Crowd portrait dynamic clustering method and device, equipment, product and storage medium

Abstract

The invention discloses a crowd portrait dynamic clustering method and device, a cloud server, user equipment, a program product and a storage medium, and the method comprises the steps: screening a first static feature parameter, a first dynamic feature parameter and a first gait feature parameter from a first data set, the first data set comprises a plurality of static feature parameters, a plurality of dynamic feature parameters and a plurality of gait feature parameters of the crowd; determining a crowd portrait clustering result according to the first static feature parameter, the first dynamic feature parameter and the first gait feature parameter; a first parameter set of the crowd portrait clustering result is determined, the crowd portrait clustering result and the first parameter set are sent to user equipment, and the crowd portrait clustering result and the first parameter set are used for the user equipment to determine individual dynamic and static portraits.

Description

Dynamic clustering methods, devices, equipment, products, and storage media for crowd profiling Technical Field

[0001] This application relates to the field of data service technology, and in particular to a method and apparatus for dynamic clustering of population profiles, a cloud server, user equipment, program products, and storage media. Background Technology

[0002] The depiction of user profiles using digital twins based on vital signs has a stronger dynamic requirement than traditional user profiling. In the internet age, user profiles are typically based on internet user behavior habits, including shopping, browsing, and searching, as well as demographic data, including age and gender. They play a crucial role in the accurate recommendation of services by internet companies and operators. However, the user data on which internet user profiles are based, whether demographic data such as age and gender or behavioral data, is based on a period of time. This limits their ability to understand the real-time state of users. That is, a user's state and needs may differ in different real-time states such as running, working, or relaxing, leaving room for further refinement and optimization. Summary of the Invention

[0003] To address the aforementioned technical issues, this application provides a method and apparatus for dynamic clustering of population profiles, a cloud server, user equipment, program products, and storage media.

[0004] Firstly, the dynamic clustering method for user profiles provided in this application is applied to a cloud server, and the method includes:

[0005] The first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter are filtered from the first dataset, which includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of the population.

[0006] The population profile clustering results are determined based on the first static feature parameter, the first dynamic feature parameter, and the first step dynamic feature parameter.

[0007] The first parameter set of the crowd profile clustering result is determined, and the crowd profile clustering result and the first parameter set are sent to the user device. The crowd profile clustering result and the first parameter set are used by the user device to determine the dynamic and static profiles of individuals.

[0008] Secondly, the dynamic clustering method for user profiles provided in this application is applied to user devices, and the method includes:

[0009] Obtain the user's first static feature parameter data, first dynamic feature parameter data, and first step state feature parameter data;

[0010] Determine the individual static profile template based on the first static feature parameter data;

[0011] The individual's static and dynamic profiles are determined based on the individual static profile template, the first dynamic feature parameter data, and the first dynamic feature parameter data.

[0012] Thirdly, the dynamic clustering method for user profiles provided in this application is applied to user devices, and the method includes:

[0013] A second dataset of subjects was obtained, which included subjects' age data, subjects' dynamic characteristic parameter data, and subjects' gait characteristic parameter data;

[0014] The first and second cluster categories are determined based on age data, and the second parameter set of the first and second cluster categories is obtained.

[0015] The subject's first score is determined based on the subject's gait characteristic parameter data and a second parameter set.

[0016] Fourthly, the dynamic clustering device for user profiles provided in this application is applied to a cloud server, and the device includes:

[0017] The first filtering unit is used to filter the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter from the first dataset. The first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of the population.

[0018] The first determining unit is used to determine the population profile clustering result based on the first static feature parameter, the first dynamic feature parameter, and the first step dynamic feature parameter; and to determine the first parameter set of the population profile clustering result.

[0019] The first sending unit is used to send the crowd profile clustering results and the first parameter set to the user equipment. The crowd profile clustering results and the first parameter set are used by the user equipment to determine the dynamic and static profiles of individuals.

[0020] Fifthly, the dynamic clustering device for user profiles provided in this application is applied to user equipment, and the device includes:

[0021] The first acquisition unit is used to acquire the user's first static feature parameter data, first dynamic feature parameter data, and first step state feature parameter data;

[0022] The second determining unit is used to determine an individual static portrait template based on the first static feature parameter data; and to determine an individual dynamic and static portrait based on the individual static portrait template, the first dynamic feature parameter data, and the first step dynamic feature parameter data.

[0023] Sixthly, the dynamic clustering device for user profiles provided in this application is applied to user equipment, and the device includes:

[0024] The second acquisition unit is used to acquire the subject's second dataset, which includes the subject's age data, the subject's dynamic characteristic parameter data, and the subject's gait characteristic parameter data.

[0025] The third determining unit is used to determine the first cluster category and the second cluster category based on age data, and to obtain the second parameter set of the first cluster category and the second cluster category; and to determine the subject's first score based on the subject's gait characteristic parameter data and the second parameter set.

[0026] Seventhly, the cloud server provided in this application includes: a first processor, a first memory, and a first communication bus; the first communication bus is used to realize the communication connection between the first processor and the first memory; the first processor is used to execute one or more computer programs stored in the first memory to implement the above-mentioned dynamic clustering method for crowd profiles applied to the cloud server.

[0027] Eighthly, the user equipment provided in this application includes: a processor and a memory, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to execute the above-described dynamic clustering method for user profiles applied to the user equipment.

[0028] Ninthly, this application provides a computer program product comprising: a computer program that, when executed by a processor, implements any of the methods described above.

[0029] Tenthly, the computer-readable storage medium provided in this application is used to store a computer program that causes a computer to perform any of the methods described above.

[0030] In the technical solution of this application, a first static feature parameter, a first dynamic feature parameter, and a first gait feature parameter are selected from a first dataset. The first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of the population. The population profile clustering result is determined based on the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter. A first parameter set of the population profile clustering result is determined, and the population profile clustering result and the first parameter set are sent to the user equipment. The population profile clustering result and the first parameter set are used by the user equipment to determine the individual's static and dynamic profiles. Thus, by selecting a portion of the static feature parameters, dynamic feature parameters, and gait feature parameters from a massive amount of parameters, the population dynamic profile clustering result is obtained, ensuring the accuracy of the population dynamic profile clustering result. It also fully utilizes the real-time body area network capabilities of the operator's edge-cloud collaboration, reducing system overhead. Attached Figure Description

[0031] Figure 1 is a schematic diagram of a static user profile in a hospital pilot project provided in an embodiment of this application;

[0032] Figure 2 is a schematic diagram of a complete user profile of medical needs provided in the embodiments of this application;

[0033] Figure 3 is a flowchart illustrating the dynamic clustering method for population profiles provided in this embodiment of the application.

[0034] Figure 4 is a flowchart of the dynamic clustering method for population profiles provided in the embodiments of this application.

[0035] Figure 5 is a flowchart illustrating the dynamic clustering method for population profiles provided in this embodiment of the application.

[0036] Figure 6 is a schematic diagram of the process of calculating dynamic parameters by a miniature long standby sensor on the sole of the foot according to an embodiment of this application;

[0037] Figure 7 is a schematic diagram of the gait attitude calculation process provided in the embodiments of this application;

[0038] Figure 8 is a schematic diagram of the gait posture calculation results provided in the embodiments of this application;

[0039] Figure 9 is a schematic diagram of the data acquisition device provided in an embodiment of this application;

[0040] Figure 10 is a schematic diagram of the division of daily activities of a person provided in an embodiment of this application;

[0041] Figure 11 is a schematic diagram of a complete user profile combined with wearable vital signs sensors provided in an embodiment of this application;

[0042] Figure 12 is a schematic diagram of the dynamic parameters provided in the embodiments of this application observed by the naked eye using existing methods;

[0043] Figure 13 is a body area network architecture diagram provided in an embodiment of this application;

[0044] Figure 14 is a schematic diagram of a dynamic clustering method for a wearable system-based twin population profile based on vital signs, provided in an embodiment of this application.

[0045] Figure 15 is a schematic diagram of the structure of the dynamic clustering device for crowd profiling provided in an embodiment of this application.

[0046] Figure 16 is a schematic diagram of the structure of the dynamic clustering device for crowd profiling provided in an embodiment of this application.

[0047] Figure 17 is a schematic diagram of the structure of the dynamic clustering device for population profiling provided in an embodiment of this application.

[0048] Figure 18 is a schematic diagram of the structure of a cloud server provided in an embodiment of this application.

[0049] Figure 19 is a schematic structural diagram of a user equipment provided in an embodiment of this application;

[0050] Figure 20 is a schematic structural diagram of a chip according to an embodiment of this application. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0052] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0053] It should also be noted that the terms "first," "second," and "third" used in the embodiments of this application are only used to distinguish similar objects and do not represent a specific order of objects. It is understood that "first," "second," and "third" can be interchanged in a specific order or sequence where permissible, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein. The terms "system" and "network" are often used interchangeably herein. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship. It should also be understood that the "instruction" mentioned in the embodiments of this application can be a direct instruction, an indirect instruction, or an indication of an association relationship. For example, A instructing B can mean that A directly instructs B, for example, B can be obtained through A; it can also mean that A indirectly instructs B, for example, A instructs C, and B can be obtained through C; or it can mean that there is an association relationship between A and B. It should also be understood that the term "correspondence" mentioned in the embodiments of this application may indicate a direct or indirect correspondence between the two, or an association between the two, or a relationship of instruction and being instructed, configuration and being configured, etc.

[0054] In the internet age, user profiles are typically based on internet user behavior habits, including shopping, browsing, and searching; and demographic data, including age and gender. They play a crucial role in the accurate recommendation of services by internet companies and operators. For example, internet giants like Toutiao (ByteDance's news aggregator) have proposed "What you care about is what's on Toutiao," emphasizing precise push notifications based on user profiles. However, the user data used for internet user profiles, whether demographic data like age and gender or behavioral data, is based on a specific time period. This limits their ability to understand the user's real-time state. A user's state and needs may differ depending on whether they are running, at work, or relaxing, requiring further refinement. For instance, even the same person may experience real-time situations such as being unavailable, inconvenient, or in a bad mood. Further refinement of real-time states contributes to the accuracy of user profiles. The depiction of population profiles using digital twins of vital signs has a stronger dynamic requirement than traditional user profiles. With the development of the Internet of Things (IoT), especially body area networks (BANs) represented by wearable systems that monitor human vital signs in real time, important means are provided to compensate for the lack of real-time capabilities in internet user profiles. Figure 1 is a schematic diagram of a static user profile in a hospital pilot project provided in this application embodiment.

[0055] Taking medical and health applications as an example, motor behavior ability is an important feature for adjusting rehabilitation and treatment plans for chronic diseases such as cardiovascular and cerebrovascular diseases and nervous system diseases. Taking Parkinson's disease as an example, the patient's stride length, gait speed, stability, and balance when walking need to be considered not only based on demographic information such as age, gender, and weight, which do not change over a period of time, but also in combination with their current real-time status, such as walking, running, uphill, and turning. Figure 2 is a complete user profile diagram of medical needs provided in the embodiments of this application. As shown in Figure 2, part 2(a) shows the segmentation and recombination of various data in multiple dimensions, including group age, weight, height, step frequency, sleep duration, etc. Part 2(b) shows the correlation analysis between high blood pressure measurement and some behavioral habits. Correlation analysis with a certain health indicator can be achieved based on the segmentation and recombination of various data.

