Diver safety monitoring method and system based on data fusion analysis

Abstract

The present application relates to the technical field of diving monitoring, solves the problem that the diver cannot be effectively monitored in the prior art, and provides a diver safety monitoring method and system based on data fusion analysis. The method comprises: acquiring real-time physical data of a diver and real-time environmental data; analyzing the real-time environmental data to obtain an environmental influence factor reflecting the influence of the current environment on the physical data; correcting the real-time physical data according to the environmental influence factor to obtain target physical data after correction; performing fusion analysis on the target physical data to obtain a real-time state analysis result of the diver; comparing the real-time state analysis result with a preset diver health state template corresponding to the real-time environmental data and performing safety evaluation to obtain a safety evaluation result. The present application can effectively monitor the safety of divers.

Description

Technical Field

[0001] This invention relates to the field of diving monitoring technology, and in particular to a method and system for monitoring diver safety based on data fusion analysis. Background Technology

[0002] In underwater operations, scientific expeditions, military diving, and deep-sea rescue missions, divers often face complex and ever-changing underwater environments, such as rapid changes in water temperature, sudden increases in water pressure, and extreme conditions like deep-sea hypoxia. These environmental changes can easily have a strong impact on the divers' physiological state, potentially inducing health risks such as abnormal heart rate, difficulty breathing, and decreased blood oxygen saturation, and even leading to life-threatening emergencies such as fainting and loss of consciousness. Therefore, to ensure the safety of divers and the successful completion of missions, it is urgent to establish a safety monitoring mechanism that can acquire and analyze divers' vital signs and environmental data in real time, dynamically assessing the divers' health status and providing safety warnings.

[0003] In existing technologies, some safety monitoring solutions only monitor divers using univariate methods by setting vital sign thresholds, such as monitoring whether heart rate is too high or blood oxygen is too low, triggering an alarm when the threshold is exceeded. These methods ignore the combined impact of underwater environmental conditions (such as water temperature, pressure, and depth) on vital sign data, and fail to consider the varying adaptability of divers to different diving tasks, easily leading to false alarms or missed alarms, and are unable to provide individualized health assessments. Furthermore, existing methods lack effective data fusion mechanisms, failing to model and analyze the relationship between multiple vital sign data and complex environmental factors, resulting in inaccurate safety judgments or delayed responses in diving scenarios with high task intensity or large environmental fluctuations.

[0004] Therefore, how to consider the impact of the environment on the measurement of vital signs in complex underwater environments, and how to effectively monitor divers' safety and accurately issue safety warnings by integrating multidimensional vital signs data are urgent problems to be solved. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for monitoring the safety of divers based on data fusion analysis, in order to solve the problem that the prior art cannot effectively monitor the safety of divers.

[0006] The technical solution adopted in this invention is:

[0007] In a first aspect, the present invention provides a method for monitoring diver safety based on data fusion analysis, the method comprising:

[0008] The system acquires real-time vital signs data and real-time environmental data of the diver, including real-time water temperature, real-time water pressure, and real-time water depth, and real-time vital signs data including real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation.

[0009] The real-time environmental data is analyzed to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data;

[0010] Based on the environmental impact factors, the real-time vital signs data are corrected to obtain the corrected target vital signs data;

[0011] The target vital signs data are fused and analyzed to obtain the diver's real-time status analysis results;

[0012] The real-time status analysis results are compared and a safety assessment is performed with a preset diver health status template corresponding to the real-time environmental data to obtain a safety assessment result.

[0013] Preferably, the analysis of real-time environmental data to obtain environmental impact factors reflecting the degree of influence of the current environment on vital sign data includes:

[0014] An environmental template database is constructed based on the type of diving mission, diving time, and individual characteristics of the diver. The environmental template database includes ideal environmental parameter range information corresponding to multiple diving mission types. The ideal environmental parameter range information includes safe water temperature range, safe water pressure range, and safe water depth range.

[0015] The real-time environmental data is compared with the environmental template database to determine whether the real-time water temperature is within the safe water temperature range, whether the real-time water pressure is within the safe water pressure range, and whether the real-time water depth is within the safe water depth range.

[0016] When the real-time water temperature is within the safe water temperature range, the real-time water pressure is within the safe water pressure range, and the real-time water depth is within the safe water depth range, a mapping model between environmental data and environmental impact factors is established based on pre-collected historical environmental data and historical vital sign data. The real-time environmental data is input into the mapping model to obtain the environmental impact factors, wherein the environmental impact factors include a first impact factor corresponding to heart rate, a second impact factor corresponding to respiratory rate, and a third impact factor corresponding to blood oxygen saturation.

[0017] Preferably, the step of establishing a mapping model between environmental data and environmental impact factors based on pre-collected historical environmental data and historical vital sign data includes:

[0018] Collect historical diving data samples, which include historical environmental data such as historical water temperature, historical water pressure, and historical water depth, as well as corresponding historical vital signs data such as historical heart rate, historical respiratory rate, and historical blood oxygen saturation. The collection conditions for the historical diving data samples include different diving mission types, different diving times, and different individual characteristics of divers.

[0019] The historical vital signs data collected from the historical diving data sample are normalized to obtain normalized historical vital signs data.

[0020] Normalized historical vital signs data and corresponding historical environmental data are statistically analyzed. Correlation coefficient analysis, covariance analysis and mutual information analysis are used to calculate the response degree of various vital signs data to various environmental data, forming a sensitivity matrix between environment and vital signs. Each element of the matrix represents the sensitivity weight value of various vital signs data to various environmental data.

[0021] Based on the sensitivity matrix, a multi-channel modeling framework is constructed, and mapping sub-models are constructed for each type of vital sign data. The mapping sub-models take water temperature, water pressure, and water depth as inputs and environmental impact factors corresponding to each type of vital sign data as outputs. The sub-models include weighted regression models, neural network models, or ensemble learning models.

[0022] Preferably, the step of statistically analyzing the normalized historical vital sign data and the corresponding historical environmental data, using correlation coefficient analysis, covariance analysis, and mutual information analysis, to calculate the responsiveness of various vital sign data to various environmental data, and forming a sensitivity matrix between the environment and vital signs, includes:

[0023] For each data sample in the normalized historical vital signs data, according to its collection time point, the water temperature value, water pressure value and water depth value at the corresponding time point are extracted from the normalized historical environmental data to form vital signs environmental sample pairs.

[0024] The vital signs and environmental samples were classified according to the data type of vital signs, and the paired sample sets between each type of vital signs data and each type of environmental data were extracted.

[0025] Response intensity indices are calculated for each of the paired sample sets. The response intensity indices include Pearson correlation coefficient, covariance value and mutual information value, which are used to characterize the linear or nonlinear correlation between vital signs and environmental parameters.

[0026] The response intensity indicators are weighted and fused to obtain the response sensitivity scores of various vital signs data to various environmental data.

[0027] The response sensitivity scores are arranged and combined to construct an initial sensitivity matrix between the vital signs and the environment. Each row of the initial sensitivity matrix corresponds to a type of vital sign data, each column corresponds to a type of environmental data, and each matrix element is the sensitivity score of the corresponding vital sign to the corresponding environment.

[0028] The elements in the initial sensitivity matrix are linearly normalized to convert the sensitivity scores into a standard weight value range, forming a normalized sensitivity matrix between the environment and the physical characteristics.

[0029] Preferably, the step of correcting the real-time vital sign data according to the environmental impact factors to obtain the corrected target vital sign data includes:

[0030] Based on the first influencing factor, the real-time heart rate is corrected to obtain the target heart rate;

[0031] The real-time respiratory rate is corrected based on the second influencing factor to obtain the target respiratory rate;

[0032] Based on the third influencing factor, the real-time blood oxygen saturation is corrected to obtain the target blood oxygen saturation.

[0033] The target vital signs data are determined based on the target heart rate, target respiratory rate, and target blood oxygen saturation.

[0034] Preferably, the step of fusing and analyzing the target vital signs data to obtain the diver's real-time status analysis results includes:

[0035] Obtain the current diving mission type information, and according to the type information, call the corresponding fusion configuration information from the preset fusion analysis parameter library. The fusion configuration information includes the conditional probability distribution information of each vital sign data under each diver's state level, as well as the prior probability information of the state level.

[0036] The target heart rate, target respiratory rate, and target blood oxygen saturation are extracted from the corrected target vital signs data and used as the observation input data for Bayesian inference.

[0037] Based on the conditional probability distribution information defined in the fusion configuration information, the likelihood probabilities of target heart rate, target respiratory rate and target blood oxygen saturation under each diver state level are calculated.

[0038] The likelihood probability of the target vital signs data at each diver state level is multiplied by the prior probability of the corresponding diver state level to obtain the unnormalized posterior probability corresponding to each diver state level.

[0039] The unnormalized posterior probabilities are normalized to obtain the normalized posterior probability distribution for each diver's state level.

[0040] The diver's status level corresponding to the highest probability in the posterior probability distribution is selected as the current real-time health status level of the diver, and the level and its probability confidence are output to form the real-time status analysis result.

[0041] Preferably, the step of calling the corresponding fusion configuration information from a preset fusion analysis parameter library according to the type information includes:

[0042] Obtain a historical diving mission dataset, which includes multiple diving mission types, environmental condition information for each diving mission, individual diver characteristic data, and actual diver status level labels;

[0043] The historical dataset is classified according to the type of diving mission to obtain multiple task type subsets, and each task type subset corresponds to the historical data of a type of diving mission.

[0044] The vital signs data in each task type subset are classified and statistically analyzed according to the corresponding state level labels. The statistical characteristics of various vital signs data under various state levels are calculated, including expected value, variance, and distribution pattern, in order to construct a conditional probability distribution model.

[0045] Based on the aforementioned statistical characteristics, a conditional probability distribution model for various vital signs data under different state levels is established using a parameter fitting method. The conditional probability distribution model includes the probability distribution functions of target heart rate, target respiratory rate, and target blood oxygen saturation under each state level.

[0046] The historical frequency of different state levels under each task type is statistically analyzed, and the prior probability of each state level is calculated based on the frequency distribution to form a set of prior probabilities for each state level.

[0047] The conditional probability distribution model of vital signs and the prior probability set of state levels corresponding to each diving mission type are used as fusion configuration information, indexed by mission type, and stored in the fusion analysis parameter library to complete the construction of the fusion analysis parameter library.

