Frequency dynamic adjustment state monitoring method and system and medium

Abstract

The invention discloses a frequency dynamic adjustment state monitoring method and system and a medium, and relates to the technical field of sleep monitoring, and the method comprises the steps: obtaining a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; traversing the first sleep monitoring log sub-sequence and the second sleep monitoring log sub-sequence to perform sleep feature analysis, performing front-back association iterative analysis on analysis results according to a sequence, and determining a first association interaction sleep monitoring feature set and a second association interaction sleep monitoring feature set; adjusting the preset first monitoring frequency and the preset second monitoring frequency to obtain a first adjusting monitoring frequency and a second adjusting monitoring frequency, and monitoring the sleep state of the target user through the sleep monitoring device. The technical problems that in the prior art, the sleep monitoring frequency is fixed, the monitoring efficiency is low, and monitoring cannot be conducted by adapting to the sleep state in real time are solved, and the technical effect of improving the sleep monitoring precision and efficiency is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sleep monitoring, in particular to a state monitoring method and system with dynamic frequency adjustment and a medium. BACKGROUND

[0002] With the continuous improvement of people's living standards and the enhancement of health awareness, sleep quality has become an important health indicator widely concerned in modern society. Especially under the promotion of intelligent wearable devices and health management systems, sleep monitoring technology has developed rapidly. However, with the increase of individual differences and the complexity of sleep patterns, traditional sleep monitoring methods face great challenges in accuracy and efficiency. Existing sleep monitoring technologies mostly use fixed monitoring frequencies or preset monitoring modes, which cannot adapt to sleep states in real time for monitoring. SUMMARY

[0003] The present application provides a state monitoring method and system with dynamic frequency adjustment and a medium, which is used to solve the technical problems of fixed sleep monitoring frequency, low monitoring efficiency, and inability to adapt to sleep states in real time in existing technologies.

[0004] In view of the above problems, the present application provides a state monitoring method and system with dynamic frequency adjustment and a medium.

[0005] The first aspect of the present application provides a state monitoring method with dynamic frequency adjustment, which comprises: Obtaining logs of a sleep monitoring device monitoring a target user's sleep state according to a preset first monitoring frequency and a preset second monitoring frequency in a preset monitoring window, obtaining a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; traversing the first sleep monitoring log subsequence and the second sleep monitoring log subsequence for sleep feature analysis, and performing forward and backward correlation iterative analysis on the analysis results in order to determine a first correlation interactive sleep monitoring feature set and a second correlation interactive sleep monitoring feature set; adjusting the preset first monitoring frequency and the preset second monitoring frequency based on the first correlation interactive sleep monitoring feature set and the second correlation interactive sleep monitoring feature set, respectively, to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency; and the sleep monitoring device monitors the target user's sleep state according to the first adjusted monitoring frequency and the second adjusted monitoring frequency.

[0006] The second aspect of the present application provides a state monitoring system with dynamic frequency adjustment, which comprises: The monitoring log acquisition module acquires logs of monitoring the sleep state of the target user by the sleep monitoring device according to the preset first monitoring frequency and the preset second monitoring frequency respectively in a preset monitoring window, to obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; the sleep feature analysis module performs sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, and performs front-back correlation iterative analysis on the analysis results in the order, to determine a first correlation interaction sleep monitoring feature set and a second correlation interaction sleep monitoring feature set; the monitoring frequency adjustment module adjusts the preset first monitoring frequency and the preset second monitoring frequency based on the first correlation interaction sleep monitoring feature set and the second correlation interaction sleep monitoring feature set respectively, to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency; and the sleep state monitoring module performs sleep state monitoring on the target user by the sleep monitoring device according to the first adjusted monitoring frequency and the second adjusted monitoring frequency.

[0007] In a third aspect, the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the state monitoring method with frequency dynamic adjustment.

[0008] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: The application obtains logs of monitoring the sleep state of a target user by a sleep monitoring device according to a preset first monitoring frequency and a preset second monitoring frequency in a preset monitoring window, obtains a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; performs sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, and performs forward and backward correlation iterative analysis on the analysis results in the order, to determine a first correlation interactive sleep monitoring feature set and a second correlation interactive sleep monitoring feature set; adjusts the preset first monitoring frequency and the preset second monitoring frequency based on the first correlation interactive sleep monitoring feature set and the second correlation interactive sleep monitoring feature set respectively, to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency; and the sleep monitoring device monitors the sleep state of the target user according to the first adjusted monitoring frequency and the second adjusted monitoring frequency. The application solves the technical problems of the prior art, such as fixed sleep monitoring frequency, low monitoring efficiency, and inability to monitor in real time according to the sleep state. The sleep state data of the target user in the preset monitoring window is collected, sleep feature analysis and forward and backward correlation iterative analysis are performed, a first correlation interactive sleep monitoring feature set and a second correlation interactive sleep monitoring feature set are obtained, the preset monitoring frequency is adjusted based on the feature sets, and the first adjusted monitoring frequency and the second adjusted monitoring frequency are obtained to monitor the sleep state, so that the technical effects of improving the accuracy and efficiency of sleep monitoring are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0010] Figure 1 A frequency dynamic adjustment state monitoring method flowchart provided by the embodiment of the present application; Figure 2 A frequency dynamic adjustment state monitoring system structure diagram provided by the embodiment of the present application.

