The denominator passes through Dynamically select "standard deviation L" b (Reflecting the statistical characteristics of global fluctuations) or "the difference between a single deviation and the median value" Using the scaling benchmark (reflecting local deviation characteristics) as a scaling reference, it differentiates and accurately quantifies "small-scale fine fluctuations" and "large-scale significant fluctuations," solving the problem that traditional statistics "lack adaptability to handling fluctuations of different scales." The correlation between scene blocks depends not only on the similarity of type characteristics but also on the fluctuation stability of their respective data (e.g., the correlation between "highly fluctuating block A" and "stable block B," and the correlation between "stable block C" and "stable block D" need to be treated differently). Simultaneously, in this embodiment, the type fluctuation coefficient B, as a quantitative indicator of the "self-fluctuation stability" of the scene block, is introduced in the correlation calculation, making the correlation... This approach integrates information from two dimensions: "type feature similarity" and "fluctuation stability matching degree." This effectively improves the accuracy of correlation estimation. Furthermore, the accuracy of scene labels directly determines the adaptability of the conditioning chamber's control parameters. Through precise correlation calculation, it can more reliably distinguish between "scene blocks with genuine correlation" and "abnormal blocks without correlation" (such as type value mutation blocks caused by equipment failure), avoiding interference from abnormal blocks in scene tree construction. The generated scene labels are purer and more closely match the user's actual sub-health state and conditioning needs, providing a more accurate basis for subsequent adaptive control of parameters such as temperature, humidity, and posture in the conditioning chamber, thus improving the precision of control.

[0109] Specifically, the correlation between scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block, including:

[0110] Retrieve the type value vector for each scene block;

[0111] The cosine similarity between any two scene blocks is obtained using the type value vector of each scene block.

[0112] Retrieve the type fluctuation coefficient corresponding to each scene block;

[0113] The type fluctuation coefficients corresponding to each pair of scene blocks are coupled to obtain the type fluctuation coupling coefficients of each pair of scene blocks.

[0114] The type fluctuation coupling coefficient between every two scene blocks is obtained by the following formula:

[0115]

[0116] Where K represents the type fluctuation coupling coefficient between every two scene blocks; B 01 and B 02 These represent the type fluctuation coefficients corresponding to the two scene blocks, respectively.

[0117] The correlation between each pair of scene blocks is obtained by using the cosine similarity and type fluctuation coupling coefficient corresponding to each pair of scene blocks.

[0118] The correlation between each pair of scene blocks is obtained using the following formula:

[0119]

[0120] Where S represents the correlation between two scene blocks; C represents the cosine similarity between two scene blocks; and K represents the type fluctuation coupling coefficient between two scene blocks.

[0121] Traditional methods (such as using only cosine similarity) only focus on the feature similarity of type value vectors, ignoring the impact of the "individual fluctuation stability" of scene blocks on the association. In this embodiment, the above scheme uses K to quantify the "matching degree of fluctuation characteristics of two blocks," so that the association degree S simultaneously carries feature similarity and fluctuation characteristic matching degree, more comprehensively reflecting the real association logic between scene blocks (such as the association between "stable fluctuation of environmental parameters" and "stable change of user state" in sub-health conditioning, and the association between "large environmental fluctuations" and "large fluctuations of user state" need to be differentiated). At the same time, the type fluctuation coupling coefficient K is obtained through... A logarithmic function is used to smoothly and sensitively quantify the "difference in fluctuation coefficients between two blocks"—it increases slowly for small differences and accelerates for large differences, avoiding excessive impact of abrupt changes in difference on the correlation; the denominator is calculated using... Normalize the "overall fluctuation magnitude of the two blocks" to prevent large fluctuation blocks from distorting K due to their numerical scale, ensuring that K accurately captures the "fluctuation characteristic matching" rather than the "absolute value of fluctuation"; correlation. In the middle K, "fluctuation matching degree weighting" is applied to C - if the fluctuation characteristics of the two blocks match (K is large), S is significantly enhanced, and otherwise it is moderately weakened, so that the correlation is more in line with the dynamic characteristics of the actual scenario (such as the fluctuation coordination between user state and environmental parameters).