[0048] Obtain the current diving mission type information, and use the type information as the search keyword to retrieve the fusion configuration information corresponding to the current mission type from the fusion analysis parameter library.

[0049] Preferably, the step of comparing the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and conducting a safety assessment to obtain the safety assessment results includes:

[0050] Based on the current diving mission type information and real-time environmental data, the indexing conditions of the diver's health status template are determined. The health status template includes status level threshold settings and coping measures information under different environmental conditions.

[0051] From the preset health status template library, the target health status template that matches the index condition is called. The target health status template records the security recommendations and risk warning strategies corresponding to each status level.

[0052] Based on the real-time status analysis results, obtain the real-time health status level and its corresponding probability confidence level based on Bayesian inference output;

[0053] The real-time health status level is compared with the status level definition in the target health status template to confirm whether the status level belongs to the warning level or the danger level category.

[0054] When the health status level is a warning level or a danger level, a safety intervention suggestion for the current diver is generated based on the response strategy for the corresponding status level in the target health status template. The safety intervention suggestion includes surfacing suggestion, enhanced monitoring, or mission termination.

[0055] The health status level, probability confidence value, and generated safety intervention recommendations are output as the current safety assessment result for the diver.

[0056] Secondly, the present invention provides a diver safety monitoring system based on data fusion analysis, the system comprising:

[0057] The data acquisition module is used to acquire the diver's real-time vital signs data and real-time environmental data. The real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth. The real-time vital signs data includes real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation.

[0058] The environmental data analysis module is used to analyze the real-time environmental data to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data.

[0059] The vital signs data correction module is used to correct the real-time vital signs data according to the environmental impact factors to obtain the corrected target vital signs data.

[0060] The fusion analysis module is used to perform fusion analysis on the target vital signs data to obtain the diver's real-time status analysis results;

[0061] The safety assessment module is used to compare the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and to conduct a safety assessment, thereby obtaining a safety assessment result.

[0062] At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in the first aspect.

[0063] In summary, the beneficial effects of the present invention are as follows:

[0064] The present invention provides a method and system for monitoring diver safety based on data fusion analysis. The method includes: acquiring real-time vital sign data and real-time environmental data of the diver, wherein the real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth, and the real-time vital sign data includes real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation; analyzing the real-time environmental data to obtain an environmental impact factor reflecting the degree of influence of the current environment on the vital sign data; correcting the real-time vital sign data according to the environmental impact factor to obtain corrected target vital sign data; performing fusion analysis on the target vital sign data to obtain a real-time status analysis result of the diver; and comparing the real-time status analysis result with a preset diver health status template corresponding to the real-time environmental data for safety assessment to obtain a safety assessment result. This invention addresses the challenges posed by complex underwater environments to diver vital sign measurement and safety monitoring. It proposes a safety monitoring method based on data fusion analysis, focusing on resolving errors in vital sign measurement caused by environmental interference and the failure of safety monitoring based on single indicators. First, it collects real-time diver vital sign data (such as heart rate, respiratory rate, and blood oxygen saturation) and environmental data (such as water temperature, water pressure, and water depth). By constructing an environmental template database and establishing a sensitivity matrix of the relationship between the environment and vital sign responses, it analyzes the degree of environmental influence on vital signs and generates corresponding environmental impact factors. Next, by establishing a multi-channel mapping model, these environmental impact factors are used to correct the original vital sign data, obtaining target vital sign data that better reflects the actual physiological state. Then, using Bayesian inference methods and considering the current diving mission type, it fuses and analyzes the vital sign data from various dimensions to generate the diver's real-time status level. Finally, the analysis results are compared with a preset health status template, and a safety assessment result is generated based on the status level and confidence level. Specific risk intervention suggestions (such as surfacing or mission termination) are output when necessary. Overall, by employing a technical approach that combines environmental perception modeling, vital sign correction, probabilistic fusion reasoning, and template comparison, accurate identification and early warning of divers' safety status have been achieved. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, and these are all within the protection scope of the present invention.

[0066] Figure 1 This is a schematic diagram of the overall working process of the diver safety monitoring method based on data fusion analysis in Embodiment 1 of the present invention;

[0067] Figure 2This is a flowchart illustrating the process of fusing and analyzing the target vital signs data to obtain the real-time status analysis results of the diver in Embodiment 1 of the present invention.

[0068] Figure 3 This is a structural block diagram of the diver safety monitoring system based on data fusion analysis in Embodiment 2 of the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. In the description of the present invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, the element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Where there is no conflict, embodiments of the present invention and the various features thereof can be combined with each other, all of which are within the scope of protection of the present invention.

[0070] Example 1

[0071] Please see Figure 1 Embodiment 1 of the present invention discloses a method for monitoring the safety of divers based on data fusion analysis, the method comprising:

[0072] The system acquires real-time vital signs data and real-time environmental data of the diver, including real-time water temperature, real-time water pressure, and real-time water depth, and real-time vital signs data including real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation.

[0073] Specifically, real-time vital signs data refers to the diver's vital signs information collected in real time through wearable or wearable sensor devices. This mainly includes heart rate (the number of heartbeats per unit time, usually expressed in bpm), respiratory rate (the number of breaths per unit time, reflecting the activity level of the respiratory system), and blood oxygen saturation (SpO2, representing the proportion of oxygen carried by hemoglobin in the blood, usually expressed as a percentage; the normal value is generally above 95%). These data reflect the diver's current vital signs and are key indicators for assessing their physical condition. Real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth. Real-time water temperature refers to the water temperature at the current diving depth; real-time water pressure represents the pressure exerted by the water on the body, increasing with depth; and real-time water depth refers to the diver's position relative to the surface. These environmental parameters have a direct impact on human physiological state. For example, excessively low real-time water temperature may lead to a decrease in heart rate and slower breathing, while excessive real-time water pressure may cause changes in gas exchange efficiency. The main purpose of this step is to build the foundational data support for subsequent multimodal fusion analysis, achieving a unified model of the human-environment state by simultaneously collecting physiological and environmental variables. Due to the extreme complexity of the diving environment, the performance and meaning of vital signs data may vary under different depths, temperatures, and pressures. Relying solely on uncorrected vital signs data will make it difficult to accurately determine health status. Therefore, environmental information must be incorporated as a prerequisite for subsequent correction, modeling, analysis, and safety warnings. In implementation, smart wearable devices integrating multiple types of sensors are used, such as diving wristbands or wetsuit embedded modules integrating heart rate monitors, pulse oximeters, and respiratory sensors. These devices fit snugly against the diver's skin, measuring heart rate and blood oxygen saturation using photoplethysmography (PPG) and detecting respiratory rate using chest band inductance or volume change methods. Real-time environmental information is acquired through depth gauges, water temperature sensors, and pressure sensors, typically integrated into diving helmets, dials, or backpack equipment. All collected parameters undergo preliminary filtering (such as moving averages and wavelet denoising) to remove signal noise, and are then transmitted in real-time to a host computer or diving safety monitoring platform via low-power wireless communication modules (such as Bluetooth or ultrasonic communication). By simultaneously collecting real-time environmental data and real-time vital sign data, dynamic vital sign modeling under environmental perception can be achieved, providing a data foundation for subsequent correction and judgment.

[0074] The real-time environmental data is analyzed to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data;

[0075] Specifically, by analyzing the collected real-time environmental data, environmental impact factors are constructed to quantify the degree of influence of current environmental conditions on divers' real-time vital signs data. In particular, different environmental variables can interfere with normal vital sign readings to varying degrees; for example, low water temperature may lead to a decrease in heart rate and respiratory rate, while high water pressure may affect the efficiency of blood gas exchange. To quantitatively describe these potential interferences, based on historical diving data samples, statistical methods (such as correlation analysis and mutual information analysis) are used to calculate the sensitivity of environmental variables to vital sign indicators. A mapping model is then constructed to convert real-time environmental inputs into corresponding impact factors. This analysis provides a basis for subsequent environmental adaptation corrections to vital sign data, effectively avoiding misjudgments caused by environmental anomalies and improving the accuracy and reliability of diver health status assessments.

[0076] Preferably, the analysis of real-time environmental data to obtain environmental impact factors reflecting the degree of influence of the current environment on vital sign data includes:

[0077] An environmental template database is constructed based on the type of diving mission, diving time, and individual characteristics of the diver. The environmental template database includes ideal environmental parameter range information corresponding to multiple diving mission types. The ideal environmental parameter range information includes safe water temperature range, safe water pressure range, and safe water depth range.

[0078] Specifically, the environmental template database refers to a knowledge base used to store the ideal environmental conditions required for diving missions. This database uses mission type (e.g., rescue, scientific research, training), mission duration (e.g., short-duration, long-duration), and diver individual characteristics (e.g., age, fitness level, physical adaptability) as indexing criteria to record the ideal environmental parameter ranges for different mission types, including safe water temperature, safe water pressure, and safe water depth intervals. The main purpose of this step is to provide a reference benchmark for subsequent judgment of whether the current environment is within an acceptable range, thereby avoiding misjudgment of the diver's physical condition in abnormal environments. Because the tolerance to different environments varies significantly depending on the specific diving mission and diver individual characteristics, a uniform standard cannot be used for judgment; therefore, a refined condition template database must be established to achieve dynamic adaptability assessment. In the implementation process, historical mission archives and actual diving records are first collected, and the environmental ranges corresponding to successful dives under different mission types are statistically analyzed. Combined with physiological adaptability research results, the extreme tolerance capabilities of different individual divers are modeled, and the safe parameter ranges for different mission types are defined. Finally, this template information is uniformly and structurally stored in the environmental template database to support subsequent real-time comparison and indexing. By combining task characteristics with individual differences, graded and categorized environmental templates are formed, which not only improves the personalization and accuracy of the system's judgment, but also provides a standard reference for pre-dive task planning and dynamic monitoring, and enhances the system's adaptability to sudden environmental changes.

[0079] The real-time environmental data is compared with the environmental template database to determine whether the real-time water temperature is within the safe water temperature range, whether the real-time water pressure is within the safe water pressure range, and whether the real-time water depth is within the safe water depth range.

