Dynamic regulation and control method and device for multi-dimensional physiological data, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as financial science and technology and medical health, and discloses a dynamic regulation and control method, device and equipment for multi-dimensional physiological data and a medium, and the method comprises the steps: obtaining real-time physiological data of at least two different dimensions from a regulated and controlled object, and recording subjective sign data; performing fusion processing on the real-time physiological data and the subjective sign data to generate fusion data with aligned time sequences; constructing a dynamic association model based on the fusion data, and determining a state vector representing the comprehensive physiological state of the regulated and controlled object; generating a dynamic intervention scheme through a decision network based on the state vector; and obtaining physiological state feedback data after the intervention scheme, and updating the decision network. According to the method, the multi-dimensional real-time physiological data and the subjective sign data are fused, the dynamic correlation model and the decision network are combined, a closed-loop regulation and control mechanism is formed, and the accuracy and the real-time performance of physiological state regulation and control are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a multi-dimensional physiological data dynamic regulation method and device, equipment and storage medium. BACKGROUND

[0002] In the field of medical and health services, the existing technology mainly relies on single type of biological data acquisition and static index analysis, generally adopts methods such as flora detection, single collection of metabolites or questionnaire survey of physical signs, and lacks the ability of collaborative analysis of multi-source and multi-dimensional real-time physiological data. Different data sources are inconsistent in acquisition time, monitoring frequency and data structure, resulting in lag and one-sidedness in physiological state evaluation, which makes it difficult to fully and accurately reflect the real comprehensive physiological state of the regulated object, and limits the scientificity and timeliness of personalized health management and dynamic intervention.

[0003] In the field of financial technology, there are problems such as isolation of health data, lag in updating and insufficient accuracy of risk identification in health insurance, health finance and health rights and interests products. The existing risk control and health management strategies are mostly based on single-dimensional health records or static health scores, lacking dynamic state perception and real-time risk adjustment mechanism based on multi-modal physiological data. Different health data lack fusion processing means, which cannot support dynamic health risk identification and generation of personalized health rights and interests scheme, reducing the overall effect of health-related financial products in accurate pricing, dynamic risk control and user incentive management.

[0004] From the overall technical system, the existing multi-modal biological data collaborative acquisition, fusion processing and dynamic state modeling capability are weak, there is a lack of effective dynamic correlation and real-time linkage between physiological state changes and intervention strategies, and the structured results reflecting the comprehensive physiological state cannot be accurately generated based on multi-source real-time data, which limits the intelligent, real-time and personalized level of health management, risk control and dynamic regulation scheme. SUMMARY

[0005] The main purpose of the present application is to provide a multi-dimensional physiological data dynamic regulation method, device, equipment and storage medium, which aims to solve the technical problem that the existing technology lacks the ability of fusion processing based on multi-dimensional real-time physiological data and subjective physical sign data, and constructing a dynamic correlation model to accurately reflect the comprehensive physiological state of the regulated object.

[0006] To achieve the above-mentioned purpose, the present application provides a multi-dimensional physiological data dynamic regulation method, comprising:

[0007] acquiring at least two real-time physiological data from different dimensions representing physiological state from the regulated object, and recording the subjective physical sign data of the regulated object;

[0008] fuse the real-time physiological data and the subjective sign data to generate fusion data in time sequence alignment;

[0009] construct a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data;

[0010] determine a state vector representing a current comprehensive physiological state of the regulated object according to the dynamic correlation model;

[0011] generate a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network based on the state vector;

[0012] acquire physiological state feedback data after the dynamic intervention scheme is executed, and update the decision network according to the physiological state feedback data.

[0013] Further, to achieve the above object, the present application provides a dynamic regulation device for multi-dimensional physiological data, comprising:

[0014] a data acquisition module configured to acquire at least two real-time physiological data from different dimensions representing physiological states from a regulated object, and record subjective sign data of the regulated object;

[0015] a data fusion module configured to fuse the real-time physiological data and the subjective sign data to generate fusion data in time sequence alignment;

[0016] a correlation modeling module configured to construct a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data;

[0017] a state evaluation module configured to determine a state vector representing a current comprehensive physiological state of the regulated object according to the dynamic correlation model;

[0018] an intervention generation module configured to generate a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network based on the state vector;

[0019] a feedback optimization module configured to acquire physiological state feedback data after the dynamic intervention scheme is executed, and update the decision network according to the physiological state feedback data.

[0020] Further, to achieve the above object, the present application also provides a computer device comprising a memory, a processor, and a dynamic regulation program for multi-dimensional physiological data stored on the memory and executable on the processor, wherein the dynamic regulation program for multi-dimensional physiological data, when executed by the processor, implements the steps of the dynamic regulation method for multi-dimensional physiological data as described above.

[0021] Further, to achieve the above object, the present application also provides a computer readable storage medium, wherein the storage medium stores a multi-dimensional physiological data dynamic regulation program, and the multi-dimensional physiological data dynamic regulation program implements the steps of the multi-dimensional physiological data dynamic regulation method when executed by a processor.

[0022] Beneficial effects: The present application relates to the field of artificial intelligence technology, and can be applied to business scenarios such as financial technology and medical health, and discloses a multi-dimensional physiological data dynamic regulation method, device, equipment and medium, which comprises the following steps: acquiring at least two kinds of real-time physiological data from different dimensions representing physiological states from a regulated object, and recording subjective sign data of the regulated object; performing fusion processing on the real-time physiological data and the subjective sign data to generate time sequence aligned fusion data; constructing a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data; determining a state vector representing the current comprehensive physiological state of the regulated object according to the dynamic correlation model; generating a dynamic intervention scheme for adjusting the physiological state of the regulated object based on the state vector through a decision network; acquiring physiological state feedback data after the dynamic intervention scheme is executed, and updating the decision network according to the physiological state feedback data. The present application fuses multiple real-time physiological data and subjective sign data of different dimensions, constructs a dynamic correlation model, reflects the comprehensive physiological state of the regulated object in real time, generates a dynamic intervention scheme in combination with a decision network, and updates the decision network through physiological state feedback data, thereby forming a closed loop of data perception, state analysis and intervention optimization, and improving the accuracy and timeliness of physiological state regulation. BRIEF DESCRIPTION OF DRAWINGS

[0023] The present application will be further described below in combination with the drawings and embodiments, wherein:

[0024] Figure 1 An application environment schematic diagram of the multi-dimensional physiological data dynamic regulation method in an embodiment of the present application;

[0025] Figure 2 A flow schematic diagram of the multi-dimensional physiological data dynamic regulation method in an embodiment of the present application;

[0026] Figure 3 A functional module schematic diagram of the multi-dimensional physiological data dynamic regulation device in a preferred embodiment of the present application;

[0027] Figure 4 A structure schematic diagram of the computer device in an embodiment of the present application;

[0028] Figure 5 Another structure schematic diagram of the computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0029] It is to be understood that the specific embodiments described herein are merely illustrative of the present application and do not limit the scope of the application.

[0030] The multi-dimension physiological data dynamic regulation method provided by the embodiments of the present application can be applied in application environments such as Figure 1 , wherein a user terminal communicates with a server through a network. The server can acquire at least two kinds of real-time physiological data from different dimensions representing physiological states of a regulated object through the user terminal, and record subjective sign data of the regulated object; perform fusion processing on the real-time physiological data and the subjective sign data to generate time-series aligned fusion data; construct a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data; determine a state vector representing a current comprehensive physiological state of the regulated object according to the dynamic correlation model; generate a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network based on the state vector; acquire physiological state feedback data after the dynamic intervention scheme is executed, and update the decision network according to the physiological state feedback data. The present application reflects the comprehensive physiological state of the regulated object in real time by fusing multiple kinds of real-time physiological data and subjective sign data from different dimensions, constructing a dynamic correlation model, generating a dynamic intervention scheme through a decision network, and updating the decision network through physiological state feedback data, thereby forming a closed loop of data perception, state analysis and intervention optimization, and improving the accuracy and timeliness of physiological state regulation. The user terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers. The present application will be described in detail through specific embodiments.

[0031] Please refer to Figure 2 , Figure 2 for a flowchart of an embodiment of the multi-dimension physiological data dynamic regulation method provided by the present application. It should be noted that although a logical sequence is shown in the flowchart, the steps shown or described herein can be executed in a different order in some cases.

[0032] As shown in Figure 2 , the multi-dimension physiological data dynamic regulation method proposed by the present application includes the following steps:

[0033] S10, acquiring at least two kinds of real-time physiological data from different dimensions representing physiological states of a regulated object, and recording subjective sign data of the regulated object;

[0034] In this embodiment, the process of acquiring at least two real-time physiological data from different dimensions representing the physiological state of the regulated object refers to the simultaneous or short-interval collection of multi-source data reflecting the physiological state of the regulated object using multiple independent detection dimensions or monitoring channels. The sources of these data are not limited to biological fluids, tissue samples, external physiological sensing parameters, etc., and the key is that the data dimensions involved are independent of each other and the mechanisms reflecting the physiological state are different. For example, one dimension can obtain information on the concentration change of metabolites in the body through sweat detection, another dimension can analyze the microbial community structure through fecal sample analysis, and further can be extended to respiratory gas component monitoring, blood biochemical index analysis, skin electrical response monitoring, heart rate variability detection, etc. As long as different data dimensions can provide information revealing the individual state from different physiological perspectives. Real-time physiological data refers to physiological parameter data with continuity, short cycle or immediacy, which can reflect the dynamic change process of the physiological state of the regulated object. The meaning of real-time is not limited to strict millisecond-level data acquisition, and can include second-level, minute-level or hour-level data acquisition frequency according to the physiological parameter change speed in the specific application scenario. For example, the acquisition frequency of sweat metabolite data can be adjusted according to the sensor performance and the rhythm of metabolite secretion, the fecal microbial data can be updated in real time based on the excretion cycle, the respiratory component data can be continuously monitored at a frequency of seconds by a portable detection device, and the blood glucose or lactic acid index can be dynamically acquired at a high frequency by an implanted device.