[0056] In hospitals, doctors typically monitor patients' real-time condition through visual observation. The aforementioned chronic diseases require long-term home-based rehabilitation, often necessitating observation of medication effects multiple times a day, several days a week, or even over several weeks, to adjust medication type and dosage, and to provide warnings. Hospital observation, on the one hand, places a heavy burden on both medical staff and patients; on the other hand, the "observer effect" (also known as the white coat effect, where a patient's gait may not fully reflect their relaxed, natural gait under the doctor's gaze due to subjective inhibition) exists. Therefore, home monitoring methods based on wearable smart systems have become an important research direction.

[0057] The distributed collaborative capabilities of the operator's body area network (BAN) at the edge and cloud are conducive to more accurate and precise characterization of user profiles. With the development of wearable devices and body area network digital twin technology, it is possible to address the shortcomings of existing user profile technologies in terms of real-time data acquisition and to achieve the aforementioned refinement and dynamism. The technical problems to be solved in this application are: on the one hand, proposing dynamic user profiles and supplementing existing static user profiles to improve the accuracy and real-time performance of user profile recommendations; on the other hand, taking into account the limited capabilities of wearable systems, in the process of integrating dynamic and static profiles, achieving on-demand selection of useful information while avoiding the introduction of massive amounts of real-time data, which would bring excessive overhead to the system. In hospital scenarios, medical staff observe the real-time behavior of patients to confirm the quality of the collected data and manually provide rehabilitation suggestions. In home scenarios, without medical staff observation, directly combining the static user profile of the user scenario with the recommendation model without analyzing the real-time status may lead to a decrease in recommendation accuracy and introduce a large amount of interfering dirty data, increasing the resource consumption of the cloud system.

[0058] Furthermore, the massive dynamic digital twins generated by body area networks (BNBs) face challenges related to dynamic and hierarchical requirements. With the development of BNBs, the sensors carried by the human body provide more real-time and comprehensive data for "digital twins" of people. Besides analyzing traditional static profiles of people (relatively stable age, height, and other behavioral habits over a period of time), it also enables more precise analysis of real-time dynamic profiles (subtle physical characteristics under different movement states). However, currently, facing massive dynamic parameters, determining which indicators to use to describe different groups of people at different times, and how to define the range of each indicator, are current problems. Therefore, how to accurately achieve dynamic clustering of population profiles becomes a problem that needs to be considered. To address this, the following technical solutions from embodiments of this application are proposed.

[0059] To facilitate understanding of the technical solutions of the embodiments of this application, the technical solutions of this application are described in detail below through specific embodiments. The above-mentioned related technologies are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. The embodiments of this application include at least some of the following contents.

[0060] Figure 3 is a flowchart illustrating the dynamic clustering method for user profiles provided in this embodiment of the application, applied to a cloud server. As shown in Figure 3, the dynamic clustering method for user profiles includes the following steps:

[0061] Step 301: Filter the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter from the first dataset. The first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of the population.

[0062] Step 302: Determine the population profile clustering result based on the first static feature parameter, the first dynamic feature parameter, and the first dynamic feature parameter.

[0063] Step 303: Determine the first parameter set of the crowd profile clustering results, and send the crowd profile clustering results and the first parameter set to the user device. The crowd profile clustering results and the first parameter set are used by the user device to determine the dynamic and static profiles of individuals.

[0064] In some implementations, the first dataset is data from the original dataset, which represents data from healthy individuals. These healthy individuals can include people of different genders, ages, heights, and other physical characteristics, and this application does not limit this. The cloud server filters first static feature parameters, first dynamic feature parameters, and first gait feature parameters from the first dataset. Based on the filtered feature parameters, it determines the population profile clustering result, and then determines the first parameter set of the population profile clustering result. The first parameter set and the population profile clustering result are sent to the user device. The user device determines the dynamic and static profiles of individuals based on the population profile clustering result and the first parameter data. Thus, by filtering some static, dynamic, and gait parameters from a massive amount of parameters, the accuracy of the population dynamic profile clustering result is ensured. It also fully utilizes the real-time body area network capabilities of the operator's edge-cloud collaboration, reducing system overhead.

[0065] In some implementations, static feature parameters include age, height, gender, BMI, etc., while dynamic feature parameters include parameters generated during the user's movement, such as walking, running, jumping, going upstairs, going downstairs, etc. Gait feature parameters include stride length, cadence, etc. It can be understood that each dynamic feature parameter contains dynamic data of the monitored object under a dynamic indicator, and each gait feature parameter contains gait data of the monitored object under a gait indicator.

[0066] In some implementations, a crowd gait profiling clustering model is used to obtain clustering results. A first static feature parameter, a first dynamic feature parameter, and a first gait feature parameter are used as inputs to obtain the clustering results.

[0067] In some implementations, the first parameter set consists of the center and boundary values ​​of the gait characteristic parameters of healthy individuals within the 80%-99% confidence interval of the clustering results. For example: based on the age of the subjects, a gait profile category is determined. All data under that category are retrieved from the healthy population clustering gait characteristic database. Then, the data within the 95% confidence interval is saved, and the mean of the gait characteristic data under that category is calculated as follows:

[0068]

[0069] Among them, Y i This represents the i-th data point under the selected gait profile category, N is the number of data points under the selected gait profile category, and μ is the mean of the gait feature data under this category;

[0070] Then, the variance of the gait feature data under this category is calculated as follows:

[0071]

[0072] Where, σ 2 Let σ represent the variance of the gait characteristic data for this category, and σ be the standard error. Then, use the following formula to calculate the 95% confidence interval for the gait characteristic data of healthy individuals:

[0073]

[0074] Where 1.96 is the coefficient of the standard error corresponding to the 95% confidence interval, and Confidence_Interval 0.95 This represents the 95% confidence interval of the calculated gait characteristic data of healthy individuals; the center value of the data within the 95% confidence interval is μ, and the boundary values ​​are the upper and lower limits of the confidence interval.

[0075] In some implementations, the first gait feature parameter includes a second gait feature parameter and a third gait feature parameter; filtering the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter from the first dataset includes: filtering the first static feature parameter and the second gait feature parameter from multiple static feature parameters and multiple gait feature parameters based on the correlation between multiple static feature parameters and multiple gait feature parameters; and filtering the first dynamic feature parameter and the third gait feature parameter from multiple dynamic feature parameters and multiple gait feature parameters based on the correlation between the first static feature parameter, multiple dynamic feature parameters, and multiple gait feature parameters.

[0076] In some implementations, demographic static index data and gait feature parameter data are first loaded; that is, multiple static feature parameters and multiple gait feature parameters from a first dataset are loaded. The correlation between the multiple static feature parameters and the multiple gait feature parameters is calculated for each static feature parameter, and a first static feature parameter and a second gait feature parameter are obtained. Then, a first dynamic feature parameter and a third gait feature parameter are selected based on the correlation between the first static feature parameter and multiple dynamic feature parameters and multiple gait feature parameters in the first dataset. For example, the Pearson correlation coefficient can be used to calculate the correlation. It can be understood that the first static feature parameter related to the gait feature parameter is first selected, and after confirming the first static feature parameter, the first dynamic feature parameter and the first gait feature parameter are selected under that static feature parameter.

[0077] In some implementations, selecting a first static feature parameter and a second gait feature parameter from multiple static feature parameters and multiple gait feature parameters based on the correlation between multiple static feature parameters and multiple gait feature parameters includes: determining a first set of correlation coefficients for multiple static feature parameters and multiple gait feature parameters; selecting a first static feature parameter from multiple static feature parameters based on the first set of correlation coefficients; and selecting a second gait feature parameter from multiple gait feature parameters based on the first static parameter and the first set of correlation coefficients.

[0078] In some implementations, the Pearson correlation coefficients of multiple static feature parameters and multiple gait feature parameters are calculated as follows:

[0079]

[0080] Among them, X i，k For the k-th data point of the i-th demographic indicator, Y j，k For the k-th data of the j-th gait feature parameter, Let be the average value of the demographic information data of the i-th individual. Let r be the average value of the j-th gait feature parameter, N be the number of data in the database, and r be the average value of the j-th gait feature parameter. i，j Let be the Pearson correlation coefficient between the i-th population informatics index and the j-th gait feature parameter.

[0081] In some implementations, the set of Person correlation coefficients between the calculated static feature parameters and gait feature parameters is defined as the first correlation coefficient set. The mean of all correlation coefficients for each static feature parameter is calculated, and the static feature parameter corresponding to the largest correlation coefficient is selected as the first static feature parameter. Alternatively, the static feature parameter whose correlation coefficient satisfies certain conditions is selected as the first static feature parameter. The specific conditions that the correlation coefficients must satisfy can be determined based on actual circumstances, and this application does not impose specific limitations. It should be noted that the first static feature parameter is at least one static feature parameter. The static feature parameter selected according to the selection conditions is the first static feature parameter, which is the most relevant static feature parameter. For example, the mean is calculated as follows:

[0082]

[0083] Where L is the number of gait feature parameters. Let be the mean correlation coefficient of the i-th population informatics index.

[0084] In some implementations, filtering a second gait feature parameter from multiple gait feature parameters based on a first static parameter and a first set of correlation coefficients includes: obtaining a second set of correlation coefficients related to the first static parameter, wherein the first set of correlation coefficients includes the second set of correlation coefficients; and filtering the second gait feature parameter from multiple gait feature parameters based on the second set of correlation coefficients and a first threshold.

[0085] In some implementations, after selecting the first static feature parameters, a second set of correlation coefficients related to the first static feature parameters is selected from the first set of correlation coefficients. Correlation coefficients with correlation coefficients greater than a first threshold are then selected from the second set of correlation coefficients, and the gait feature parameters corresponding to these correlation coefficients are designated as the second gait feature parameters. For example, the first threshold can be 0.3-0.5. A larger range of correlation coefficients greater than 0.3 can be selected to avoid overlooking some rare diseases. The specific setting of the first threshold can be determined according to the actual situation, and this application does not impose specific limitations on it.

[0086] In some implementations, selecting a first dynamic feature parameter and a third gait feature parameter from multiple dynamic feature parameters and multiple gait feature parameters based on the correlation between a first static feature parameter, multiple dynamic feature parameters, and multiple gait feature parameters includes: determining a third set of correlation coefficients for multiple dynamic feature parameters and multiple gait feature parameters under the first static parameter; selecting a first dynamic feature parameter from multiple dynamic feature parameters based on the third set of correlation coefficients; and selecting a second gait feature parameter from multiple gait feature parameters based on the first dynamic feature parameter and the third set of correlation coefficients.

[0087] In some implementations, for each selected first static feature parameter, the Pearson correlation coefficient between multiple dynamic feature parameters and multiple gait feature parameters is calculated as follows:

[0088]

[0089] Among them, X i，k For the k-th data point of the i-th dynamic indicator parameter, Y j，k For the k-th data of the j-th gait feature parameter, Let be the average value of the i-th dynamic indicator. Let r be the average value of the j-th gait feature parameter, N be the number of data in the database, and r be the average value of the j-th gait feature parameter. i，j Let be the Pearson correlation coefficient between the i-th dynamic index parameter and the j-th gait feature parameter.

[0090] In some implementations, the set of Person correlation coefficients of the calculated dynamic feature parameters and gait feature parameters is used as a third correlation coefficient set. The mean of all correlation coefficients for each dynamic feature parameter is calculated, and the dynamic feature parameter corresponding to the largest correlation coefficient is selected as the first dynamic feature parameter. Alternatively, the dynamic feature parameter whose correlation coefficient meets certain conditions is selected as the first dynamic feature parameter. The specific conditions that the correlation coefficients meet can be determined according to actual circumstances, and this application does not impose specific limitations. It should be noted that the first dynamic feature parameter is at least one dynamic feature parameter. The dynamic feature parameter selected according to the selection conditions is the first dynamic feature parameter, which is the most relevant dynamic feature parameter. For example, the mean is calculated as follows:

[0091]

[0092] Where L is the number of gait feature parameters. Let be the average correlation coefficient of the i-th dynamic indicator parameter.