[0122] With a more comprehensive and accurate correlation degree S, the "real correlation" and "false / weak correlation" between scene blocks can be more reliably distinguished:

[0123] It can effectively filter out false associations that are "accidentally similar in features but do not match in fluctuations" (such as blocks with similar features but abnormal fluctuations caused by temporary equipment failure).

[0124] This makes the subsequent construction of the scenario tree closer to the user's actual sub-health state and conditioning needs, thereby allowing the conditioning cabin to more accurately adapt the control parameters (such as cabin temperature, humidity, user posture, etc.) generated based on scenario tags, and improve the effect of sub-health conditioning (such as for the precise scenario of "sleep needs + stable environmental fluctuations + semi-reclining posture", the control parameters are more in line with the user's physiological rhythm).

[0125] Meanwhile, the type fluctuation coefficient B itself is calculated based on "historical data collection" (reflecting dynamic fluctuation patterns), while the calculation of K and S further incorporates "dynamic coupling of fluctuation characteristics" into the correlation degree: it not only inherits the fluctuation statistical characteristics of historical data, but also responds in real time to the fluctuation changes of newly collected data, solving the problem of poor adaptability of traditional "static correlation methods" in the scenario of "dynamic collection and fluctuation changes" on big data cloud platforms, and providing technical support for the "real-time adaptive control" of the conditioning cabin (such as after a user enters the conditioning cabin, the fluctuations of the environment and their own state can be correlated in real time and fed back to the control logic).

[0126] In this embodiment, the user's basic data includes age, gender, type of sub-health symptoms, etc.

[0127] In this embodiment, the types of each scenario block are as follows: the environmental state is divided into two categories: normal and abnormal; the user characteristics are divided into two categories: basic and special; the conditioning needs are divided into three levels: mild, moderate and severe; and the activity intensity is divided into two levels: low and medium.

[0128] In this embodiment, the type value of each scene block is as follows:

[0129] The temperature range of 22-26℃ is mapped to the value "1", and the temperature range of 26-28℃ is mapped to "2".

[0130] Map a seated state to "1" and light activity to "2";

[0131] The need for sleep improvement is mapped to "1", fatigue relief to "2", and metabolic regulation to "3".

[0132] In this embodiment, normal environment is usually strongly correlated with light demand and low activity level, abnormal environment is strongly correlated with heavy demand and medium activity level, and normal environment is a contradictory combination with heavy demand and low activity level.

[0133] In this embodiment, the associated scene tag descriptions are as follows: around the normal environment, associated scene tag descriptions can be generated such as normal environment, normal physical condition of a 60-year-old male, mild need to relieve drowsiness, low activity level or normal environment, normal physical condition of a 60-year-old female, mild need to relieve drowsiness, and low activity level.

[0134] In this embodiment, the scenario classification unit divides scenarios into multiple dimensions, including the cabin environment, user needs, and activity status. A multi-dimensional feature matching algorithm ensures classification accuracy. Through scenario segmentation, numerical processing, correlation analysis, and scenario tree construction, key features are extracted layer by layer to generate standardized scenario labels. Historical data is also used to correct deviations, ensuring a high degree of consistency between the labels and actual scenarios. This provides accurate scenario data support for the cloud platform analysis module, avoiding deviations caused by vague scenario descriptions. It better matches the user's actual usage scenarios, calibrates user physiological parameters to eliminate individual difference errors, and makes the multi-source dataset output by the sub-health data collection module more accurate. This provides a precise scenario foundation for the subsequent construction of initial control parameters, solves the problem of data distortion caused by individual differences, and improves the adaptability of the conditioning cabin to different usage scenarios.

[0135] In this embodiment, the specific process of the conditioning scene adaptation module constructing the initial control parameters includes:

[0136] Based on the current scenario type and the final scenario label, retrieve the historical treatment data sample library for the corresponding scenario. The historical treatment data sample library contains the baseline range of user physiological parameters, treatment parameter configurations, and treatment effect data for users with different basic attributes in the same scenario.

[0137] The baseline values ​​of the current user's physiological parameters are determined by matching the mean of the calibrated physiological parameter data with the baseline range of the user's physiological parameters.