[0080] Specifically, the real-time water temperature, pressure, and depth values ​​collected by sensors are matched item by item with the ideal parameter ranges defined in the environmental template database to determine whether the current dive is within the safety boundary. This rapid identification of potential risks in the current diving environment serves as a prerequisite for continuing state analysis. If the environment significantly exceeds the safe range, any correction or analysis of vital signs may become meaningless; therefore, a judgment must be made at this point to improve the reliability of subsequent analysis. In practice, the corresponding safe range template is retrieved from the database based on the diving mission type and individual diver information, and then the real-time environmental data is compared with its threshold. The comparison process generally uses a floating-point range judgment algorithm to ensure compatibility with minor measurement errors. If any environmental data exceeds the set range, an environmental anomaly marker is immediately triggered, skipping the vital signs correction step and directly entering the risk warning process.

[0081] When the real-time water temperature is within the safe water temperature range, the real-time water pressure is within the safe water pressure range, and the real-time water depth is within the safe water depth range, a mapping model between environmental data and environmental impact factors is established based on pre-collected historical environmental data and historical vital sign data. The real-time environmental data is input into the mapping model to obtain the environmental impact factors, wherein the environmental impact factors include a first impact factor corresponding to heart rate, a second impact factor corresponding to respiratory rate, and a third impact factor corresponding to blood oxygen saturation.

[0082] Specifically, a mapping model refers to a mathematical or machine learning model used to map environmental parameters (such as water temperature, water pressure, and water depth) to their degree of influence on vital sign data. Examples include multivariate linear regression models, feedforward neural network models, or ensemble regression tree models. Environmental impact factors are intermediate quantities output by these models, representing the degree of interference of a particular environmental combination on heart rate, respiratory rate, and blood oxygen saturation. By quantifying the influence of complex environmental variables, a scientific basis is provided before subsequent environmental interference correction processing of vital sign data. This avoids misjudging normal physiological responses to the body from the environment as health abnormalities, contributing to more accurate individual condition assessments. In implementation, based on historical diving data samples, statistical modeling and multi-channel learning techniques are used to establish independent predictive sub-models for each vital sign indicator. The model inputs are historical water temperature, historical water pressure, and historical water depth, and the output is the estimated interference value for a specific vital sign. During real-time monitoring, newly acquired environmental data is input into the corresponding sub-model, outputting three environmental impact factors corresponding to heart rate, respiratory rate, and blood oxygen saturation, respectively. By enabling quantitative modeling and rapid inference of environmental effects, it significantly improves the ability to interpret the sources of abnormal vital sign data, reduces false alarms, and lays the foundation for personalized health monitoring. Simultaneously, it supports adaptive learning capabilities, continuously optimizing assessment accuracy as new data is accumulated, a crucial element in ensuring the long-term applicability of the monitoring system.

[0083] Preferably, the step of establishing a mapping model between environmental data and environmental impact factors based on pre-collected historical environmental data and historical vital sign data includes:

[0084] Collect historical diving data samples, which include historical environmental data such as historical water temperature, historical water pressure, and historical water depth, as well as corresponding historical vital signs data such as historical heart rate, historical respiratory rate, and historical blood oxygen saturation. The collection conditions for the historical diving data samples include different diving mission types, different diving times, and different individual characteristics of divers.

[0085] Specifically, historical diving data samples refer to complete data records collected during past diving missions. These records include environmental data (such as water temperature, water pressure, and depth) and corresponding vital signs (such as heart rate, respiratory rate, and blood oxygen saturation) at different points in time during the dive. These data samples cover various diving scenarios, such as shallow-water photography missions, deep-sea scientific exploration missions, and high-physical-training missions, using different dive times (such as 10 minutes, 30 minutes, and 1 hour) and the physical conditions of different individuals (such as young amateur divers versus older professional divers) as variables. This categorized collection approach maximizes the reflection of the differences in physiological responses to the human body in real-world diving environments. A representative and comprehensive historical data repository is constructed as the core basis for subsequent modeling. Data quality and diversity directly affect the accuracy and universality of environmental impact factors; therefore, it is essential to establish complete specifications for classification dimensions, coverage boundaries, and recording frequencies at the source of data collection. This step also lays the foundation for data dimensional consistency for subsequent normalization and sensitivity modeling. During implementation, historical data is collected through the following channels: historical mission logs automatically recorded by existing diving equipment (such as integrated diving watches and smart diving suits); diving test data accumulated by research institutions or the military; and manually entered mission documents and manual monitoring results. Before data collection, a standardized data format should be established, for example, using a minute-per-minute recording time step to ensure a one-to-one correspondence between environmental and vital sign data. Furthermore, data cleaning rules need to be established (such as removing missing fields or abnormal jumps) to ultimately form a structured, labeled training sample set. The key technical advantage of this step lies in establishing a multi-dimensional, high-quality sample data source, which is beneficial for the model to fully learn the real relationship between the environment and vital signs under different conditions during the training phase, avoiding a decline in generalization ability due to insufficient samples or bias. In the subsequent system deployment phase, this database can also be continuously updated to achieve long-term adaptive evolution of the model.

[0086] The historical vital signs data collected from the historical diving data sample are normalized to obtain normalized historical vital signs data.

[0087] Specifically, normalization here refers to mapping vital sign data with different dimensions (such as heart rate in beats per minute and blood oxygen saturation in percentage) to the same numerical range using a unified mathematical transformation method (such as maximum / minimum normalization and Z-score standardization). For example, heart rate values ​​may originally be between 60 and 180 beats per minute, but after normalization, they are transformed into values ​​between 0 and 1 for consistency in subsequent calculations. The purpose of this step is to eliminate computational bias caused by scale differences between different vital sign data, ensuring that subsequent statistical analysis or model inputs do not dominate the overall trend due to the excessively large range of a single indicator value. At the same time, normalization also helps improve the convergence speed and stability of model training, and is an indispensable preprocessing step, especially when multivariate analysis is involved. Implementing normalization typically involves two steps: the first step is the statistical analysis stage, where the system scans the maximum, minimum, mean, and standard deviation of each vital sign dimension in all historical samples; the second step is the numerical transformation stage, where all vital sign values ​​are transformed one by one. Normalization should be consistent, meaning subsequent models should also use the same normalization parameters for real-time data preprocessing. By improving the comparability and quality of vital sign data, interference from different units can be avoided and mislead the analytical model. This not only supports data cleaning in the modeling phase but also provides a unified data structure for the subsequent sensitivity matrix construction, helping to form a more accurate environment-vital sign response relationship.

[0088] Normalized historical vital signs data and corresponding historical environmental data are statistically analyzed. Correlation coefficient analysis, covariance analysis and mutual information analysis are used to calculate the response degree of various vital signs data to various environmental data, forming a sensitivity matrix between environment and vital signs. Each element of the matrix represents the sensitivity weight value of various vital signs data to various environmental data.

[0089] Specifically, the sensitivity matrix is ​​a two-dimensional matrix structure where rows represent different vital signs (such as heart rate, respiratory rate, and blood oxygen saturation), and columns represent different environmental data (such as water temperature, water pressure, and water depth). Each element in the matrix represents the response weight (i.e., sensitivity) of that vital sign to that environmental variable. For example, a value of 0.72 in the matrix indicates that the heart rate has a high degree of response to water pressure, with significant changes. The goal of this step is to extract the influence patterns of environmental variables on vital signs, providing a mathematical basis for the subsequent construction of an environmental impact factor model. Because the influence of environmental changes on human physiological states is non-linear and varies greatly among individuals, it is necessary to combine multiple analytical methods to comprehensively assess the strength of the correlation, thereby establishing a more robust explanatory model. In the implementation process, firstly, correlation coefficient analysis is used to measure the strength of linear relationships, suitable for preliminary analysis; secondly, covariance analysis is used to identify whether environmental changes lead to a consistent trend in the direction of vital sign fluctuations; finally, mutual information analysis is introduced to assess non-linear correlations, which is particularly suitable for identifying data pairs with unclear implicit relationships but statistical dependence. Ultimately, the results of the three methods are merged to calculate a weighted sensitivity index, forming a complete sensitivity matrix. This step, through multi-dimensional statistical extraction techniques, not only improves the modeling accuracy of the environment-physical characteristic relationship but also enables parameter sensitivity ranking and variable priority management. This helps the system dynamically adjust parameter collection priorities according to task requirements during subsequent deployments. Simultaneously, the sensitivity matrix can also be used in risk assessment systems to provide early warnings for potentially highly sensitive populations and high-risk environments, enhancing the system's preventative response capabilities.

[0090] Preferably, the step of statistically analyzing the normalized historical vital sign data and the corresponding historical environmental data, using correlation coefficient analysis, covariance analysis, and mutual information analysis, to calculate the responsiveness of various vital sign data to various environmental data, and forming a sensitivity matrix between the environment and vital signs, includes:

[0091] For each data sample in the normalized historical vital signs data, according to its collection time point, the water temperature value, water pressure value and water depth value at the corresponding time point are extracted from the normalized historical environmental data to form vital signs environmental sample pairs.

[0092] Specifically, a vital signs-environment sample pair refers to the combination of normalized vital signs data and environmental data corresponding to the same point in time. For example, at the 15th minute of a historical diving record, the recorded normalized heart rate is 0.64, water temperature is 0.72, water pressure is 0.81, and water depth is 0.65. These data constitute a complete vital signs-environment sample pair. Time point matching is crucial to ensure a one-to-one correspondence, providing a clear input structure for subsequent related analyses. By synchronizing time, the vital signs data are precisely paired with the environmental influences they experienced at that time, providing a data foundation for calculating the degree of environmental influence on vital signs in the next step of the analysis. If the vital signs data and environmental data are not synchronized, it will lead to analysis distortion. Therefore, time alignment is a prerequisite for establishing a reasonable causal chain. In the implementation process, all normalized vital signs samples are traversed, and the water temperature, water pressure, and water depth values ​​at the same time point are searched and extracted from the normalized environmental data based on their timestamps (e.g., minute t, minute t+1). This process can be accelerated using two pointers or index hashing to improve matching efficiency. When environmental data is missing at a certain point in time, it should be filled in using methods such as forward imputation and linear interpolation to avoid omissions. Ultimately, each vital sign sample is accompanied by a set of environmental data values, forming a standardized analysis pair. This step constructs a sample pair set with unified data dimensions and consistent temporal semantics, greatly improving the consistency and effectiveness of data input. Simultaneously, it provides a clearly structured basic unit for subsequent dimensional analysis, response modeling, and matrix calculations, facilitating automated processing and visualization.

[0093] The vital signs and environmental samples were classified according to the data type of vital signs, and the paired sample sets between each type of vital signs data and each type of environmental data were extracted.