[0035] The process of recording the subjective sign data of the regulated object mainly involves actively collecting individual self-perception, physiological discomfort, symptom expression, and other non-objective measurement data through interactive devices. Subjective sign data includes but is not limited to pain score, fatigue level, gastrointestinal discomfort degree, sleep quality self-evaluation, anxiety level, etc. These data are derived from the active feedback information of the regulated object based on actual feelings. The recording form can adopt various human-computer interaction channels such as smart phones, tablet terminals, wearable devices, voice assistants, etc., combined with standardized electronic questionnaires, sliding bar scoring, voice recognition input, expression selection, text input, etc. Multiple feedback methods to ensure the operability and standardization of the acquisition of subjective sign data. For example, the pain level can be quantified by visual analog scale (VAS), the abdominal distension degree can be obtained by combining the simplified symptom score tool, the sleep state can be filled in with the self-sleep quality evaluation form, and the psychological stress can be presented by digital score or multiple-choice psychological evaluation tool. The above process is expandable and not limited to specific devices or data types, and is suitable for collecting subjective sign information in various scenarios.

[0036] In the specific implementation process, obtaining real-time physiological data from at least two different dimensions of the regulated object can be achieved in the following way. First, the wearable sweat sensor is attached to the skin surface of the regulated object, and the micro-sweat is collected through the microfluidic chip. Combined with the electrochemical detection module, the content change of the metabolite index in the sweat, especially the content change of tryptophan derivative, is monitored in real time. Second, through the intelligent excrement monitoring equipment with microbial spectrum analysis capability, the abundance of bacteria in the excrement sample of the regulated object is analyzed in real time, and the fecal microbial community structure characteristics are extracted based on high-throughput sequencing or spectrum fingerprint technology. In addition, the metabolomics information in the breath sample can be collected by a portable breath analyzer, and the blood biochemical parameters or tissue metabolic indicators can be continuously obtained by an implantable physiological monitoring device. The combination of these data dimensions can be flexibly adjusted according to specific needs to ensure that different physiological state information can be captured in multiple dimensions in real time.

[0037] In the specific implementation process of recording subjective sign data, the regulated object's smart terminal device can push a standardized questionnaire or voice interaction task to regularly guide the object to feedback its perceived physiological state. For example, the system can regularly push an electronic form containing fatigue perception, digestive discomfort, psychological emotion, etc. dimensions to the object every day, and the object can complete the information filling by sliding bar or clicking options, and the system automatically records the corresponding score data. In combination with the touch screen built-in wearable device, the object can be guided to actively feedback specific subjective sign information before and after exercise, diet, and sleep, forming a standardized and structured data recording process. In order to improve the credibility of subjective data, multiple repeated measurements, data trend analysis or cross-validation mechanism combined with objective physiological data can be introduced in the implementation process to further optimize the effectiveness of subjective sign data.

[0038] Example: In the medical and health business field, for individuals with chronic intestinal function disorders, through the wearable sweat metabolite monitoring device, combined with real-time analysis of fecal microbial community, the individual's intestinal metabolism and microbial balance state data are dynamically obtained, supplemented by regularly recorded subjective sign information such as abdominal distension, abdominal pain perception, and defecation comfort, to comprehensively evaluate the intestinal health level and provide real-time physiological basis for precise dietary or drug intervention programs.

[0039] In the financial technology business field, combined with employee health management in high-intensity work environment, the physiological stress state of employees is dynamically mastered through real-time monitoring of their sweat metabolite indicators and respiratory gas metabolomics information, and the employees are guided to regularly feedback subjective information such as fatigue perception, sleep quality, and anxiety emotion, to establish an employee health risk data model for enterprises, and assist financial institutions in implementing customized health intervention measures to reduce production efficiency fluctuations and labor risks caused by health problems.

[0040] The embodiment can break through the limitations of physiological state monitoring under a single data source by synchronously acquiring real-time physiological data of multiple different dimensions and comprehensively recording subjective sign data, and can improve the perception comprehensiveness and accuracy of the physiological change process of the regulated object. The multi-dimensional data source constructs a complete physiological state information chain, ensures that the logical correlation between physiological data and subjective perception is presented, and avoids deviation of the regulation scheme caused by a single detection result. Based on the dynamic monitoring of the physiological state change of the regulated object, the process forms multi-element, real-time and systematic data support, which significantly improves the scientificity and adaptability of the subsequent regulation process.

[0041] S20, fusing the real-time physiological data and the subjective sign data to generate time-series aligned fusion data;

[0042] In the embodiment, the real-time physiological data and the subjective sign data are fused, specifically, based on time, data content or source dimension, data of different sources but representing physiological states of the regulated object are uniformly integrated and processed and analyzed to form joint data results having logical corresponding relationship in the same time frame. The real-time physiological data refers to physiological parameter data obtained from different detection channels, having dynamic and time-effectiveness, including but not limited to metabolite indicators, microbiome information, physiological signal parameters, etc. These data are collected on a continuous time axis by respective monitoring devices, having different sampling periods, data resolutions and time stamp information. The subjective sign data refers to information actively fed back by the regulated object related to its own physiological perception, covering structured or semi-structured data content such as symptom expression, perception score, sign level, etc. These data are usually obtained from the input of the regulated object based on standardized evaluation tools, having characteristics such as non-fixed acquisition frequency and relatively subjective time point.

[0043] The fusion processing refers to that, through various data preprocessing techniques, the time stamp information of different data sources is first compared and corrected to eliminate time errors caused by different equipment synchronization, data loss or subjective delay. On this basis, data interpolation, time window sliding, rhythm alignment and other means are used to realize the structural integration of different source data under a unified time reference framework. The process is not limited to simple time series splicing, but also includes logical mapping and physiological correlation analysis of data content, to ensure that the final fusion data is continuous in time, unified in structure and has multi-dimensional information expression capability in content.

[0044] The time-aligned fusion data refers to a data set based on a unified time coordinate system formed after the above fusion processing, wherein each time point contains complete data segments from different data sources, and can synchronously reflect the multi-dimensional physiological state of the regulated object at the time point. The fusion data has the characteristics of time sequence continuity, data structure standardization and information expression integrity, and is the basic data input for subsequent analysis, modeling and intervention decision-making.

[0045] In actual operation, the fusion processing of real-time physiological data and subjective sign data can be realized through the following steps. First, real-time physiological data is collected through technical means, including sweat metabolite data, fecal microbial data, respiratory metabolism information, blood physiological parameters, etc., ensuring that each data source is accompanied by accurate timestamp information. At the same time, the intelligent terminal device is used to guide the regulated object to input the subjective sign data, and record the time point and specific content of each input. Subsequently, for the time error existing in each data source, a dynamic time compensation algorithm is used to correct and synchronize different data sequences on the time axis. Specifically, based on the standardized time reference signal, multi-element data interpolation, sliding window statistics and other technologies can be used to fill in the missing data, balance the data granularity difference, and realize the unification of the data time dimension.

[0046] Further, to improve the data integration effect, the compensated sweat metabolite data and fecal microbial data can be mapped to the physiological rhythm coordinate system, i.e. according to the circadian rhythm, physiological peak or specific time window, adjusting the data time axis, so that the change trend reflected by different physiological parameters is aligned under the rhythm framework. Finally, the mapped physiological data of various types and subjective sign data are structurally combined to generate time-aligned fusion data that meets the requirements. The data can adopt a multi-dimensional matrix, a data cube or other data structure forms suitable for multi-source heterogeneous data integration, ensuring that the data has good accessibility, scalability and analysis adaptability.

[0047] Example: In the medical health business field, for individuals with intestinal microecological imbalance or metabolic abnormalities, by real-time acquisition of the change of tryptophan metabolite ratio in sweat and the information of genus abundance in fecal samples, combined with the subjective sign information such as the degree of abdominal distension and the level of defecation discomfort actively fed back by the regulated object every day, the time compensation, rhythm mapping and structure integration technology are used to generate fusion data with time alignment characteristics. This data provides a high-credibility data basis for subsequent dynamic correlation modeling, state anomaly identification and intervention strategy formulation.

[0048] In the field of financial technology business, around the health risk monitoring of employees in a high-intensity work environment, real-time collection of employee sweat metabolic indicators, respiratory metabolomics parameters and blood glucose dynamic change data, while guiding employees to regularly feedback fatigue perception, sleep quality, psychological stress level, through fusion processing to form a unified time sequence alignment fusion data, to build a data foundation for enterprise management, to support real-time evaluation of employee health status and development of dynamic optimization strategy, to reduce the instability of labor and loss of production efficiency caused by health risks.

[0049] The embodiment can break the information island problem caused by fragmented data sources by fusion processing of real-time physiological data and subjective sign data, and establish a complete, continuous and dynamic data chain reflecting the physiological state of the controlled object. The fusion data has strict alignment standards in the time dimension, ensuring the logical consistency and expression integrity of multiple data within the same time frame. Compared with single-dimensional data or data that has not been fused, the time sequence aligned fusion data can improve the accuracy and timeliness of subsequent dynamic modeling, state recognition and intervention decision-making, reduce analysis bias caused by inconsistent data time and content mismatch, and significantly enhance the scientificity and practicality of the overall physiological state regulation system.

[0050] S30, based on the fusion data, a dynamic correlation model between the real-time physiological data and the subjective sign data is constructed;

[0051] In the embodiment, based on the fusion data, a dynamic correlation model between the real-time physiological data and the subjective sign data is constructed, specifically, after completing the time alignment and structure integration of different source data, further through the process of data organization, information extraction and correlation rule learning, a mathematical expression structure is established which can dynamically reflect the interaction relationship and change pattern between the real-time physiological data and the subjective sign data. The fusion data refers to the data set formed after time correction, rhythm mapping and structure merging processing, which has a unified time frame and multi-dimensional content expression ability, including but not limited to metabolite indicators, microbial community information and sign score data, which can comprehensively describe the multi-dimensional physiological state of the controlled object at different times.

