A method and platform for dynamically matching popular science content with real-time physiological trends

Abstract

This application provides a method and platform for dynamically matching popular science content based on real-time physiological trends, relating to the field of information push technology. It receives multi-dimensional physiological time-series data from target users in real time; performs activity state trend analysis on the multi-dimensional physiological time-series data to obtain real-time activity stages; calls adjacent activity stages and performs continuity verification; if the continuity verification fails, it traverses the activity record module to call adjacent activity duration intervals and obtains historical related physiological data; it performs quantitative evaluation of the activity status of the historical related physiological data to generate activity quantification coefficients; and matches and pushes popular science content based on the activity quantification coefficients and adjacent activity stages, feeding back to the user's terminal. This application solves the technical problem in existing technologies where the lack of analysis of users' real-time physical condition leads to the inability to accurately match users' immediate health needs with popular science content, improving the accuracy and real-time nature of popular science content push, while also enhancing user satisfaction.

Description

Technical Field

[0001] This application relates to the field of information push technology, specifically to a method and platform for dynamically matching popular science content that integrates real-time physiological trends. Background Technology

[0002] Health education is an important means of raising public health awareness and promoting health management and disease prevention. With the development of information technology, the delivery of health education content is gradually shifting from traditional media to intelligent and personalized approaches.

[0003] Currently, many health education platforms use big data analytics and artificial intelligence to push health knowledge to users based on static information such as age, gender, and medical history. However, these methods are mostly based on basic user attributes and historical data, lacking dynamic perception of users' real-time health status and failing to meet the needs of personalized health management. Existing methods for pushing health education content often ignore the multi-dimensional dynamic characteristics of users' physiological data, failing to effectively capture users' actual health needs under different activity states. Furthermore, with the widespread use of health monitoring devices, these methods fail to fully utilize real-time collected physiological data for in-depth analysis, resulting in content that cannot match users' immediate physiological conditions. This may lead to low content relevance and unsuitability for specific scenarios, ultimately causing a decline in user acceptance and satisfaction. Summary of the Invention

[0004] This application provides a method and platform for dynamically matching popular science content that integrates real-time physiological trends. It solves the technical problem that the lack of analysis of users' real-time physical condition in existing technologies leads to the inability of health popular science content push to accurately match users' immediate health needs. It achieves the technical effect of improving the accuracy and real-time performance of popular science content push, while enhancing user satisfaction.

[0005] In view of the above problems, this application provides a method for dynamically matching popular science content that integrates real-time physiological trends. The method includes: receiving physiological data of a target user in real time through a physiological monitoring module to obtain multi-dimensional physiological time-series data; performing activity state trend analysis on the multi-dimensional physiological time-series data to obtain real-time activity stages; calling adjacent activity stages from an activity recording module based on the end timestamp of the real-time activity stage, and performing continuity verification on the real-time activity stage and adjacent activity stages; if the continuity verification fails, using the adjacent activity stages to traverse the activity recording module to call for continuous duration of each stage, obtaining adjacent activity duration intervals; using the adjacent activity duration intervals to select historically related physiological data from the physiological monitoring module; using the adjacent activity stages as constraints to quantitatively evaluate the activity status of the historically related physiological data, generating an activity quantification coefficient; matching the pushed content based on the activity quantification coefficient and adjacent activity stages to obtain real-time pushed popular science content; and displaying the real-time pushed popular science content on the target user's mobile terminal.

[0006] On the other hand, this application also provides a dynamic matching platform for popular science content that integrates real-time physiological trends. The platform includes: a physiological data acquisition unit, used to receive real-time physiological data of the target user through a physiological monitoring module to obtain multidimensional physiological time-series data; an activity trend analysis unit, used to perform activity state trend analysis on the multidimensional physiological time-series data to obtain real-time activity stages; a continuity verification unit, used to retrieve adjacent activity stages from the activity recording module based on the end timestamp of the real-time activity stage, and perform continuity verification on the real-time activity stage and the adjacent activity stages; and an adjacent activity duration interval determination unit, used to determine the duration interval of adjacent activities if the continuity verification fails, by using the adjacent activity stages... The system iterates through the activity recording module to call the continuous duration of each stage, obtaining the duration interval of adjacent activities; a historical associated physiological data acquisition unit is used to select historical associated physiological data from the physiological monitoring module using the adjacent activity duration intervals; an activity status quantification evaluation unit is used to quantify the activity status of the historical associated physiological data with the adjacent activity stages as constraints, generating an activity quantification coefficient; a push content matching unit is used to match push content according to the activity quantification coefficient and adjacent activity stages to obtain real-time push science popularization content; a science popularization content feedback unit is used to feed back the real-time push science popularization content to the target user's mobile terminal for display. A continuity verification unit is also included.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The physiological monitoring module receives real-time physiological data from target users, obtaining multi-dimensional physiological time-series data to establish a dynamic perception of the user's current physiological state, providing data support for subsequent analysis. Activity trend analysis is performed on the multi-dimensional physiological time-series data to obtain real-time activity stages, clarifying the user's current activity state and laying the contextual foundation for content push notifications. Adjacent activity stages are retrieved from the activity record module based on timestamps, and a consistency comparison is performed between the real-time activity stages and adjacent activity stages to detect the continuity of the user's activity state, ensuring accurate matching between pushed content and user behavior. If inconsistencies are found, the activity record module is iterated through using the adjacent activity stages to retrieve the continuous duration of each stage, obtaining the adjacent activity duration interval. Historical related physiological data is then selected from the physiological monitoring module using this adjacent activity duration interval. Retrospective analysis of historical data improves the reliability and consistency of push notifications. Using the adjacent activity stages as constraints, the activity status of the historical related physiological data is quantitatively evaluated, generating activity quantification coefficients to provide a quantitative basis for the selection and matching of pushed content. Based on the activity quantification coefficient and adjacent activity stages, push content is matched to obtain real-time pushed science popularization content. This ensures that the pushed health science popularization content is highly relevant to the user's current state, improving the personalization and targeting of science popularization content. The real-time pushed science popularization content is then displayed on the target user's mobile terminal, achieving instant feedback and information delivery.