[0094] Specifically, classifying samples by data type in this step involves splitting sample pairs according to their vital sign variables (such as heart rate, respiratory rate, and blood oxygen saturation). Each type of vital sign data will form an independent sample set with each environmental variable (such as water temperature, water pressure, and water depth). For example, the system will extract all data pairs related to heart rate and water pressure to form one subset, respiratory rate and water temperature to form another subset, and so on, forming a total of 3×3=9 paired sample sets. The purpose of this step is to establish a dedicated statistical sample pool for each type of vital sign and environmental variable combination, facilitating independent analysis of their correlation and sensitivity in subsequent steps. Through classification, cross-interference between different vital signs can be avoided, making the analysis results more targeted and interpretable. In practice, the vital sign and environmental sample pairs are traversed, and each record is extracted and classified by field, with the 9 sets of samples, such as "heart rate and water temperature," "heart rate and water pressure," and "heart rate and water depth," placed into the corresponding cache sets. Nested dictionaries (such as `dict vital sign type environment type`) can be used in the data structure for efficient access. This classification process supports not only parallel computing but also subsequent independent visualization operations. The resulting paired sample set significantly improves the clarity and accuracy of statistical analysis. The system can formulate specific strategies for each particular combination, such as using a nonlinear model for "blood oxygen and water depth" and a linear fit for "heart rate and water temperature." The advantage of the classification strategy lies in providing flexible modeling space.

[0095] Response intensity indices are calculated for each of the paired sample sets. The response intensity indices include Pearson correlation coefficient, covariance value and mutual information value, which are used to characterize the linear or nonlinear correlation between vital signs and environmental parameters.

[0096] Specifically, response intensity indicators are statistical indicators that quantify whether a certain vital sign variable is affected by a certain environmental variable. These include: the Pearson correlation coefficient, which measures the strength of a linear correlation and ranges from -1 to 1; the covariance, which reflects the directional consistency trend between two variables, with larger values ​​indicating stronger consistency; and the mutual information, which measures the degree of nonlinear information sharing between two variables, with higher values ​​indicating stronger correlation, even without a linear relationship. By comprehensively measuring the response behavior characteristics of vital signs to the environment through multiple statistical dimensions, biases arising from relying on a single indicator are avoided. Parallel linear and nonlinear analysis can help the system identify different types of influencing mechanisms, providing a high-dimensional data foundation for constructing sensitivity scores and final modeling. In the implementation process, each vital sign-environment sample set will be input into three analysis models, calculating three sets of indicators respectively. Efficient implementation can be achieved using numerical computing libraries such as NumPy and SciPy. It should be noted that when the sample size is small, a bias-corrected Pearson formula should be used; when calculating mutual information, discretization or kernel density estimation should be used to improve accuracy. The results of each set of indicators will be uniformly recorded for subsequent weighted fusion processing. This step covers the complete correlation structure from explicit linear to implicit nonlinear, ensuring that both obvious correlations and subtle implicit relationships are captured. The results can be used for model selection (such as linear regression or tree models) and also contribute to enhancing model interpretability and risk control analysis.

[0097] The response intensity indicators are weighted and fused to obtain the response sensitivity scores of various vital signs data to various environmental data.

[0098] Specifically, weighted fusion refers to combining three obtained response intensity indicators (Pearson coefficient, covariance, and mutual information) according to preset or trained weighting factors to calculate a comprehensive score, called the response sensitivity score. For example, in the analysis of heart rate and water pressure, the three indicators are 0.6, 0.4, and 0.7, respectively. The fusion weights are 0.3, 0.2, and 0.5, so the fusion score is 0.3×0.6+0.2×0.4+0.5×0.7. Integrating multi-dimensional statistical indicators into unified and comparable values ​​gives the response intensity of vital signs and the environment a unified dimension, facilitating ranking and matrix construction. The weighting strategy can also reflect the priority of different analysis dimensions in real-world scenarios. For example, in strongly nonlinear environments, the weight of mutual information can be set higher. In the implementation process, static weights (i.e., manually set empirical weights) or dynamic weights (automatically learned weights through data-driven methods such as principal component analysis and information gain method) can be used. In the initial stage of system deployment, a static weighting strategy is generally used; after sufficient training data, a dynamic fusion strategy can be adopted to improve generalization ability. The fusion process should retain all original indicators to facilitate subsequent anomaly analysis and source tracing. This step integrates complex statistical indicators, improving decision-making efficiency and comparative clarity. Especially when constructing the matrix, a unified score effectively avoids the problem of cross-weighting multiple indicators, improving model stability and practicality.

[0099] The response sensitivity scores are arranged and combined to construct an initial sensitivity matrix between the vital signs and the environment. Each row of the initial sensitivity matrix corresponds to a type of vital sign data, each column corresponds to a type of environmental data, and each matrix element is the sensitivity score of the corresponding vital sign to the corresponding environment.

[0100] Specifically, the initial sensitivity matrix is ​​a two-dimensional matrix with dimensions equal to the number of symptom variables multiplied by the number of environmental variables. Each element represents the response sensitivity score. This matrix format not only clearly reflects the relationships between each variable but also serves as the input basis for subsequent modeling or feature selection. Systematizing and structuring the dispersed sensitivity scores constructs a complete symptom-environment response map, supporting both overall system evaluation and providing a basis for anomaly detection and alerting functions in individual paths. During implementation, the system can predefine lists of symptom variables and environmental variables, initialize an empty matrix (e.g., 3×3), and sequentially fill in the fusion scores from the previous stage. The matrix storage structure can use a two-dimensional array or a PandasDataFrame for easy subsequent normalization, sorting, and visualization. This step establishes an efficient index structure and a unified data view, providing structured input for the AI ​​model and facilitating the automatic generation and optimized training of the model's feature matrix. It also provides an intuitive data foundation for expert visual analysis, parameter tuning, and early warning strategy formulation.

[0101] The elements in the initial sensitivity matrix are linearly normalized to convert the sensitivity scores into a standard weight value range, forming a normalized sensitivity matrix between the environment and the physical characteristics.

[0102] Specifically, linear normalization refers to mapping all values ​​in the sensitivity matrix to a standardized range, commonly 0 and 1, to improve the comparability of the response intensity of different characteristics to different environmental parameters. The purpose of this step is to eliminate the influence of different numerical scales on subsequent modeling or weighted analysis, ensuring all sensitivity scores are within a uniform range, facilitating subsequent modeling weight allocation, priority ranking, and visualization rendering (such as heatmaps). In implementation, the system first iterates through all elements in the matrix, recording the maximum and minimum values. Then, it applies the linear normalization formula to each element. Note that in the special case where the maximum and minimum values ​​are equal, division by zero errors should be avoided; in such cases, the value can be uniformly set to 0.5 or skipped. It is recommended to cache intermediate states during the normalization process, retaining the original matrix for comparative analysis. The normalized sensitivity matrix not only has better computational stability but is also easier to integrate into multi-source model systems as input features or adjustment factors.

[0103] Based on the sensitivity matrix, a multi-channel modeling framework is constructed, and mapping sub-models are constructed for each type of vital sign data. The mapping sub-models take water temperature, water pressure, and water depth as inputs and environmental impact factors corresponding to each type of vital sign data as outputs. The sub-models include weighted regression models, neural network models, or ensemble learning models.

[0104] Specifically, the multi-channel modeling framework refers to breaking down vital sign data such as heart rate, respiratory rate, and blood oxygen saturation into multiple independent modeling channels. The mapping sub-model is a mathematical tool that transforms three environmental quantities—water temperature, water pressure, and water depth—into the magnitude of disturbance to the corresponding vital signs. For example, the sub-model of the "heart rate channel" specifically outputs the disturbance amount of heart rate. The purpose of this design is to use prior information from the sensitivity matrix—"which environmental quantity is more sensitive to which vital sign"—to estimate the specific disturbance value in real time and with precision. In implementation, historical vital signs are first mapped one-to-one with environmental data using timestamps. Then, outliers are removed by denoising using moving averages and wavelet thresholding. The most effective environmental input is selected based on the sensitivity matrix and scaled between zero and one. Subsequently, a suitable model is selected for each channel—if linear relationships dominate, weighted least squares regression is used with sensitivity weights embedded in the coefficients for initialization; if nonlinearity is present, a three-layer feedforward neural network is used with weight penalties added to the loss function; if the sample size is small and variable coupling is complex, random forests or gradient boosting trees are used with weight penalties introduced in the split gain. Each channel is optimized using five-fold cross-validation and then solidified for inference services. This channel-specific, weighted, and regularized modeling process ensures the accuracy and interpretability of the model and facilitates independent updates and maintenance for each channel, thereby improving the adaptability and reliability of the entire monitoring system in various diving environments.

[0105] Based on the environmental impact factors, the real-time vital signs data are corrected to obtain the corrected target vital signs data;

[0106] Specifically, the environmental impact factor is the disturbance quantity output in real time by the sub-model in the previous step, used to quantify the specific impact of current water temperature, water pressure, and water depth on various vital sign readings; the target vital sign data is the more realistic value obtained after removing these disturbances from the original readings. Its core purpose is to distinguish between "environmental adaptive changes" and "physiological abnormalities," avoiding misjudging a decrease in heart rate caused by low temperature or a deviation in blood oxygen readings caused by high pressure as a health problem. The implementation process is as follows: The monitoring platform receives raw vital signs and environmental data from the wearable device every second. Environmental parameters are first processed using normalized parameters identical to those used in the training phase, and then fed into the corresponding channel model to calculate heart rate interference, respiratory interference, and blood oxygen interference. At the data level, the raw vital sign signals are smoothed using a second-order unbiased Kalman filter, and compensation is performed according to the type of vital sign using addition or multiplication. For example, the interference value is directly subtracted from heart rate and respiratory rate, while blood oxygen is inversely converted as a percentage. The corrected value is then compared with an individual threshold database. If the threshold is exceeded, a voice or vibration alarm is triggered, and the four-tuple of "raw value, environmental quantity, interference quantity, and corrected value" is written to the local cache and synchronized to the mother ship server via underwater communication. If the interference quantity remains high for dozens of consecutive sampling cycles, the system automatically lowers the threshold and prompts the diver to surface. This link requires only lightweight mathematical operations and small model inference, with a total latency of less than fifty milliseconds. It can significantly reduce the false alarm rate and improve the detection rate of true anomalies, providing fast, accurate, and traceable health assurance for diving operations.