[0052] The dynamic correlation model refers to the use of multi-dimensional structured organization, feature extraction and cross-dimensional correlation rule learning to generate a data structure or model expression form that can reflect the mutual dependence relationship and dynamic evolution law between the metabolite dimension, microbial dimension and time dimension. Specifically, the following key steps are included:

[0053] Firstly, the fusion data is restructured according to the metabolite dimension, microorganism dimension and time dimension to form a three-dimensional data structure. The metabolite dimension corresponds to the sweat metabolite data, respiratory metabolism parameters and other indicators reflecting the in vivo metabolic state. The microorganism dimension corresponds to the fecal microorganism data, intestinal flora abundance and diversity information. The time dimension corresponds to the continuous time sequence of data collection. The three-dimensional data structure can be in the form of multi-dimensional array, tensor or other forms that can express the spatial structure of multi-source heterogeneous data, ensuring the integrity and consistency of the data in each dimension.

[0054] Secondly, based on the three-dimensional data structure, multivariate statistical analysis, matrix decomposition, tensor decomposition or machine learning methods are applied to extract core factors containing metabolite-microorganism-time correlation characteristics. This process selects key parameters or feature combinations that can effectively reflect the internal relationship between dimensions through mathematical dimension reduction, feature aggregation or pattern recognition techniques, reduces data redundancy, and improves information expression efficiency and analysis accuracy. The core factors represent a set of representative features in the data structure and reflect the dynamic interaction information between metabolic state, microbial ecology and time variation in the physiological logic.

[0055] Further, based on the extracted core factors, a cross-dimension association rule library is constructed to clarify the statistical association, dynamic dependency and potential causal structure between different dimension features. The association rule library can be established through data mining, pattern recognition or deep learning, and has attributes such as expression strength, directionality and time dynamic characteristics, which can dynamically reflect the change trend and coordination pattern between various indicators in the physiological state evolution process of the controlled object.

[0056] Finally, the association rule library is defined as a dynamic association model to form an overall data expression and relationship reasoning framework. The dynamic association model not only has the characteristics of strong structure and rich expression dimension, but also supports the introduction of new data, dynamically adjusts internal parameters based on the preset weight update mechanism, continuously optimizes the expression ability and adaptability of the model, and ensures that the model can accurately reflect the dynamic association relationship between real-time physiological data and subjective sign data for a long time, stably and accurately.

[0057] In actual operation, multiple technical routes and implementation means can be adopted to construct the dynamic correlation model. For example, for the medical health business scenario, the collected fusion data includes sweat metabolite data, fecal microorganism data, and subjective sign score information. First, the data from different sources is reorganized into a three-dimensional data matrix based on the time series standard, the metabolite dimension includes KYN / 5-HT ratio and CRP level, the microorganism dimension includes genus abundance and intestinal diversity index, and the time dimension is divided according to hours or days. Subsequently, by applying Tucker decomposition, principal component analysis or other matrix decomposition techniques, the core factor combination with a singular value greater than a certain threshold is extracted, which represents the dominant change pattern and time-dependent characteristics between the metabolite and microorganism dimensions.

[0058] Further, by using the Apriori algorithm, structural equation model or deep neural network, a cross-dimension correlation rule library is constructed to determine the statistical correlation between different metabolite indicators and microbial characteristics, the change trend and its influence law on the subjective sign data. The rule library is updated in real time by a dynamic weight updating mechanism, combined with newly collected data to adjust the correlation strength and dependence pattern, ensuring that the model always fits the actual physiological state changes of the regulated object.

[0059] In the financial technology business field, for the health monitoring of employees in a high-intensity work environment, the collected fusion data includes sweat metabolite indicators, blood glucose dynamic changes, subjective fatigue perception scores and other information. Through the same data structure reorganization and correlation model construction process, a model framework is generated that can reflect the dynamic relationship between work load, metabolic stress and sign changes, assisting in realizing real-time monitoring and trend warning of employee health risks.

[0060] The embodiment can break through the limitations of traditional static physiological state analysis by constructing a dynamic correlation model based on fusion data, and realize unified modeling and dynamic correlation expression of multi-dimensional, multi-source physiological data and subjective sign information. Compared with the way of not establishing such a model, the dynamic correlation model has the advantages of strong expression ability, high dynamic adaptability, and excellent real-time updating effect, can continuously, accurately and comprehensively reflect the multi-dimensional data relationship of the regulated object in the actual physiological state change process, improve the scientificity and pertinence of subsequent state recognition and intervention strategy formulation, and enhance the stability, adaptability and intelligent level of the overall regulation system.

[0061] S40, determining a state vector representing the current comprehensive physiological state of the regulated object according to the dynamic correlation model;

[0062] In this embodiment, the state vector representing the current comprehensive physiological state of the regulated object is determined according to the dynamic correlation model. Specifically, based on the dynamic correlation model constructed and updated in real time, a multi-dimensional vector expression that can comprehensively reflect the overall physiological state of the regulated object is generated through correlation rule reasoning and feature combination extraction for the multi-source fusion data at the current time. The dynamic correlation model is a mathematical expression framework established by multi-dimensional data structure organization, core factor extraction and cross-dimensional association rule learning in the previous step, which contains the dynamic dependence relationship and change pattern between metabolites, microorganisms and time dimension.

[0063] The state vector is a set of physiological state quantitative results extracted based on the current time fusion data under the reasoning logic of the dynamic correlation model, including but not limited to metabolite state features, microorganism state features and sign state features, which comprehensively reflect the comprehensive physiological state of the regulated object at a specific time node. The state vector has the characteristics of clear structure, complete information and strong expression ability, which supports the formulation of subsequent control strategies and the generation of intervention actions.

[0064] The specific implementation process includes the following contents: First, extract the metabolite-microorganism-time correlation features at the current time from the dynamic correlation model. The correlation features reflect the comprehensive relationship structure between metabolite indicators, microbial community characteristics and subjective sign information under the background of time dynamic change, which are derived from the core factors and association rule library extracted in the previous step, and have the properties of cross-dimension, dynamic and quantitative expression.

[0065] Second, the metabolite-microorganism-time correlation features are further decomposed into metabolite state features, microorganism state features and sign state features. Metabolite state features refer to a set of quantitative indicators that can reflect the metabolic level, internal environment homeostasis or metabolic function state of the regulated object, for example, including sweat tryptophan metabolite ratio, KYN / 5-HT index, lactic acid content, etc. Microorganism state features refer to a set of indicators that reflect the community structure, functional diversity and ecological balance state of the intestinal or other microecological system, such as genus abundance, diversity index, functional flora ratio, etc. The sign state feature is a set of indicators that reflect the physiological perception, symptom manifestation or subjective sign change of the regulated object, which is derived from the symptom quantitative score of the user interaction interface, the sign evaluation result, etc.

[0066] Finally, the metabolite state features, microorganism state features and sign state features are integrated into a state feature set according to the preset combination rule, and the state feature set is defined as a state vector to form a complete physiological state expression output. The state vector structure usually adopts an ordered numerical array, a matrix form or other data structure form that can support machine learning and decision reasoning, to ensure that the subsequent system can efficiently receive, process and apply the physiological state expression result.

[0067] In practical applications, determining the state vector can be based on different types of data sources and flexible adjustment according to scene requirements. For example, in the medical health business field, the fusion data includes sweat metabolite data, fecal microorganism data, and subjective sign score information. Through dynamic correlation model reasoning, the tryptophan metabolic pathway feature, the intestinal flora ecological state feature, and the symptom quantitative index at the current time are extracted, respectively, to generate metabolite state features such as KYN / 5-HT ratio, microorganism state features such as genus abundance distribution, and sign state features such as diarrhea frequency score, and finally combine to form a state vector containing 8-12 indicators, which comprehensively reflects the comprehensive physiological state of the regulated object.

[0068] In the financial technology business field, the fusion data can include sweat metabolites, blood glucose levels, stress-related physiological parameters, and subjective fatigue perception scores. The correlation features extracted by the dynamic correlation model are further decomposed into metabolite state features such as electrolyte metabolism level, microorganism state features such as oral flora structure, and sign state features such as fatigue score index, and combined to form a state vector, providing real-time data support for employee health state monitoring, risk assessment, and personalized intervention.

[0069] The embodiment determines the state vector based on the dynamic correlation model, which can dynamically integrate multi-dimensional biological data and subjective sign information into a unified and structured physiological state expression, breaking through the limitations of isolated physiological indicators, fragmented information, and single expression in traditional methods. The state vector has the advantages of rich information, clear structure, and comprehensive expression, which can improve the accuracy, real-time performance, and dynamic adaptability of physiological state evaluation while maintaining data integrity, and enhance the perception ability of the overall regulation system to the physiological changes of the regulated object and the scientific nature of the intervention decision.

[0070] S50, based on the state vector, generating a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network;

[0071] In this embodiment, based on the state vector, a dynamic intervention scheme for adjusting the physiological state of the regulated object is generated through a decision network, specifically, after the state vector has been accurately determined, a decision network that has been pre-constructed and dynamically updated is used to combine the comprehensive physiological state information expressed by the state vector to infer and generate an intervention measure set with pertinence, real-time performance, and dynamic adjustment capability, so as to actively adjust the physiological state of the regulated object. The state vector, as a structured expression result of the comprehensive fusion of multi-modal biological data and subjective sign information in the system, directly reflects the overall physiological condition of the regulated object, and has high information integrity and expression accuracy.

[0072] The decision network is an intelligent reasoning and decision-making module in the system for intervention scheme generation, usually with a multi-layer structure, which can include functional substructures such as policy network, biological constraint filtering layer and evaluation network. Among them, the policy network is used to output an initial intervention action set according to the state vector, the biological constraint filtering layer is used to screen out intervention actions that do not meet the physiological safety requirements or exceed the biological tolerance range, to ensure the physiological adaptability and risk controllability of the intervention measures, and the evaluation network is used to quantify the comprehensive value of each candidate intervention action, and finally select the optimal intervention action according to the value score to form a dynamic intervention scheme with dynamic adaptability and individualization characteristics.

[0073] The specific implementation process includes the following: First, input the state vector into the policy network in the decision network, and the policy network generates an initial intervention action set according to the internal training parameters and reasoning logic, combined with the metabolite state characteristics, microbial state characteristics and sign state characteristics expressed by the state vector. The initial intervention action refers to the feasible intervention measure proposed by the system in the theoretical space for the current physiological state of the regulated object, for example, including diet ratio adjustment, nutrition supplement suggestion, lifestyle optimization measure, etc.