[0093] In some implementations, selecting a second gait feature parameter from multiple gait feature parameters based on a first dynamic feature parameter and a third set of correlation coefficients includes: obtaining a fourth set of correlation coefficients related to the first dynamic feature parameter, wherein the third set of correlation coefficients includes the fourth set of correlation coefficients; and selecting a third gait feature parameter from multiple gait feature parameters based on the fourth set of correlation coefficients and a second threshold.

[0094] In some implementations, after selecting the first dynamic feature parameter, a fourth set of correlation coefficients related to the first dynamic feature parameter is selected from the third set of correlation coefficients. From the fourth set of correlation coefficients, correlation coefficients with a correlation coefficient greater than a second threshold are selected, and the gait feature parameters corresponding to these correlation coefficients are designated as the third gait feature parameters. The second threshold is greater than the first threshold and is used for subsequent processing. For example, a range with correlation coefficients greater than 0.5 can be selected. The specific setting of the second threshold can be determined according to actual circumstances, and this application does not impose specific limitations on it.

[0095] In some implementations, if the user equipment determines that the subject is a diseased population based on the clustering results obtained from the embodiments of this application, the first threshold is adjusted to be equal to the second threshold, the population gait profile clustering model is retrained, and the aforementioned steps are repeated.

[0096] Understandably, clustering results group gait data of healthy individuals with similar static human characteristics into one category.

[0097] Understandably, the clustering result based on the selected static feature parameters is a relatively large classification result. Under the selected static feature parameters, the clustering result based on the selected dynamic feature parameters and gait feature parameters is a refinement of the previous classification result.

[0098] In some implementations, the method further includes: receiving a second dataset, individual dynamic and static profiles, and health information of the subjects sent by a user device; and updating the population profile clustering results based on the second dataset, individual dynamic and static profiles, and health information of the subjects.

[0099] In some implementations, once the user device determines the profile of each user based on the clustering results, the second dataset of the subject, the individual dynamic and static profiles, and the health information are stored, and the population profile clustering results are updated based on the received second dataset, individual dynamic and static profiles, and the health information of the subject.

[0100] In some implementations, if the user device determines that the subject belongs to the disease population, the size of the first threshold is adjusted, the aforementioned clustering model is retrained, and the aforementioned operation is repeated. In this way, the problem of inaccurate results due to inaccurate data can be avoided.

[0101] In some implementations, if a subject is determined to be healthy, the subject's data is stored to update the aforementioned clustering results and / or clustering model.

[0102] In some implementations, a second dataset, individual dynamic and static profiles, and health information of the subjects are received from the user equipment; the population profile clustering model is updated based on the second dataset, individual dynamic and static profiles, and health information of the subjects.

[0103] The technical solution provided in this application involves filtering first static feature parameters, first dynamic feature parameters, and first gait feature parameters from a first dataset. The first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of a population. The population profile clustering result is determined based on the first static feature parameters, first dynamic feature parameters, and first gait feature parameters. A first parameter set for the population profile clustering result is determined, and the population profile clustering result and the first parameter set are sent to a user device. The population profile clustering result and the first parameter set are used by the user device to determine individual static and dynamic profiles. Thus, by filtering a portion of static feature parameters, dynamic feature parameters, and gait feature parameters from a massive amount of parameters, a dynamic population profile clustering result is obtained, ensuring the accuracy of the dynamic population profile clustering result. It also fully utilizes the real-time body area network capabilities of the operator's edge-cloud collaboration, reducing system overhead.

[0104] Figure 4 is a second flowchart illustrating the dynamic clustering method for user profiles provided in this application embodiment, applied to a user device. As shown in Figure 4, the dynamic clustering method for user profiles includes the following steps:

[0105] Step 401: Obtain the user's first static feature parameter data, first dynamic feature parameter data, and first step state feature parameter data;

[0106] Step 402: Determine the individual static profile template based on the first static feature parameter data;

[0107] Step 403: Determine the individual's static and dynamic portrait based on the individual static portrait template, the first dynamic feature parameter data, and the first step dynamic feature parameter data.

[0108] In some implementations, the user equipment obtains the first static feature parameter data, first dynamic feature parameter data, and first step gait feature parameter data corresponding to the user based on the first static feature parameter data, first dynamic feature parameter data, and first step gait feature parameter data determined by the cloud server. The user's static profile template is determined based on the first static feature parameter data, and then the user's individual dynamic and static profile is determined based on the user's individual static profile template, first dynamic feature parameter data, and first step gait feature parameter data. In this way, the user's dynamic and static profiles can be accurately determined based on the user's static feature parameters, dynamic feature parameters, and gait feature parameters, reducing system overhead.

[0109] In some implementations, the user device may be a mobile phone, tablet, etc.

[0110] In some implementations, before determining the individual static profile template based on the first static feature parameter data, the method further includes: obtaining the crowd profile clustering result sent by the cloud server and the first parameter set corresponding to the crowd profile clustering result; determining the individual static profile template based on the first static feature parameter data includes: determining the individual static profile template based on the first static feature parameter data, the crowd profile clustering result and the first parameter set.

[0111] In some implementations, the user device inputs demographic static features, such as age, height, gender, and BMI, and obtains the corresponding population profile clustering results and the first parameter set corresponding to the population profile clustering results from the cloud server. Based on the static features input by the user device, the corresponding clustering result is determined. This clustering result is a twin group static profile, which is used as an individual static profile template.

[0112] In some implementations, determining an individual's static and dynamic profile based on an individual static profile template, first dynamic feature parameter data, and first step state feature parameter data includes: determining an error value based on the individual static profile template and first step state feature parameter data; determining a corrected error prediction value based on the error value; determining an individual static and dynamic profile template based on the first dynamic feature parameter data and first step state feature parameter data; and determining an individual static and dynamic profile based on the individual static and dynamic profile template and the corrected error prediction value.

[0113] In some implementations, based on all gait feature parameter data within the individual static portrait template, the comprehensive gait feature parameter of the group static portrait cluster center under that cluster category is calculated using the following formula:

[0114]

[0115] Among them, c i For the i-th data of the j-th gait feature parameter, N i The number of gait feature data; each gait feature has a gait feature comprehensive parameter.

[0116] In some implementations, the distance between the gait feature parameters of the crowd portrait generated from the individual static portrait template and the actual gait feature parameters corresponding to the static information is calculated, and error correction is performed to obtain the error value. For example, the formula for calculating the error value is as follows:

[0117]

[0118] Where, Δ i，j，k The prediction error of cluster centers for individual twin static portraits and individual static portrait templates.

[0119] In some implementations, a corrected error prediction value is determined based on the error value. For example, the error value is input into a deep learning error correction algorithm model to obtain the corrected error prediction value, as follows:

[0120]

[0121] Where x represents the static information of the individual twin static profile, F(·) is the deep learning error correction algorithm, and Δpred is the predicted value of the correction error. This is the error value.

[0122] In some implementations, based on the user's first dynamic feature parameter data and first step dynamic feature parameter data, a corresponding clustering result is selected from the clustering results on the cloud server, i.e., a group twin dynamic and static profile, as the individual dynamic and static profile template for the user. This can be denoted as c. initial .

[0123] In some implementations, the individual dynamic and static portrait is determined based on the individual dynamic and static portrait template and the corrected error prediction value. For example, the dynamic and static information and the individual twin static portrait are input into a deep learning error correction algorithm to obtain the corrected error prediction value, and then the gait features of the individual twin dynamic and static portrait after the first correction based on the static information are calculated. The specific calculation method is as follows:

[0124] c corrected =c initial +Δ pred

[0125] Among them, c initial To create dynamic and static portraits of individuals within a population, Δ pred To correct the error prediction value, c corrected Gait characteristics for creating dynamic and static portraits of individual twins.

[0126] In some implementations, after determining the corrected error prediction value based on the error value, the method further includes: determining a loss value based on the corrected error prediction value and the error value; and determining whether a first model should stop training based on the loss value, wherein the first model represents the model that has obtained the corrected error prediction value.

[0127] In some implementations, the loss value is calculated based on the corrected error prediction value and the actual error value output by the first model. For example, the loss value is calculated as follows:

[0128]

[0129] in, Δ represents the k-th gait feature parameter in the corrected error prediction value of the subject.k This represents the corresponding true gait feature error, where K is the number of clusters, N is the number of gait feature parameters, and L is the final loss function.

[0130] In some implementations, the first model can be a deep learning error correction algorithm model, and the correction error value is obtained based on the model, and the individual dynamic and static portrait is determined by combining the individual dynamic and static portrait template.

[0131] In some implementations, determining whether the first model should stop training based on the loss value includes: obtaining the loss value of the first model during the Nth training iteration, where N is a positive integer; obtaining the loss values ​​of the first model during the previous M training iterations, where N = M + 1, and M is a positive integer; and determining whether the first model should stop training based on the loss value of the Nth iteration and the loss values ​​of the previous M iterations.

[0132] In some implementations, it is determined whether the currently calculated loss value has decreased compared to the average loss value of the previous M iterations. If the loss value has decreased, it indicates that the first model is still converging. The parameters of the deep learning error correction model are then updated using gradient descent. To avoid excessively long training times, a maximum number of training epochs is set to Z, meaning training stops after reaching the maximum number of training epochs. M and Z are positive integers, and their specific values ​​are determined based on actual conditions; this application does not impose specific limitations on this. For example: setting M to 10 and Z to 1000, if the loss value no longer decreases after 10 calculations, it indicates that the model can no longer converge, and model updates are stopped. This is specifically expressed as follows:

[0133]

[0134] Among them, L t This is the loss value calculated during the current training. The average of the loss values ​​for the first 10 iterations is denoted by 'count', which is the consecutive count of the loss value that no longer decreases during training, or the stop condition that is triggered when the loss value reaches 10.

[0135] In some implementations, the method further includes: acquiring the amount of data for individual dynamic and static portraits; when the amount of data exceeds a third threshold, updating the first model based on the dataset corresponding to the individual dynamic and static portraits.

[0136] In some implementations, when the amount of individual dynamic and static portrait data exceeds a third threshold, a model update mechanism needs to be activated. This involves retraining the crowd gait portrait clustering model and the deep learning-based error correction algorithm model with all existing data to optimize the algorithm. As the amount of data increases, iteration refines the crowd gait portrait, continuously reducing the prediction error of gait features. Here, x is a positive integer, and the specific value of x can be determined according to actual conditions; this application does not impose specific limitations on this. For example, x can be set to 200. The error correction algorithm is updated by the user device, while the crowd gait portrait clustering model is updated by the cloud server. Therefore, the cloud server also needs to obtain the individual dynamic and static portrait data from the user device. The crowd gait portrait clustering model is the model used by the cloud server to obtain the clustering results, i.e., the model for obtaining the group static portrait template and the individual dynamic and static portrait template.

[0137] In some implementations, it can also be determined whether the first model needs to be updated based on whether the amount of data for the category of individual static and dynamic portraits is greater than a third threshold.

[0138] In some implementations, it is determined whether the aforementioned crowd gait profile clustering model needs to be updated based on the amount of data in the category of individual dynamic and static profiles or based on the amount of data in individual dynamic and static profiles.