[0011] Explanation of reference signs: monitoring log acquisition module 11, sleep feature analysis module 12, monitoring frequency adjustment module 13, sleep state monitoring module 14. DETAILED DESCRIPTION

[0012] The application provides a frequency dynamically adjusted state monitoring method, system and medium, aiming at solving the technical problems that the sleep monitoring frequency is fixed, the monitoring efficiency is low, and the sleep state cannot be monitored in real time in the prior art. The sleep state data of a target user in a preset monitoring window is collected, sleep feature analysis and before-and-after correlation iterative analysis are performed, a first correlation interaction sleep monitoring feature set and a second correlation interaction sleep monitoring feature set are obtained, the preset monitoring frequency is adjusted based on the feature sets, and an optimized first adjusted monitoring frequency and a second adjusted monitoring frequency are obtained to perform sleep state monitoring, so that the technical effect of improving the accuracy and efficiency of sleep monitoring is achieved.

[0013] The technical solutions in the embodiments of the application will be clearly and completely described in connection with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0014] It should be noted that any variation of the terms "comprise" and "have" is intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to the process, method, product or device.

[0015] Embodiment one, as shown in the application provides a frequency dynamically adjusted state monitoring method, the method comprises: Figure 1 Step S100: obtaining logs of sleep monitoring devices respectively monitoring the sleep state of a target user according to a preset first monitoring frequency and a preset second monitoring frequency in a preset monitoring window, to obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence.

[0016] In the embodiments of the application, in order to obtain the sleep data of the target user, a preset monitoring window is first set. The monitoring window refers to a specific time range for collecting sleep state information of the target user, for example, 8 hours of night sleep time. Through the monitoring window, the continuity and integrity of data collection in time are ensured.

[0017] Next, the sleep monitoring device collects data of the sleep state of the target user according to the preset first monitoring frequency and the preset second monitoring frequency. Here, the monitoring frequency refers to the number of times of collecting data by the device in unit time.

[0018] ​After the data collection is completed, the collected data is classified and arranged according to the frequency. The data belonging to the first monitoring frequency is extracted to generate a first sleep monitoring log subsequence, and the data belonging to the second monitoring frequency is extracted to generate a second sleep monitoring log subsequence. The log subsequence is a segmented collection of original data, containing all monitoring records at this frequency, such as heart rate, respiratory rate, and other parameters.

[0019] Further, the method provided by the application embodiment further comprises: obtaining a historical sleep monitoring log sequence of the target user in a historical monitoring window; traversing the historical sleep monitoring log sequence to extract abnormal logs, and obtaining a plurality of historical abnormal sleep monitoring logs, wherein each historical abnormal sleep monitoring log includes a historical node; extracting abnormal features based on the plurality of historical abnormal sleep monitoring logs, and obtaining a plurality of historical abnormal sleep feature sets; performing same-type aggregation on the plurality of historical abnormal sleep monitoring logs according to the plurality of historical abnormal sleep feature sets, and obtaining a plurality of clustered historical abnormal sleep monitoring log sets; and performing monitoring frequency analysis according to the plurality of clustered historical abnormal sleep monitoring log sets and corresponding historical nodes, and obtaining the preset first monitoring frequency and the preset second monitoring frequency.

[0020] In the application embodiment, first, a historical sleep monitoring log sequence of a target user in a historical monitoring window is obtained. The historical sleep monitoring log sequence refers to all sleep state data recorded by the sleep monitoring device of the target user during the pre-set historical monitoring window. The log sequence contains physiological feature data (such as heart rate, respiratory rate, body movement, etc.) of the user and time sequence information.

[0021] Next, the historical sleep monitoring log sequence is traversed to extract abnormal logs. Specifically, based on statistical method-based anomaly detection techniques, such as standard deviation analysis, log data deviating from the normal sleep pattern is identified. For example, if the fluctuation amplitude of the heart rate or respiratory rate of the user exceeds 3 times the standard deviation of the normal value, these data will be marked as abnormal logs. Each abnormal log contains a historical node, i.e., the specific time point of the anomaly. Through this process, a plurality of historical abnormal sleep monitoring logs are obtained.

[0022] Then, abnormal features are extracted based on multiple historical abnormal sleep monitoring logs. This step uses time series analysis methods, such as the Autoregressive Integral Moving Average (ARIMA) model, to extract key features from the abnormal logs. These features include the fluctuation amplitude, duration, and periodic changes at the time of the abnormality, used to describe the specific nature of each abnormality. Through this process, the abnormal logs are transformed into a structured set of multiple historical abnormal sleep features.

[0023] Next, multiple historical abnormal sleep monitoring logs are clustered based on several sets of historical abnormal sleep features. This process uses the K-means clustering method to group abnormal logs with similar characteristics into the same category. For example, abnormal logs with similar heart rate fluctuations are grouped into one cluster, while abnormal logs with sleep apnea are grouped into another. This process yields multiple clustered sets of historical abnormal sleep monitoring logs, each representing a similar type of abnormality.

[0024] Finally, based on the historical abnormal sleep monitoring log sets of multiple clusters and their corresponding historical nodes, a monitoring frequency analysis is performed. In this process, the time interval between logs in each cluster is calculated, and the monitoring frequency for each type of anomaly is determined by the minimum and maximum anomaly intervals. For example, if the minimum interval for a certain type of anomaly is 30 minutes and the maximum interval is 2 hours, the recommended monitoring frequency should be within this range to ensure the capture of both frequent and occasional anomalies. Ultimately, a preset first monitoring frequency and a preset second monitoring frequency are obtained.