[0074] Secondly, the initial intervention action is screened through the biological constraint filtering layer, and the filtering layer screens the initial intervention action one by one according to the set biological safety threshold, physiological adaptation standard and risk control rule, eliminates the actions that do not meet the physiological safety, ethical norms or technical implementation conditions, and retains the filtered intervention action set that meets the multi-dimensional restriction conditions, to ensure that the final scheme has biological feasibility and implementation safety.

[0075] Then, the filtered intervention action is input into the evaluation network, and the evaluation network evaluates the physiological regulation effect, long-term health impact and system adaptability of each intervention action according to the built-in value function, reinforcement learning mechanism and historical feedback data, and calculates the value score of each intervention action. The value score reflects the comprehensive application value and effect expectation of each intervention action under a certain physiological state.

[0076] Finally, select the intervention action with the highest value score from the filtered intervention action as the dynamic intervention scheme, which is directly implemented for the regulated object, aiming to optimize, regulate or improve its overall physiological state in real time, and improve the health level and physiological function.

[0077] In the field of medical and health services, the process of generating a dynamic intervention plan based on a state vector can be flexibly adjusted in combination with different disease management needs and intervention modes. For problems such as intestinal function disorder, metabolic abnormalities, or microecological imbalance, the state vector may contain tryptophan metabolite ratio, intestinal flora diversity index, and symptom score information. The strategy network generates initial intervention actions including dietary structure adjustment, functional flora supplementation, and lifestyle optimization suggestions based on this information. The biological constraint filtering layer filters compliant actions based on patient individual allergy history, physiological tolerance, and medical guidance standards. The evaluation network determines the personalized dietary or nutritional intervention plan with the highest value score based on clinical effect data and individual feedback information, and finally outputs and implements the dynamic intervention plan to promote dynamic improvement of physiological state.

[0078] In the field of financial technology services, for the needs of employee health management, stress control, and production efficiency improvement, the state vector may integrate sweat metabolism data, stress-related physiological indicators, and subjective fatigue scores. The initial intervention actions output by the strategy network may include work rhythm adjustment, nutrition supplementation plan, or health management suggestions. The biological constraint filtering layer ensures that the intervention measures do not affect work continuity and physiological safety. The evaluation network calculates the comprehensive value of each intervention action based on employee health data, work performance feedback, and system simulation results, and finally determines and implements the dynamic intervention plan to achieve dynamic optimization and risk control of employee health status.

[0079] This embodiment can realize real-time linkage between intervention measures and individual physiological state by dynamically generating intervention plans based on state vectors and combining decision network structures, breaking through the problem of measure lag and lack of dynamic adjustment capability in traditional static intervention mode. This method uses the organic combination of structured physiological state expression, intelligent decision reasoning, and biological safety constraint mechanism to improve the accuracy, adaptability, and implementation effect of intervention plans, has good dynamic regulation capability and health risk prevention effect, and significantly enhances the regulation efficiency and health management level of the overall system for individual physiological state.

[0080] S60, acquiring physiological state feedback data after executing the dynamic intervention plan, and updating the decision network according to the physiological state feedback data.

[0081] In this embodiment, the physiological state feedback data after executing the dynamic intervention plan is acquired, and the decision network is updated according to the physiological state feedback data. After the implementation of the dynamic intervention plan ends, the system real-time collects and records information reflecting the changes of the physiological state of the regulated object, and uses this information to adjust and optimize the internal structure or parameter configuration of the decision network in reverse, so as to improve the adaptability and physiological effect of the subsequent intervention plan. The physiological state feedback data is derived from multi-modal data collection after the actual effect of the dynamic intervention plan, has high timeliness and pertinence, and usually includes new real-time physiological data and new subjective sign data.

[0082] The new real-time physiological data refers to the data set related to the physiological indicators obtained by the sensing device after the dynamic intervention scheme is executed. The sources can include but are not limited to sweat metabolism data, fecal microbial data or other real-time physiological parameters. The new subjective sign data refers to the subjective perception information of the regulated object after the intervention scheme is implemented, such as symptom quantification indicators, sign score results or self-state evaluation information. The two together constitute the physiological state feedback data, which comprehensively reflects the implementation effect of the intervention scheme and the physiological change trend.

[0083] The decision network is updated according to the physiological state feedback data. Specifically, the system generates a reward value using a reward function calculation method based on the obtained feedback data. The reward value is used to quantify the implementation effect of the current dynamic intervention scheme and the health improvement level. The higher the value, the more significant the intervention effect and the more obvious the physiological state improvement. The decision network includes a strategy network and an analysis network. The strategy network is used for generating and reasoning intervention actions, and the analysis network is used for predicting the effect and evaluating the value of the intervention scheme. The updating process includes adjusting the parameter structure of the strategy network and the analysis network based on the reward value, so as to realize the dynamic optimization of the intervention scheme generation and effect prediction ability, and form a data-driven and real-time adaptive closed-loop regulation mechanism.

[0084] The specific implementation process includes the following contents: first, after the dynamic intervention scheme is implemented, the system obtains new real-time physiological data through a multi-modal biological data acquisition device, and combines the new subjective sign data recorded by the user interaction interface to form a complete physiological state feedback data set. The feedback data content usually involves the change of the intervention target indicator, the improvement degree of the subjective sign and the related auxiliary information.

[0085] Secondly, the system calculates the reward value based on the physiological state feedback data. The reward value reflects the specific influence degree of the intervention scheme on the comprehensive physiological state of the regulated object. The numerical calculation can combine the dynamic adjustment effect of the physiological indicator change amplitude, the subjective score improvement situation or other key indicators to ensure the comprehensiveness and objectivity of the evaluation result.

[0086] Then, the system updates the decision network according to the reward value, adjusts the internal parameters of the strategy network and the analysis network, optimizes the generation logic of the intervention action by updating the parameters of the strategy network, and improves the matching degree and effectiveness of the subsequent intervention scheme and the individual physiological state. The parameter update of the analysis network enhances the accuracy of the intervention effect prediction and the risk assessment ability, so as to ensure that the system gradually improves the stability and physiological adaptability of the overall intervention scheme in the process of multiple feedback optimization.

[0087] In the medical health business field, the process of obtaining physiological state feedback data and updating the decision network can be flexibly adapted to different needs such as intestinal function regulation, metabolic balance remodeling, or microecological environment optimization. As new real-time physiological data, sweat metabolism data and fecal microbial data, and as new subjective symptom data, the system calculates a reward value based on the change in the index and the subjective perception improvement after the implementation of the intervention scheme. The system dynamically adjusts the strategy network parameters and analysis network parameters in combination with the reward value, optimizes subsequent personalized nutritional intervention measures, functional flora supplement schemes, or lifestyle adjustment suggestions, and forms a continuous, data-driven health status dynamic management process.

[0088] In the financial technology business field, it is suitable for employee health management, production environment adaptability adjustment, and risk control under high-pressure working conditions. The system obtains new real-time physiological data and new subjective symptom data after the implementation of the dynamic intervention scheme, reflects the changes in physiological load, stress level, or fatigue state, evaluates the intervention effect through the reward value, dynamically optimizes the decision network parameter configuration, and improves the health risk prevention and work efficiency maintenance capability.

[0089] The embodiment obtains physiological state feedback data after the implementation of the dynamic intervention scheme in real time, and dynamically updates the decision network in combination with the reward value mechanism. The system can continuously optimize the intervention scheme generation and effect prediction capability, establish a data closed loop and a dynamically adaptive health status regulation mechanism, significantly improve the precision, effectiveness, and safety of the intervention measures, and promote the continuous improvement and dynamic stability of the overall physiological state, breaking through the technical limitations of one-way control and lack of feedback adjustment in the traditional intervention mode.

[0090] The present application relates to the field of artificial intelligence technology, which can be applied to financial technology and medical health business scenarios, and discloses a multi-dimensional physiological data dynamic regulation method, device, equipment and medium, comprising: obtaining at least two kinds of real-time physiological data from different dimensions representing physiological state from a regulated object, and recording subjective symptom data of the regulated object; performing fusion processing on the real-time physiological data and the subjective symptom data to generate time sequence aligned fusion data; based on the fusion data, a dynamic correlation model between the real-time physiological data and the subjective symptom data is constructed; according to the dynamic correlation model, a state vector representing the current comprehensive physiological state of the regulated object is determined; based on the state vector, a dynamic intervention scheme for regulating the physiological state of the regulated object is generated through a decision network; physiological state feedback data after the implementation of the dynamic intervention scheme is obtained, and the decision network is updated according to the physiological state feedback data. The present application fuses multiple different dimensions of real-time physiological data and subjective symptom data, constructs a dynamic correlation model, reflects the comprehensive physiological state of the regulated object in real time, generates a dynamic intervention scheme in combination with the decision network, and updates the decision network through the physiological state feedback data, forming a closed loop of data perception, state analysis and intervention optimization, and improving the accuracy and timeliness of physiological state regulation.

[0091] In one embodiment, the step S10 described above comprises:

[0092] S101, collecting, by a wearable sweat sensor, sweat metabolic data of the regulated object containing tryptophan metabolite ratio;

[0093] S102, collecting, by a feces detection device with spectral analysis function, fecal microbial data of the regulated object containing genus abundance;

[0094] S103, collecting, by a user terminal interactive interface, symptom quantitative indicators of the regulated object, and taking the symptom quantitative indicators as subjective sign data;

[0095] S104, synchronously transmitting the sweat metabolic data, fecal microbial data and subjective sign data to a data processing center.

[0096] In the embodiment, at least two kinds of real-time physiological data representing physiological state from different dimensions are obtained from the regulated object, and subjective sign data of the regulated object is recorded. In the data collection process, the same regulated object is faced, data sets with physiological state indicating significance are obtained based on different biological levels and different source channels, and individual subjective perception information is recorded synchronously, forming a multi-modal and multi-dimensional physiological information input basis.