[0139] The technical solution provided in this application involves acquiring a user's first static feature parameter data, first dynamic feature parameter data, and first step-state feature parameter data; determining an individual static profile template based on the first static feature parameter data; and determining an individual dynamic and static profile based on the individual static profile template, the first dynamic feature parameter data, and the first step-state feature parameter data. In this way, the user device can determine the corresponding individual dynamic and static profile for itself based on the individual static profile template and actual data, enabling timely viewing of the user's own dynamic and static profile and facilitating the assessment of the user's health status based on this profile.

[0140] Figure 5 is a flowchart illustrating the third step of the dynamic clustering method for user profiles provided in this application embodiment, applied to a user device. As shown in Figure 5, the dynamic clustering method for user profiles includes the following steps:

[0141] Step 501: Obtain the second dataset of the subjects, which includes the subjects' age data, dynamic characteristic parameter data, and gait characteristic parameter data.

[0142] Step 502: Determine the first cluster category and the second cluster category based on the age data, and obtain the second parameter set of the first cluster category and the second cluster category.

[0143] Step 503: Determine the subject's first score based on the subject's gait characteristic parameter data and the second parameter set.

[0144] In some implementations, a second dataset of the subjects is obtained, which includes the subjects' age data, dynamic characteristic parameter data, and gait characteristic parameter data, i.e., the subjects' static characteristic parameter data, dynamic characteristic parameter data, and gait characteristic parameter data. A first cluster category and a second cluster category are determined based on the age data, and a second parameter set for each cluster category is obtained. The first cluster category can be a cluster category obtained from data on a patient population, and the second cluster category can be a cluster category obtained from data on a healthy population. A first score for the subjects is determined based on the gait characteristic parameter data and the second parameter set. This first score is the subjects' health score, used to assess their health status.

[0145] In some implementations, before obtaining the second dataset of the subjects, the method includes: obtaining a third dataset, which includes the patients' age data and the patients' second gait characteristic parameter data; obtaining the population profile clustering results sent by the cloud server; dividing the first data according to the population profile clustering results and the age data to obtain the division result, wherein the first data includes the age data in the third dataset and the age data in the population profile clustering results.

[0146] In some implementations, a third dataset needs to be obtained before obtaining the second dataset of the subjects. The third dataset includes the patients' age data and the patients' second gait characteristic parameter data, that is, the third dataset is the patient dataset. The population profile clustering results sent by the cloud server are obtained. The first data is divided according to the population profile clustering results and the age data to obtain the division results. The first data includes the age data in the third dataset and the age data in the population profile clustering results. That is, the first data includes the age data in the patient dataset and also includes the age data of healthy people.

[0147] In some implementations, the age data of patients and the age data in the population profile clustering results are divided based on the age data in the third dataset and the age boundary of the healthy population cluster to obtain the age data division results. Based on the division results and the age data of the subjects, the category to which the subjects belong is determined. All patient data in the same cluster category as the subjects are extracted, including static feature parameter data, dynamic feature parameter data and gait feature parameter data in that category. Based on the aforementioned calculation formula of gait feature mean, standard error and 80%-99% confidence interval, the relevant indicators of the patient gait feature parameter data in that category are calculated.

[0148] In some implementations, determining a first cluster category and a second cluster category based on age data includes: determining a first cluster category based on the age data of the subjects and the segmentation results; and determining a second cluster category based on the age data of the subjects and the clustering results of the population profile.

[0149] In some implementations, a first cluster category is determined based on the subjects' age data and the segmentation results. This first cluster category corresponds to the cluster category of the patient dataset. A second cluster category is determined based on the subjects' age data and the population profile clustering results. This second cluster category corresponds to the cluster category of the healthy population. This is used to determine the distribution of the subjects' gait characteristic parameters based on the subjects' gait characteristic parameters, as well as the first and second cluster categories, and to calculate the corresponding score.

[0150] In some implementations, after determining the subject's first score based on the subject's gait characteristic parameter data and a second parameter set, the method further includes: determining the subject's health status information based on the first score; and sending the second dataset, the first score, and the health status information to a cloud server.

[0151] In some implementations, determining the distribution of gait characteristic parameters of a subject among gait characteristic parameters of healthy individuals and patients includes the following three scenarios:

[0152] Scenario 1: If the subject's gait characteristic parameters fall within the range of healthy individuals' gait characteristic values, then calculate the subject's risk score in healthy gait:

[0153] First, a linear score function is generated based on the center and boundary values ​​of the gait characteristics of healthy individuals:

[0154]

[0155] Where Y represents the gait characteristic data of the subjects. μ represents the maximum value within the range of gait characteristics of healthy individuals. H This represents the center value of the numerical range of gait characteristics in healthy individuals.

[0156] By inputting the subject's gait parameters into the scoring function, the subject's risk level score is obtained.

[0157] Scenario 2: If the subject's gait characteristic parameters fall within the range of the patient's gait characteristic values, then calculate the subject's risk score in the patient's gait:

[0158] First, a linear score function is generated based on the patient's gait feature center and boundary values:

[0159]

[0160] Where Y represents the gait characteristic data of the subjects. μ represents the maximum value within the range of patient gait characteristics. D The center value of the numerical range of the patient's gait characteristics;

[0161] By inputting the subject's gait parameters into the scoring function, the severity score of the subject is obtained.

[0162] Scenario 3: When the subject's gait characteristic parameters fall at the intersection of the gait characteristic value ranges of healthy individuals and patients, both risk score and severity score are calculated simultaneously to achieve a more comprehensive assessment of the test subject's gait health status.

[0163] In some implementations, 100, 60, and 0 are respectively the health cluster center, the boundary between health and patients, and the patient cluster center. The results can be converted based on the first score to determine the condition of the subject.

[0164] In some implementations, the method further includes: acquiring the amount of data for individual dynamic and static portraits; when the amount of data is greater than a third threshold, updating the first model based on the dataset corresponding to the individual dynamic and static portraits, wherein the first model represents the model for acquiring the predicted value of the correction error.

[0165] In some implementations, when the amount of individual dynamic and static portrait data exceeds a third threshold, a model update mechanism needs to be activated. This involves retraining the crowd gait portrait clustering model and the deep learning-based error correction algorithm model with all existing data to optimize the algorithm. As the amount of data increases, iteration refines the crowd gait portrait, continuously reducing the prediction error of gait features. Here, x is a positive integer, and the specific value of x can be determined according to actual conditions; this application does not impose specific limitations on this. For example, x can be set to 200. The error correction algorithm is updated by the user device, while the crowd gait portrait clustering model is updated by the cloud server. Therefore, the cloud server also needs to obtain the individual dynamic and static portrait data from the user device. The crowd gait portrait clustering model is the model used by the cloud server to obtain the clustering results, i.e., the model for obtaining the group static portrait template and the individual dynamic and static portrait template.

[0166] In some implementations, it can also be determined whether the first model needs to be updated based on whether the amount of data for the category of individual static and dynamic portraits is greater than a third threshold.

[0167] In some implementations, the need to update the aforementioned gait profile clustering model is determined based on the amount of data in the category of individual dynamic and static profiles.

[0168] In some implementations, the first model can be a deep learning error correction algorithm model, and the correction error value is obtained based on the model, and the individual dynamic and static portrait is determined by combining the individual dynamic and static portrait template.

[0169] The technical solution of this application embodiment involves obtaining a second dataset of the subject, which includes the subject's age data, dynamic characteristic parameter data, and gait characteristic parameter data; determining a first cluster category and a second cluster category based on the age data, and obtaining a second parameter set for the first cluster category and the second cluster category; and determining a first score for the subject based on the subject's gait characteristic parameter data and the second parameter set. In this way, the subject's health status can be fully assessed based on the first score.

[0170] The technical solutions of the embodiments of this application are illustrated below with specific application examples.

[0171] Firstly, in the internet age, user profiles are typically based on internet user behavior habits, including shopping, browsing, and searching; and demographic data, including age and gender. These profiles play a crucial role in the accurate recommendation of services by internet companies and telecom operators. Therefore, the depiction of user profiles using biometric digital twins has a stronger dynamic requirement than traditional user profiles. For example, internet giants like Toutiao (ByteDance's news aggregator) have proposed "What you care about is what's on Toutiao," emphasizing precise push notifications based on user profiles. However, the user data used in internet user profiles, whether demographic data such as age and gender, or behavioral data, is based on a specific time period. This limits their ability to understand the user's real-time state. That is, a user's state and needs may differ depending on whether they are running, at work, or relaxing, requiring further refinement. For instance, even the same person may experience real-time situations such as being unavailable, inconvenient, or in a bad mood. Further refinement of real-time states contributes to the accuracy of user profiles. With the development of the Internet of Things (IoT), especially body area networks (BANs) such as wearable systems that monitor human vital signs in real time, important means are provided to compensate for the lack of real-time capabilities in internet user profiles.

[0172] Taking healthcare applications as an example, motor skills are a crucial characteristic for adjusting rehabilitation and treatment plans for chronic diseases such as cardiovascular and cerebrovascular diseases and nervous system disorders. Taking Parkinson's disease as an example, a patient's stride length, speed, stability, and balance while walking need to be considered not only based on demographic data such as age, gender, and weight, which remain constant over a period of time, but also in conjunction with their current real-time condition, such as walking, running, uphill walking, and turning. In hospitals, doctors typically observe patients' real-time condition visually. These chronic diseases often require long-term home rehabilitation, frequently necessitating observation of medication effects multiple times a day, several days a week, or even several weeks, to adjust medication types and dosages and provide warnings. Hospitalization, on the one hand, places a heavy burden on both medical staff and patients; on the other hand, the "observer effect" (also known as the white coat effect, where a patient's gait may not fully reflect the relaxed, natural gait under the doctor's gaze due to subjective inhibition) makes home monitoring methods based on wearable smart systems an important research direction.

[0173] Secondly, the distributed collaborative capabilities of the operator's body area network (BAN) at the edge and cloud levels facilitate more accurate and nuanced user profiling. With the development of wearable devices and body area network digital twin technology, addressing the shortcomings of existing user profiling technologies in terms of real-time data acquisition becomes possible, thus resolving the aforementioned issues of refinement and dynamism. The technical problems addressed in this application are: firstly, proposing dynamic user profiles to supplement existing static user profiles, improving the accuracy and real-time performance of user profile recommendations; secondly, considering the limited capabilities of wearable systems, achieving on-demand selection of useful information while avoiding the introduction of massive amounts of real-time data that would impose excessive overhead on the system during the fusion of dynamic and static profiles. In hospital settings, medical staff observe patients' real-time behavioral states to confirm the quality of collected data and manually provide rehabilitation suggestions. In home settings, without medical staff observation, directly combining static user profiles with real-time data without analyzing the real-time state may lead to decreased recommendation accuracy and introduce a large amount of interfering, dirty data, increasing cloud system resource consumption.

[0174] Finally, the challenge facing massive dynamic digital twins in body area networks is the need for dynamic and hierarchical data. With the development of body area networks, the sensors carried by the human body bring more real-time and comprehensive data to the "digital twins" of "people". In addition to analyzing the traditional static profiles of people (relatively stable age, height and other behavioral habits over a period of time), it is also possible to analyze the real-time dynamic profiles of people more accurately (subtle physical signs under different movement states).