[0025] Furthermore, the method provided in the application embodiments also includes: A first set of historical abnormal sleep features is randomly selected from the plurality of historical abnormal sleep feature sets and used as the first clustering target. The plurality of historical abnormal sleep feature sets are then aggregated toward the first clustering target according to a preset clustering similarity threshold to obtain a first cluster of historical abnormal sleep monitoring log sets. The first cluster of historical abnormal sleep monitoring log sets is then removed from the plurality of historical abnormal sleep feature sets. A second set of historical abnormal sleep features is then randomly selected from the removed set of historical abnormal sleep features and aggregated with the first clustering target to obtain a second cluster of historical abnormal sleep monitoring log sets. This process is repeated multiple times until all the historical abnormal sleep feature sets are aggregated to obtain multiple clusters of historical abnormal sleep monitoring log sets.

[0026] In this embodiment, a first set of historical abnormal sleep features is randomly selected from multiple sets of historical abnormal sleep features as the first clustering target. Specifically, a set of historical abnormal features is selected through random sampling as the benchmark for subsequent clustering analysis. The selected first set of historical abnormal sleep features represents a specific type of abnormal sleep feature, such as a heart rate abnormality with large fluctuations.

[0027] Next, multiple sets of historical abnormal sleep features are aggregated towards the first clustering target according to a preset clustering similarity threshold to form the first cluster of historical abnormal sleep monitoring log sets. This process uses a clustering algorithm based on similarity metrics, such as K-means clustering, to calculate the similarity between sets of historical abnormal sleep features (e.g., using Euclidean distance or cosine similarity), aggregating abnormal feature sets similar to the first clustering target together. The preset clustering similarity threshold determines the similarity requirement between feature sets; only sets with similarity higher than this threshold are clustered together. Finally, through this process, a cluster containing one type of abnormal features is obtained, called the first cluster of historical abnormal sleep monitoring log sets.

[0028] After constructing the first cluster, the historical abnormal sleep monitoring log set of the first cluster is removed from multiple historical abnormal sleep feature sets, i.e., the already clustered feature set is removed. This operation ensures that subsequent clustering does not repeatedly consider processed data points. After removal, the remaining historical abnormal sleep feature set still contains unclustered feature sets, and clustering continues from these remaining sets.

[0029] Next, a second set of historical abnormal sleep features is randomly selected from the previously removed sets of historical abnormal sleep features, serving as the new clustering target. This process repeats the steps of the first clustering, selecting a new set of historical abnormal sleep features and using it as a benchmark for the next round of clustering. At this point, this new clustering target will be compared with other abnormal feature sets, and those feature sets similar to it will be aggregated together to form a second cluster of historical abnormal sleep monitoring log sets.

[0030] By repeatedly aggregating these data multiple times until all historical abnormal sleep feature sets are clustered, multiple clustered historical abnormal sleep monitoring log sets are ultimately obtained. Each cluster contains a set of sleep monitoring logs with similar abnormal features, representing a pattern or trend of a certain type of sleep abnormality. This process, through iterative clustering, ensures that all abnormal feature sets can ultimately be reasonably classified and assigned to different clusters.

[0031] Furthermore, in the method provided in the application embodiment, the plurality of historical abnormal sleep feature sets are aggregated towards the first clustering target according to a preset clustering similarity threshold to obtain a first cluster of historical abnormal sleep monitoring log sets, including: The similarity between the first clustering target and each of the multiple historical abnormal sleep feature sets is calculated to obtain multiple first clustering target similarities; the historical abnormal sleep monitoring logs corresponding to the first clustering target similarities that are greater than or equal to the preset clustering similarity threshold are added to the first cluster historical abnormal sleep monitoring log set.

[0032] In this embodiment, the similarity between the first clustering target and each of the multiple historical abnormal sleep feature sets is first calculated. In this process, the similarity between the multiple first clustering targets is obtained by using Euclidean distance or cosine similarity calculation.

[0033] Next, the obtained first cluster target similarities are compared with the preset cluster similarity threshold. When the first cluster target similarity of a certain historical abnormal sleep feature set is greater than or equal to the preset cluster similarity threshold, the historical abnormal sleep monitoring log corresponding to the historical abnormal sleep feature set is added to the first cluster historical abnormal sleep monitoring log set.

[0034] Furthermore, the method provided in the application embodiments also includes: Based on the multiple clustered historical abnormal sleep monitoring log sets and corresponding historical nodes, multiple minimum abnormal intervals and multiple maximum abnormal intervals are determined; the minimum value of the multiple minimum abnormal intervals is taken as the preset first monitoring frequency; and the maximum value of the multiple maximum abnormal intervals is taken as the preset second monitoring frequency.

[0035] In this embodiment, the time difference is first calculated for the historical abnormal sleep monitoring logs in each cluster to obtain the time interval between each pair of adjacent historical nodes, i.e., the abnormal interval. This step is achieved by calculating the time difference between adjacent abnormal events using a timestamp difference calculation method. Then, the minimum and maximum abnormal intervals for each cluster are determined. The minimum abnormal interval is the shortest time interval between adjacent abnormal events in the cluster, while the maximum abnormal interval is the longest time interval between adjacent abnormal events in the cluster.