[0009] In summary, this application accurately identifies a user's current activity status and health needs through real-time multidimensional physiological data collection, activity phase trend analysis, and quantitative evaluation based on historical data correlation. Combined with a personalized content matching algorithm, it pushes health science content that is highly relevant to the user's actual status, significantly improving the accuracy and real-time nature of health science content delivery, while also enhancing user experience and satisfaction, and facilitating the dissemination of personalized health management knowledge.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0011] Figure 1 A flowchart illustrating the method for dynamically matching popular science content that integrates real-time physiological trends, as provided in this application embodiment.

[0012] Figure 2The flowchart of the method for dynamically matching popular science content that integrates real-time physiological trends provided in this application embodiment is as follows: the activity trend analysis of the preprocessing results is performed by the activity stage judgment model to obtain the real-time activity stage.

[0013] Figure 3 A flowchart illustrating the process of obtaining the activity quantification coefficient in the dynamic matching method for popular science content that integrates real-time physiological trends, as provided in the embodiments of this application.

[0014] Figure 4 A schematic diagram of the structure of the science popularization content dynamic matching platform that integrates real-time physiological trends provided in the embodiments of this application.

[0015] Figure labeling: Physiological data acquisition unit 10, activity trend analysis unit 20, continuity verification unit 30, adjacent activity duration interval determination unit 40, historical correlation physiological data acquisition unit 50, activity status quantitative evaluation unit 60, push content matching unit 70, popular science content feedback unit 80. Detailed Implementation

[0016] This application provides a method and platform for dynamically matching popular science content that integrates real-time physiological trends. It utilizes a physiological monitoring module to receive physiological data from target users, obtaining multidimensional physiological time-series data. Activity trend analysis is performed on this multidimensional physiological time-series data to determine real-time activity stages. By comparing the data with adjacent activity stages in the activity recording module, discrepancies are determined. When discrepancies are found, the continuous duration of the adjacent activity stages is retrieved from the activity recording module to obtain the duration interval of the adjacent activities, and historical related physiological data is obtained from the physiological monitoring module. Using adjacent activity stages as constraints, the activity status of the historical related physiological data is quantitatively evaluated to generate an activity quantification coefficient. Finally, content is matched and pushed based on the activity quantification coefficient and adjacent activity stages, and the real-time popular science content is displayed on the target user's mobile terminal.

[0017] Example 1, as Figure 1 As shown in the embodiments of this application, a method for dynamically matching popular science content that integrates real-time physiological trends is provided. The method includes:

[0018] Step S1: Receive the target user's physiological data in real time through the physiological monitoring module to obtain multidimensional physiological time-series data.

[0019] Specifically, the physiological monitoring module is a communication terminal connected to one or more physiological monitoring devices, capable of acquiring the target user's physiological data from these devices. These physiological monitoring devices can be smartwatches, health monitoring bracelets, etc., which collect multi-dimensional physiological data such as heart rate, blood pressure, respiratory rate, and blood oxygen saturation through sensors. For example, a heart rate sensor records the number of heartbeats per minute, a blood oxygen sensor measures the oxygen saturation in the blood, and a skin conductance sensor monitors sweat gland activity. The acquired physiological data of the target user is then sorted according to time series to obtain multi-dimensional physiological time-series data.

[0020] By collecting multidimensional physiological time-series data, we can perceive users' physical state in real time, providing data support for subsequent analysis and ensuring that popular science content is aligned with users' real-time health needs.

[0021] Step S2: Perform activity state trend analysis on the multidimensional physiological time series data to obtain the real-time activity stage.

[0022] Specifically, real-time activity stage refers to the user's current behavioral or physiological state. In-depth analysis of multidimensional physiological time-series data reveals the changing trends of various physiological indicators over time, thereby determining the user's current activity state. For example, analyzing changes in heart rate and step count can determine whether the user is sleeping, engaging in light exercise, or strenuous exercise. This facilitates subsequent targeted actions based on the user's activity state.

[0023] Step S3: Based on the end timestamp of the real-time activity stage, retrieve the adjacent activity stage from the activity record module, and perform a continuity check on the real-time activity stage and the adjacent activity stage.

[0024] Specifically, the activity recording module is a system module that stores historical activity stage information, recording the user's activity data over a period of time, including the start time and duration of each activity stage, facilitating subsequent retrieval and analysis. Each time new physiological data is received from the physiological monitoring module, a corresponding timestamp is generated. This timestamp marks the specific point in time when the multidimensional physiological time-series data was generated. Using the timestamp as an index, the activity recording module searches for previous activity stages adjacent to the current real-time activity stage, then compares the activity type, duration, and physiological indicators of the two to determine if they are similar or identical. By comparing with adjacent activity stages, it is possible to detect whether the user's activity state has undergone a sudden change. This comparison helps to promptly detect transitions from one activity state to another; for example, a sudden change from sedentary to walking interrupts continuity.

[0025] Step S4: If the continuity check fails, the activity record module is traversed through the adjacent activity stages to call the stage continuity duration and obtain the duration interval of the adjacent activities.

[0026] Specifically, when a discrepancy is found between the real-time activity phase and adjacent activity phases in step S3, the historical activity records are traversed based on the adjacent activity phases. The duration of the adjacent activity phase is then retrieved from the activity record module, i.e., the adjacent activity duration interval. By obtaining the adjacent activity duration interval, a comprehensive understanding of the user's activity history and changes in physical condition can be obtained, providing supplementary information for further in-depth analysis.

[0027] Step S5: Obtain historical associated physiological data by selecting from the physiological monitoring module using the adjacent activity duration intervals.

[0028] Specifically, a time window is defined using the duration interval of adjacent activities. The physiological monitoring module then filters out physiological data associated with the user's activity status within this time range—that is, historically correlated physiological data. This historical data reflects the user's physiological condition during similar past activity phases, providing a historical perspective for a comprehensive assessment of the user's physical condition. This, in turn, provides a more accurate and reliable basis for subsequent quantitative evaluation of activities.

[0029] Step S6: Using the adjacent activity stages as constraints, perform a quantitative evaluation of the activity status of the historical associated physiological data to generate an activity quantification coefficient.