[0175] However, the drawback of existing technologies is the lack of methods for selecting descriptive indicators and quantifying specific ranges for dynamic profiling. Firstly, given the massive amount of dynamic parameters, which indicators should be used to describe different populations and different times? Taking gait as an example, 164 gait parameters can be calculated through sensor calibration, preprocessing, feature extraction, coordinate system transformation, and posture calculation. This is relevant to the massive amounts of data collected in real-time by body area networks, which are characterized by their large quantity and dynamic nature. Firstly, there is the large quantity; taking gait as an example, 164 gait parameters are calculated through sensor calibration, preprocessing, feature extraction, coordinate system transformation, and posture calculation. Figure 6 is a schematic diagram of the process for calculating dynamic parameters using a miniature long-standby sensor on the sole of the foot, as provided in this embodiment. As shown in Figure 6, the IMU first acquires data, then preprocesses the acquired data, and then performs spatial posture calculation and temporal event calculation to obtain the final dynamic parameters. These dynamic parameters include parameters in the time domain, spatial domain, frequency domain, and energy domain. Figure 7 is a schematic diagram of the gait posture calculation process provided in this embodiment. As shown in Figure 7, it includes the following steps:

[0176] Step 1: Acquiring raw IMU data, including acceleration and angular velocity.

[0177] Step 2: IMU data preprocessing, including missing value compensation, outlier detection, and high-frequency noise filtering.

[0178] Step 3: Spatiotemporal feature analysis. This includes gait spatial posture features and gait temporal event features. Gait spatial posture features are obtained using a complementarity algorithm and quaternions, while gait temporal event features are based on peak and threshold methods, as well as gait period determination.

[0179] Step 4: Gait 3D Localization and Gait Phase Recognition. This is determined based on a zero-velocity joint zero-position gait 3D localization method and a turn recognition method based on artificial intelligence error correction.

[0180] Step 5: Calculate gait parameters to obtain parameters in the spatial, temporal, frequency, and energy domains.

[0181] Figure 8 is a schematic diagram of the gait posture calculation results provided in an embodiment of this application. As shown in Figure 8, the dynamic parameters in the time domain include the average gait period, period duration, etc., and the dynamic parameters in the spatial domain include stride length, left fore-and-aft displacement, etc. The dynamic parameters in the energy domain include driving capability, control capability, etc., and the dynamic parameters in the frequency domain include toe strike count, etc.

[0182] The number of dynamic indicators required by wearable systems is closely related to the scenario, such as step length in running, walking, and climbing stairs. If these indicators are not filtered according to the specific activity scenario, the massive amount of dirty data from the large number of parameters will affect the accuracy of cluster analysis. Figure 9 is a schematic diagram of the data acquisition device provided in an embodiment of this application. Figure 10 is a schematic diagram of the division of daily activities of a person provided in an embodiment of this application. Figure 11 is a schematic diagram of a complete user profile combined with wearable vital sign sensors provided in an embodiment of this application.

[0183] Secondly, what should be the threshold range for each indicator for different populations and at different times? Dynamic vital signs parameters, including various indicators such as stride length, stability, and balance in different scenarios, serve as external motor behavior representations of diseases and are important bases for diagnosing cardiovascular, cerebrovascular, and neurological chronic diseases such as stroke and Parkinson's disease. In hospitals, dynamic parameters are subjectively observed by medical staff, and doctors subjectively judge and correct for population differences and whether the collection process conforms to the standards based on their experience. For example, for stride length indicators, doctors will subjectively make certain adjustments for people of different heights; in different scenarios such as walking straight, uphill, and up stairs, doctors will also subjectively judge whether the collection does not meet the requirements of the collection standards and require re-collection. The current diagnostic process for doctors, taking Parkinson's disease as an example, involves subjective observation and completion of the Unified Parkinson's Disease Rating Scale (UPDRS). This scale is completed subjectively by the doctor, who selects the severity of the patient's condition and categorizes it into five levels from normal to severe. The scale assesses stride length, gait speed, foot lift, heel strike during walking, balance, turning, arm swing, frozen gait (difficulty and hesitation in initiating, turning, and approaching a target), resting tremor, and "off-period" gait due to reduced medication effectiveness. Figure 12 is a schematic diagram illustrating the dynamic parameters provided in this application's embodiment through visual observation using existing methods.

[0184] The relapse rate of related chronic diseases is high after hospital discharge, requiring long-term monitoring and adjustments to medication dosage and type. Both doctors and patients have a relatively urgent need for home-based information systems. In scenarios involving home and office work without human doctor intervention, wearable intelligent analysis information systems that collaborate across edge and cloud are needed to determine whether health conditions are exceeding acceptable limits, requiring emergency alerts and rehabilitation assistance, etc. Existing home-based information systems lack methods and mechanisms for determining dynamic parameters and health ranges based on specific scenarios.

[0185] Based on this, this application aims to provide a method for sorting and filtering portrait features and defining the range of healthy population features, combining the needs of digital twins of human body characteristics with the scenario requirements of both static and dynamic portraits. It also leverages the real-time body area network capabilities of the operator's edge-cloud collaboration to balance the computational resource demands under massive demands. This helps medical service institutions and related wearable system service providers to reduce costs and improve portability while ensuring the accuracy of gait recognition in post-diagnosis health management scenarios by optimizing the placement of sensors throughout the body. Figure 13 is a body area network architecture diagram provided in this application embodiment. As shown in Figure 13, the body area network includes an application network composed of terminals connected by a body area soft gateway, the body area soft gateway itself, and core nodes. First, relevant data is collected, including data related to external body shape and internal body structure. This data is then aggregated to the body area network core node and interacts with a cloud server, which is used for modeling applications.

[0186] Based on this, the dynamic clustering method for crowd profiles provided in the embodiments of this application will be further described. Figure 14 is a schematic diagram of a dynamic clustering method for crowd profiles based on a wearable system using a vital sign twin, provided in an embodiment of this application. As shown in Figure 14, it mainly includes a cloud-based crowd modeling application server and a human body domain mobile convergence center. In this embodiment, the cloud-based crowd modeling application server is simply referred to as the cloud, i.e., the aforementioned cloud server, and the human body domain mobile convergence center is a mobile phone, i.e., the aforementioned user device. The specific description is as follows:

[0187] Level 1 (Cloud): Filtering candidate static parameters and clustering healthy individuals according to static parameter profiles, including the following steps:

[0188] Step 1401: Model initialization training.

[0189] This model is a crowd gait profiling clustering model used to obtain clustering results.

[0190] Step 1402: Extract gait feature data from the static feature data domain in the database and filter static features.

[0191] Load the static demographic index data and gait feature parameter data, and iterate through the static demographic index data. This iteration involves calculating the correlation coefficients between all static demographic index data and all gait feature parameters to filter static features. The specific filtering process is as follows:

[0192] The Pearson correlation coefficient between each pair of demographic indicators and gait characteristic parameters is calculated as follows:

[0193]

[0194] Among them, Xi，k For the k-th data point of the i-th demographic indicator, Y j，k For the k-th data of the j-th gait feature parameter, Let be the average value of the demographic information data of the i-th individual. Let r be the average value of the j-th gait feature parameter, N be the number of data in the database, and r be the average value of the j-th gait feature parameter. i，j Let be the Pearson correlation coefficient between the i-th population informatics index and the j-th gait feature parameter.

[0195] For each demographic indicator, the mean of all correlation coefficients is calculated, and the demographic indicator corresponding to the largest correlation coefficient is selected as the most relevant indicator, i.e., the selected static parameter. The specific calculation process for the mean correlation coefficient is as follows:

[0196]

[0197] Where L is the number of gait feature parameters. Let be the mean correlation coefficient of the i-th population informatics index;

[0198] Under the most relevant index, gait features with a correlation coefficient greater than the first parameter threshold are selected, where the first parameter threshold is between 0.3 and 0.5, for subsequent data processing; taking the example of the embodiment, a larger range of thresholds with a correlation coefficient greater than 0.3 are selected to avoid missing some rare diseases.

[0199] Step 1403: Set the clustering parameters: number of clusters is k, maximum number of iterations is M.

[0200] Step 1404: Clustering model training module (calculate the Euclidean distance to the cluster center).

[0201] Step 1405: Train the stop judgment module.

[0202] If you need to stop, proceed to step 1406; otherwise, proceed to step 1404.

[0203] Step 1406: Evaluation of clustering model performance, calculation of population twin effect evaluation coefficient.

[0204] Step 1407: Determine that the evaluation coefficient of the group twin effect is greater than the first threshold.

[0205] The first threshold is used to judge the clustering effect. The specific setting can be determined according to the actual situation, and this application does not make a specific limitation on it. If the evaluation coefficient of the group twinning effect is greater than the first threshold, then step 1408 is executed; otherwise, the clustering model is retrained and step 1403 is executed.

[0206] Step 1408: Obtain the first clustering result.

[0207] Based on the selected static features, the corresponding first clustering result is determined.

[0208] The second level (cloud-based): This corresponds to dynamic scenario parameters, establishing dynamic health boundaries. For the selected static parameters, further filtering is performed on relevant dynamic indicator parameters.

[0209] Step 1409: Extract the dynamic feature data corresponding to the static features in the database and filter the dynamic feature data.

[0210] The specific screening process is as follows:

[0211] The static demographic index data and corresponding gait feature parameter data are loaded sequentially, and the dynamic demographic index data are traversed separately.

[0212] Under the selected static index parameters, the Pearson correlation coefficient between the dynamic index parameters and gait characteristic parameters is calculated as follows:

[0213]

[0214] Among them, X i，k For the k-th data point of the i-th dynamic indicator parameter, Y j，k For the k-th data of the j-th gait feature parameter, Let be the average value of the i-th dynamic indicator. Let r be the average value of the j-th gait feature parameter, N be the number of data in the database, and r be the average value of the j-th gait feature parameter. i，j Let be the Pearson correlation coefficient between the i-th dynamic index parameter and the j-th gait feature parameter.

[0215] For each dynamic indicator parameter, calculate the average of all correlation coefficients and select the dynamic indicator parameter corresponding to the largest correlation coefficient as the most relevant indicator. The calculation method for the average correlation coefficient is as follows:

[0216]

[0217] Where L is the number of gait feature parameters. Let be the average correlation coefficient of the i-th dynamic indicator parameter.

[0218] Under the most relevant metric, gait features with a correlation coefficient greater than the second parameter threshold are selected, where the second parameter is greater than the aforementioned first parameter threshold, for subsequent data processing. For example, the second parameter threshold is 0.5, but this application does not impose specific limitations on it, as the second parameter threshold can be determined according to actual circumstances.

[0219] Step 1410: Set the clustering parameters: number of clusters k, maximum number of iterations M

[0220] Step 1411: Clustering model training module (calculate the Euclidean distance to the cluster center).

[0221] Step 1412: Train the stop judgment module.

[0222] If you need to stop, proceed to step 1413; otherwise, proceed to step 1411.

[0223] Step 1413: Evaluation of clustering model performance, calculation of population twin effect evaluation coefficient.

[0224] Step 1414: Determine if the evaluation coefficient of the group twin effect is greater than the second threshold.

[0225] The threshold should be used to judge the clustering effect. The specific setting can be determined according to the actual situation, and this application does not make specific limitations on it. If the evaluation coefficient of the group twinning effect is greater than the threshold, then proceed to step 1415; otherwise, retrain the clustering model and proceed to step 1410.

[0226] Step 1415: Obtain the second clustering result.

[0227] Based on the selected static features, the corresponding second clustering results are determined according to the selected dynamic features and gait features. It can be understood that the second clustering results obtained here are a refinement of the first clustering results obtained at the first level.

[0228] Step 1416: Calculate the gait characteristic parameters of the cluster categories for healthy individuals, including the 95% interval center value and upper and lower boundaries of the cluster.

[0229] The parameters of the second cluster category in the second clustering result are calculated. In this embodiment, the center and boundary values ​​of the healthy person gait characteristic parameters of the cluster category are calculated within the 80%-99% confidence interval. Taking the embodiment as an example: based on the gait profile category of the population determined by the subject's age, all data under that category are retrieved from the healthy person clustering gait characteristic database, and then the data within the 95% confidence interval are saved.