[0036] Next, the minimum anomaly intervals of all clusters are compared, and the minimum value is selected as the final preset first monitoring frequency. This means that frequent monitoring is required within the shortest possible time interval during the monitoring period. Similarly, the maximum anomaly intervals of all clusters are compared, and the maximum value is selected as the final preset second monitoring frequency. This indicates that the monitoring frequency can be reduced when there are no anomalies for a longer period of time.

[0037] Step S200: Traverse the first sleep monitoring log subsequence and the second sleep monitoring log subsequence to perform sleep feature analysis, and perform sequential correlation iterative analysis on the analysis results to determine the first correlation interaction sleep monitoring feature set and the second correlation interaction sleep monitoring feature set.

[0038] In this embodiment, a sleep feature recognition network layer is first invoked to perform sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, respectively. Through the processing of the feature recognition network layer, a first sleep monitoring feature set subsequence is extracted from the first log subsequence, and a second sleep monitoring feature set subsequence is extracted from the second log subsequence.

[0039] Next, iterative analysis of the correlation between each feature set subsequence is performed. In the first sleep monitoring feature set subsequence, the t-th feature set and the (t+1)-th feature set are obtained sequentially, and the correlation between them is calculated. A new feature set is generated through an iterative process, namely the t-th first iteration sleep monitoring feature set. This process starts from the first feature set of the subsequence and iterates step by step until all feature sets are processed. The feature set obtained in the last iteration is taken as the first correlated interactive sleep monitoring feature set.

[0040] Similarly, in the second sleep monitoring feature set subsequence, the q-th feature set and the (q+1)-th feature set are obtained sequentially, and the same correlation iterative analysis is performed to generate the second iterative sleep monitoring feature set. This process starts from the first feature set of the subsequence and iterates step by step to the end, with the feature set obtained in the last iteration serving as the second correlation interaction sleep monitoring feature set.

[0041] Furthermore, the method provided in the application embodiments also includes: The sleep feature recognition network layer is invoked to perform sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, respectively, to obtain the first sleep monitoring feature set subsequence and the second sleep monitoring feature set subsequence. The t-th and t+1-th first sleep monitoring feature sets in the first sleep monitoring feature set subsequence are obtained and subjected to iterative correlation analysis to obtain the t-th first iteration sleep monitoring feature set. After multiple iterations, until the end of the first sleep feature set subsequence is reached, the n-th first iteration sleep monitoring feature set obtained in the last iteration is taken as the first correlated interactive sleep monitoring feature set, where 1 ≤ t. ≤n+1, where n is the total number of logs in the first sleep monitoring log subsequence minus 1, and n is an integer greater than or equal to 1; obtain the q-th second sleep monitoring feature set and the (q+1)-th second sleep monitoring feature set in the second sleep monitoring feature set subsequence and perform iterative analysis to obtain the q-th second iteration sleep monitoring feature set. After multiple iterations, until the end of the second sleep monitoring feature set subsequence is reached, the m-th second iteration sleep monitoring feature set obtained in the last iteration is taken as the second associated interactive sleep monitoring feature set, where 1≤q≤m+1, m is the total number of logs in the second sleep monitoring log subsequence minus 1, and m is an integer greater than or equal to 1.

[0042] In this embodiment, a sleep feature recognition network layer is first used to analyze the sleep features of the first and second sleep monitoring log subsequences. The sleep feature recognition network layer, after training, is able to extract key features during sleep based on input multi-channel physiological signals (such as electrocardiograms, motion data, etc.) and identify different sleep stages (such as deep sleep, light sleep, and REM sleep). During training, the network learns how to associate physiological signals with sleep stages from labeled sleep data. After training, the network can identify the first and second sleep monitoring feature set subsequences, which correspond to the sleep feature information in the first and second sleep monitoring log subsequences, respectively.

[0043] Next, the t-th and t+1-th first sleep monitoring feature sets are obtained from the first sleep monitoring feature set subsequence, and iterative analysis of their correlation is performed. Through multiple iterations, the t-th first iterative sleep monitoring feature set is gradually generated. The iteration starts from the first feature set and continues until the end of the first sleep monitoring feature set subsequence is reached, finally obtaining the n-th first iterative sleep monitoring feature set, which is used as the first correlated interactive sleep monitoring feature set. Here, the value of t is in the range of 1 ≤ t ≤ n+1, where n is the total number of logs in the first sleep monitoring log subsequence minus 1, and n is an integer greater than or equal to 1.

[0044] Similarly, for the second sleep monitoring feature set subsequence, the q-th and q+1-th second sleep monitoring feature sets are obtained, and the same steps are followed for iterative analysis to generate the q-th second iterative sleep monitoring feature set. This process iterates from the first feature set until the end of the second sleep monitoring feature set subsequence, ultimately obtaining the m-th second iterative sleep monitoring feature set, which serves as the second associated interactive sleep monitoring feature set. Here, the value of q is in the range of 1 ≤ q ≤ m+1, where m is the total number of logs in the second sleep monitoring log subsequence minus 1, and m is an integer greater than or equal to 1.

[0045] Finally, through iterative analysis of the subsequences of the first and second sleep monitoring feature sets, respectively, the first associated interactive sleep monitoring feature set and the second associated interactive sleep monitoring feature set were generated.