[0030] Specifically, using the adjacent activity stages determined in step S3 as constraints, the historical related physiological data obtained in step S5 is input into a pre-set quantitative evaluation algorithm. This algorithm calculates these data according to specific rules, ultimately obtaining a numerical value representing the activity status, namely, the activity quantification coefficient. For example, for the historical related physiological data of the "running" stage, a quantification coefficient representing the running effect is calculated by comparing indicators such as heart rate and exercise speed with standard values. This coefficient is an important indicator for comprehensively evaluating the user's physical activity status, providing a quantitative basis for matching subsequent popular science content.

[0031] Step S7: Match the pushed content with the activity quantification coefficient and adjacent activity stages to obtain real-time pushed science popularization content.

[0032] Specifically, the activity quantification coefficient obtained in step S6 and the adjacent activity stages in step S3 are used as matching conditions. The system then filters through a pre-prepared health science content library to find content most suitable for the user's current state, i.e., real-time push of science content. This ensures that the pushed content is highly relevant to the user's physical condition and activity stage. For example, if the activity quantification coefficient indicates excessive exercise intensity, and the adjacent activity stage is "running," then articles on how to reasonably control running intensity will be found in the health science content library.

[0033] Step S8: The real-time pushed science popularization content is fed back to the target user's mobile terminal for display.

[0034] Specifically, the real-time push science popularization content obtained in step S7 is sent to the target user's mobile terminal via a network communication protocol, and then displayed on the mobile terminal's screen, allowing the user to receive personalized health science popularization information on the mobile terminal. For example, an article about reasonably controlling running intensity can be displayed in a smartphone's health application.

[0035] Furthermore, step S2 in this embodiment of the application also includes:

[0036] Step S21: Pre-construct an activity stage judgment model, wherein the activity stage judgment model includes multiple activity stage judgment branches connected in parallel, and the multiple activity stage judgment branches have multiple sample activity stage identifiers.

[0037] Step S22: After preprocessing the multidimensional physiological time series data using a multi-level window mechanism, the activity trend analysis of the preprocessing results is performed using the activity phase judgment model to obtain the real-time activity phase.

[0038] Specifically, an activity stage determination model is constructed to identify the user's activity stage based on multidimensional physiological data. This model consists of multiple parallel decision branches, each identifying a different type of activity stage and labeled with a corresponding sample activity stage identifier, such as "sleep," "rest," or "exercise." For example, the activity stage determination model comprises three decision branches: a sleep stage branch, an exercise stage branch, and a rest stage branch. These branches have corresponding input physiological data types set according to the characteristics of their respective activity stages. The input physiological data types for the sleep stage branch include heart rate, respiratory rate, and acceleration data; the input physiological data types for the exercise stage branch include acceleration, cadence, and heart rate; and the input physiological data types for the rest stage branch include EDA, HRV, and heart rate acceleration.

[0039] After the model is built, a multi-level window mechanism is applied to divide the time windows into different levels to preprocess the multi-dimensional physiological time series data. For example, the multi-level window mechanism sets short-term windows (e.g., 10 minutes) to focus on recent data changes and long-term windows (e.g., 1 hour) to grasp the overall trend. The data within the different time windows divided by the multi-level window mechanism are input into multiple judgment branches of the activity stage judgment model for activity trend analysis. The activity stage judgment model combines the short-term and long-term data results to determine the user's most likely current activity stage, i.e., the real-time activity stage.

[0040] Furthermore, step S21 also includes:

[0041] Step S211: Interact to obtain multiple sets of associated physiological indicators for the various sample activity stages.

[0042] Step S212: Collect sample data with the first associated physiological index group as a constraint to obtain multiple sample physiological data groups and multiple sample stage probability coefficients.

[0043] Step S213: Use the multiple sample physiological data sets and multiple sample stage probability coefficients as training data to construct a first activity stage judgment unit for judging the activity stage of the first sample.

[0044] Step S214: Load the data filter according to the first associated physiological indicator group, obtain the first adaptive data filter, and configure the first adaptive data filter at the front end of the first activity stage judgment unit to complete the construction of the first activity stage judgment branch.

[0045] Step S215: By analogy, sample data is collected and model is constructed using the multiple sets of related physiological indicators as constraints to obtain the multiple activity stage judgment branches.

[0046] Step S216: Complete the construction of the activity stage judgment model by connecting the multiple activity stage judgment branches in parallel.

[0047] Specifically, through information exchange, the types and numerical ranges of physiological indicators corresponding to different activity stages are determined from medical research findings and expert experience, i.e., associated physiological indicator groups. For example, for the sleep activity stage, associated physiological indicator groups include heart rate (50 to 70 beats per minute) and respiratory rate (12 to 20 breaths per minute).

[0048] One set of related physiological indicators is randomly selected from multiple sets, designated as the first set, corresponding to the first sample activity stage. Using the first set as a constraint, physiological data conforming to the data type and numerical range of the first set are collected from historical data to obtain multiple sample physiological data sets. During the data collection process, the collected data are analyzed and processed to calculate the probability coefficient of each sample physiological data set belonging to the first sample activity stage. For example, for a sample physiological data set containing data such as heart rate and respiratory rate, its sample stage probability coefficient is calculated based on the degree of matching with the "sleep" related physiological indicator set, by calculating the similarity and deviation between the data and standard indicators.

[0049] Multiple sets of collected physiological data and corresponding probability coefficients for different sample stages are used as input data. A first activity stage determination unit is constructed using machine learning algorithms or rule-based programming logic. This first activity stage determination unit is used to determine whether the target user is in the first sample activity stage. For example, using a decision tree algorithm in machine learning, each physiological indicator in the sample physiological data set is used as input features, and the sample stage probability coefficients are used as labeling results to train a decision tree model. This model serves as the first activity stage determination unit for determining the first sample activity stage ("sleep").

[0050] Since the physiological data collected by the physiological monitoring module typically contains many dimensions, and the first activity stage judgment unit can only process data of the same type as the first associated physiological indicator group, a data filter is constructed based on the data type of the first associated physiological indicator group to filter out data that does not conform to the type of the first associated physiological indicator group. This filter is denoted as the first adaptive data filter. After constructing the first adaptive data filter, it is placed at the front end of the first activity stage judgment unit, forming the first activity settlement judgment branch. When data is input, it first undergoes preliminary screening through this filter. Only data that conforms to the data type of the first associated physiological indicator group will enter the first activity stage judgment unit for further judgment.