[0230]

[0231] Among them, Y i This represents the i-th data point under the selected gait profile category, N is the number of data points under the selected gait profile category, and μ is the mean of the gait feature data under this category;

[0232]

[0233] Where, σ 2 This represents the variance of the gait feature data for that category, where σ is the standard error.

[0234]

[0235] Where 1.96 is the coefficient of the standard error corresponding to the 95% confidence interval, and Confidence_Interval 0.95 This represents the 95% confidence interval of the calculated gait characteristic data of healthy individuals; the center value of the data within the 95% confidence interval is μ, and the boundary values ​​are the upper and lower limits of the confidence interval.

[0236] The third level (mobile device): Download the corresponding static profile and make static adjustments to adapt to the individual. Download and iterate the static vital sign twin parameter profile model of the individual's healthy population, as well as the corresponding dynamic model profile, and iterate to improve it.

[0237] Step 1417: Fill in the subject's demographic static data.

[0238] Step 1418: Periodically download the dynamic scene profile feature boundaries under the static profile model corresponding to static features (age, gender, BMI, etc.).

[0239] Step 1419: Select static parameters based on the group twin static profile model.

[0240] Step 1420: Wearable sensors.

[0241] Subjects wear wearable sensors to acquire gait characteristic parameter data.

[0242] Step 1421: Gait characteristic parameters of the subject.

[0243] Step 1422: Given the dynamic profile features of the tester (walking, running, jumping, etc.), classify the tester into a dynamic population cluster category based on the cluster boundaries.

[0244] Step 1423: Category Information Feature Extraction Module.

[0245] Step 1424: Feature fusion module.

[0246] Step 1425: Error correction prediction module.

[0247] Step 1426: Correct the error prediction value.

[0248] Step 1427: Deep learning error correction.

[0249] Step 1428: Loss Calculation.

[0250] The loss is calculated based on the corrected error prediction and the deep learning error correction.

[0251] Step 1429: Determine if the loss is no longer decreasing.

[0252] If the decrease continues, proceed to steps 1430 to 1431; otherwise, proceed to step 1432.

[0253] Step 1430: Determine if the number of training rounds is less than 1000.

[0254] If the number of iterations is less than 1000, proceed to step 1431; otherwise, proceed to step 1432.

[0255] Step 1431: Update model parameters using gradient descent algorithm.

[0256] Return to step 1425.

[0257] Specifically, steps 1424 to 1431 constitute the process of the deep learning error correction algorithm and the early stopping mechanism during model training. The details are as follows:

[0258] The mobile phone periodically downloads static and dynamic images of the corresponding healthy twin groups, i.e., the aforementioned clustering results, as templates for individual twin images. The error is calculated based on the static twin image of the individual and the actual gait feature parameters.

[0259] (1) Input static demographic features (age, height, gender, BMI) via mobile phone;

[0260] (2) Download the corresponding twin group static portrait from the cloud as an individual twin static portrait template;

[0261] (3) Based on all gait feature data within the individual twin static portrait template, calculate the comprehensive gait feature parameters of the twin group static portrait cluster centers under the cluster category. The calculation formula is as follows:

[0262]

[0263] Among them, c i Let be the i-th data of the j-th gait feature parameter, and N be the number of gait feature data; each gait feature has a gait feature comprehensive parameter.

[0264] (4) Calculate the distance based on the gait feature reference of the generated crowd profile and the actual gait feature parameters corresponding to the static information, and perform error correction:

[0265]

[0266] Where, Δ i，j，k The prediction error of cluster centers between individual twin static portraits and individual twin static portrait templates;

[0267] Loss function calculation, deep learning error correction algorithm model training, and early stopping mechanism design:

[0268] (1) Input the prediction error into the deep learning error correction algorithm model to obtain the corrected error prediction value;

[0269] Δ pred =F(x, Δ)

[0270] Where x represents the static information of the individual twin static portrait, F(·) is the deep learning error correction algorithm, and Δ pred To correct the error prediction value;

[0271] (2) Calculate the loss value based on the corrected error prediction value and the actual error value output by the model:

[0272]

[0273] in, Δ represents the k-th gait feature parameter in the corrected error prediction value of the subject. k This represents the corresponding true gait feature error, where K is the number of clusters, N is the number of gait feature parameters, and L is the final loss function.

[0274] (3) Determine whether the current calculated loss value has decreased compared with the average loss value of the previous m times: If the loss value has decreased, it indicates that the model is still converging. Update the parameters of the deep learning error correction model by gradient descent. In order to avoid the model training time being too long, set the maximum training rounds to n, that is, stop training after reaching the maximum training rounds. Here, m and n are positive integers and are set by the user.

[0275] Taking the example: If m is set to 10 and n to 1000, and the loss value no longer decreases after 10 calculations, it indicates that the model can no longer converge, and model updates are stopped.

[0276]

[0277] Among them, L t This is the loss value calculated during the current training. The average of the loss values ​​for the first 10 iterations is denoted by 'count', which is the consecutive count of the loss value that no longer decreases during training, or the stop condition that is triggered when the loss value reaches 10.

[0278] For example, the specific steps for obtaining the individual twin static portrait and the individual twin dynamic and static portrait are as follows:

[0279] Based on the K-Means gait profile clustering model and cluster boundaries, the test subjects were classified and individual twin static profiles were calculated.

[0280] (1) Given the static information data of the test subject (gender, age, height, BMI);

[0281] (2) Based on the cluster boundaries, the testers are divided into individual twin static profiles;

[0282] Given the dynamic profile features of the test taker (walking, running, jumping, etc.), the test taker is divided into dynamic profile cluster categories based on the cluster boundaries;

[0283] (1) Based on the cluster boundaries, the testers are divided into a dynamic population profile cluster category;

[0284] (2) Calculate the mean of the gait characteristic parameters of this group of people in the 80%-99% confidence interval, and use the calculation results as individual twin static and dynamic portraits of the gait cluster category;

[0285] The dynamic and static information and the individual twin static portrait are input into the deep learning error correction algorithm to obtain the correction error prediction value, and then the gait features of the individual twin dynamic and static portrait after the first correction based on the static information are calculated.

[0286] C corrected =C initial +Δ pred

[0287] Among them, c initial To create dynamic and static portraits of individuals within a population, Δ pred To correct the error prediction value, c corrected Gait characteristics for creating dynamic and static portraits of individual twins;

[0288] The correlation between the subject's gait characteristics and dynamic and static information is calculated, and the gait parameters are sorted accordingly to obtain the real-time dynamic feature parameters corresponding to the individual's relevant static feature profile scenario.

[0289] Level 4 (Mobile): Health risk score based on boundaries and center. This involves determining the distribution of gait characteristic parameters of the subject (patient), differentiating between three scenarios, and deriving a gait health score, including the following steps:

[0290] Step 1432: Determine which range the subject is located in.

[0291] Step 1433: Generate an inverse linear score function based on the gait feature center value and boundary feature value of healthy persons and / or patients.

[0292] Specifically, based on the judgment result of step 1432, the corresponding steps are executed. If the subject is determined to be a healthy individual, step 1433a is executed, which involves generating an inverse linear score function based on the gait feature center value and boundary value of a healthy person. If the subject is determined to be from an overlapping region, step 1433b is executed, which involves generating an inverse linear score function based on the gait feature center value and boundary value of both a healthy person and a patient. If the subject is determined to be a patient, step 1433c is executed, which involves generating a positive linear score function based on the gait feature center value and boundary value of a patient.

[0293] Step 1434: Input the subject's gait parameters into the scoring function.

[0294] Step 1435: Obtain the subject's risk and / or severity score.

[0295] Specifically, based on the aforementioned judgment results, the corresponding steps are executed. If the subject is determined to be a healthy individual, step 1435a is executed, which involves obtaining the subject's risk level score; if the subject is determined to be from an overlapping region, step 1435b is executed, which involves obtaining the subject's risk level score and severity score; if the subject is determined to be a patient, step 1435c is executed, which involves obtaining the subject's severity score.

[0296] Step 1436: 100, 60, and 0 represent the health cluster center, the boundary between health and patients, the patient cluster center, and the score conversion, respectively.

[0297] The specific description of the fourth level is as follows:

[0298] First, patient data needs to be loaded and the center and boundary values ​​of the gait feature parameters for healthy individuals in each cluster need to be calculated:

[0299] (1) Load the patient dataset, which includes the patients' age data and gait characteristic parameter data;

[0300] (2) Based on the age data in the patient dataset and the age boundary of the healthy population cluster, extract the patient data in the same cluster category as the subjects;

[0301] (3) Referring to the calculation formulas for the mean, standard error and 80%-99% confidence interval of gait characteristics in (1), calculate the relevant indicators of the gait characteristic parameter data of patients in this category.

[0302] Secondly, the distribution of the subjects' gait characteristic parameters was determined, and different processing algorithms were used to calculate the gait health score. The distribution of the subjects' gait characteristic parameters among healthy individuals and patients was determined, and was divided into the following three cases.

[0303] Scenario 1: If the subject's gait characteristic parameters fall within the range of healthy individuals' gait characteristic values, then calculate the subject's risk score in healthy gait:

[0304] First, a linear score function is generated based on the center and boundary values ​​of the gait characteristics of healthy individuals:

[0305]

[0306] Where Y represents the gait characteristic data of the subjects. μ represents the maximum value within the range of gait characteristics of healthy individuals. H This represents the center value of the numerical range of gait characteristics in healthy individuals.

[0307] The risk level score of the subject is obtained by substituting the subject's gait parameters into the scoring function;

[0308] Scenario 2: If the subject's gait characteristic parameters fall within the range of the patient's gait characteristic values, then calculate the subject's risk score in the patient's gait:

[0309] First, a linear score function is generated based on the patient's gait feature center and boundary values:

[0310]

[0311] Where Y represents the gait characteristic data of the subjects. μ represents the maximum value within the range of patient gait characteristics. D The center value of the numerical range of the patient's gait characteristics;

[0312] The severity score of the subject is obtained by substituting the subject's gait parameters into the scoring function;

[0313] Scenario 3: When the subject's gait characteristic parameters fall at the intersection of the gait characteristic value ranges of healthy individuals and patients, both risk score and severity score are calculated simultaneously to achieve a more comprehensive assessment of the test subject's gait health status.

[0314] Fifth level (mobile phone): (Triggering feedback mechanism, individual twin model, health index boundary feedback to the cloud, iterative optimization of cloud group twin database, including step eighteen)

[0315] Step 1437: New data is added to the database.

[0316] Step 1438: Increase the amount of new data by 1.

[0317] Step 1439: Determine if the amount of new data has reached 200.

[0318] If the number exceeds 200, proceed to step 1440; otherwise, end the current process.

[0319] Step 1440: Retrain the model.

[0320] The tester's data is stored, and a model update trigger mechanism is set up to save the newly added data to the database and count the newly added data. When the number of data reaches a threshold, the model update mechanism is triggered, which can ensure the real-time optimization of the model and continuously reduce the prediction error and the actual error.

[0321] (1) During the test, record the dynamic and static information data and the actual gait characteristic data of the subjects and save them to the database, and record the amount of data of the newly added dynamic and static individual twin portraits.

[0322] (2) When the newly added data reaches x, the model update mechanism is activated, and the crowd gait profile clustering model and the deep learning-based error correction algorithm model are retrained using all existing datasets to optimize the algorithm. As the amount of data increases, the crowd gait profile is refined through iteration, and the prediction error of gait features is continuously reduced. Here, x is a positive integer and can be defined by the user. This application does not make a specific limitation on this. For example, x is 200. In some implementations, the training model can be re-updated after the amount of data of individual dynamic and static profile categories exceeds 200.