[0046] Furthermore, the method provided in the application embodiments also includes: Calculate the inner product mapping similarity of each sleep monitoring feature in the t-th first sleep monitoring feature set and the (t+1)-th first sleep monitoring feature set to obtain the t-th first inner product mapping similarity set; traverse the t-th first inner product mapping similarity set, perform normalization processing, and embed the processing result into an initially empty matrix to obtain the t-th iterative correlation matrix; perform convolution calculation on the (t+1)-th first sleep monitoring feature set and the t-th iterative correlation matrix to obtain the t-th first iterative sleep monitoring feature set.

[0047] In this embodiment, the t-th and t+1-th first sleep monitoring feature sets are first extracted from the first sleep monitoring log subsequence. Each feature set includes multidimensional feature data, such as heart rate, respiratory rate, and activity level. These features reflect the user's sleep state within a specific time period. To measure the similarity between these two sets, the inner product mapping similarity method is used to calculate their correlation. The inner product similarity is calculated by pairwise multiplication of each feature value to obtain a similarity score between the two sets of features. The t-th first inner product mapping similarity set is obtained through the aforementioned process.

[0048] Next, the obtained t-th first inner product mapping similarity set is normalized to transform the data to a uniform scale range, typically the [0, 1] interval. In this step, by traversing all similarity values ​​in the similarity set, the min-max normalization method is applied to compress the similarity values ​​to a uniform range. After this normalization, all similarity values ​​are on the same scale, ensuring the fairness of subsequent analysis. Finally, the normalized similarity set is embedded into an initially empty matrix to obtain the t-th iterative correlation matrix.

[0049] Subsequently, the trained convolutional computation network layer is used to perform convolutional computation on the (t+1)th first sleep monitoring feature set and the tth iterative correlation matrix. The convolutional computation extracts local feature relationships, capturing deeper dynamic correlations between features at different time points, and ultimately generates the tth first iterative sleep monitoring feature set.

[0050] Furthermore, the method provided in the application embodiments also includes: Multiple sample sleep monitoring feature sets and multiple sample iterative correlation matrices are obtained, along with the corresponding multiple sample first iterative sleep monitoring feature sets, as training sets. The convolutional network layer is trained using the training set, and the network parameters are updated according to the training results until the training converges, thus obtaining the trained convolutional computation network layer. Based on the convolutional computation network layer, the (t+1)th first sleep monitoring feature set and the tth iterative correlation matrix are convolved to obtain the tth first iterative sleep monitoring feature set.

[0051] In this embodiment, multiple sample data points are first extracted from a historical sleep monitoring database using database query technology. The sample data includes multiple sets of sleep monitoring features and iterative correlation matrices calculated by an algorithm, which record the correlation between each pair of features in the time series. Furthermore, a first iterative sleep monitoring feature set for each sample is extracted, representing the sleep state after integrating the correlations of features before and after the iteration. Then, by dividing the model into training, validation, and test sets, the scientific nature and generalization ability of the model training are ensured.

[0052] Next, a Convolutional Neural Network (CNN) is used to train the training data. The CNN calculates the prediction results for the input data through forward propagation and uses a loss function (such as mean squared error) to calculate the difference between the model's predictions and the actual values. Based on the calculated loss value, the weights and biases of the convolutional layers are adjusted through backpropagation to minimize the loss function and optimize the network parameters. After multiple rounds of iterative training, the convolutional network completes the parameter updates, ultimately training a precise convolutional computation network layer that performs excellently in capturing local features and correlations.

[0053] Once training is complete, the newly collected sleep monitoring data is analyzed using the trained convolutional neural network layers. Specifically, the (t+1)th first sleep monitoring feature set and the t-th iteration correlation matrix are input into the trained convolutional network for convolutional computation. The convolutional network extracts features from the input data using a sliding window and captures the feature correlation information between the current time point and previous and subsequent time points using convolution operations. Finally, after convolutional computation, the t-th first iteration sleep monitoring feature set is obtained.

[0054] Step S300: Based on the first associated interactive sleep monitoring feature set and the second associated interactive sleep monitoring feature set, adjust the preset first monitoring frequency and the preset second monitoring frequency respectively to obtain the first adjusted monitoring frequency and the second adjusted monitoring frequency.

[0055] In this embodiment, sleep monitoring feature sets and corresponding monitoring frequency records for multiple samples are first obtained from a historical database. The sleep monitoring feature sets include various sleep state characteristic data of the user at specific time points, such as deep sleep, light sleep, respiratory rate, and heart rate, representing changes in sleep state. The monitoring frequency records contain preset monitoring frequencies and adjusted monitoring frequencies corresponding to these feature data, indicating how the monitoring frequency is dynamically adjusted based on different sleep state changes.

[0056] When using this data as input to train a neural network model, the input data includes a set of sleep monitoring features and a preset monitoring frequency, while the output data is the adjusted monitoring frequency. During the training process, a feedforward neural network is employed. Through multiple rounds of training, the network learns the non-linear relationship between the input data (i.e., the sleep monitoring feature set and the preset monitoring frequency) and the output data (i.e., the adjusted monitoring frequency). The training process uses the backpropagation algorithm to continuously adjust the weights in the network, ultimately resulting in a trained neural network model. This model can predict the adjusted monitoring frequency based on real-time input sleep feature data and the preset monitoring frequency.