[0051] Following the same steps as constructing the first activity stage judgment branch, sample data is collected for other related physiological indicator groups, sample stage probability coefficients are calculated, activity stage judgment units are constructed, and data filters are loaded, resulting in multiple activity stage judgment branches. These multiple constructed activity stage judgment branches are combined in parallel to form a complete activity stage judgment model. When physiological data is input, these judgment branches can simultaneously analyze and judge the data, and the results of multiple branches are combined to determine the user's activity stage.

[0052] Furthermore, such as Figure 2 As shown, step S22 in this embodiment further includes:

[0053] Step S221: Divide the time window according to the multi-level window mechanism to obtain the long-term monitoring window and the short-term monitoring window.

[0054] Step S222: Using the short-term monitoring window as a constraint, select the first-stage time series data from the physiological monitoring module.

[0055] Step S223: In the activity stage judgment model, after pre-screening the first stage time series data through multiple adaptive data filters, the filtering results are analyzed through the multiple activity stage judgment branches to output multiple real-time stage probability coefficients.

[0056] Step S224: Using the short-term monitoring window as a constraint, cyclically select the phase time series data of the physiological monitoring module, and perform real-time phase probability judgment through the activity phase judgment model until the real-time reception duration of the physiological data of the physiological monitoring module meets the long-term monitoring window, and output M sets of real-time phase probability coefficients.

[0057] Step S225: Perform activity trend analysis based on the M groups of real-time stage probability coefficients to obtain the real-time activity stages.

[0058] Specifically, based on the established multi-level window mechanism, long-term monitoring windows and short-term monitoring windows are divided. For example, the long-term monitoring window is set to 2 hours, and the short-term monitoring window is set to 10 minutes. By dividing the time window into different time scales, physiological data can be observed from different time scales. The long-term window can reflect the long-term trend of physiological data, while the short-term window can capture rapid changes in physiological data, so as to achieve rapid response to real-time activity status.

[0059] Using a short-term monitoring window as a constraint, the most recent data is selected from the multidimensional physiological time-series data received by the physiological monitoring module and recorded as the first-stage time-series data. For example, data such as heart rate and steps recorded by a smart bracelet in the most recent 10 minutes are extracted.

[0060] The first-stage time-series data is input into the activity stage judgment model. Each judgment branch in the model first uses its own adaptive data filter to perform preliminary screening of the first-stage time-series data, filtering out unnecessary data. Then, the filtered data is input into its respective judgment unit for analysis, outputting multiple real-time stage probability coefficients representing the likelihood of different activity stages. These real-time stage probability coefficients reflect the probability that the target user is currently in the corresponding activity stage.

[0061] The system continuously selects new time-series data from the physiological monitoring module, using short-term monitoring windows as units. After each selection, a real-time stage probability determination is performed using an activity stage judgment model. This process is repeated until the total duration of the collected physiological data meets the requirements of a long-term monitoring window. At this point, the model outputs M sets of real-time stage probability coefficients, where M is a positive integer, calculated by dividing the long-term monitoring window duration by the short-term time window duration and rounding up. These M sets of real-time stage probability coefficients reflect the probability that the target user is in each activity stage at different time points.

[0062] Activity trend analysis is performed based on the obtained M groups of real-time stage probability coefficients. By analyzing the changing trends of these probability coefficients, the current activity stage of the user is determined. This allows for accurate identification of the user's activity status, and corresponding health education content is provided based on this information.

[0063] Furthermore, step S225 of this application embodiment also includes:

[0064] Step S225-1: By decomposing the M groups of real-time stage probability coefficients, obtain multiple stage time-series distributions corresponding to the various sample activity stages.

[0065] Step S225-2: Interact to obtain historical activity stage data, and perform stage transition probability analysis on the various sample activity stages based on the historical activity stage data, and construct a stage transition probability matrix based on the analysis results.

[0066] Step S225-3: Calculate the hourly probability distribution of the multiple stage time series distributions using the stage transition probability matrix to obtain multiple dynamic stage probability sequences.

[0067] Step S225-4: Reorganize the multiple dynamic stage probability sequences according to the M collection times of the M groups of real-time stage probability coefficients to obtain the M groups of node dynamic probabilities.

[0068] Step S225-5: Maximize the dynamic probabilities of the M groups of nodes to obtain the M dominant activity stages and the M dominant activity probabilities.

[0069] Step S225-6: After sorting the M dominant activity stages according to the M collection times, perform stage transition stability analysis based on the probabilities of the M dominant activities, and output the real-time activity stages.

[0070] Specifically, the M sets of real-time stage probability coefficients are decomposed according to their order and the type of sample activity stage, resulting in multiple stage time-series distributions. These time-series distributions represent the probability of each sample activity stage at different time points. Each sample activity stage may have a different time-series distribution, which describes the duration and frequency of different activity stages. For example, if there are three sample activity stages (such as sleep, movement, and stillness), then for each set of real-time stage probability coefficients, the probability values ​​of these three activity stages at the corresponding time points can be decomposed, thus obtaining three stage time-series distributions.

[0071] By interacting with modules in physiological monitoring devices (such as smartwatches and health monitoring bracelets) that store historical activity phase data, this method obtains historical activity phase data for the target user over a period of time (e.g., a week, a month). This data reflects, to some extent, the target user's activity patterns, such as waking up at 8 am on a weekday and then running for one hour. Statistical analysis software is used to analyze this historical data, determining the transition probabilities between different activity phases. For example, what is the probability of transitioning from a sleep state to a light exercise state, and what is the probability of transitioning from a light exercise state to a vigorous exercise state? Based on the analysis results, a phase transition probability matrix is ​​constructed, where each element represents the probability of transitioning from one activity phase to another.

[0072] Using the constructed stage transition probability matrix, the probability distributions of the previously obtained stage time series distributions are calculated hourly. That is, for each time point, based on the current stage time series distribution and the stage transition probability matrix, the probability distributions of each activity stage at the next time point are calculated, thus obtaining multiple dynamic stage probability sequences. This sequence reflects the changes in the target user's activity stages within a certain time period.