[0107] Preferably, the step of correcting the real-time vital sign data according to the environmental impact factors to obtain the corrected target vital sign data includes:

[0108] Based on the first influencing factor, the real-time heart rate is corrected to obtain the target heart rate;

[0109] Specifically, the first environmental impact factor refers to the estimated disturbance to the diver's heart rate under current environmental conditions, such as the natural decrease in heart rate caused by a drop in water temperature or an increase in water pressure. The purpose of this step is to isolate this non-pathological change caused by the external physical environment from the original heart rate reading, obtaining a target heart rate value that more closely approximates the diver's true physiological state. In practice, the system first filters the original heart rate signal to reduce short-term fluctuations, then inputs the current water temperature, water pressure, and water depth into the heart rate mapping model to obtain the first environmental impact factor. Next, the system subtracts this disturbance value from the original heart rate, or performs a proportional correction according to the model's recommendations, ultimately generating the target heart rate. To prevent abnormal deviations, the corrected heart rate is compared with the individual's historical average heart rate to ensure it remains within an acceptable range. This processing method effectively avoids false alarms caused by adaptive heart rate changes due to environmental factors, helping to more accurately determine whether a genuine health risk exists.

[0110] The real-time respiratory rate is corrected based on the second influencing factor to obtain the target respiratory rate;

[0111] Specifically, the second environmental impact factor reflects the degree of interference with breathing rate under current environmental conditions. For example, a high-pressure environment in deep water may cause the breathing rate to naturally increase or decrease. The purpose of this correction step is to eliminate rhythmic changes caused by environmental factors, ensuring that the system's judgment of a higher or lower breathing rate is not influenced by environmental factors, thus more accurately reflecting the diver's lung and neurological state. In implementation, the platform acquires the raw breathing rate through a chest strap or fitted sensors, then sends real-time environmental data into the mapping model corresponding to the breathing rate to calculate the second environmental impact factor. The correction method can be subtracting the factor value from the raw frequency or adjusting it proportionally; the corrected result is the target breathing rate. This process also includes outlier removal, range limiting, and data write-back operations to ensure stable and reliable results. By introducing a breathing rate correction mechanism, the system can still make reasonable judgments under different water depths, temperatures, or gas concentrations, improving the stability of health status identification.

[0112] Based on the third influencing factor, the real-time blood oxygen saturation is corrected to obtain the target blood oxygen saturation.

[0113] Specifically, the third environmental impact factor refers to an estimated value reflecting the impact of current environmental conditions on blood oxygen saturation measurements. This impact may stem from variations in water temperature, pressure fluctuations affecting oxygen exchange efficiency, or interference with photoelectric measurement accuracy. The core purpose of this correction is to eliminate false decreases or increases in blood oxygen readings caused by environmental factors, thereby more accurately determining whether a diver is at risk of hypoxia. In practice, the system first performs wavelet denoising and averaging filtering on the raw blood oxygen signal, then inputs the real-time environment into the blood oxygen mapping model to obtain the third environmental impact factor. Since blood oxygen data is typically expressed as a percentage, the correction process often employs a multiplicative compensation method, multiplying the original percentage value by a correction factor before standardization to obtain the final target blood oxygen saturation. Simultaneously, this result is cross-validated with the individual's historical range to ensure that model errors do not cause false alarms or missed alarms. This correction strategy effectively improves the anti-interference capability of blood oxygen assessment, significantly reducing error sources, especially in complex underwater lighting and pressure environments.

[0114] The target vital signs data are determined based on the target heart rate, target respiratory rate, and target blood oxygen saturation.

[0115] Specifically, target vital sign data refers to a set of core physiological state indicators formed after correcting for environmental disturbances in three key vital signs. These indicators collectively constitute the basis for judging the diver's current true health status. The main purpose of this step is to unify and aggregate the three independently corrected data points of heart rate, respiratory rate, and blood oxygen saturation into a comprehensive data package that can be directly used for health assessment and risk warning. During implementation, the system encapsulates the target heart rate, target respiratory rate, and target blood oxygen saturation in a unified format, while recording related metadata such as timestamps, dive mission numbers, and diver identification, forming a complete data structure. Subsequently, this target vital sign data will be used in the status analysis model, anomaly detection logic, and visualization platform display. This unified processing not only facilitates subsequent algorithm processing but also supports medical personnel in tracking and analyzing the diver's status at every moment and making remote intervention decisions, further improving the system's real-time response capability and overall health management level.

[0116] The target vital signs data are fused and analyzed to obtain the diver's real-time status analysis results;

[0117] Specifically, the fusion analysis of target vital sign data refers to integrating the three core indicators—corrected heart rate, respiratory rate, and blood oxygen saturation—along with their changing trends, into a holistic health profile. Based on this profile, real-time status labels such as "safe," "mild warning," or "severe warning" are then output. The purpose of this approach is to overcome the limitation of single indicators being susceptible to occasional fluctuations, improving the accuracy and robustness of judgments through cross-verification of multiple indicators. In practice, the system first calculates derived features such as short-term averages, fluctuation amplitudes, and rates of increase or decrease for each vital sign. Then, thresholds and weights are set according to diving operation specifications, and a hierarchical weighted algorithm is used to synthesize a comprehensive score. If sufficient on-site data is available, a classification model based on decision trees or Naive Bayes can be introduced. Within the model, the mapping relationship between various indicator combinations and health status is automatically learned, and the judgment results are continuously updated using a sliding time window. Finally, the platform performs dual verification of the comprehensive score and the model output, generating real-time status analysis results, and providing prompts to divers or command terminals via voice or vibration. Because this fusion analysis considers instantaneous values, trend values, and the coupling characteristics of multiple indicators, it not only significantly reduces single-point failures and false alarm rates, but also enables the early detection of slowly accumulating risks, making rescue and intervention more targeted and timely.

[0118] Preferably, please refer to Figure 2 The fusion analysis of the target vital signs data to obtain the diver's real-time status analysis results includes:

[0119] Obtain the current diving mission type information, and according to the type information, call the corresponding fusion configuration information from the preset fusion analysis parameter library. The fusion configuration information includes the conditional probability distribution information of each vital sign data under each diver's state level, as well as the prior probability information of the state level.

[0120] Specifically, the system first identifies the type of diving mission, such as scientific sampling, equipment maintenance, or emergency rescue. Then, using this as a retrieval key, it retrieves the corresponding fusion configuration information from the fusion analysis parameter library. This fusion configuration information refers to a set of parameters pre-calculated based on historical data from a large amount of similar missions. This includes the conditional probability distributions of heart rate, respiratory rate, and blood oxygen saturation at various warning levels (safe, mild warning, severe warning), as well as the prior probability of each warning level in the given mission scenario. The purpose of this is to ensure that subsequent state determination calculations fully consider the differences in mission intensity and risk, enabling intelligent threshold adjustments tailored to the specific mission. In practice, the mission type identifier is automatically passed in by the mission management module before diving or immediately upon entry. The parameter library uses a key-value index to quickly retrieve the corresponding entry and loads it into memory as input to the real-time fusion algorithm. If the mission is switched mid-dive (e.g., from reconnaissance to rescue), a hot update is triggered to reload the new configuration. The beneficial effects of this technology are as follows: on the one hand, it can finely characterize the correspondence between vital signs and state levels through conditional probability distribution, which can improve the judgment accuracy of Bayesian inference or weighted scoring models; on the other hand, it can achieve sensitive response to special high-risk tasks through the scenario weights inherent in prior probability, thereby significantly reducing false alarms and false negatives and ensuring the real-time safety of divers in variable operating scenarios.

[0121] Preferably, the step of calling the corresponding fusion configuration information from a preset fusion analysis parameter library according to the type information includes:

[0122] Obtain a historical diving mission dataset, which includes multiple diving mission types, environmental condition information for each diving mission, individual diver characteristic data, and actual diver status level labels;

[0123] Specifically, the historical diving mission dataset refers to a comprehensive data package collected and archived during past actual dives. This includes the mission type (e.g., scientific sampling, equipment maintenance, emergency rescue), environmental conditions such as temperature, pressure, and depth, as well as vital signs data of divers during the mission, such as heart rate, respiratory rate, and blood oxygen saturation, and the actual status level (safe, mild warning, severe warning) marked by medical personnel. The purpose of this step is to provide a rich, authentic, and traceable sample source for subsequent statistical modeling, ensuring the probabilistic model is sufficiently representative. In practice, the project management platform exports raw files from diving computers, smart wearable devices, and operation logs, performs time interpolation to complete missing items, manually reviews and removes obvious outliers, and stores them in the analysis database using standardized field formatting. Simultaneously, mission numbers and divers' identities are anonymized to protect privacy. This ensures that subsequent analysis is based on high-quality data, reduces model bias, and improves the reliability of safety assessments.

[0124] The historical dataset is classified according to the type of diving mission to obtain multiple task type subsets, and each task type subset corresponds to the historical data of a type of diving mission.

[0125] Specifically, task type classification refers to splitting mixed historical records into multiple subsets according to predefined task scenario labels, such as storing data from scientific research tasks separately from data from rescue tasks. The aim is to address the significant differences in the impact of different task intensities and work postures on vital signs, making subsequent statistical features more homogeneous. In implementation, data entries can be batch-filtered using the "task type" field or keywords in the task management system, and then a script can automatically write the filtered results into separate data tables or folders. If task types are missing or confused, log records are manually checked for supplementary labeling or merging. The categorized subsets allow the model to more accurately learn the specific relationship between vital signs and risks under each task scenario, improving the targeting of the analysis.

[0126] The vital signs data in each task type subset are classified and statistically analyzed according to the corresponding state level labels. The statistical characteristics of various vital signs data under various state levels are calculated, including expected value, variance, and distribution pattern, in order to construct a conditional probability distribution model.

[0127] Specifically, within each task type subset, based on the three status levels of "safe," "mild warning," and "severe warning" in medical labeling, centralized and discrete statistical analyses are performed on heart rate, respiratory rate, and blood oxygen saturation to generate a summary table of features such as expected value, variance, and skewness. The purpose of this is to extract the core features of the distribution of vital signs under the same task type and risk level, using them as a priori morphological reference for conditional probability distributions. The specific operations include: first, aggregating by status level using database grouping statements or data analysis scripts; then, calling statistical functions to calculate the mean, variance, and distribution pattern (e.g., whether it approximates a normal distribution) of each indicator; if the sample size is insufficient, a bootstrap sampling method is used to expand the confidence interval. The statistical results provide a quantitative basis for subsequent model fitting, reducing the uncertainty caused by loose parameters.