[0057] After the model training is complete, the first set of associated interactive sleep monitoring features and the preset first monitoring frequency will be used as input data and fed into the trained neural network to obtain the first adjusted monitoring frequency. The second set of associated interactive sleep monitoring features and the preset second monitoring frequency will be used as input data and fed into the trained neural network to obtain the second adjusted monitoring frequency.

[0058] Step S400: The sleep monitoring device monitors the sleep state of the target user according to the first adjusted monitoring frequency and the second adjusted monitoring frequency.

[0059] In this embodiment, the sleep monitoring device monitors the sleep state of the target user based on a first adjusted monitoring frequency and a second adjusted monitoring frequency. Specifically, the sleep monitoring device uses both adjusted monitoring frequencies simultaneously for data collection, ensuring efficient and accurate capture of changes in the user's sleep state throughout the monitoring process.

[0060] In summary, the embodiments of this application have at least the following technical effects: This application obtains logs of a sleep monitoring device monitoring the sleep state of a target user within a preset monitoring window, according to a preset first monitoring frequency and a preset second monitoring frequency, to obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; it iterates through the first and second sleep monitoring log subsequences to perform sleep feature analysis, and performs sequential correlation iterative analysis on the analysis results to determine a first set of correlated interactive sleep monitoring features and a second set of correlated interactive sleep monitoring features; based on the first and second sets of correlated interactive sleep monitoring features, it adjusts the preset first monitoring frequency and the preset second monitoring frequency to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency; the sleep monitoring device monitors the sleep state of the target user according to the first and second adjusted monitoring frequencies. This invention addresses the technical problems of existing sleep monitoring technologies, such as fixed monitoring frequency, low monitoring efficiency, and inability to adapt to sleep states in real time. By collecting sleep state data of target users within a preset monitoring window, performing sleep feature analysis and iterative correlation analysis, a first set of correlated interactive sleep monitoring features and a second set of correlated interactive sleep monitoring features are obtained. Based on these feature sets, the preset monitoring frequency is adjusted to obtain optimized first and second adjusted monitoring frequencies for sleep state monitoring, thereby improving the accuracy and efficiency of sleep monitoring.

[0061] Example 2, based on the same inventive concept as the frequency dynamic adjustment state monitoring method in the foregoing examples, such as... Figure 2 As shown, this application provides a frequency dynamic adjustment state monitoring system. The system and method embodiments in this application are based on the same inventive concept. The system includes: The system comprises: a monitoring log acquisition module 11, which acquires logs from the sleep monitoring device monitoring the sleep state of the target user within a preset monitoring window according to a preset first monitoring frequency and a preset second monitoring frequency, respectively, to obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence; a sleep feature analysis module 12, which traverses the first and second sleep monitoring log subsequences to perform sleep feature analysis, and performs sequential correlation iterative analysis on the analysis results to determine a first and a second set of correlated interactive sleep monitoring features; a monitoring frequency adjustment module 13, which adjusts the preset first and second monitoring frequencies based on the first and second sets of correlated interactive sleep monitoring features, respectively, to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency; and a sleep state monitoring module 14, which monitors the sleep state of the target user through the sleep monitoring device according to the first and second adjusted monitoring frequencies.

[0062] Furthermore, the system is also used to implement the following functions: The sleep feature recognition network layer is invoked to perform sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, respectively, to obtain the first sleep monitoring feature set subsequence and the second sleep monitoring feature set subsequence. The t-th and t+1-th first sleep monitoring feature sets in the first sleep monitoring feature set subsequence are obtained and subjected to iterative correlation analysis to obtain the t-th first iteration sleep monitoring feature set. After multiple iterations, until the end of the first sleep feature set subsequence is reached, the n-th first iteration sleep monitoring feature set obtained in the last iteration is taken as the first correlated interactive sleep monitoring feature set, where 1 ≤ t. ≤n+1, where n is the total number of logs in the first sleep monitoring log subsequence minus 1, and n is an integer greater than or equal to 1; obtain the q-th second sleep monitoring feature set and the (q+1)-th second sleep monitoring feature set in the second sleep monitoring feature set subsequence and perform iterative analysis to obtain the q-th second iteration sleep monitoring feature set. After multiple iterations, until the end of the second sleep monitoring feature set subsequence is reached, the m-th second iteration sleep monitoring feature set obtained in the last iteration is taken as the second associated interactive sleep monitoring feature set, where 1≤q≤m+1, m is the total number of logs in the second sleep monitoring log subsequence minus 1, and m is an integer greater than or equal to 1.

[0063] Furthermore, the system is also used to implement the following functions: Calculate the inner product mapping similarity of each sleep monitoring feature in the t-th first sleep monitoring feature set and the (t+1)-th first sleep monitoring feature set to obtain the t-th first inner product mapping similarity set; traverse the t-th first inner product mapping similarity set, perform normalization processing, and embed the processing result into an initially empty matrix to obtain the t-th iterative correlation matrix; perform convolution calculation on the (t+1)-th first sleep monitoring feature set and the t-th iterative correlation matrix to obtain the t-th first iterative sleep monitoring feature set.