[0073] The previously obtained multiple dynamic stage probability sequences are recombined according to the collection time of the M groups of real-time stage probability coefficients to obtain M groups of node dynamic probabilities. These node dynamic probabilities represent the probability of the user's activity stage at each specific time point. For example, if the M groups of real-time stage probability coefficients are collected every 10 minutes, then the dynamic stage probability sequence is divided into M groups of node dynamic probabilities at 10-minute intervals, with each group of node dynamic probabilities corresponding to a collection time point.

[0074] For each group of node dynamic probabilities, the activity phase with the highest probability value is identified as the dominant activity phase of the current node. This dominant activity phase represents the most likely activity state of the target user at the current node, and the probability value corresponding to the dominant activity phase is the dominant activity probability. This operation is performed on each of the M groups of node dynamic probabilities to obtain M dominant activity phases and M dominant activity probabilities. Each dominant activity phase and probability corresponds to a data collection time point.

[0075] The M dominant activity phases are sorted according to M collection times, resulting in a chronological sequence of dominant activity phases. Then, the stability of transitions between these dominant activity phases is analyzed by combining the M dominant activity probabilities. For example, if a dominant activity phase maintains a high dominant activity probability across multiple consecutive collection times, then this activity phase can be considered to have high stability. Through this stability analysis, the user's real-time activity phase is ultimately determined and output.

[0076] Furthermore, the embodiment of this application includes the following steps before step S6:

[0077] Step S-1: Predefine multiple standard durations for the various sample activity phases.

[0078] Step S-2: Call the first standard duration of the first sample activity phase from the multiple standard durations.

[0079] Step S-3: Collect sample data with the first associated physiological indicator group and 1 / M of the first standard duration as constraints to obtain multiple sample physiological monitoring groups and multiple sample activity quantification coefficients.

[0080] Step S-4: Use the multiple sample physiological monitoring groups and multiple sample activity quantification coefficients as training data to construct the first activity evaluation model.

[0081] Step S-5: By analogy, construct multiple activity evaluation models for the various sample activity stages, and complete the construction of the activity evaluation model library by associating and storing the various sample activity stages and multiple activity evaluation models.

[0082] Specifically, an activity assessment model is a model that evaluates activity phases based on specific physiological data and standard durations. By inputting different physiological indicators, the activity assessment model can determine whether a particular activity phase meets predetermined standards, or whether changes in its physiological characteristics are within the normal range.

[0083] Based on expert experience and statistical analysis of population activities, standard durations for each activity phase are pre-defined. For example, the standard duration for the exercise phase is set at 30 minutes, and the standard duration for the sleep phase is set at 7 hours. These standard durations provide a benchmark for subsequent activity evaluation to identify whether the activity is normal.

[0084] The first standard duration corresponding to the first sample activity phase is retrieved from multiple standard durations. Sample data is collected using the first associated physiological indicator group and 1 / M of the first standard duration as constraints, resulting in multiple sample physiological monitoring groups. Here, M is the ratio of the long-term monitoring window to the short-term monitoring window in the multi-level window mechanism. During the data collection process, a sample activity quantification coefficient is calculated for each collected sample physiological monitoring group. For example, for a 10-minute running sample physiological monitoring group (including heart rate, steps, etc.), the average, maximum, variance, or standard deviation of each physiological indicator is calculated. Then, based on factors such as the ratio of heart rate to the normal running heart rate range and the ratio of steps to normal running steps, a weighted average method is used to calculate a sample activity quantification coefficient representing the running effect.

[0085] Multiple physiological monitoring groups and their corresponding activity quantification coefficients are collected as training data. A primary activity evaluation model is constructed using machine learning or statistical analysis methods. For example, if regression analysis is used in machine learning, physiological indicators from the monitoring groups are used as independent variables, and activity quantification coefficients are used as dependent variables to construct a regression model for primary activity evaluation. This model can predict the activity quantification coefficients of the first sample based on the input physiological data, thereby evaluating the activity status.

[0086] Following the method used to construct the first activity evaluation model, the same model construction operation is performed sequentially for other sample activity stages to obtain multiple activity evaluation models corresponding to multiple sample activity stages. All the constructed activity evaluation models for various sample activity stages are stored according to their relationships, forming an activity evaluation model library. This activity evaluation model library allows for the quantitative evaluation of each stage of a user's activities.

[0087] Furthermore, such as Figure 3 As shown, step S6 in this embodiment further includes:

[0088] Step S61: Schedule real-time activity evaluation models from the activity evaluation model library according to the adjacent activity stages.

[0089] Step S62: Call the real-time standard duration from the multiple standard durations according to the adjacent activity phases.

[0090] Step S63: Segment the historical associated physiological data according to 1 / M of the duration of the real-time standard to obtain multiple historical associated subsets.

[0091] Step S64: Use the real-time activity evaluation model to evaluate the multiple historical correlation subsets and output multiple activity quantification sub-values.

[0092] Step S65: Obtain the activity quantization coefficient by performing time-series weighting on the multiple activity quantization sub-values.

[0093] Specifically, after a target user's previous activity ends, the system evaluates the user's activity in real time to obtain a corresponding activity quantification coefficient. This coefficient is then used to push health education content. For example, after a user wakes up, the system evaluates the user's sleep activity based on sleep phase monitoring data and then pushes health advice about sleep.

[0094] Based on the type of adjacent activity phases, a search is performed in the activity evaluation model library to find a matching activity evaluation model, which is denoted as the real-time activity evaluation model. Simultaneously, based on the type of adjacent activity phases, the real-time standard duration corresponding to that activity phase is selected from multiple standard durations.

[0095] Based on the duration of the real-time standard, the historically associated physiological data is divided into multiple historically associated subsets at a ratio of 1 / M. For example, historically associated physiological data consists of a series of physiological data such as heart rate and steps recorded in chronological order. Dividing the data into 10-minute intervals, each of these data segments constitutes a historically associated subset.

[0096] Each historical correlation subset is input into the real-time activity evaluation model to obtain the corresponding activity quantification value. This activity quantification value reflects the user's activity during that activity phase within a short time period.

[0097] Multiple activity quantification sub-values ​​are weighted based on the temporal relationships of historically related subsets to obtain a comprehensive activity quantification coefficient. This coefficient can comprehensively represent the intensity and persistence of the target user throughout the entire activity phase, such as sleep quality and fat burning during exercise, thus providing important data support for subsequent health assessments and feedback. For example, activity quantification sub-values ​​closer to the middle stage can be assigned higher weights, and these activity quantification sub-values ​​can be calculated using linear weighting to ultimately obtain a comprehensive activity quantification coefficient.