[0128] Based on the aforementioned statistical characteristics, a conditional probability distribution model for various vital signs data under different state levels is established using a parameter fitting method. The conditional probability distribution model includes the probability distribution functions of target heart rate, target respiratory rate, and target blood oxygen saturation under each state level.

[0129] Specifically, a conditional probability distribution model refers to using mathematical functions to describe the probability of a certain value of a vital sign indicator at a given state level. Its purpose is to enable the system to infer the risk level from the vital sign values ​​during real-time analysis. In implementation, firstly, a distribution type consistent with the statistical pattern is selected: if the heart rate is approximately normal, a normal distribution function is used; if the respiratory rate has a right-tailed long tail, a log-normal or gamma distribution is used. Then, the distribution parameters are solved using maximum likelihood estimation or least squares. To prevent overfitting, cross-validation or the Bayesian information criterion can be used to select the optimal model. After completion, three combinations of probability density functions for heart rate, respiratory rate, and blood oxygen saturation are obtained for different state levels. This model enables the system to quickly calculate the posterior risk using methods such as Bayesian inference, improving the accuracy of real-time judgment.

[0130] The historical frequency of different state levels under each task type is statistically analyzed, and the prior probability of each state level is calculated based on the frequency distribution to form a set of prior probabilities for each state level.

[0131] Specifically, prior probability refers to the proportion of each risk level that has occurred independently in a specific task type throughout history, reflecting the inherent risk of the task. The purpose is to assign higher or lower initial risk weights during real-time inference, ensuring the results closely reflect reality. The implementation process is as follows: count the frequency of occurrence of the three state levels within each task subset, then divide by the total number of records in that subset to obtain the percentage; if the data distribution is highly imbalanced, a Laplace correction can be introduced to prevent the probability from being zero. The resulting prior probability set can be directly used with Bayes' theorem, improving the overall model's robustness to extreme cases.

[0132] The conditional probability distribution model of vital signs and the prior probability set of state levels corresponding to each diving mission type are used as fusion configuration information, indexed by mission type, and stored in the fusion analysis parameter library to complete the construction of the fusion analysis parameter library.

[0133] Specifically, the conditional probability distribution model of vital signs and the prior probability of state level corresponding to each of the aforementioned task types are combined into a fusion configuration record, and written into the parameter library using the task type as the index key. The purpose of this library is to provide a standard template for one-click invocation for real-time monitoring, avoiding delays caused by ad-hoc modeling. In implementation, a document-oriented database or a key-value database can be used to serialize model parameters into structured data; version numbers and timestamps are also recorded to support rolling model upgrades. The existence of the parameter library enables the system to load on demand and take effect immediately, significantly reducing the configuration overhead during task switching.

[0134] Obtain the current diving mission type information, and use the type information as the search keyword to retrieve the fusion configuration information corresponding to the current mission type from the fusion analysis parameter library.

[0135] Specifically, before a diver enters the water, the task management module sends a current task type identifier to the monitoring system. Using this identifier as a search keyword, the system quickly retrieves the corresponding fusion configuration from the parameter library, including the conditional probability distribution functions for the three types of vital signs and the prior probabilities for the three risk levels, and loads this information into memory for the real-time fusion algorithm to use. This design aims to implement different thresholds for different tasks, ensuring that the evaluation criteria are synchronized with the intensity of the operation. This can be achieved using a message queue-triggered loading and cache hot-update mechanism to ensure that the switching latency is controlled within milliseconds. Through this step, the system can maintain a balance between judgment sensitivity and false alarm rate in various scenarios such as scientific research and rescue, significantly improving the intelligence level of diving safety assurance.

[0136] The target heart rate, target respiratory rate, and target blood oxygen saturation are extracted from the corrected target vital signs data and used as the observation input data for Bayesian inference.

[0137] Specifically, the target heart rate, target respiratory rate, and target blood oxygen saturation are three core physiological indicators that have been filtered out for environmental interference in the previous steps. For example, a diver's current heart rate is 78 beats per minute, respiratory rate is 17 breaths per minute, and blood oxygen saturation is 97%. The purpose of this step is to convert vital sign readings into usable observables for subsequent probability calculations, ensuring that the analysis focuses only on real physiological fluctuations rather than external noise. In practice, these three values ​​are directly read from the corrected data structure within each sampling period, and timestamps and diver identification numbers are simultaneously added to ensure alignment of subsequent multi-source data. By clearly defining the observation entry point, redundant data parsing is reduced, and rapid retrieval is facilitated, improving the timeliness and traceability of real-time inference.

[0138] Based on the conditional probability distribution information defined in the fusion configuration information, the likelihood probabilities of target heart rate, target respiratory rate and target blood oxygen saturation under each diver state level are calculated.

[0139] Specifically, likelihood probability refers to the probability of observing the current vital signs value when assuming the diver is in a certain state level. The aim is to quantify the fit between observed values ​​and each risk level using a statistical model. In implementation, the conditional probability distribution function of the three vital signs indicators for each state level is first obtained based on the fusion configuration retrieved according to the current task type. Then, the extracted target heart rate, target respiratory rate, and target blood oxygen saturation are substituted into the corresponding functions to obtain three values. The three results are then multiplied (assuming independent conditions) or obtained by looking up a table to obtain the joint likelihood. This approach fully utilizes historical statistical patterns, avoids misjudgments caused by excessively coarse thresholds, and makes the assessment of the diver's state level more consistent with objective laws.

[0140] The likelihood probability of the target vital signs data at each diver state level is multiplied by the prior probability of the corresponding diver state level to obtain the unnormalized posterior probability corresponding to each diver state level.

[0141] Specifically, prior probabilities reflect the proportion of occurrence of each risk level within the same type of task; for example, in rescue missions, mild or severe warnings naturally account for a higher proportion. The purpose of this step is to combine the inherent risk level with the degree of matching with current vital signs to form the initial support for each state level. The specific method is as follows: The prior probabilities of the three risk levels are read one by one, and they are multiplied by the corresponding likelihood probabilities obtained in the previous step to obtain three unnormalized posterior probabilities. This operation follows Bayesian principles, allowing rare but dangerous states to receive appropriate weight in high-risk tasks, reducing underreporting.

[0142] The unnormalized posterior probabilities are normalized to obtain the normalized posterior probability distribution for each diver's state level.

[0143] Specifically, the unnormalized posterior probabilities are normalized. Normalization involves dividing the three unnormalized posterior probabilities by their sum simultaneously, ensuring the sum equals one, resulting in a standardized posterior probability distribution. The purpose is to provide directly comparable probability values ​​for the next step of judgment and to guarantee mathematical consistency. In implementation, this only requires summing the three probabilities and scaling them proportionally; to prevent the denominator from approaching zero, a very small constant is added for smoothing in extreme cases. The normalized probability distribution makes the credibility of each state level immediately apparent and facilitates a visual representation of risk trends on a graphical interface.

[0144] The diver's status level corresponding to the highest probability in the posterior probability distribution is selected as the current real-time health status level of the diver, and the level and its probability confidence are output to form the real-time status analysis result.

[0145] Specifically, the state level corresponding to the maximum value in the normalized posterior distribution is identified as the diver's real-time health status. The purpose of this step is to make the most reasonable diagnosis based on the principle of maximizing probability, while simultaneously providing a confidence level for command personnel to quickly determine whether intervention is necessary. In practice, the platform records the selected level, corresponding probability, and sampling time, and triggers different levels of audio-visual or vibration alerts. If the confidence level is lower than a preset threshold, it will be marked as uncertain and a retest will be recommended. This probability-based automatic judgment method significantly reduces false alarms and missed alarms caused by simple threshold methods, improving the efficiency of underwater operations safety assurance.

[0146] The real-time status analysis results are compared and a safety assessment is performed with a preset diver health status template corresponding to the real-time environmental data to obtain a safety assessment result.

[0147] Specifically, comparing real-time status analysis results with a preset health status template for safety assessment means that after determining the diver's current health status level (e.g., "safe," "mild warning," or "severe warning"), the system further refers to a health status standard template that matches the current environmental conditions to confirm whether the status falls within the normal, acceptable, or high-risk range in the current environment, thus arriving at a more robust safety assessment result. The health status template is a pre-established rule base based on extensive historical diving data, listing the safety range levels of corresponding vital signs under different water depths, temperatures, and pressures. For example, in environments deeper than 30 meters and with temperatures below 15 degrees Celsius, a slightly elevated heart rate and respiratory rate may be a normal compensatory response, while the same values ​​in shallow, high-temperature conditions may indicate fatigue or overheating. The purpose of this step is to prevent the system from misjudging risk solely based on vital sign levels, thereby incorporating environmental adaptability into the comprehensive assessment and improving the accuracy and interpretability of the safety analysis. In its implementation, after completing status identification, the platform simultaneously extracts real-time environmental data such as current water temperature, water pressure, and water depth, and matches the corresponding standard template entries from the health template library. It then compares the currently identified status level with the recommended status level range indicated in the template to determine if it falls within an acceptable range. If the analysis result is higher than the template's recommended level, the system outputs "Abnormal status, environmental adaptation risk warning"; if it is consistent with or lower than the standard, it outputs a status comment such as "Normal" or "High but Acceptable," and records it in the safety assessment report. This dual cross-comparison mechanism, combining environmental conditions and physiological status, effectively improves the accuracy of diving safety assessments, significantly reducing the possibility of false alarms and overlooked risks, especially in complex environments or special missions.

[0148] Preferably, the step of comparing the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and conducting a safety assessment to obtain the safety assessment results includes:

[0149] Based on the current diving mission type information and real-time environmental data, the indexing conditions of the diver's health status template are determined. The health status template includes status level threshold settings and coping measures information under different environmental conditions.