[0064] Furthermore, the system is also used to implement the following functions: Multiple sample sleep monitoring feature sets and multiple sample iterative correlation matrices are obtained, along with the corresponding multiple sample first iterative sleep monitoring feature sets, as training sets. The convolutional network layer is trained using the training set, and the network parameters are updated according to the training results until the training converges, thus obtaining the trained convolutional computation network layer. Based on the convolutional computation network layer, the (t+1)th first sleep monitoring feature set and the tth iterative correlation matrix are convolved to obtain the tth first iterative sleep monitoring feature set.

[0065] Furthermore, the system is also used to implement the following functions: Obtain the historical sleep monitoring log sequence of the target user within the historical monitoring window; traverse the historical sleep monitoring log sequence to extract abnormal logs, obtaining multiple historical abnormal sleep monitoring logs, wherein each historical abnormal sleep monitoring log includes a historical node; extract abnormal features based on the multiple historical abnormal sleep monitoring logs to obtain multiple historical abnormal sleep feature sets; aggregate the multiple historical abnormal sleep monitoring logs according to the multiple historical abnormal sleep feature sets to obtain multiple clustered historical abnormal sleep monitoring log sets; perform monitoring frequency analysis based on the multiple clustered historical abnormal sleep monitoring log sets and the corresponding historical nodes to obtain the preset first monitoring frequency and the preset second monitoring frequency.

[0066] Furthermore, the system is also used to implement the following functions: A first set of historical abnormal sleep features is randomly selected from the plurality of historical abnormal sleep feature sets and used as the first clustering target. The plurality of historical abnormal sleep feature sets are then aggregated toward the first clustering target according to a preset clustering similarity threshold to obtain a first cluster of historical abnormal sleep monitoring log sets. The first cluster of historical abnormal sleep monitoring log sets is then removed from the plurality of historical abnormal sleep feature sets. A second set of historical abnormal sleep features is then randomly selected from the removed set of historical abnormal sleep features and aggregated with the first clustering target to obtain a second cluster of historical abnormal sleep monitoring log sets. This process is repeated multiple times until all the historical abnormal sleep feature sets are aggregated to obtain multiple clusters of historical abnormal sleep monitoring log sets.

[0067] Furthermore, the system is also used to implement the following functions: The similarity between the first clustering target and each of the multiple historical abnormal sleep feature sets is calculated to obtain multiple first clustering target similarities; the historical abnormal sleep monitoring logs corresponding to the first clustering target similarities that are greater than or equal to the preset clustering similarity threshold are added to the first cluster historical abnormal sleep monitoring log set.

[0068] Furthermore, the system is also used to implement the following functions: Based on the multiple clustered historical abnormal sleep monitoring log sets and corresponding historical nodes, multiple minimum abnormal intervals and multiple maximum abnormal intervals are determined; the minimum value of the multiple minimum abnormal intervals is taken as the preset first monitoring frequency; and the maximum value of the multiple maximum abnormal intervals is taken as the preset second monitoring frequency.

[0069] In Embodiment 3, based on the same inventive concept as the frequency dynamic adjustment state monitoring method in the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program, which, when executed, implements the steps of any one of the methods in Embodiment 1 above.

[0070] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0071] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0072] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for monitoring the state of dynamic frequency adjustment, characterized in that, The method includes: Obtain logs from the sleep monitoring device monitoring the sleep state of the target user within a preset monitoring window, according to a preset first monitoring frequency and a preset second monitoring frequency, and obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence. The sleep feature analysis is performed by traversing the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, and the analysis results are analyzed in sequence to determine the first associated interactive sleep monitoring feature set and the second associated interactive sleep monitoring feature set. Based on the first set of associated interactive sleep monitoring features and the second set of associated interactive sleep monitoring features, the preset first monitoring frequency and the preset second monitoring frequency are adjusted respectively to obtain the first adjusted monitoring frequency and the second adjusted monitoring frequency. The sleep monitoring device monitors the sleep state of the target user according to the first and second adjusted monitoring frequencies.

2. The frequency dynamic adjustment state monitoring method as described in claim 1, characterized in that, include: The sleep feature recognition network layer is invoked to perform sleep feature analysis on the first sleep monitoring log subsequence and the second sleep monitoring log subsequence, respectively, to obtain the first sleep monitoring feature set subsequence and the second sleep monitoring feature set subsequence; The first sleep monitoring feature set t and the first sleep monitoring feature set (t+1) in the first sleep monitoring feature set subsequence are obtained and subjected to iterative analysis to obtain the first iterative sleep monitoring feature set t. After multiple iterations, the end of the first sleep monitoring feature set subsequence is reached. The first iterative sleep monitoring feature set n obtained in the last iteration is taken as the first associated interactive sleep monitoring feature set, where 1≤t≤n+1, n is the total number of logs in the first sleep monitoring log subsequence minus 1, and n is an integer greater than or equal to 1. The q-th and q+1-th second sleep monitoring feature sets in the second sleep monitoring feature set subsequence are obtained and subjected to iterative analysis to obtain the q-th second iterative sleep monitoring feature set. After multiple iterations, until the end of the second sleep monitoring feature set subsequence is reached, the m-th second iterative sleep monitoring feature set obtained in the last iteration is taken as the second associated interactive sleep monitoring feature set, where 1≤q≤m+1, m is the total number of logs in the second sleep monitoring log subsequence minus 1, and m is an integer greater than or equal to 1.