[0098] Furthermore, step S7 in this embodiment of the application also includes:

[0099] Step S71: Interact to obtain multiple sets of sample push content sets for multiple activity quantification intervals of the various sample activity stages.

[0100] Step S72: Associate and store the various sample activity stages, multiple activity quantification intervals, and multiple sample push content sets to generate a push material library.

[0101] Step S73: Match the push content in the push material library according to the activity quantification coefficient and the adjacent activity stages to obtain a set of backup push content.

[0102] Step S74: Interact with the target user to obtain their historical reading logs, and perform a tendency feature analysis based on the historical reading logs to obtain the target tendency features.

[0103] Step S75: Obtain the real-time push science popularization content from the backup push content set according to the target tendency characteristics.

[0104] Specifically, the system interactively retrieves sample content sets for different quantitative intervals of various sample activity stages from a health knowledge database. For example, for the "sleep" sample activity stage, it searches the health knowledge database for content such as "how to improve sleep quality" and "the dangers of staying up late" for different quantitative intervals of sleep quality activities, forming sample content sets for each interval.

[0105] Different activity stages, activity quantification intervals, and their corresponding push content are associated and stored in a database to form a push content library. For example, in a relational database, three tables can be created: one table stores sample activity stage information, one table stores activity quantification interval information, and another table stores sample push content set information. Then, these three tables are linked together through foreign key relationships to generate the push content library.

[0106] The target activity quantification range is determined based on the activity quantification coefficient. Then, combined with information from adjacent activity stages, a search and matching process is performed in the push content library to find multiple sets of corresponding push content, forming a backup push content set.

[0107] Historical reading logs are retrieved from databases or log files storing the target users' reading history. These logs are then analyzed using data mining and analysis techniques to determine the target users' preferences for the pushed content. For example, the frequency with which users read different types of content can be counted, and patterns in user reading across different time periods can be analyzed.

[0108] Based on the obtained target interest characteristics, the content is filtered and retrieved from the backup push content set. Keyword matching or category matching is used to determine the real-time push science popularization content. For example, if the target interest characteristic is "sports injury prevention," then content related to sports injury prevention is searched in the backup push content set and identified as the real-time push science popularization content.

[0109] By following the steps above, the most relevant content can be selected and pushed to users in real time based on their activity data and reading preferences, thereby improving user satisfaction and enhancing the effectiveness of information delivery.

[0110] In summary, the method for dynamically matching popular science content based on real-time physiological trends provided in this application has the following technical effects:

[0111] This application's embodiments comprehensively consider multiple factors such as multidimensional physiological data, activity trends, and historically related physiological data, enabling more accurate matching of health education content suitable for the user's current physical state and activity stage. This improves the effectiveness, accuracy, and real-time nature of the pushed content. Personalized pushed content and timely responses to changes in the user's activity status ensure that the health education content received by the user better meets their needs, thereby increasing user acceptance and satisfaction with the health education content and enhancing the user experience. Precise health education content delivery ensures that users receive health education knowledge relevant to their own physical condition and activity status, helping them better understand and apply this knowledge, thus improving the actual effectiveness of health education.

[0112] Example 2, as Figure 4 As shown in the embodiment of this application, a dynamic matching platform for popular science content that integrates real-time physiological trends is provided. The platform includes:

[0113] Physiological data acquisition unit 10 is used to receive the physiological data of the target user in real time through the physiological monitoring module and obtain multi-dimensional physiological time-series data.

[0114] Activity trend analysis unit 20 is used to perform activity state trend analysis on the multidimensional physiological time series data to obtain real-time activity stages.

[0115] The continuity verification unit 30 is used to call adjacent activity stages from the activity record module according to the timestamp, and to perform continuity verification on the real-time activity stage and the adjacent activity stages.

[0116] The adjacent activity duration interval determination unit 40 is used to obtain the adjacent activity duration interval by traversing the activity record module through the adjacent activity stages and calling the stage continuous duration if the continuity check fails.

[0117] Historical associated physiological data acquisition unit 50 is used to obtain historical associated physiological data from the physiological monitoring module by using the adjacent activity duration interval.

[0118] Activity status quantification evaluation unit 60 is used to perform activity status quantification evaluation on the historical associated physiological data with the adjacent activity stages as constraints, and generate activity quantification coefficients.

[0119] The push content matching unit 70 is used to match push content according to the activity quantification coefficient and adjacent activity stages to obtain real-time push science popularization content.

[0120] The science popularization content feedback unit 80 is used to feed back the real-time pushed science popularization content to the target user's mobile terminal for display.

[0121] Furthermore, the activity trend analysis unit 20 in this embodiment is also used to perform the following steps:

[0122] A pre-constructed activity stage judgment model is used, which includes multiple parallel activity stage judgment branches, each branch having various sample activity stage identifiers. After preprocessing the multidimensional physiological time series data using a multi-level window mechanism, the activity trend analysis of the preprocessed results is performed using the activity stage judgment model to obtain the real-time activity stage.

[0123] Furthermore, the activity trend analysis unit 20 in this embodiment is also used to perform the following steps:

[0124] The process involves: interactively obtaining multiple sets of associated physiological indicators for various sample activity stages; collecting sample data with a first set of associated physiological indicators as constraints to obtain multiple sets of sample physiological data and multiple sample stage probability coefficients; using the multiple sets of sample physiological data and multiple sample stage probability coefficients as training data to construct a first activity stage judgment unit for judging the first sample activity stage; loading a data filter based on the first set of associated physiological indicators to obtain a first adaptive data filter, and configuring the first adaptive data filter at the front end of the first activity stage judgment unit to complete the construction of the first activity stage judgment branch; and so on, collecting sample data and building the model with the multiple sets of associated physiological indicators as constraints to obtain the multiple activity stage judgment branches; and completing the construction of the activity stage judgment model by connecting the multiple activity stage judgment branches in parallel.