[0323] If the overall judgment result indicates a disease population, the model is retrained and the first parameter threshold is adjusted to be equal to the second parameter threshold. The steps from the second to the fifth level are repeated.

[0324] The technical solution of this application provides a method for medical patients and researchers to dynamically present a dynamic user profile of vital signs by combining a wearable system, which involves hierarchical screening of a large number of static and dynamic vital sign parameters and the range of each parameter. Specifically, it provides an end-to-end overall process method for hierarchical screening of a large number of static and dynamic vital sign parameters for dynamically presenting a user profile of vital signs based on a wearable system, with each parameter determining the health range of three population groups. The method involves screening from a massive number of indicators, including a dynamic screening step for massive vital sign parameters used in the profile depiction. This includes static parameter screening and internal dynamic parameter screening with different levels of thresholds. The initial threshold is statically lenient and dynamically strict; if the final score indicates a disease, it will be adjusted to a uniformly strict threshold and iterated again. The threshold setting is beneficial for the dual judgment of training process iteration effect and resource consumption rounds, achieving a balance between scarce population samples to avoid omissions and the wasteful load of massive dynamic vital signs of the population. The system employs a dual-judgment approach, considering both training iteration effectiveness and resource consumption rounds, to balance the scarcity of population samples and avoid omissions, while also mitigating the wasteful burden of handling massive dynamic vital signs. The ranges of the three indicators are unified into a single calculation method, influencing the initial indicator screening thresholds. Specifically, the process of determining the range of indicator parameters used for profile creation after screening includes merging different calculations of characteristic parameters for healthy individuals, diseased individuals, and individuals with both health and disease conditions into a unified, integrated assessment range, taking into account the business characteristics related to disease risk. If the final score indicates a disease condition, both thresholds in the initial screening steps are adjusted to strictly enforced thresholds, and the process iterates again.

[0325] Figure 15 is a schematic diagram of the structure of the dynamic clustering device for crowd profiling provided in an embodiment of this application, applied to a cloud server. As shown in Figure 15, the dynamic clustering device for crowd profiling includes:

[0326] The first filtering unit 1501 is used to filter the first static feature parameter, the first dynamic feature parameter and the first gait feature parameter from the first dataset. The first dataset includes multiple static feature parameters, multiple dynamic feature parameters and multiple gait feature parameters of the population.

[0327] The first determining unit 1502 is used to determine the crowd profile clustering result based on the first static feature parameter, the first dynamic feature parameter, and the first step dynamic feature parameter; and to determine the first parameter set of the crowd profile clustering result;

[0328] The first sending unit 1503 is used to send the crowd profile clustering results and the first parameter set to the user equipment. The crowd profile clustering results and the first parameter set are used by the user equipment to determine the dynamic and static profiles of individuals.

[0329] In some embodiments, the apparatus further includes: a first receiving unit 1504; the first receiving unit 1504 is configured to receive a second dataset, individual dynamic and static portraits, and health information of the subject sent by a user device; the apparatus further includes: a first updating unit 1505; the first updating unit 1505 is configured to update the population portrait clustering results based on the second dataset, individual dynamic and static portraits, and health information of the subject.

[0330] In some embodiments, the first gait feature parameter includes a second gait feature parameter and a third gait feature parameter; the first filtering unit 1501 is used to filter a first static feature parameter and a second gait feature parameter from multiple static feature parameters and multiple gait feature parameters based on the correlation between multiple static feature parameters and multiple gait feature parameters; and to filter a first dynamic feature parameter and a third gait feature parameter from multiple dynamic feature parameters and multiple gait feature parameters based on the correlation between the first static feature parameter, multiple dynamic feature parameters and multiple gait feature parameters.

[0331] In some embodiments, the first screening unit 1501 is configured to determine a first set of correlation coefficients for a plurality of static feature parameters and a plurality of gait feature parameters; screen a first static feature parameter from the plurality of static feature parameters according to the first set of correlation coefficients; and screen a second gait feature parameter from the plurality of gait feature parameters according to the first static parameter and the first set of correlation coefficients.

[0332] In some implementations, the first filtering unit 1501 is used to obtain a second set of correlation coefficients related to the first static parameter, the first set of correlation coefficients including the second set of correlation coefficients; and to filter the second gait feature parameters from multiple gait feature parameters according to the second set of correlation coefficients and a first threshold.

[0333] In some embodiments, the first screening unit 1501 is used to determine a third set of correlation coefficients for a plurality of dynamic feature parameters and a plurality of gait feature parameters under a first static parameter; to screen a first dynamic feature parameter from the plurality of dynamic feature parameters according to the third correlation coefficient set; and to screen a second gait feature parameter from the plurality of gait feature parameters according to the first dynamic feature parameter and the third correlation coefficient set.

[0334] In some implementations, the first filtering unit 1501 is used to obtain a fourth set of correlation coefficients related to the first dynamic feature parameter, the third set of correlation coefficients includes the fourth set of correlation coefficients; and to filter the third gait feature parameter from multiple gait feature parameters according to the fourth set of correlation coefficients and a second threshold.

[0335] Those skilled in the art should understand that the functions of each unit in the dynamic crowd profiling clustering device shown in Figure 15 can be understood with reference to the relevant description of the aforementioned method. The functions of each unit in the dynamic crowd profiling clustering device shown in Figure 15 can be implemented by a program running on a processor, or by specific logic circuits.

[0336] Figure 16 is a schematic diagram of the structure of the dynamic clustering device for crowd profiling provided in an embodiment of this application, applied to a user equipment. As shown in Figure 16, the dynamic clustering device for crowd profiling includes:

[0337] The first acquisition unit 1601 is used to acquire the user's first static feature parameter data, first dynamic feature parameter data and first step state feature parameter data;

[0338] The second determining unit 1602 is used to determine an individual static portrait template based on the first static feature parameter data; and to determine an individual dynamic and static portrait based on the individual static portrait template, the first dynamic feature parameter data, and the first step dynamic feature parameter data.

[0339] In some implementations, the first acquisition unit 1601 is used to acquire the crowd profile clustering result sent by the cloud server and the first parameter set corresponding to the crowd profile clustering result before determining the individual static profile template based on the first static feature parameter data; the second determination unit 1602 is used to determine the individual static profile template based on the first static feature parameter data, the crowd profile clustering result and the first parameter set.

[0340] In some embodiments, the second determining unit 1602 is used to determine an error value based on the individual static portrait template and the first step state feature parameter data; determine a corrected error prediction value based on the error value; determine an individual dynamic and static portrait template based on the first dynamic feature parameter data and the first step state feature parameter data; and determine an individual dynamic and static portrait based on the individual dynamic and static portrait template and the corrected error prediction value.

[0341] In some embodiments, the second determining unit 1602 is used to determine a loss value based on the corrected error prediction value and the error value after determining the corrected error prediction value based on the error value; the device further includes: a first judging unit 1603; the first judging unit 1603 is used to judge whether the first model stops training based on the loss value, the first model representing the model that has obtained the corrected error prediction value.

[0342] In some implementations, the first judgment unit 1603 is used to obtain the loss value of the first model during the Nth training iteration, where N is a positive integer; obtain the loss value of the first model during the previous M training iterations, where N = M + 1, and M is a positive integer; and determine whether the first model should stop training based on the loss value of the Nth iteration and the loss values ​​of the previous M iterations.

[0343] In some embodiments, the first acquisition unit 1601 is used to acquire the amount of data of individual dynamic and static portraits; the device further includes: a second update unit 1604; the second update unit 1604 is used to update the first model according to the dataset corresponding to the individual dynamic and static portraits when the amount of data is greater than a third threshold.

[0344] Those skilled in the art should understand that the functions of each unit in the dynamic crowd profiling clustering device shown in Figure 16 can be understood with reference to the relevant description of the aforementioned method. The functions of each unit in the dynamic crowd profiling clustering device shown in Figure 16 can be implemented by a program running on a processor, or by specific logic circuits.

[0345] Figure 17 is a schematic diagram of the structure of the dynamic clustering device for crowd profiling provided in this embodiment of the application, which is applied to user equipment. As shown in Figure 17, the dynamic clustering device for crowd profiling includes:

[0346] The second acquisition unit 1701 is used to acquire the subject's second dataset, which includes the subject's age data, the subject's dynamic characteristic parameter data, and the subject's gait characteristic parameter data.

[0347] The third determining unit 1702 is used to determine the first cluster category and the second cluster category based on age data, and to obtain the second parameter set of the first cluster category and the second cluster category; and to determine the subject's first score based on the subject's gait characteristic parameter data and the second parameter set.

[0348] In some implementations, the second acquisition unit 1701 is used to acquire a third dataset before acquiring the second dataset of the subject, the third dataset including the patient's age data and the patient's second gait characteristic parameter data; acquire the population profile clustering results sent by the cloud server; divide the first data according to the population profile clustering results and the age data to obtain the division result, the first data including the age data in the third dataset and the age data in the population profile clustering results.

[0349] In some implementations, the third determining unit 1702 is used to determine a first cluster category based on the subject's age data and the segmentation results; and to determine a second cluster category based on the subject's age data and the population profile clustering results.

[0350] In some embodiments, the device further includes: a second judgment unit 1703; the second judgment unit 1703 is used to determine the subject's health status information based on the first score after determining the subject's first score based on the subject's gait characteristic parameter data and the second parameter set; the device further includes: a second sending unit 1704; the second sending unit 1704 is used to send the second dataset, the first score and the health status information to a cloud server.

[0351] In some embodiments, the second acquisition unit 1701 is used to acquire the amount of data of individual dynamic and static portraits; the device further includes: a third update unit 1705; the third update unit 1705 is used to update the first model according to the dataset corresponding to the individual dynamic and static portraits when the amount of data is greater than a third threshold, the first model representing the model for acquiring the predicted value of the correction error.

[0352] Those skilled in the art should understand that the functions of each unit in the dynamic crowd profiling clustering device shown in Figure 17 can be understood with reference to the relevant description of the aforementioned method. The functions of each unit in the dynamic crowd profiling clustering device shown in Figure 17 can be implemented by a program running on a processor, or by specific logic circuits.

[0353] This application provides a cloud server. Figure 18 is a schematic diagram of the structure of a cloud server provided in this application embodiment. As shown in Figure 18, in this embodiment, the cloud server includes: a first processor 1801, a first memory 1802, and a first communication bus 1803;

[0354] The first communication bus 1803 is used to realize the communication connection between the first processor and the first memory;

[0355] The first processor 1801 is used to execute one or more computer programs stored in the first memory 1802 to implement the above-mentioned dynamic clustering processing method for crowd profiles applied to a cloud server.

[0356] Figure 19 is a schematic structural diagram of a user equipment 1900 provided in an embodiment of this application. The user equipment 1900 shown in Figure 19 includes a processor 1910, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0357] Optionally, as shown in FIG19, the user equipment 1900 may further include a memory 1920. The processor 1910 may retrieve and run computer programs from the memory 1920 to implement the methods described in the embodiments of this application.

[0358] The memory 1920 can be a separate device independent of the processor 1910, or it can be integrated into the processor 1910.

[0359] Optionally, as shown in FIG19, the user equipment 1900 may further include a transceiver 1930, and the processor 1910 may control the transceiver 1930 to communicate with other devices. Specifically, it may send information or data to other devices or receive information or data sent by other devices.