[0097] The at least two kinds of real-time physiological data representing physiological state from different dimensions specifically include sweat metabolic data collected by a wearable sweat sensor and fecal microbial data collected by a feces detection device with spectral analysis function. The sweat metabolic data refers to the dynamic detection result based on the metabolite composition in the sweat of the regulated object, and the tryptophan metabolite ratio is the core content in the sweat metabolic data, reflecting the real-time metabolism level of the tryptophan metabolic pathway in the body and its correlation with the overall physiological state. The tryptophan metabolite ratio can include the ratio of kynurenine to 5-hydroxytryptamine, and 5-hydroxytryptamine is also called serotonin, which is widely involved in the regulation process of the gut-brain axis, the immune system and metabolic balance. Kynurenine, as a degradation product of tryptophan, is involved in oxidative stress, immune regulation and neural function regulation, and the ratio of the two can reflect the risk trend of metabolic abnormalities, inflammatory reactions or intestinal function disorders.

[0098] Fecal microbiome data refers to a collection of data reflecting the structure of intestinal microbial community, functional state and metabolic activity obtained based on excretion detection. As a specific indicator in fecal microbiome data, genus abundance represents the relative proportion or quantity level of a specific microorganism in the intestinal flora. Changes in genus abundance can indicate the stability of intestinal microecology, metabolic potential or disease risk. Common genera include Bacteroides, Lactobacillus, Bifidobacterium, etc. The abundance of these genera is closely related to intestinal barrier function, inflammation level and overall metabolic state. The excretion detection device with spectral analysis function can non-invasively identify the microbiome information in the excretion through near-infrared spectroscopy, Raman spectroscopy or other spectral methods suitable for biological sample analysis, extract genus abundance indicators, and realize dynamic monitoring of the microecological environment.

[0099] Subjective sign data refers to quantitative information actively input by the regulated object based on their own physiological experience, health perception or symptom changes through the user terminal interaction interface. Symptom quantification indicators, as an important part of subjective sign data, can include but are not limited to abdominal distension score, abdominal pain score, defecation regularity, stool type classification, etc. A standardized and structured indicator system is used to ensure data consistency and comparability. Collecting subjective sign data through the user terminal interaction interface improves the real-time and compliance of data acquisition, and facilitates the synchronous integration with objective physiological data.

[0100] Synchronizing sweat metabolism data, fecal microbiome data and subjective sign data to the data processing center refers to after data collection is completed, real-time and complete multi-modal physiological information is aggregated to a unified data processing platform through network communication modules, mobile terminals or other information transmission means, ensuring the continuity and timeliness of the data link, and providing an information base for subsequent data fusion, dynamic modeling and individualized regulation.

[0101] This embodiment realizes real-time acquisition of sweat metabolism data and fecal microbiome data based on data collection means representing physiological state in different dimensions, and synchronously records subjective sign data. The system forms a multi-modal, physiological level rich data input structure, breaking the limitations of data isolation and information fragmentation in traditional physiological monitoring, and improving the comprehensiveness and dynamics of physiological state description. The acquisition of tryptophan metabolite ratio fills the technical gap of sweat monitoring at the metabolite spectrum level, the detection of genus abundance expands the real-time and functionality of microecological monitoring, and the quantitative input of subjective sign data enhances the adaptability of data and individual perception. Synchronous transmission of multi-source data ensures the timeliness of information integration, and builds an efficient data base for dynamic perception of physiological state, significantly improving the individualization, scientificity and real-time response capability of subsequent regulation measures.

[0102] In one embodiment, the above step S20 comprises:

[0103] S201, identify the timestamp deviation between the sweat metabolism data and the fecal microbial data in the real-time physiological data;

[0104] S202, perform dynamic time compensation processing on the sweat metabolism data and the fecal microbial data according to the timestamp deviation, to generate compensated sweat metabolism data and compensated fecal microbial data;

[0105] S203, map the compensated sweat metabolism data and the compensated fecal microbial data to a physiological rhythm coordinate system, to generate mapped sweat metabolism data and mapped fecal microbial data;

[0106] S204, combine the mapped sweat metabolism data, the mapped fecal microbial data, and the subjective sign data to generate time-aligned fusion data.

[0107] In this embodiment, the fusion processing of real-time physiological data and subjective sign data to generate time-aligned fusion data means that based on multi-source heterogeneous physiological information, data time sequence calibration, dynamic compensation and coordinate mapping operations are performed to finally realize data structure unification, time dimension alignment and fusion result output. Real-time physiological data includes sweat metabolism data and fecal microbial data, and subjective sign data is derived from the autonomous input of the regulated object. The fusion processing needs to consider the time attribute, spatial attribute and physiological rhythm characteristics of the data to ensure the scientificity and accuracy of the data integration.

[0108] Identifying the timestamp deviation between the sweat metabolism data and the fecal microbial data in the real-time physiological data means that based on the respective data recording times, the asynchrony or delay characteristics of the two in the time dimension are analyzed in the process of obtaining the sweat metabolism data and the fecal microbial data. The timestamp deviation may be caused by differences in response speed of data collection devices, dispersion of collection time nodes or data transmission delay. If not corrected, it may lead to data fusion distortion and affect the accuracy of overall physiological state modeling. The identification process can use timestamp comparison, event marker synchronization, data packet time sequence analysis and other methods to ensure high precision of deviation detection.

[0109] Performing dynamic time compensation processing on the sweat metabolism data and the fecal microbial data according to the timestamp deviation to generate compensated sweat metabolism data and compensated fecal microbial data means that based on the identified timestamp deviation information, the corresponding relationship of the two types of data on the time axis is adjusted to eliminate the time inconsistency. Dynamic time compensation processing can be performed in various ways, including but not limited to multi-order interpolation method, dynamic time warping algorithm or time sequence correction method based on physiological rhythm model. The compensated data is synchronized with the actual physiological process in the time dimension, forming a data set with time integrity and logical continuity.

[0110] mapping the compensated sweat metabolite data and the compensated fecal microbiome data to a circadian coordinate system to generate mapped sweat metabolite data and mapped fecal microbiome data refers to constructing a standardized time-space system based on human circadian rhythm characteristics, and uniformly mapping the compensated multi-source data to the coordinate system. The circadian coordinate system can be established based on the circadian rhythm, metabolic cycle or intestinal microecological fluctuation rule. Common forms include time partitions divided by 24-hour cycles. The mapping process preserves the original numerical structure and rhythm dynamic characteristics of the data, ensuring the biological interpretability of the data structure and the standardization of comparative analysis.

[0111] merging the mapped sweat metabolite data, the mapped fecal microbiome data and the subjective sign data to generate time series aligned fusion data refers to integrating physiological data of different sources and different categories in the circadian coordinate system to form multi-dimensional, time-consistent and structurally unified data output results. In the merging process, the data can be organized using structured data tables, tensor structures or multi-level data models. The fusion data has time continuity, physiological index diversity and individual perception information integrity, providing an information base for subsequent dynamic modeling, state analysis and personalized intervention.

[0112] The present embodiment effectively solves the problem of time desynchronization of multi-source physiological data in the collection and transmission process by identifying the timestamp deviation between sweat metabolite data and fecal microbiome data and combining dynamic time compensation processing, thereby ensuring the accurate alignment of data on the time axis. The introduction of the circadian coordinate system improves the biological interpretability and rhythm characteristic preservation ability of data integration, avoiding rhythm deviation and data distortion caused by time misalignment. The time series aligned fusion data formed ultimately has multi-modal information richness, time structure uniformity and physiological rhythm synchronicity, enhancing the overall data quality and the reliability of dynamic modeling, and significantly improving the accuracy and practicality of subsequent physiological state analysis and intervention measure design.

[0113] In one embodiment, the above step S30 comprises:

[0114] S301, organizing the fusion data into a three-dimensional data structure comprising metabolite dimension, microbiome dimension and time dimension;

[0115] S302, extracting core factors comprising metabolite-microbiome-time correlation characteristics from the three-dimensional data structure;

[0116] S303, constructing a cross-dimension correlation strategy library based on the core factors;

[0117] S304, defining the cross-dimension correlation strategy library as a dynamic correlation model, and updating the weight values of the dynamic correlation model according to the new data.

[0118] In this embodiment, the fusion data is a data set integrated from multiple physiological information, with multi-dimensional structure and time synchronization characteristics, organized into a three-dimensional data structure composed of metabolite dimension, microbial dimension and time dimension, which helps to express the complex relationship between different physiological indicators in a unified framework. The metabolite dimension refers to the metabolomics characteristics in the sweat metabolite data, which can include tryptophan metabolites, short-chain fatty acids and other numerical information reflecting metabolic status. The microbial dimension refers to the distribution characteristics of the fecal microbial data, such as genus abundance, microbial diversity index, and relative proportion of specific functional flora. The time dimension is based on the timestamp or physiological rhythm mapping results in the fusion data, providing positioning reference for various data in time series. By integrating different sources and types of information through a three-dimensional data structure, a complete data cube can be constructed to ensure logical rigor and analysis expandability of data organization.

[0119] Extracting metabolite-microbe-time correlation features from the three-dimensional data structure is essentially using mathematical dimension reduction, tensor decomposition or feature extraction algorithms to identify important parameter sets that affect the interaction between metabolic status, microbial community and time dynamics. This parameter set is the core factor, which can comprehensively reflect the key change trends and high-order correlation patterns in multi-modal data, reduce redundant information interference, highlight the essential relationship within the data, and facilitate the construction of more compact and accurate correlation expression models.

[0120] Based on the core factor, a cross-dimensional correlation strategy library is established, which involves using statistical modeling, machine learning or graph structure methods to systematically describe the quantitative relationships and logical structures between metabolite indicators, microbial community characteristics and time changes. The essence of the strategy library is a parameterized and structured multi-dimensional correlation rule set that can support fast reasoning and flexible invocation, with real-time scalability and data adaptability, providing structured support for dynamic modeling and prediction analysis in different scenarios.