[0150] Specifically, health status templates are pre-defined, structured rule sets used to guide divers' health assessments and risk responses. Templates typically include threshold ranges for vital signs, health status classifications (e.g., normal, warning, dangerous), and corresponding response measures (e.g., reducing intensity, surfacing, or terminating the mission) for different environmental conditions such as water depth, pressure, and temperature. For example, a template might specify a moderate warning state as water temperature <10°C and heart rate above 120 bpm. The purpose of this step is to quickly select the most suitable health assessment template based on the actual diving mission scenario (e.g., deep-sea salvage, equipment maintenance, diving training) and current environmental conditions (e.g., high pressure and low temperature), providing an accurate reference system for subsequent real-time health judgments. Different missions have different risk tolerances and warning mechanisms; therefore, templates must be indexable and mission-sensitive. During implementation, a health status template library needs to be established, and index fields need to be defined for each template, such as mission type = training, water temperature range = 5-15°C, and water pressure range = 1.2-3 atm. Once the current task type and environmental data are obtained, the template that best matches the current conditions is located using a conditional matching algorithm (such as a multi-field combined hash index or a conditional filtering expression tree), serving as a reference for the next step of the judgment. This step realizes the transition from general data to personalized assessment strategies, ensuring that the system can dynamically adjust health judgment rules according to the actual complexity and risk level of the diving task. This dynamic adaptation mechanism improves the flexibility and practicality of the assessment system, avoiding the risk of misjudgment or underreporting.

[0151] From the preset health status template library, the target health status template that matches the index condition is called. The target health status template records the security recommendations and risk warning strategies corresponding to each status level.

[0152] Specifically, template retrieval refers to extracting a template from the established health status template library that perfectly matches or is closest to matching the current index conditions, and loading its content for subsequent real-time status comparison and decision support. The target health status template not only includes the vital sign threshold ranges corresponding to each health status level (such as normal, mild warning, severe warning, and danger), but also includes corresponding safety recommendations and risk handling measures, such as immediately suspending the task and escalating to a higher level under the danger category.

[0153] The purpose of this step is to quickly read and load the complete template content after selecting suitable template index conditions, enabling subsequent real-time health level assessment and risk response to be based on clear, explicit, and structured rules. The template invocation mechanism provides the system with standardized and modular risk assessment criteria. In implementation, template content in the target format can be retrieved from the template database through path mapping or index referencing. The template structure typically includes three main fields: 1) environmental adaptation range, 2) a vital signs and level correspondence table, and 3) level-corresponding recommendations. For example, the template might specify that "heart rate > 130 bpm and water pressure > 2.5 atm" corresponds to a dangerous level and recommend "task abort + rapid ascent." This template is loaded into the cache or model module for subsequent use. This step ensures that the health assessment process has standardized criteria, enabling the entire assessment system to support task customization and strategy reasoning capabilities, while providing the necessary foundation for achieving a closed-loop risk response. Template-based design also facilitates subsequent updates and expansions, such as adding extreme cold diving templates or high-pressure experiment templates.

[0154] Based on the real-time status analysis results, obtain the real-time health status level and its corresponding probability confidence level based on Bayesian inference output;

[0155] Specifically, Bayesian inference is a probability-based reasoning method used to calculate the probability of each health status level (such as normal, warning, danger) occurring under current observed vital signs data and environmental conditions, and outputs the most likely status level and its corresponding confidence level. For example, the system calculates that the current state is "warning" with a confidence level of 87%. The purpose of this step is to input actual observed vital signs (such as real-time heart rate and respiratory rate) into the model, fuse them with historical data distributions and thresholds in the state template, and obtain a dynamic, data-driven health status assessment result. Using Bayesian methods can integrate prior knowledge and current data, providing results with confidence assessments rather than absolute value judgments, thus enhancing the system's robustness. In implementation, the system uses the current vital signs as the observed variable, calls a Bayesian classifier or Bayesian network model trained based on prior probability distributions, and outputs the posterior probability of each state level. For example: Normal (10%), Warning (87%), Danger (3%), and determines the current state level as "Warning" based on the Maximum A posteriori (MAP) principle. This step transforms traditional rule-based judgment into probabilistic modeling, enhancing the system's resilience in complex environments and noisy data, and improving the reliability and interpretability of the diagnosis. The probability output also facilitates integration with subsequent risk strategy systems, such as setting intervention intensity or reporting mechanisms based on confidence levels.

[0156] The real-time health status level is compared with the status level definition in the target health status template to confirm whether the status level belongs to the warning level or the danger level category.

[0157] Specifically, the status level comparison refers to matching the health level output by the Bayesian model with the status levels defined in the template according to rules to determine whether it is a level requiring intervention (such as a warning level or a danger level). For example, if the output level is "severe warning," but the template specifies that it falls within the "immediate intervention required" range, then this level will be marked as an "intervention level." The purpose of this step is to determine whether the current health status triggers a warning or intervention threshold, which is a key judgment node in the risk closed-loop process. Only when the system confirms that the current status is a warning or danger level will the corresponding strategy generation logic be initiated; otherwise, it can be considered a safe status, and only routine records are made. In implementation, the system compares the output level label (e.g., level=2) with the level number table in the template (e.g., 0=normal, 1=mild warning, 2=severe warning, 3=dangerous), and reads the "whether to trigger intervention" boolean flag corresponding to the level, or directly determines whether it is greater than the set threshold (e.g., level≥2). The judgment result will serve as the trigger signal for whether to generate suggestions subsequently. This step ensures the system maintains high sensitivity while avoiding unnecessary overreaction, thus guaranteeing the stability and reliability of the health assessment system. The comparison mechanism can also be designed to support individualized strategies; for example, the trigger threshold can be lowered for high-risk individuals to achieve customized risk control.

[0158] When the health status level is a warning level or a danger level, a safety intervention suggestion for the current diver is generated based on the response strategy for the corresponding status level in the target health status template. The safety intervention suggestion includes surfacing suggestion, enhanced monitoring, or mission termination.

[0159] Specifically, safety intervention recommendations refer to the automatic invocation of corresponding strategies from templates within the system based on the health status level and its meaning, generating specific operational suggestions to guide divers or the command system in taking action. For example, if the level is "dangerous," the system might recommend "immediately and slowly ascend to a safe depth + abort the current mission." The purpose of this step is to establish a closed-loop risk response, automating the process from status assessment to strategy output, ensuring that the system can provide actionable recommendations to protect personnel safety when health risks arise during a diving mission. In implementation, the system reads the response strategy field from the target template that matches the current level. The strategy content can include multiple dimensions: action suggestions (ascend, monitor, terminate mission); priority (high, medium, low); execution method (manual operation, system prompts, remote commands); and auxiliary information (such as estimated ascent time, target depth). The recommendation generation module formats this content, outputting structured commands or graphical prompts. Through this step, the system achieves intelligent intervention capabilities from data perception to proactive feedback, significantly reducing the risk of accidents caused by delayed reactions or human misjudgment. The modular design of the intervention recommendations also supports interface integration with subsequent smart bracelets, wearable devices, remote control platforms, and other devices.

[0160] The health status level, probability confidence value, and generated safety intervention recommendations are output as the current safety assessment result for the diver.

[0161] Specifically, the safety assessment result is a comprehensive, structured output of the diver's current health status, containing three core fields: current status level (e.g., "Warning"), status judgment confidence level (e.g., 87%), and recommended safety strategy (e.g., "ascend immediately and exit the mission"), forming a risk notification output for mission executors or monitoring platforms. The purpose of this step is to integrate the aforementioned judgment and strategy information into a standardized result, enabling the system to provide clear and actionable health assessment information for decision support, automated control, or record archiving. During implementation, each data field is formatted into a JSON structure or UI-displayable data and transmitted to the front-end interface, console, or remote monitoring platform. Some systems can also synchronize the assessment results to the diver's worn equipment, enabling real-time prompts and vibration alarms, enhancing scenario adaptability. This result output mechanism builds an interactive bridge between the system and the user, promoting the rapid implementation of health judgment results in actual operations, and also providing a complete data loop for subsequent statistical analysis, model optimization, and risk review.

[0162] Example 2

[0163] Please see Figure 3 Embodiment 2 of the present invention also provides a diver safety monitoring system based on data fusion analysis, the system comprising:

[0164] The data acquisition module is used to acquire the diver's real-time vital signs data and real-time environmental data. The real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth. The real-time vital signs data includes real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation.

[0165] The environmental data analysis module is used to analyze the real-time environmental data to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data.

[0166] The vital signs data correction module is used to correct the real-time vital signs data according to the environmental impact factors to obtain the corrected target vital signs data.

[0167] The fusion analysis module is used to perform fusion analysis on the target vital signs data to obtain the diver's real-time status analysis results;

[0168] The safety assessment module is used to compare the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and to conduct a safety assessment, thereby obtaining a safety assessment result.

[0169] At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in Example 1.

[0170] Specifically, the diver safety monitoring system based on data fusion analysis provided in this embodiment of the invention includes: a data acquisition module for acquiring real-time vital sign data and real-time environmental data of the diver, wherein the real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth, and the real-time vital sign data includes real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation; an environmental data analysis module for analyzing the real-time environmental data to obtain an environmental impact factor reflecting the degree of influence of the current environment on the vital sign data; a vital sign data correction module for correcting the real-time vital sign data according to the environmental impact factor to obtain corrected target vital sign data; a fusion analysis module for performing fusion analysis on the target vital sign data to obtain the diver's real-time status analysis result; and a safety assessment module for comparing the real-time status analysis result with a preset diver health status template corresponding to the real-time environmental data and conducting a safety assessment to obtain a safety assessment result. This system is designed for monitoring the safety of divers in complex underwater environments. To address the challenges of measuring and monitoring divers' vital signs and ensuring safety, this paper proposes a safety monitoring method based on data fusion analysis. This method focuses on resolving errors in vital sign measurement caused by environmental interference and the failure of safety monitoring based on single indicators. First, real-time collection of divers' vital sign data (such as heart rate, respiratory rate, and blood oxygen saturation) and environmental data (such as water temperature, water pressure, and water depth) is performed. An environmental template database and a sensitivity matrix relating the environment to vital sign responses are constructed to analyze the degree of environmental influence on vital signs and generate corresponding environmental impact factors. Next, a multi-channel mapping model is established to use these environmental impact factors to correct the original vital sign data, obtaining target vital sign data that better reflects the actual physiological state. Then, using Bayesian inference methods and considering the current diving mission type, the vital sign data from various dimensions are fused and analyzed to generate the diver's real-time status level. Finally, the analysis results are compared with a preset health status template, and a safety assessment result is generated based on the status level and confidence level. Specific risk intervention suggestions (such as surfacing or mission termination) are output when necessary. Overall, by employing a technical approach that combines environmental perception modeling, vital sign correction, probabilistic fusion reasoning, and template comparison, accurate identification and early warning of divers' safety status were achieved.

[0171] Specifically, the processor may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement embodiments of the present invention.