[0138] Set the data collection cycle for user physiological parameters and cabin operation status data according to the treatment need level corresponding to the scene label;

[0139] By integrating the baseline values ​​of physiological parameters with the data acquisition cycle and combining the segmented features of environmental parameters in the scene tags, the initial control parameters of the conditioning chamber are generated, and the association mapping relationship between the initial control parameters and scene type and scene tag is established.

[0140] In this embodiment, the specific process by which the feature processing module generates runtime feature data and physiological feature data includes:

[0141] Based on the set data acquisition cycle, each acquisition cycle is evenly divided into several consecutive time nodes, and each time node is synchronized with the acquisition cycle.

[0142] Two adjacent time points are selected to form a monitoring period, and the rate of change of the operational status data and the rate of change of the user's physiological parameters are calculated within each monitoring period.

[0143] The rate of change of operational status data and the rate of change of user physiological parameters for all monitoring periods within the same data collection cycle are summarized to generate operational characteristic data and physiological characteristic data.

[0144] In this embodiment, the running feature data and physiological feature data also include associated acquisition period ID, scene label, scene type and feature anomaly identifier, such as an abnormal fluctuation identifier where the rate of change exceeds ±20%.

[0145] In this embodiment, the parameter adjustment instruction generation process of the adaptive control module includes:

[0146] The target parameters in the conditioning plan, including the operating parameters of each actuator, the control parameters of the cabin environment, and the initial control parameters, are compared with the difference to obtain the parameter adjustment amount of each actuator. The adjustment level (fine-tuning level, normal level, and strong level) is then classified according to the absolute value of the parameter adjustment amount.

[0147] For different adjustment levels, a corresponding adjustment strategy is matched, and the adjustment strategy is converted into a drive signal (voltage, frequency, power adjustment signal) for the actuator, generating an executable parameter adjustment command.

[0148] In this embodiment, an absolute value of adjustment of <10% is the fine-tuning level, 10%-30% is the normal level, and >30% is the powerful level. The fine-tuning level uses PID closed-loop control to achieve smooth adjustment, the normal level uses segmented step-by-step adjustment to gradually approach the target parameter at preset time intervals, and the powerful level uses a composite adjustment method of fast adjustment followed by fine adjustment to quickly reduce parameter deviation and then stabilize parameter value through fine control.

[0149] In this embodiment, by generating initial control parameters and establishing a mapping relationship with the scenario, the initial parameters can accurately match the user's physiological state and scenario needs, avoiding blind parameter settings that are detached from historical and valid data. At the same time, the data collection cycle is set according to the level of treatment needs, ensuring the efficiency and effectiveness of data collection and laying a precise foundation for subsequent control. The adjustment levels are divided according to the absolute value of the adjustment amount and matched with differentiated adjustment strategies. The system can be flexibly controlled according to the importance of parameters and the adjustment range, avoiding sudden parameter changes from affecting the user experience. At the same time, it ensures that the operating parameters of the treatment chamber accurately match the needs of the treatment plan, solving the problem of insufficient adaptability between parameter adjustment and the actuator, and improving the overall accuracy and stability of the system's control.

[0150] In this embodiment, the big data cloud platform is also used to learn from the received evaluation data and feedback information, specifically including:

[0151] The quantitative indicators in the effect evaluation report are associated and stored with the corresponding treatment plan, scene tag, and scene type. The historical treatment data sample library is updated, and treatment effect cases under the same scene are added.

[0152] Time-series analysis was performed on multiple treatment data of the same user to extract the trend characteristics of the user's physiological indicators, and the treatment plan for the user was optimized based on the trend characteristics.

[0153] The transfer learning algorithm is used to transfer conditioning data from new scenarios that are not included in the historical conditioning data sample library to the training of the sub-health conditioning adaptive regulation model in similar scenarios, thereby supplementing the model training samples and improving the model's adaptability to new scenarios.

[0154] In this embodiment, the big data cloud platform achieves self-learning through associated storage, time series analysis, and transfer learning. It supplements existing conditioning solutions with cases in the same scenario, optimizes exclusive solutions for individual users, expands the adaptability of the sub-health conditioning adaptive regulation model to new scenarios, and enables the system to be continuously upgraded as it is used, continuously improving the conditioning effect and scenario adaptability, and extending the system life cycle.