[0121] Defining the strategy library as a dynamic correlation model and dynamically updating the weight values based on new data reflects the adaptive ability and continuous optimization characteristics of the correlation model. New data includes real-time acquisition or external supplementary fusion data, and the update process can use techniques such as incremental learning, weight adjustment, and parameter re-optimization to ensure that the model dynamically absorbs the latest information while maintaining the original logical structure, improving the accuracy, reliability and scenario adaptability of the overall correlation expression, and achieving multi-source information dynamic fusion expression for complex physiological systems.

[0122] The embodiment realizes the unified expression of the time, space and logical relationship between different physiological indicators by the organization and correlation feature extraction of multi-dimensional structure, breaking through the limitations of traditional single indicator analysis. The extraction of core factors reduces the data complexity and improves the subsequent model operation efficiency and expression clarity. The construction of the strategy library forms a callable and adjustable rule set of metabolism, microorganisms and time information, enhancing the systematicness and flexibility of data analysis. The dynamic weight updating mechanism ensures that the correlation model has real-time learning and self-optimization ability, which can effectively adapt to physiological state changes, data structure adjustment and application scene switching, and improve the response ability of the overall system to complex physiological states and the scientificity of intervention strategy generation.

[0123] In one embodiment, the above step S40 comprises:

[0124] S401, extracting a metabolite-microorganism-time correlation feature at the current time from the dynamic correlation model;

[0125] S402, decomposing the metabolite-microorganism-time correlation feature into a metabolite state feature, a microorganism state feature and a sign state feature;

[0126] S403, combining the metabolite state feature, the microorganism state feature and the sign state feature to form a state feature set, and taking the state feature set as a state vector.

[0127] In the embodiment, the dynamic correlation model has high-dimensional expression ability for real-time physiological data and subjective sign data after multi-source data fusion and structure construction, and can map the correlation pattern of metabolite indicators, microorganism features and time changes. Extracting the metabolite-microorganism-time correlation feature at the current time from the dynamic correlation model is essentially obtaining a key parameter set reflecting the current physiological state according to the mapping node of the real-time data corresponding to the current time point in the dynamic correlation model. The parameter set is derived from the correlation strategy library of the dynamic correlation model, and specifically includes metabolic pathway regulation parameters, microbial abundance change factors, time window dynamic response coefficients, etc., which are used to comprehensively describe the metabolism-microbiota-time interaction relationship of the current physiological state.

[0128] The metabolite-microorganism time correlation features extracted at the current time are further decomposed into metabolite state features, microorganism state features, and sign state features, which is a structured classification of different physiological indicators. The metabolite state features refer to a set of physiological indicators based on sweat metabolite data, which can include tryptophan metabolite ratios, short-chain fatty acid concentrations, metabolic fluctuation indexes, etc. The microorganism state features refer to the parameters of the flora structure extracted from the fecal microorganism data, such as genus abundance, diversity index, and functional flora proportion. The sign state features refer to the quantitative results of symptoms extracted from subjective sign data, including pain scores, abdominal distension degrees, emotional state scores, and other subjective evaluation values. The metabolite state features, microorganism state features, and sign state features are combined according to data types, indicator attributes, and statistical structures to form a state feature set, ensuring clear and compatible multi-dimensional and multi-type expression structures of physiological states.

[0129] The state feature set is defined as a state vector to convert complex physiological state parameters into standardized, continuous, and operable numerical vectors through a unified data structure. As the input basis of the subsequent decision network, the state vector needs to have good structural compatibility and information integrity to support quantitative calculation and structural modeling of multiple types of indicators in the same expression space. By mapping the state feature set to the state vector, metabolite features, microorganism features, and sign features can be fully considered in the decision-making process, which is beneficial to dynamic analysis and strategy generation for the current comprehensive physiological state.

[0130] The present embodiment can capture the comprehensive change trend of physiological state in real time by extracting the current time correlation features from the dynamic correlation model, improving the timeliness of state recognition. By decomposing the correlation features into metabolite state features, microorganism state features, and sign state features, it helps to refine the expression structure of physiological state and ensures the effective use of different physiological data types. Combining the decomposed state features into a state feature set and mapping it to a state vector effectively realizes the unified quantitative expression of complex physiological state, enhancing the input adaptability of the subsequent decision network and the physiological state analysis capability.

[0131] In one embodiment, the above step S50 includes:

[0132] S501, inputting the state vector into a policy network in the decision network to generate an initial intervention action;

[0133] S502, filtering the initial intervention action through a biological constraint filtering layer in the decision network to generate a filtered intervention action;

[0134] S503, determining a value score of the filtered intervention action through an analysis network in the decision network;

[0135] S504, taking the intervention action with the highest value score in the filtered intervention actions as the dynamic intervention scheme.

[0136] In this embodiment, the state vector is a standardized expression form of fused multi-dimensional data features of metabolites, microorganisms, and signs, has a complete, continuous, and calculable data structure, and can provide physiological state input basis for subsequent intelligent decision-making. Inputting the state vector into the policy network in the decision network is essentially based on the comprehensive physiological parameters in the state vector to call the policy network structure to generate a preliminary adjustment scheme for the current physiological state. The policy network is generally implemented through a deep neural network or other nonlinear mapping structure, has the ability to generate intervention actions based on historical training data and current state features. Intervention actions usually include dietary adjustment suggestions, nutrient intake schemes, lifestyle intervention parameters, etc.

[0137] Filtering the initial intervention action through the biological constraint filtering layer is to introduce physiological safety limits and scientific rationality verification after generating the action, to avoid generating intervention actions that violate biological laws or exceed the individual physiological bearing range. The biological constraint filtering layer can set hard screening rules based on multi-dimensional information such as medical knowledge base, individual historical data, physiological limit parameters, etc., to eliminate unreasonable, unsafe, or unexecutable schemes in the initial intervention action, and only retain actions that meet physiological feasibility and intervention scientificity, and output filtered intervention actions.

[0138] Determining the value score of the filtered intervention action through the analysis network is to calculate the expected effect index of the action for improving the current physiological state based on historical effect data, individual feedback information, and model prediction mechanism for each feasible intervention action, to form a value score as a quantitative reference for the advantages and disadvantages of the action. The analysis network is generally implemented through a value function or reward function structure, which can comprehensively evaluate the effectiveness, stability, and long-term impact of the action according to the multi-dimensional input of the physiological state and the specific content of the intervention action.

[0139] From all the filtered intervention actions, selecting the intervention action with the highest value score as the dynamic intervention scheme is to output the intervention strategy with the highest physiological improvement potential, the optimal effect, and the lowest risk based on quantitative evaluation results, to ensure that the intervention scheme has dynamic, targeted, and optimal execution effect. The dynamic intervention scheme as the direct basis for subsequent execution links can realize real-time, individualized, and closed-loop dynamic adjustment of the physiological state of the controlled object.

[0140] By inputting the state vector into the policy network, the embodiment can intelligently generate a preliminary intervention action conforming to the current physiological state by combining multi-dimensional physiological characteristics and historical model experience. By filtering the preliminary intervention action through a biological constraint filtering layer, the physiological safety and execution feasibility of the generated scheme are ensured, and the physiological risk is reduced. By analyzing the network to calculate the value score of each filtered intervention action, the actual improvement potential of the intervention scheme can be quantitatively evaluated, and the prediction accuracy of the intervention effect is improved. Finally, based on the value score, the optimal intervention action is selected as the dynamic intervention scheme, effectively realizing the optimal decision output in the physiological state adjustment process, promoting the continuous optimization and adjustment of the individual physiological state towards the health goal, and overall enhancing the intelligence, scientificity and effectiveness of the dynamic regulation process.

[0141] In one embodiment, the above step S60 comprises:

[0142] S601, collecting new real-time physiological data and new subjective sign data after the execution of the dynamic intervention scheme, and taking the new real-time physiological data and the new subjective sign data as physiological state feedback data;

[0143] S602, determining a reward value of the decision network based on the physiological state feedback data;

[0144] S603, updating parameters of a policy network and an analysis network in the decision network according to the reward value, to obtain an updated decision network.

[0145] In the embodiment, the new real-time physiological data collected after the execution of the dynamic intervention scheme generally includes real-time detection results based on sweat metabolites, fecal microorganisms and the like, for reflecting the immediate physiological response of the regulated object under the action of the intervention scheme. Such data can be continuously obtained through wearable sensors, intelligent sanitary devices and the like, and has the ability of real-time and dynamic reflection of the metabolic state change of the individual. The new subjective sign data collected generally includes subjective perception of the regulated object, such as symptom change, physical discomfort index or sign quantitative information actively fed back through the user interaction terminal, for supplementing objective physiological indicators and enhancing the completeness and individual difference adaptation ability of the feedback data. The new real-time physiological data and the new subjective sign data are taken together as the physiological state feedback data, which can comprehensively and dynamically depict the physiological response effect after the execution of the intervention scheme.

[0146] The reward value of the decision network is determined based on the physiological state feedback data, which is essentially to evaluate the improvement of the current intervention effect on the target physiological state, and to generate a quantitative index for guiding the adaptive optimization of the decision network. The reward value can be calculated by setting the target physiological parameter interval, referring to the best state data in the past, combining the health risk model output, etc., which can reflect the pros and cons and trend of the intervention effect. The design of the reward value needs to consider the comprehensive measurement of short-term index improvement and long-term physiological stability to ensure the objectivity, dynamics and guiding value of the feedback signal.

[0147] Updating the strategy network and analysis network parameters in the decision network according to the reward value means dynamically optimizing the structure parameters in the decision network responsible for action generation and action evaluation, improving the adaptability of the network to complex physiological states and the accuracy of the intervention scheme output. The strategy network parameter update optimizes the mapping relationship of the state vector to the intervention action, enhances the scientificity and individual matching degree of the generated action. The analysis network parameter update improves the evaluation accuracy of the action value and reduces the output of invalid or suboptimal schemes. The whole realizes a closed-loop self-optimization mechanism driven by physiological feedback data, forming a dynamic, continuous and personalized physiological state regulation capability.