3. The frequency dynamic adjustment state monitoring method as described in claim 2, characterized in that, include: Calculate the inner product mapping similarity of each sleep monitoring feature in the t-th first sleep monitoring feature set and the (t+1)-th first sleep monitoring feature set to obtain the t-th first inner product mapping similarity set; The t-th first inner product mapping similarity set is traversed and normalized, and the processing result is embedded into an initially empty matrix to obtain the t-th iterative correlation matrix; The first sleep monitoring feature set (t+1) and the t-th iterative correlation matrix are convolved to obtain the t-th first iterative sleep monitoring feature set.

4. The frequency dynamic adjustment state monitoring method as described in claim 3, characterized in that, include: Obtain multiple sets of sleep monitoring features from multiple samples and multiple iterative correlation matrices from multiple samples, as well as the corresponding sets of sleep monitoring features from the first iteration of multiple samples, as training sets; The convolutional network layer is trained using the training set, and the network parameters are updated based on the training results until the training converges, thus obtaining the trained convolutional computation network layer. Based on the convolutional computation network layer, the (t+1)th first sleep monitoring feature set and the tth iterative correlation matrix are convolved to obtain the tth first iterative sleep monitoring feature set.

5. The state monitoring method for dynamic frequency adjustment as described in claim 1, characterized in that, Obtain logs from the sleep monitoring device monitoring the target user's sleep state at preset first and second monitoring frequencies within a preset monitoring window, respectively, to obtain a first sleep monitoring log subsequence and a second sleep monitoring log subsequence, which also includes: Obtain the historical sleep monitoring log sequence of the target user within the historical monitoring window; The historical sleep monitoring log sequence is traversed to extract abnormal logs, resulting in multiple historical abnormal sleep monitoring logs, each of which includes a historical node. Based on the multiple historical abnormal sleep monitoring logs, abnormal features are extracted to obtain multiple sets of historical abnormal sleep features. Based on the multiple sets of historical abnormal sleep features, the multiple historical abnormal sleep monitoring logs are aggregated in the same category to obtain multiple clustered sets of historical abnormal sleep monitoring logs. Based on the multiple clustered historical abnormal sleep monitoring log sets and corresponding historical nodes, a monitoring frequency analysis is performed to obtain the preset first monitoring frequency and the preset second monitoring frequency.

6. The frequency dynamic adjustment state monitoring method as described in claim 5, characterized in that, include: A first set of historical abnormal sleep features is randomly selected from the plurality of historical abnormal sleep feature sets and used as the first clustering target. The multiple sets of historical abnormal sleep features are aggregated toward the first clustering target according to a preset clustering similarity threshold to obtain a first cluster of historical abnormal sleep monitoring logs. The first cluster of historical abnormal sleep monitoring logs is removed from the plurality of historical abnormal sleep feature sets. Then, a second historical abnormal sleep feature set is randomly selected from the plurality of historical abnormal sleep feature sets after the removal. The first clustering target is then aggregated into the same category to obtain the second cluster of historical abnormal sleep monitoring logs. After multiple aggregations, until all sets of historical abnormal sleep features are aggregated, multiple clustered sets of historical abnormal sleep monitoring logs are obtained.

7. The frequency dynamic adjustment state monitoring method as described in claim 6, characterized in that, The multiple sets of historical abnormal sleep features are aggregated towards the first clustering target according to a preset clustering similarity threshold to obtain a first cluster of historical abnormal sleep monitoring log sets, including: The similarity between the first clustering target and each of the multiple historical abnormal sleep feature sets is calculated to obtain the similarity between the first clustering target and the multiple historical abnormal sleep feature sets. The historical abnormal sleep monitoring logs corresponding to the first cluster target similarity values ​​that are greater than or equal to the preset clustering similarity threshold are added to the first cluster historical abnormal sleep monitoring log set.

8. The frequency dynamic adjustment state monitoring method as described in claim 5, characterized in that, include: Based on the multiple clustered historical abnormal sleep monitoring log sets and corresponding historical nodes, multiple minimum abnormal intervals and multiple maximum abnormal intervals are determined; The minimum value of the plurality of minimum anomaly intervals is taken as the preset first monitoring frequency; The maximum value of the plurality of maximum anomaly intervals is used as the preset second monitoring frequency.

9. A state monitoring system for dynamic frequency adjustment, characterized in that, The system includes: The monitoring log acquisition module acquires logs from the sleep monitoring device monitoring the sleep state of the target user within a preset monitoring window according to a preset first monitoring frequency and a preset second monitoring frequency, and obtains a first sleep monitoring log subsequence and a second sleep monitoring log subsequence. The sleep feature analysis module traverses the first sleep monitoring log subsequence and the second sleep monitoring log subsequence to perform sleep feature analysis, and performs sequential correlation iterative analysis on the analysis results to determine the first correlation interaction sleep monitoring feature set and the second correlation interaction sleep monitoring feature set. A monitoring frequency adjustment module adjusts the preset first monitoring frequency and the preset second monitoring frequency based on the first associated interactive sleep monitoring feature set and the second associated interactive sleep monitoring feature set, respectively, to obtain a first adjusted monitoring frequency and a second adjusted monitoring frequency. A sleep state monitoring module, wherein the sleep state monitoring module monitors the sleep state of the target user through the sleep monitoring device according to the first adjustment monitoring frequency and the second adjustment monitoring frequency.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a state monitoring method for dynamic frequency adjustment as described in any one of claims 1-8.

Images