[0125] Furthermore, the activity trend analysis unit 20 in this embodiment is also used to perform the following steps:

[0126] The time window is divided according to a multi-level window mechanism to obtain a long-term monitoring window and a short-term monitoring window. Using the short-term monitoring window as a constraint, the first-stage time-series data is selected from the physiological monitoring module. In the activity stage judgment model, after pre-screening the first-stage time-series data through multiple adaptive data filters, the filtering results are analyzed through multiple activity stage judgment branches to output multiple real-time stage probability coefficients. Using the short-term monitoring window as a constraint, the stage time-series data of the physiological monitoring module is iteratively selected, and real-time stage probability judgment is performed through the activity stage judgment model until the real-time reception duration of the physiological data of the physiological monitoring module meets the long-term monitoring window, outputting M sets of real-time stage probability coefficients. Activity trend analysis is performed based on the M sets of real-time stage probability coefficients to obtain the real-time activity stage.

[0127] Furthermore, the activity trend analysis unit 20 in this embodiment is also used to perform the following steps:

[0128] By decomposing the M sets of real-time stage probability coefficients, multiple stage time-series distributions corresponding to the various sample activity stages are obtained; historical activity stage data is interactively obtained, and stage transition probability analysis of the various sample activity stages is performed based on the historical activity stage data, and a stage transition probability matrix is ​​constructed based on the analysis results; the stage transition probability matrix is ​​used to calculate the hourly probability distribution of the multiple stage time-series distributions to obtain multiple dynamic stage probability sequences; the multiple dynamic stage probability sequences are reorganized according to the M collection times of the M sets of real-time stage probability coefficients to obtain M sets of node dynamic probabilities; the maximum value of the M sets of node dynamic probabilities is called to obtain M dominant activity stages and M dominant activity probabilities; after sorting the M dominant activity stages according to the M collection times, stage transition stability analysis is performed according to the M dominant activity probabilities, and the real-time activity stages are output.

[0129] Furthermore, the platform described in this application embodiment is also used to perform the following steps:

[0130] Multiple standard durations for the various sample activity stages are predefined; the first standard duration of the first sample activity stage is retrieved from the multiple standard durations; sample data is collected with constraints of the first associated physiological indicator group and 1 / M of the first standard duration to obtain multiple sample physiological monitoring groups and multiple sample activity quantification coefficients; the multiple sample physiological monitoring groups and multiple sample activity quantification coefficients are used as training data to construct a first activity evaluation model; and so on, multiple activity evaluation models for the various sample activity stages are constructed, and the activity evaluation model library is completed by associating and storing the various sample activity stages and multiple activity evaluation models.

[0131] Furthermore, the activity status quantification evaluation unit 60 in this embodiment is also used to perform the following steps:

[0132] The real-time activity evaluation model is scheduled from the activity evaluation model library according to the adjacent activity stages; the real-time standard duration is called from the multiple standard durations according to the adjacent activity stages; the historical associated physiological data is segmented according to 1 / M of the real-time standard duration to obtain multiple historical associated subsets; the real-time activity evaluation model is used to evaluate the multiple historical associated subsets and output multiple activity quantification sub-values; the activity quantification coefficient is obtained by performing time-series weighting on the multiple activity quantification sub-values.

[0133] Furthermore, in this embodiment of the application, the content matching unit 70 is also used to perform the following steps:

[0134] The system interactively obtains multiple sets of sample push content across multiple activity quantification intervals for various sample activity stages; it associates and stores the various sample activity stages, multiple activity quantification intervals, and multiple sets of sample push content to generate a push material library; it matches push content in the push material library based on the activity quantification coefficient and adjacent activity stages to obtain a backup push content set; it interactively obtains the target user's historical reading logs and performs preference feature analysis based on the historical reading logs to obtain target preference features; and it retrieves the real-time push science popularization content from the backup push content set based on the target preference features.

[0135] Through the foregoing detailed description of the method for dynamically matching popular science content that integrates real-time physiological trends, those skilled in the art can clearly understand that the platform for dynamically matching popular science content that integrates real-time physiological trends in this embodiment, as disclosed in Embodiment 2, corresponds to the method disclosed in Embodiment 1 and has corresponding functional units and beneficial effects. For relevant details, please refer to the description in the method section.

[0136] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for dynamically matching popular science content with real-time physiological trends, characterized in that, The method comprises: Real-time physiological data of a target user is received by a physiological monitoring module to obtain multi-dimensional physiological time series data; Activity state trend analysis is performed on the multi-dimensional physiological time series data to obtain a real-time activity stage; An adjacent activity stage is called from an activity record module according to an end timestamp of the real-time activity stage, and continuity verification is performed on the real-time activity stage and the adjacent activity stage; If the continuity verification fails, the adjacent activity stage is used to traverse the activity record module to call a stage continuity duration, and an adjacent activity duration interval is obtained; The adjacent activity duration interval is used to frame and obtain historical associated physiological data from the physiological monitoring module; The adjacent activity stage is used as a constraint to quantitatively evaluate the activity condition of the historical associated physiological data, and an activity quantitative coefficient is generated; According to the activity quantitative coefficient and the adjacent activity stage, push content matching is performed to obtain real-time push popular science content; The real-time push popular science content is fed back to a mobile terminal display of the target user; Activity state trend analysis is performed on the multi-dimensional physiological time series data to obtain a real-time activity stage, and the method comprises: An activity stage judgment model is pre-constructed, wherein the activity stage judgment model comprises a plurality of activity stage judgment branches in parallel, and the plurality of activity stage judgment branches have a plurality of sample activity stage identifiers; After the multi-dimensional physiological time series data is preprocessed by applying a multi-level window mechanism, activity trend analysis is performed on the preprocessing result by the activity stage judgment model to obtain a real-time activity stage; An activity stage judgment model is pre-constructed, and the method comprises: A plurality of associated physiological indicator groups of the plurality of sample activity stages are interactively obtained; Sample data collection is performed with a first associated physiological indicator group as a constraint to obtain a plurality of sample physiological data groups and a plurality of sample stage probability coefficients; The plurality of sample physiological data groups and the plurality of sample stage probability coefficients are used as training data to construct a first activity stage judgment unit for first sample activity stage judgment; After a first adaptive data filter is obtained by loading a data filter according to the first associated physiological indicator group, the first adaptive data filter is configured at the front end of the first activity stage judgment unit to complete the construction of a first activity stage judgment branch; By analogy, sample data collection and model construction are performed with the plurality of associated physiological indicator groups as constraints to obtain the plurality of activity stage judgment branches; The construction of the activity stage judgment model is completed by connecting the plurality of activity stage judgment branches in parallel; After the multi-dimensional physiological time series data is preprocessed by applying a multi-level window mechanism, activity trend analysis is performed on the preprocessing result by the activity stage judgment model to obtain a real-time activity stage, and the method comprises: According to the multi-level window mechanism, time window division is performed to obtain a long-term monitoring window and a short-term monitoring window; A first stage time series data is framed from the physiological monitoring module with the short-term monitoring window as a constraint; In the activity stage judgment model, after pre-screening the first stage time series data through multiple adaptive data filters, the screening results are analyzed through the multiple activity stage judgment branches, and multiple real-time stage probability coefficients are output; With the short-term monitoring window as a constraint, the stage time series data of the physiological monitoring module is looped and framed, and real-time stage probability judgment is performed through the activity stage judgment model until the real-time receiving duration of the physiological data of the physiological monitoring module meets the long-term monitoring window, and M sets of real-time stage probability coefficients are output; According to the M sets of real-time stage probability coefficients, the activity trend is analyzed, and the real-time activity stage is obtained; According to the M sets of real-time stage probability coefficients, the activity trend is analyzed, and the real-time activity stage is obtained, and the method comprises: By decomposing the M sets of real-time stage probability coefficients, multiple stage time series distributions corresponding to the multiple sample activity stages are obtained; The historical activity stage data is interactively obtained, and after stage transition probability analysis of the multiple sample activity stages is performed according to the historical activity stage data, a stage transition probability matrix is constructed based on the analysis results; The stage transition probability matrix is used to calculate the hourly probability distribution of the multiple stage time series distributions, and multiple dynamic stage probability sequences are obtained; According to the M acquisition times of the M sets of real-time stage probability coefficients, the multiple dynamic stage probability sequences are reorganized, and M sets of node dynamic probabilities are obtained; The M sets of node dynamic probabilities are called maximum values to obtain M dominant activity stages and M dominant activity probabilities; After sorting the M dominant activity stages according to the M acquisition times, stage transition stability analysis is performed according to the M dominant activity probabilities, and the real-time activity stage is output.