[0360] The transceiver 1930 may include a transmitter and a receiver. The transceiver 1930 may further include an antenna, and the number of antennas may be one or more.

[0361] The user equipment 1900 can implement the corresponding processes of the crowd profiling dynamic clustering device applied to the user equipment in the various methods of the embodiments of this application, which will not be described in detail here for the sake of brevity.

[0362] Figure 20 is a schematic structural diagram of a chip according to an embodiment of this application. The chip 2000 shown in Figure 20 includes a processor 2010, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0363] Optionally, as shown in FIG20, chip 2000 may further include memory 2020. Processor 2010 can call and run computer programs from memory 2020 to implement the methods in the embodiments of this application.

[0364] The memory 2020 can be a separate device independent of the processor 2010, or it can be integrated into the processor 2010.

[0365] Optionally, the chip 2000 may also include an input interface 2030. The processor 2010 can control the input interface 2030 to communicate with other devices or chips; specifically, it can acquire information or data sent by other devices or chips.

[0366] Optionally, the chip 2000 may also include an output interface 2040. The processor 2010 can control the output interface 2040 to communicate with other devices or chips, specifically, to output information or data to other devices or chips.

[0367] Optionally, the chip can be applied to the cloud server in the embodiments of this application, and the chip can implement the corresponding processes implemented by the cloud server in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0368] Optionally, the chip can be applied to the user equipment in the embodiments of this application, and the chip can implement the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0369] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0370] It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0371] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0372] It should be understood that the above-described memory is exemplary and not a limiting description. For example, the memory in the embodiments of this application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM), etc. That is to say, the memory in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

[0373] This application also provides a computer program product, including a computer program.

[0374] Optionally, the computer program product can be applied to the cloud server in the embodiments of this application, and when the computer program is executed by the processor, it implements the corresponding processes implemented by the cloud server in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0375] Optionally, the computer program product can be applied to the user equipment in the embodiments of this application, and when the computer program is executed by the processor, it implements the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0376] This application also provides a computer-readable storage medium for storing computer programs.

[0377] Optionally, the computer-readable storage medium can be applied to the cloud server in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the cloud server in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0378] Optionally, the computer-readable storage medium can be applied to the user equipment in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0379] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0380] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0381] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0382] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0383] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0384] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0385] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dynamic clustering method for user profiles, characterized in that, Applied to a cloud server, the method includes: filtering a first static feature parameter, a first dynamic feature parameter, and a first gait feature parameter from a first dataset, wherein the first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of a population; determining a population profile clustering result based on the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter; determining a first parameter set of the population profile clustering result; and sending the population profile clustering result and the first parameter set to a user device, wherein the population profile clustering result and the first parameter set are used by the user device to determine an individual's static and dynamic profile.

2. The method according to claim 1, characterized in that, The method further includes: receiving a second dataset, individual dynamic and static portraits, and health information of the subjects sent by a user device; and updating the population portrait clustering results based on the second dataset, the individual dynamic and static portraits, and the health information of the subjects.

3. The method according to claim 1, characterized in that, The first gait feature parameter includes a second gait feature parameter and a third gait feature parameter; the step of filtering the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter from the first dataset includes: filtering the first static feature parameter and the second gait feature parameter from the plurality of static feature parameters and the plurality of gait feature parameters based on the correlation between the plurality of static feature parameters and the plurality of gait feature parameters; and filtering the first dynamic feature parameter and the third gait feature parameter from the plurality of dynamic feature parameters and the plurality of gait feature parameters based on the correlation between the first static feature parameter, the plurality of dynamic feature parameters, and the plurality of gait feature parameters.

4. The method according to claim 3, characterized in that, The step of selecting a first static feature parameter and a second gait feature parameter from the plurality of static feature parameters and the plurality of gait feature parameters based on the correlation between the plurality of static feature parameters and the plurality of gait feature parameters includes: determining a first set of correlation coefficients for the plurality of static feature parameters and the plurality of gait feature parameters; selecting a first static feature parameter from the plurality of static feature parameters based on the first set of correlation coefficients; and selecting a second gait feature parameter from the plurality of gait feature parameters based on the first static parameter and the first set of correlation coefficients.

5. The method according to claim 4, characterized in that, The step of filtering a second gait feature parameter from the plurality of gait feature parameters based on the first static parameter and the first correlation coefficient set includes: obtaining a second correlation coefficient set related to the first static parameter, wherein the first correlation coefficient set includes the second correlation coefficient set; and filtering a second gait feature parameter from the plurality of gait feature parameters based on the second correlation coefficient set and a first threshold.

6. The method according to claim 3, characterized in that, The step of selecting a first dynamic feature parameter and a third gait feature parameter from the plurality of dynamic feature parameters and the plurality of gait feature parameters based on the correlation between the first static feature parameter, the plurality of dynamic feature parameters, and the plurality of gait feature parameters includes: determining a third set of correlation coefficients for the plurality of dynamic feature parameters and the plurality of gait feature parameters under the first static parameter; selecting a first dynamic feature parameter from the plurality of dynamic feature parameters based on the third set of correlation coefficients; and selecting a second gait feature parameter from the plurality of gait feature parameters based on the first dynamic feature parameter and the third set of correlation coefficients.

7. The method according to claim 6, characterized in that, The step of filtering a second gait feature parameter from the plurality of gait feature parameters based on the first dynamic feature parameter and the third correlation coefficient set includes: obtaining a fourth correlation coefficient set related to the first dynamic feature parameter, wherein the third correlation coefficient set includes the fourth correlation coefficient set; and filtering a third gait feature parameter from the plurality of gait feature parameters based on the fourth correlation coefficient set and a second threshold.

8. A dynamic clustering method for user profiles, characterized in that, Applied to user equipment, the method includes: acquiring first static feature parameter data, first dynamic feature parameter data, and first step-state feature parameter data of a user; determining an individual static profile template based on the first static feature parameter data; and determining an individual dynamic and static profile based on the individual static profile template, the first dynamic feature parameter data, and the first step-state feature parameter data.

9. The method according to claim 8, characterized in that, Before determining the individual static profile template based on the first static feature parameter data, the method further includes: obtaining the crowd profile clustering result sent by the cloud server and the first parameter set corresponding to the crowd profile clustering result; determining the individual static profile template based on the first static feature parameter data includes: determining the individual static profile template based on the first static feature parameter data, the crowd profile clustering result and the first parameter set.

10. The method according to claim 9, characterized in that, The step of determining an individual dynamic and static portrait based on the individual static portrait template, the first dynamic feature parameter data, and the first step state feature parameter data includes: determining an error value based on the individual static portrait template and the first step state feature parameter data; determining a corrected error prediction value based on the error value; determining an individual dynamic and static portrait template based on the first dynamic feature parameter data and the first step state feature parameter data; and determining the individual dynamic and static portrait based on the individual dynamic and static portrait template and the corrected error prediction value.

11. The method according to claim 10, characterized in that, After determining the corrected error prediction value based on the error value, the method further includes: determining a loss value based on the corrected error prediction value and the error value; and determining whether the first model should stop training based on the loss value, wherein the first model represents the model that obtained the corrected error prediction value.

12. The method according to claim 11, characterized in that, The step of determining whether the first model should stop training based on the loss value includes: obtaining the loss value of the first model during the Nth training iteration, where N is a positive integer; obtaining the loss value of the first model during the previous M training iterations, where N = M + 1, and M is a positive integer; and determining whether the first model should stop training based on the loss value of the Nth training iteration and the loss values ​​of the previous M training iterations.

13. The method according to claim 11, characterized in that, The method further includes: acquiring the amount of data for individual dynamic and static portraits; when the amount of data is greater than a third threshold, updating the first model based on the dataset corresponding to the individual dynamic and static portraits.

14. A dynamic clustering method for user profiles, characterized in that, Applied to a user device, the method includes: acquiring a second dataset of a subject, the second dataset including the subject's age data, the subject's dynamic characteristic parameter data, and the subject's gait characteristic parameter data; determining a first cluster category and a second cluster category based on the age data, and acquiring a second parameter set for the first cluster category and the second cluster category; and determining a first score for the subject based on the subject's gait characteristic parameter data and the second parameter set.

15. The method according to claim 14, characterized in that, Before obtaining the second dataset of the subjects, the method includes: obtaining a third dataset, which includes the patients' age data and the patients' second gait characteristic parameter data; obtaining the population profile clustering results sent by the cloud server; dividing the first data according to the population profile clustering results and the age data to obtain a division result, wherein the first data includes the age data in the third dataset and the age data in the population profile clustering results.

16. The method according to claim 15, characterized in that, The step of determining the first cluster category and the second cluster category based on the age data includes: determining the first cluster category based on the age data of the subjects and the segmentation results; and determining the second cluster category based on the age data of the subjects and the population profile clustering results.

17. The method according to claim 14, characterized in that, After determining the subject's first score based on the subject's gait characteristic parameter data and the second parameter set, the method further includes: determining the subject's health status information based on the first score; and sending the second dataset, the first score, and the health status information to a cloud server.

18. The method according to claim 17, characterized in that, The method further includes: acquiring the amount of data for individual dynamic and static portraits; when the amount of data is greater than a third threshold, updating the first model based on the dataset corresponding to the individual dynamic and static portraits, wherein the first model represents the model for acquiring the predicted value of the correction error.

19. A dynamic clustering device for crowd profiling, characterized in that, The device, applied to a cloud server, includes: a first filtering unit for filtering a first static feature parameter, a first dynamic feature parameter, and a first gait feature parameter from a first dataset, wherein the first dataset includes multiple static feature parameters, multiple dynamic feature parameters, and multiple gait feature parameters of a population; a first determining unit for determining a population profile clustering result based on the first static feature parameter, the first dynamic feature parameter, and the first gait feature parameter; and determining a first parameter set of the population profile clustering result; and a first sending unit for sending the population profile clustering result and the first parameter set to a user device, wherein the population profile clustering result and the first parameter set are used by the user device to determine an individual's static and dynamic profile.

20. A dynamic clustering device for crowd profiling, characterized in that, Applied to user equipment, the device includes: a first acquisition unit, configured to acquire first static feature parameter data, first dynamic feature parameter data, and first step-state feature parameter data of a user; and a second determination unit, configured to determine an individual static profile template based on the first static feature parameter data; and to determine an individual dynamic and static profile based on the individual static profile template, the first dynamic feature parameter data, and the first step-state feature parameter data.

21. A dynamic clustering device for crowd profiling, characterized in that, The device, applied to a user equipment, includes: a second acquisition unit for acquiring a second dataset of a subject, the second dataset including the subject's age data, the subject's dynamic characteristic parameter data, and the subject's gait characteristic parameter data; a third determination unit for determining a first cluster category and a second cluster category based on the age data, and acquiring a second parameter set for the first cluster category and the second cluster category; and determining a first score of the subject based on the subject's gait characteristic parameter data and the second parameter set.

22. A cloud server, characterized in that, include: A first processor, a first memory, and a first communication bus; The first communication bus is used to establish a communication connection between the first processor and the first memory; The first processor is configured to execute one or more computer programs stored in the first memory to implement the dynamic clustering method for crowd profiling as described in any one of claims 1 to 7.

23. A user equipment, characterized in that, include: A processor and a memory for storing a computer program, the processor for calling and running the computer program stored in the memory to perform the method as claimed in any one of claims 8 to 13 or 14 to 18.

24. A computer program product, characterized in that, include: A computer program that, when executed by a processor, implements the method according to any one of claims 1 to 18.

25. A computer-readable storage medium, characterized in that, Used to store a computer program that causes a computer to perform the method as described in any one of claims 1 to 18.

Images