[0148] Example: In the field of medical health business, for the long-existing individual health management demand related to intestinal flora imbalance and chronic low-grade inflammation, a dynamic physiological state regulation process can be designed for a regulated object. First, the sweat samples and fecal samples of the regulated object are continuously collected. The sweat collection is realized by wearing a tryptophan metabolite special sweat sensor to real-time obtain the metabolite ratio of tryptophan metabolic pathway, such as monitoring the concentration change of 5-hydroxytryptophan and kynurenine. The fecal collection is realized by installing a spectrum detection module in the household sanitary equipment to analyze the abundance distribution of different genera in the sample, such as monitoring the flora ratio of Bifidobacterium and Bacteroides. At the same time, the daily subjective signs of the regulated object are recorded through the mobile phone application, including the symptom quantitative indicators such as intestinal discomfort score, abdominal distension degree, defecation habit change, etc. All the collected data are synchronously uploaded to the backend data processing center.

[0149] In the data processing center, the sweat metabolic data, fecal microbial data and subjective sign data are analyzed synchronously in time. First, the time stamp deviation between the sweat metabolic data and the fecal microbial data due to the time difference of collection is identified, and a dynamic time compensation operation is performed based on the time stamp deviation to adjust the data time alignment. The sweat metabolic data and the fecal microbial data after time compensation are further mapped to the physiological rhythm coordinate system, and through the establishment of a time framework based on the circadian rhythm, the data of different sources are unified into the dynamic physiological time axis. The data after physiological rhythm mapping and the subjective sign data are combined to generate time sequence aligned fusion data, providing a basis for subsequent modeling.

[0150] Based on the time alignment of the fusion data, a dynamic correlation model is constructed. The fusion data is arranged into a three-dimensional data structure according to the metabolite dimension, the microorganism dimension and the time dimension, and the correlation characteristics of metabolites, microorganisms and time are extracted, for example, the change rule of the correlation of tryptophan derivative metabolite ratio and specific genus abundance at different time points. Further, based on these correlation characteristics, a cross-dimension correlation strategy library is generated, and the weight values of each correlation rule in the correlation strategy library are adjusted by continuously introducing new fusion data. The weight value adjustment reflects the change in the correlation strength in the actual individual data, and the entire correlation model is dynamically adjusted through the weight updating mechanism to ensure that the model responds to the physiological state changes in real time.

[0151] The dynamic correlation model extracts the metabolite-microorganism-time correlation characteristics at the current time, and divides the extracted correlation characteristics into three parts: metabolite state characteristics, microorganism state characteristics and sign state characteristics, which respectively describe the sweat metabolite level, fecal flora structure and subjective symptom state. The three types of state characteristics are combined into a state characteristic set, and the state characteristic set is defined as a state vector, which comprehensively represents the comprehensive physiological state of the current regulated object.

[0152] The state vector is input into the decision network. First, the strategy network generates an initial intervention action based on the state vector, such as suggesting adjusting the daily protein intake ratio or suggesting taking specific probiotic supplements. The initial intervention action is filtered by the biological constraint filtering layer to eliminate actions that exceed the upper limit of dietary safety or conflict with the history of allergies, and the filtered intervention action that meets the biological constraints is obtained. Further, the analysis network calculates the value score of the filtered intervention action, and the value score evaluates the expected effect of the intervention action on optimizing the current physiological state. The intervention action with the highest value score is selected as the dynamic intervention scheme this time, and is output to the user as a personalized dietary and lifestyle adjustment suggestion.

[0153] After the regulated object executes the dynamic intervention scheme, the system continuously collects new sweat metabolite data, new fecal microorganism data and new subjective sign data as physiological state feedback data. Based on the feedback data, the reward value of this intervention effect is calculated, and the reward value is calculated based on indicators such as the improvement degree of tryptophan metabolite ratio, the improvement level of intestinal flora diversity and the improvement amplitude of subjective sign score. According to the reward value, the parameters of the strategy network and the analysis network are updated synchronously, the action generation mechanism and the action value evaluation standard are adjusted, and the optimized decision network is obtained, which provides more adaptive individual physiological state support for subsequent dynamic intervention schemes, and continuously realizes personalized dynamic intervention adjustment based on biological feedback.

[0154] In the field of financial technology business, for the needs of personalized physiological monitoring and dynamic risk adjustment in insurance customer health management, financial health scoring and dynamic policy pricing, a continuous biofeedback-driven health management process can be designed around a single subject. First, deploy a customer-oriented physiological data acquisition device to obtain real-time sweat metabolite data from the customer through a wearable sweat sensor, focusing on tryptophan metabolite ratios, such as monitoring the relative concentrations of 5-hydroxytryptophan and kynurenine, for real-time estimation of financial health risk. Collect the customer's excrement samples through embedded intelligent detection devices, analyze the changes in bacterial abundance, and monitor the dynamic proportion of intestinal flora such as Bifidobacterium and Bacteroides. At the same time, record the customer's active feedback on physical changes through the interactive interface of the insurance company's customer service platform, including fatigue score, indigestion frequency, subjective health score, and other quantitative indicators. The above physiological data and subjective physical data are transmitted synchronously to the financial data processing center to ensure that real-time health information and customer financial health records are updated synchronously.

[0155] In the data processing center, the customer's physiological data and subjective physical data are time series fused. First, identify the time deviation between sweat metabolite data and fecal microbial data, and perform dynamic time compensation operation for time stamp difference to adjust the time alignment structure of the data and avoid the influence of data island phenomenon on financial risk assessment. Further map the compensated data to a time reference based on the customer's physiological rhythm, unify the sweat metabolite data and fecal microbial data collected at different time points into the financial health dynamic time axis, and construct time series aligned fusion data to support subsequent dynamic financial risk modeling.

[0156] Based on the time series aligned fusion data, a financial health risk dynamic correlation model is constructed. The fusion data is organized into a three-dimensional data structure according to metabolites, microorganisms and time dimensions, and the customer-specific metabolite-microorganism-time correlation features are extracted to identify the correlation rules between the customer's tryptophan metabolic state and intestinal flora dynamics in financial health performance. According to the extracted correlation features, a cross-dimensional correlation strategy library is generated, and the correlation rule weight values are dynamically adjusted to ensure that the financial risk correlation model reflects the trend of the customer's real-time physiological state in a timely manner, providing dynamic support for the financial product risk control system.

[0157] Further extract the metabolite-microorganism-time correlation features at the current time, and subdivide them into metabolite state features, microorganism state features and physical state features, which respectively depict the customer's current physiological metabolic status, intestinal microecological state and physical health score. Combine the above state features into a state feature set to form a complete state vector as a real-time indicator of the customer's financial health status.

[0158] The state vector is input into a decision network deployed by the insurance company, and an initial financial intervention action is generated based on the state vector through a strategy network, such as adjusting insurance premiums for customer health risks, customizing health management services, or recommending personalized health investment portfolios. The initial financial intervention action is filtered through a biological constraint filtering layer to exclude actions that do not match the customer's historical health records or risk tolerance, generating filtered intervention actions that meet the risk rules. The value score of the filtered intervention action is calculated through an analysis network to quantify the expected value of each financial intervention action to the customer's long-term health risk reduction and financial service optimization, and the intervention action with the highest value score is selected as the current financial dynamic intervention scheme to dynamically adjust the customer's financial health rights and interests.

[0159] After the customer executes the financial dynamic intervention scheme, the financial data processing center continues to collect new sweat metabolic data, new fecal microbial data, and new subjective physical data as physiological state feedback data. The reward value of the intervention effect is calculated based on the feedback data, which refers to the customer's tryptophan metabolic ratio improvement, gut flora diversity change, and subjective health score improvement, and other financial health core indicators. The reward value is used to update the parameters of the strategy network and the analysis network in real time, adjust the financial action generation mechanism and action value evaluation standard, and obtain an optimized decision network. The optimized decision network provides continuous support for subsequent customer financial dynamic intervention, builds a dynamic financial health management closed loop driven by biological feedback, and improves the risk response capability and individualization precision of financial services.

[0160] The embodiment can comprehensively reflect the immediate physiological effect of the intervention measures and the individual perception change by collecting new real-time physiological data and new subjective physical data after the execution of the dynamic intervention scheme, and provide real and dynamic physiological state feedback information. The reward value is determined based on the physiological state feedback data, which can quantitatively measure the actual improvement level of the intervention effect, and ensure the scientificity of the decision process and the clearness of the result orientation. The parameters of the strategy network and the analysis network are updated through the reward value to form a dynamic self-adaptive optimization process based on physiological feedback, enhance the ability of the decision network to cope with complex physiological state changes, promote the continuous optimization and adjustment of the intervention scheme, and overall improve the intelligent level, intervention effect, and individual adaptability of the dynamic regulation system.

[0161] In an embodiment, a multi-dimensional physiological data dynamic regulation device is provided, which corresponds to the multi-dimensional physiological data dynamic regulation method in the above embodiments. Referring to Figure 3 , Figure 3 A functional module schematic diagram of a preferred embodiment of the multi-dimensional physiological data dynamic regulation device of the present application. The data acquisition module 10, the data fusion module 20, the correlation modeling module 30, the state evaluation module 40, the intervention generation module 50, and the feedback optimization module 60. The detailed description of each functional module is as follows:

[0162] a data acquisition module 10, configured to acquire real-time physiological data from a subject, the real-time physiological data being derived from at least two physiological states of the subject and representing different dimensions of the physiological states, and to record subjective sign data of the subject;

[0163] a data fusion module 20, configured to fuse the real-time physiological data and the subjective sign data to generate time-series aligned fusion data;

[0164] a correlation modeling module 30, configured to construct a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data;

[0165] a state evaluation module 40, configured to determine a state vector representing a current comprehensive physiological state of the subject according to the dynamic correlation model;

[0166] an intervention generation module 50, configured to generate a dynamic intervention scheme for adjusting the physiological state of the subject by a decision network based on the state vector;

[0167] a feedback optimization module 60, configured to acquire physiological state feedback data after the dynamic intervention scheme is executed, and to update the decision network according to the physiological state feedback data.