[0155] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive control system for a sub-health conditioning cabin based on a big data cloud platform, characterized in that, This includes a sub-health data collection module, a treatment scenario adaptation module, a cloud platform analysis module, an adaptive control module, and a treatment effect feedback module. These modules, along with the sub-health data collection module, cloud platform analysis module, adaptive control module, and treatment effect feedback module, construct a real-time data interaction network through a big data cloud platform, forming an adaptive control link for the sub-health treatment cabin. The sub-health data acquisition module is used to collect real-time physiological parameters and operational status data of users in the sub-health conditioning cabin. It preprocesses the collected physiological parameters and operational status data to generate multi-source datasets. The conditioning scenario adaptation module is used to divide the usage scenarios based on the user's conditioning needs, activity status and cabin environment conditions, mark the corresponding scenario tags, and construct the initial control parameters of the sub-health conditioning cabin by combining the historical conditioning data sample library in the big data cloud platform. The cloud platform analysis module is used to extract the operational and physiological characteristics data of multi-source datasets and input them into the constructed sub-health conditioning adaptive regulation model for conditioning effect diagnosis. Based on the diagnostic results of the sub-health conditioning adaptive regulation model, the corresponding conditioning plan is generated. The adaptive control module is used to match the corresponding parameter adjustment instructions based on the conditioning plan and the initial control parameters of the sub-health conditioning cabin. The sub-health conditioning cabin receives the parameter adjustment instructions and controls the corresponding actuators to perform the control operation. The treatment effect feedback module is used to collect users' physiological feedback data during and after the treatment, compare it with the baseline value before treatment and issue abnormal warnings, generate an effect evaluation report, and transmit the evaluation data and feedback information to the big data cloud platform. The conditioning scenario adaptation module includes: The scenario classification unit is used to acquire cabin environment data and user input of target conditioning needs and target activity status, divide the usage scenarios into static conditioning scenarios, dynamic intervention scenarios and emergency relief scenarios, and output the initial scenario classification results. The scene label construction unit is used to extract scene feature data corresponding to the scene classification result based on the initial scene classification result and the historical conditioning data sample library, and generate the corresponding scene label; The data calibration unit is used to retrieve the standard physiological parameter range that is the same as the scene label in the historical conditioning data sample library, calculate the deviation between the user physiological parameters in the multi-source dataset and the standard physiological parameter range, perform calibration processing on the user physiological parameters in the multi-source dataset, and output the calibrated physiological parameter data.

2. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 1, characterized in that, The sub-health data acquisition module integrates an operating status acquisition sensor and a user physiological parameter acquisition sensor. The operating status acquisition sensor is used to acquire data on cabin temperature, humidity, oxygen concentration, far-infrared radiation intensity, magnetic flux density, negative oxygen ion concentration, and cabin temperature uniformity. The user physiological parameter acquisition sensor is used to acquire data on the user's heart rate, heart rate variability, blood oxygen saturation, skin resistance, body surface temperature distribution, microcirculation velocity, and brain wave rhythm.

3. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 2, characterized in that, The specific process of generating corresponding scene tags by the scene tag building unit includes: The current scene corresponding to the initial scene classification result is analyzed to determine the scene data, which includes cabin environment data, user basic data, conditioning demand data and activity status data. The scene data is divided into multiple scene blocks according to scene category, including environmental parameter blocks, user attribute blocks, conditioning demand blocks, and activity status blocks; Obtain the type of each scene block and perform numerical processing to obtain the type value of each scene block; Calculate the correlation degree between each scene block based on the type value, retain the related scene blocks with a correlation degree less than the preset correlation degree threshold, and remove the abnormal blocks with no correlation. A scene tree is constructed based on the correlation between each related scene block and each scene block, where each scene block corresponds to a scene node in the scene tree. Key features are extracted from the scene nodes in the scene tree from top to bottom, and the extraction method of object data included in the scene nodes is determined according to the type of scene node; The feature range that meets the extraction rules in each level is determined as the target feature region, and the feature data in the target feature region is validated. After the validation is passed, an initial scene label description is generated based on the extracted node feature data. Retrieve historical processing data sample library from the big data cloud platform, perform similarity matching between the initial scene label description and the historical scene label, correct the deviation data in the initial label description, and form a standardized scene feature description; Based on the effective feature data of all scene nodes in the scene tree, several associated scene label descriptions are obtained; Based on the feature extraction order of the several associated scene tags, a queue is established, the matching degree with historical valid scene tags is calculated, and the standardized scene feature description with the highest matching degree is selected as the final scene tag of the current scene and stored in the big data cloud platform.

4. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 3, characterized in that, The specific process of calculating the correlation between each scene block based on the type value includes: Retrieve the type value for each scene block; Generate a type value vector corresponding to each scene block using the type value of each scene block; Retrieve the type value vector corresponding to the scene block for each data collection; Obtain the L2 norm of the type value vector corresponding to each data collection based on the type value vector corresponding to each data collection in each scene segment; The type fluctuation coefficient corresponding to each scene block is obtained based on the L2 norm of the type value vector determined by each data collection for each scene block. The correlation between different scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block.

5. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 4, characterized in that, The correlation between scene blocks is obtained by combining the type value vector of each scene block with the type fluctuation coefficient corresponding to each scene block, including: Retrieve the type value vector for each scene block; The cosine similarity between any two scene blocks is obtained using the type value vector of each scene block. Retrieve the type fluctuation coefficient corresponding to each scene block; The type fluctuation coefficients corresponding to each pair of scene blocks are coupled to obtain the type fluctuation coupling coefficients of each pair of scene blocks. The correlation between each pair of scene blocks is obtained by using the cosine similarity and type fluctuation coupling coefficient corresponding to each pair of scene blocks.

6. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 5, characterized in that, The specific process of constructing the initial control parameters for the conditioning scenario adaptation module includes: Based on the current scenario type and the final scenario label, retrieve the historical treatment data sample library for the corresponding scenario. The historical treatment data sample library contains the baseline range of user physiological parameters, treatment parameter configurations, and treatment effect data for users with different basic attributes in the same scenario. The baseline values ​​of the current user's physiological parameters are determined by matching the mean of the calibrated physiological parameter data with the baseline range of the user's physiological parameters. Set the data collection cycle for user physiological parameters and cabin operation status data according to the treatment need level corresponding to the scene label; By integrating the baseline values ​​of physiological parameters with the data acquisition cycle and combining the segmented features of environmental parameters in the scene tags, the initial control parameters of the conditioning chamber are generated, and the association mapping relationship between the initial control parameters and scene type and scene tag is established.

7. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 6, characterized in that, The specific process for generating the operational characteristic data and physiological characteristic data includes: Based on the set data acquisition cycle, each acquisition cycle is evenly divided into several consecutive time nodes, and each time node is synchronized with the acquisition cycle. Two adjacent time points are selected to form a monitoring period, and the rate of change of the operational status data and the rate of change of the user's physiological parameters are calculated within each monitoring period. The rate of change of operational status data and the rate of change of user physiological parameters for all monitoring periods within the same data collection cycle are summarized to generate operational characteristic data and physiological characteristic data.

8. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 7, characterized in that, The parameter adjustment instruction generation process of the adaptive control module includes: The difference between the target parameters and the initial control parameters in the conditioning plan is calculated to obtain the parameter adjustment amount for each actuator, and the adjustment level is divided according to the absolute value of the parameter adjustment amount. For different adjustment levels, a corresponding adjustment strategy is matched, and the adjustment strategy is converted into a drive signal for the actuator to generate an executable parameter adjustment instruction.

9. The adaptive control system for a sub-health conditioning cabin based on a big data cloud platform as described in claim 8, characterized in that, The big data cloud platform is also used to learn from the received evaluation data and feedback information, specifically including: The quantitative indicators in the effect evaluation report are associated and stored with the corresponding treatment plan, scene tag, and scene type. The historical treatment data sample library is updated, and treatment effect cases under the same scene are added. Time-series analysis was performed on multiple treatment data of the same user to extract the trend characteristics of the user's physiological indicators, and the treatment plan for the user was optimized based on the trend characteristics. Migrate conditioning data from new scenarios that are not included in the historical conditioning data sample library to train the sub-health conditioning adaptive regulation model in similar scenarios, thus supplementing the model training samples.

Images