2. The method of claim 1, wherein the method further comprises: With the adjacent activity stage as a constraint, the activity condition of the historical associated physiological data is quantitatively evaluated, and an activity quantization coefficient is generated, and before that, the method comprises: Predefining multiple standard duration times of the multiple sample activity stages; Calling a first standard duration time of the first sample activity stage from the multiple standard duration times; With 1 / M of the first associated physiological index group and the first standard duration time as a constraint, sample data is collected to obtain multiple sample physiological monitoring groups and multiple sample activity quantization coefficients; The multiple sample physiological monitoring groups and the multiple sample activity quantization coefficients are used as training data to construct a first activity evaluation model; By analogy, multiple activity evaluation models of the multiple sample activity stages are constructed, and the construction of the activity evaluation model library is completed by associating the multiple sample activity stages and the multiple activity evaluation models.

3. The method of claim 2, wherein the method further comprises: With the adjacent activity stage as a constraint, the activity condition of the historical associated physiological data is quantitatively evaluated, and an activity quantization coefficient is generated, and the method comprises: According to the adjacent activity stage, a real-time activity evaluation model is dispatched from the activity evaluation model library; According to the adjacent activity stage, a real-time standard duration time is called from the multiple standard duration times; According to 1 / M of the real-time standard duration time, the historical associated physiological data is segmented to obtain multiple historical associated subsets; The real-time activity evaluation model is used to evaluate the plurality of historical correlation subsets, and a plurality of activity quantitative sub-values are output; The activity quantitative coefficient is obtained by time series weighting of the plurality of activity quantitative sub-values.

4. The method of claim 1, wherein the method further comprises: According to the activity quantitative coefficient and the adjacent activity stage, the push content matching is performed, and real-time push popular science content is obtained, and the method comprises: Interactively obtaining a plurality of sets of sample push content sets of a plurality of sets of activity quantitative intervals of the plurality of sample activity stages; Correlating to store the plurality of sample activity stages, the plurality of sets of activity quantitative intervals and the plurality of sets of sample push content sets, and generating a push material library; According to the activity quantitative coefficient and the adjacent activity stage, the push content matching is performed in the push material library, and a standby push content set is obtained; Interactively obtaining the historical reading log of the target user, and performing tendency feature analysis based on the historical reading log to obtain a target tendency feature; According to the target tendency feature, the real-time push popular science content is obtained from the standby push content set.

5. A dynamic matching platform of popular science content fused with real-time physiological trends, characterized in that, The platform is used to perform the popular science content dynamic matching method fused with real-time physiological trends according to any one of claims 1-4, comprising: A physiological data acquisition unit, the physiological data acquisition unit is used for real-time receiving physiological data of a target user through a physiological monitoring module, and obtaining multi-dimensional physiological time series data; An activity trend analysis unit, the activity trend analysis unit is used for activity state trend analysis on the multi-dimensional physiological time series data, and obtaining a real-time activity stage; A continuity verification unit, the continuity verification unit is used for calling an adjacent activity stage from an activity record module according to the end timestamp of the real-time activity stage, and performing continuity verification on the real-time activity stage and the adjacent activity stage; An adjacent activity duration interval determination unit, the adjacent activity duration interval determination unit is used for calling a stage continuous duration from the activity record module by traversing the activity record module using the adjacent activity stage if the continuity verification fails, and obtaining an adjacent activity duration interval; A historical correlation physiological data acquisition unit, the historical correlation physiological data acquisition unit is used for obtaining historical correlation physiological data from the physiological monitoring module by framing using the adjacent activity duration interval; An activity condition quantitative evaluation unit, the activity condition quantitative evaluation unit is used for activity condition quantitative evaluation on the historical correlation physiological data by taking the adjacent activity stage as a constraint, and generating an activity quantitative coefficient; A push content matching unit, the push content matching unit is used for push content matching according to the activity quantitative coefficient and the adjacent activity stage, and obtaining real-time push popular science content; A popular science content feedback unit, the popular science content feedback unit is used for feeding back the real-time push popular science content to a mobile terminal display of the target user.

Images