[0172] The memory may include a large-capacity storage device for data or instructions. For example, and not limitingly, the memory may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory may include removable or non-removable (or fixed) media. Where appropriate, the memory may be internal or external to a data processing device. In a particular embodiment, the memory is a non-volatile solid-state memory. In a particular embodiment, the memory includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0173] The processor reads and executes computer program instructions stored in memory to implement any of the diver safety monitoring methods based on data fusion analysis in the above embodiments.

[0174] In summary, the embodiments of the present invention provide a method and system for monitoring the safety of divers based on data fusion analysis.

[0175] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0176] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the required tasks. The programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0177] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant locality, and corresponding operation entry points shall be provided for the user to choose to authorize or refuse.

[0178] It should also be noted that the exemplary embodiments mentioned in this invention describe methods or systems based on a series of steps or apparatus. However, this invention is not limited to the order of the steps described above; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0179] The above description is merely a specific embodiment of the present invention. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the protection scope of the present invention.

Claims

1. A method for monitoring diver safety based on data fusion analysis, characterized in that, The method includes: The system acquires real-time vital signs data and real-time environmental data of the diver, including real-time water temperature, real-time water pressure, and real-time water depth, and real-time vital signs data including real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation. The real-time environmental data is analyzed to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data; Based on the environmental impact factors, the real-time vital signs data are corrected to obtain the corrected target vital signs data; The target vital signs data are fused and analyzed to obtain the diver's real-time status analysis results; The real-time status analysis results are compared and a safety assessment is performed with a preset diver health status template corresponding to the real-time environmental data to obtain a safety assessment result. The analysis of the real-time environmental data yields environmental impact factors that reflect the degree of influence of the current environment on vital sign data, including: An environmental template database is constructed based on the type of diving mission, diving time, and individual characteristics of the diver. The environmental template database includes ideal environmental parameter range information corresponding to multiple diving mission types. The ideal environmental parameter range information includes safe water temperature range, safe water pressure range, and safe water depth range. The real-time environmental data is compared with the environmental template database to determine whether the real-time water temperature is within the safe water temperature range, whether the real-time water pressure is within the safe water pressure range, and whether the real-time water depth is within the safe water depth range. When the real-time water temperature is within the safe water temperature range, the real-time water pressure is within the safe water pressure range, and the real-time water depth is within the safe water depth range, a mapping model between environmental data and environmental impact factors is established based on pre-collected historical environmental data and historical vital sign data. The real-time environmental data is input into the mapping model to obtain the environmental impact factors, wherein the environmental impact factors include a first impact factor corresponding to heart rate, a second impact factor corresponding to respiratory rate, and a third impact factor corresponding to blood oxygen saturation.

2. The diver safety monitoring method based on data fusion analysis according to claim 1, characterized in that, The establishment of a mapping model between environmental data and environmental impact factors based on pre-collected historical environmental data and historical vital sign data includes: Collect historical diving data samples, which include historical environmental data such as historical water temperature, historical water pressure, and historical water depth, as well as corresponding historical vital signs data such as historical heart rate, historical respiratory rate, and historical blood oxygen saturation. The collection conditions for the historical diving data samples include different diving mission types, different diving times, and different individual characteristics of divers. The historical vital signs data collected from the historical diving data sample are normalized to obtain normalized historical vital signs data. Normalized historical vital signs data and corresponding historical environmental data are statistically analyzed. Correlation coefficient analysis, covariance analysis and mutual information analysis are used to calculate the response degree of various vital signs data to various environmental data, forming a sensitivity matrix between environment and vital signs. Each element of the matrix represents the sensitivity weight value of various vital signs data to various environmental data. Based on the sensitivity matrix, a multi-channel modeling framework is constructed, and mapping sub-models are constructed for each type of vital sign data. The mapping sub-models take water temperature, water pressure, and water depth as inputs and environmental impact factors corresponding to each type of vital sign data as outputs. The sub-models include weighted regression models, neural network models, or ensemble learning models.

3. The diver safety monitoring method based on data fusion analysis according to claim 2, characterized in that, The process of statistically analyzing the normalized historical vital sign data with the corresponding historical environmental data, employing correlation coefficient analysis, covariance analysis, and mutual information analysis, to calculate the responsiveness of various vital sign data to various environmental data, forming a sensitivity matrix between the environment and vital signs, includes: For each data sample in the normalized historical vital signs data, according to its collection time point, the water temperature value, water pressure value and water depth value at the corresponding time point are extracted from the normalized historical environmental data to form vital signs environmental sample pairs. The vital signs and environmental samples were classified according to the data type of vital signs, and the paired sample sets between each type of vital signs data and each type of environmental data were extracted. Response intensity indices are calculated for each of the paired sample sets. The response intensity indices include Pearson correlation coefficient, covariance value and mutual information value, which are used to characterize the linear or nonlinear correlation between vital signs and environmental parameters. The response intensity indicators are weighted and fused to obtain the response sensitivity scores of various vital signs data to various environmental data. The response sensitivity scores are arranged and combined to construct an initial sensitivity matrix between the vital signs and the environment. Each row of the initial sensitivity matrix corresponds to a type of vital sign data, each column corresponds to a type of environmental data, and each matrix element is the sensitivity score of the corresponding vital sign to the corresponding environment. The elements in the initial sensitivity matrix are linearly normalized to convert the sensitivity scores into a standard weight value range, forming a normalized sensitivity matrix between the environment and the physical characteristics.

4. The diver safety monitoring method based on data fusion analysis according to claim 1, characterized in that, The step of correcting the real-time vital sign data based on the environmental impact factors to obtain the corrected target vital sign data includes: Based on the first influencing factor, the real-time heart rate is corrected to obtain the target heart rate; The real-time respiratory rate is corrected based on the second influencing factor to obtain the target respiratory rate; Based on the third influencing factor, the real-time blood oxygen saturation is corrected to obtain the target blood oxygen saturation. The target vital signs data are determined based on the target heart rate, target respiratory rate, and target blood oxygen saturation.

5. The diver safety monitoring method based on data fusion analysis according to any one of claims 1-4, characterized in that, The fusion analysis of the target vital signs data to obtain the diver's real-time status analysis results includes: Obtain the current diving mission type information, and according to the type information, call the corresponding fusion configuration information from the preset fusion analysis parameter library. The fusion configuration information includes the conditional probability distribution information of each vital sign data under each diver's state level, as well as the prior probability information of the state level. The target heart rate, target respiratory rate, and target blood oxygen saturation are extracted from the corrected target vital signs data and used as the observation input data for Bayesian inference. Based on the conditional probability distribution information defined in the fusion configuration information, the likelihood probabilities of target heart rate, target respiratory rate and target blood oxygen saturation under each diver state level are calculated. The likelihood probability of the target vital signs data at each diver state level is multiplied by the prior probability of the corresponding diver state level to obtain the unnormalized posterior probability corresponding to each diver state level. The unnormalized posterior probabilities are normalized to obtain the normalized posterior probability distribution for each diver's state level. The diver's status level corresponding to the highest probability in the posterior probability distribution is selected as the current real-time health status level of the diver, and the level and its probability confidence are output to form the real-time status analysis result.

6. The diver safety monitoring method based on data fusion analysis according to claim 5, characterized in that, The step of calling the corresponding fusion configuration information from the preset fusion analysis parameter library according to the type information includes: Obtain a historical diving mission dataset, which includes multiple diving mission types, environmental condition information for each diving mission, individual diver characteristic data, and actual diver status level labels; The historical diving mission dataset is classified according to the type of diving mission to obtain multiple mission type subsets, and each mission type subset corresponds to the historical data of a type of diving mission. The vital signs data in each task type subset are classified and statistically analyzed according to the corresponding state level labels. The statistical characteristics of various vital signs data under various state levels are calculated, including expected value, variance, and distribution pattern, in order to construct a conditional probability distribution model. Based on the aforementioned statistical characteristics, a conditional probability distribution model for various vital signs data under different state levels is established using a parameter fitting method. The conditional probability distribution model includes the probability distribution functions of target heart rate, target respiratory rate, and target blood oxygen saturation under each state level. The historical frequency of different state levels under each task type is statistically analyzed, and the prior probability of each state level is calculated based on the frequency distribution to form a set of prior probabilities for each state level. The conditional probability distribution model of vital signs and the prior probability set of state levels corresponding to each diving mission type are used as fusion configuration information, indexed by mission type, and stored in the fusion analysis parameter library to complete the construction of the fusion analysis parameter library. Obtain the current diving mission type information, and use the type information as the search keyword to retrieve the fusion configuration information corresponding to the current mission type from the fusion analysis parameter library.

7. The diver safety monitoring method based on data fusion analysis according to claim 5, characterized in that, The step of comparing the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and conducting a safety assessment to obtain the safety assessment results includes: Based on the current diving mission type information and real-time environmental data, the indexing conditions of the diver's health status template are determined. The health status template includes status level threshold settings and coping measures information under different environmental conditions. From the preset health status template library, the target health status template that matches the index condition is called. The target health status template records the security recommendations and risk warning strategies corresponding to each status level. Based on the real-time status analysis results, obtain the real-time health status level and its corresponding probability confidence level based on Bayesian inference output; The real-time health status level is compared with the status level definition in the target health status template to confirm whether the status level belongs to the warning level or the danger level category. When the health status level is a warning level or a danger level, a safety intervention suggestion for the current diver is generated based on the response strategy for the corresponding status level in the target health status template. The safety intervention suggestion includes surfacing suggestion, enhanced monitoring, or mission termination. The health status level, probability confidence value, and generated safety intervention recommendations are output as the current safety assessment result for the diver.

8. A diver safety monitoring system based on data fusion analysis, characterized in that, The system includes: The data acquisition module is used to acquire the diver's real-time vital signs data and real-time environmental data. The real-time environmental data includes real-time water temperature, real-time water pressure, and real-time water depth. The real-time vital signs data includes real-time heart rate, real-time respiratory rate, and real-time blood oxygen saturation. The environmental data analysis module is used to analyze the real-time environmental data to obtain environmental impact factors that reflect the degree of influence of the current environment on vital sign data. The vital signs data correction module is used to correct the real-time vital signs data according to the environmental impact factors to obtain the corrected target vital signs data. The fusion analysis module is used to perform fusion analysis on the target vital signs data to obtain the diver's real-time status analysis results; The safety assessment module is used to compare the real-time status analysis results with a preset diver health status template corresponding to the real-time environmental data and to conduct a safety assessment, thereby obtaining a safety assessment result. At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in any one of claims 1-7.

Images