[0168] In an embodiment, the data acquisition module 10 is specifically configured to:

[0169] acquire sweat metabolic data containing tryptophan metabolite ratios of the subject by a wearable sweat sensor;

[0170] acquire fecal microbial data containing genus abundances of the subject by a fecal detection device with a spectral analysis function;

[0171] acquire symptom quantification indicators of the subject by a user terminal interactive interface, and use the symptom quantification indicators as the subjective sign data;

[0172] synchronously transmit the sweat metabolic data, the fecal microbial data, and the subjective sign data to a data processing center.

[0173] In an embodiment, the data fusion module 20 is specifically configured to:

[0174] identify a timestamp deviation between the sweat metabolic data and the fecal microbial data in the real-time physiological data;

[0175] perform dynamic time compensation processing on the sweat metabolic data and the fecal microbial data according to the timestamp deviation to generate compensated sweat metabolic data and compensated fecal microbial data;

[0176] mapping the compensated sweat metabolite data and the compensated fecal microbiome data to a circadian coordinate system to generate mapped sweat metabolite data and mapped fecal microbiome data;

[0177] merging the mapped sweat metabolite data, the mapped fecal microbiome data and the subjective sign data to generate time-series aligned fusion data.

[0178] In an embodiment, the association modeling module 30 is specifically configured to:

[0179] organize the fusion data into a three-dimensional data structure comprising metabolite dimension, microbiome dimension and time dimension;

[0180] extract core factors comprising metabolite-microbiome-time association features from the three-dimensional data structure;

[0181] construct a cross-dimension association strategy library based on the core factors;

[0182] define the cross-dimension association strategy library as a dynamic association model, and update weight values of the dynamic association model according to new data.

[0183] In an embodiment, the state evaluation module 40 is specifically configured to:

[0184] extract metabolite-microbiome-time association features of the current time from the dynamic association model;

[0185] decompose the metabolite-microbiome-time association features into metabolite state features, microbiome state features and sign state features;

[0186] combine the metabolite state features, the microbiome state features and the sign state features to form a state feature set, and take the state feature set as a state vector.

[0187] In an embodiment, the intervention generation module 50 is specifically configured to:

[0188] input the state vector into a policy network in a decision network to generate an initial intervention action;

[0189] filter the initial intervention action through a biological constraint filtering layer in the decision network to generate a filtered intervention action;

[0190] determine a value score of the filtered intervention action through an analysis network in the decision network;

[0191] take an intervention action with the highest value score in the filtered intervention action as a dynamic intervention scheme.

[0192] In an embodiment, the feedback optimization module 60 is specifically configured to:

[0193] collecting new real-time physiological data and new subjective sign data after the dynamic intervention scheme is executed, and taking the new real-time physiological data and the new subjective sign data as physiological state feedback data;

[0194] determining a reward value of the decision network based on the physiological state feedback data;

[0195] updating parameters of a policy network and an analysis network in the decision network according to the reward value, to obtain an updated decision network.

[0196] In an embodiment, a computer device is provided, which can be a server, and an internal structure diagram thereof can be as shown in Figure 4 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide determination and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external user terminal through a network connection. The computer program is executed by the processor to implement functions or steps of a server side of a multi-dimensional physiological data dynamic regulation method.

[0197] In an embodiment, a computer device is provided, which can be a user terminal, and an internal structure diagram thereof can be as shown in Figure 5 The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide determination and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external server through a network connection. The computer program is executed by the processor to implement functions or steps of a user terminal side of a multi-dimensional physiological data dynamic regulation method

[0198] In an embodiment, a computer device is provided, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the following steps when executing the computer program:

[0199] acquiring at least two real-time physiological data from different dimensions representing a physiological state of a regulated object, and recording subjective sign data of the regulated object;

[0200] fusing the real-time physiological data and the subjective sign data to generate time-series aligned fusion data;

[0201] constructing a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data;

[0202] determining a state vector representing a current comprehensive physiological state of the regulated object according to the dynamic correlation model;

[0203] generating a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network based on the state vector;

[0204] obtaining physiological state feedback data after the dynamic intervention scheme is executed, and updating the decision network according to the physiological state feedback data.

[0205] In one embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the following steps:

[0206] obtaining at least two real-time physiological data from different dimensions representing the physiological state of the regulated object, and recording subjective sign data of the regulated object;

[0207] fusing the real-time physiological data and the subjective sign data to generate time-series aligned fusion data;

[0208] constructing a dynamic correlation model between the real-time physiological data and the subjective sign data based on the fusion data;

[0209] determining a state vector representing a current comprehensive physiological state of the regulated object according to the dynamic correlation model;

[0210] generating a dynamic intervention scheme for adjusting the physiological state of the regulated object through a decision network based on the state vector;

[0211] obtaining physiological state feedback data after the dynamic intervention scheme is executed, and updating the decision network according to the physiological state feedback data.

[0212] It should be noted that the functions or steps that the above computer readable storage medium or computer device can implement can be referred to the related descriptions of the server side and the user side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0213] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0214] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0215] It should be noted that if non-company software tools or components appear in the embodiments of the present application, they are only used for example introduction and do not represent actual use. The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A dynamic control method for multi-dimensional physiological data, characterized in that: The following steps are involved: Acquiring at least two real-time physiological data representing physiological states from different dimensions from a controlled subject, and recording subjective physical sign data of the controlled subject; fusing the real-time physiological data and the subjective vital sign data to generate time-series aligned fused data; constructing a dynamic correlation model between the real-time physiological data and the subjective vital sign data based on the fused data; Determining a state vector representing the current comprehensive physiological state of the regulated object according to the dynamic association model; Based on the state vector, generating a dynamic intervention plan for regulating the physiological state of the regulated object through a decision network; Acquire physiological state feedback data after executing the dynamic intervention plan, and update the decision network according to the physiological state feedback data.

2. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: Acquiring at least two real-time physiological data representing physiological states from different dimensions from a controlled subject, and recording subjective physical sign data of the controlled subject, including: Collecting sweat metabolism data including tryptophan metabolite ratios of the regulated subject through a wearable sweat sensor; Collect fecal microbial data including bacterial genus abundance of the regulated subjects using an excrement detection device with spectral analysis function; Collecting symptom quantification indicators of the regulated subject through the user terminal interactive interface, and using the symptom quantification indicators as subjective physical sign data; The sweat metabolism data, fecal microbial data and subjective physical sign data are synchronously transmitted to a data processing center.

3. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: The real-time physiological data and the subjective physical sign data are fused to generate time-series aligned fused data, including: identifying a time stamp deviation between sweat metabolism data and fecal microbial data in the real-time physiological data; performing dynamic time compensation processing on the sweat metabolism data and the fecal microbial data according to the timestamp deviation to generate compensated sweat metabolism data and compensated fecal microbial data; Mapping the compensated sweat metabolic data and the compensated fecal microbial data to a physiological rhythm coordinate system to generate mapped sweat metabolic data and mapped fecal microbial data; The mapped sweat metabolism data, the mapped fecal microbial data and the subjective physical sign data are merged to generate time-series aligned fusion data.

4. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: Constructing a dynamic correlation model between the real-time physiological data and the subjective vital sign data based on the fused data, including: Organizing the fused data into a three-dimensional data structure comprising a metabolite dimension, a microbial dimension, and a time dimension; extracting core factors containing metabolite-microorganism temporal correlation features from the three-dimensional data structure; Building a cross-dimensional association strategy library based on the core factors; The cross-dimensional association strategy library is defined as a dynamic association model, and the weight value of the dynamic association model is updated according to the newly added data.

5. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: Determining a state vector representing the current comprehensive physiological state of the regulated object according to the dynamic association model includes: Extracting metabolite-microorganism temporal correlation features at the current moment from the dynamic correlation model; Decomposing the metabolite-microorganism time-related characteristics into metabolite state characteristics, microorganism state characteristics and physical sign state characteristics; The metabolite state features, microorganism state features and physical sign state features are combined to form a state feature set, and the state feature set is used as a state vector.

6. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: Based on the state vector, a dynamic intervention plan for regulating the physiological state of the regulated object is generated through a decision network, including: Inputting the state vector into the policy network in the decision network to generate an initial intervention action; filtering the initial intervention action through a biological constraint filtering layer in the decision network to generate a filtered intervention action; determining a value score of the filtered intervention action by an analysis network in the decision network; The intervention action with the highest value score among the filtered intervention actions is used as the dynamic intervention plan.

7. The dynamic control method of multi-dimensional physiological data according to claim 1, characterized in that: Acquiring physiological state feedback data after executing the dynamic intervention plan, and updating the decision network according to the physiological state feedback data, including: collecting new real-time physiological data and new subjective physical sign data after executing the dynamic intervention plan, and using the new real-time physiological data and new subjective physical sign data as physiological state feedback data; determining a reward value of the decision network based on the physiological state feedback data; The parameters of the policy network and the analysis network in the decision network are updated according to the reward value to obtain an updated decision network.

8. A dynamic control device for multi-dimensional physiological data, characterized in that: The dynamic control device for multi-dimensional physiological data includes: A data acquisition module is used to obtain real-time physiological data from at least two different dimensions representing the physiological state of the regulated subject, and record the subjective physical sign data of the regulated subject; A data fusion module, configured to fuse the real-time physiological data and the subjective vital sign data to generate time-series aligned fusion data; A correlation modeling module, configured to construct a dynamic correlation model between the real-time physiological data and the subjective vital sign data based on the fusion data; a state evaluation module, configured to determine a state vector representing a current comprehensive physiological state of the regulated subject according to the dynamic association model; An intervention generation module, configured to generate a dynamic intervention plan for regulating the physiological state of a regulated subject through a decision network based on the state vector; The feedback optimization module is used to obtain physiological state feedback data after executing the dynamic intervention plan and update the decision network according to the physiological state feedback data.

9. A computer device, characterized in that: The computer device includes a memory, a processor, and a dynamic control program for multi-dimensional physiological data stored in the memory and runnable on the processor. When the dynamic control program for multi-dimensional physiological data is executed by the processor, the steps of the dynamic control method for multi-dimensional physiological data as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The storage medium stores a dynamic control program for multi-dimensional physiological data, which, when executed by a processor, implements the steps of the dynamic control method for multi-dimensional physiological data according to any one of claims 1 to 7.

Images