Data evaluation method and device, equipment, medium and program product

Abstract

The invention provides a data evaluation method and device, equipment, a medium and a program product. The method comprises the steps of obtaining static information, motion information and user requirements of a to-be-tested user and sensor data of a worn acquisition sensor; according to the static information and the motion information, carrying out portrait matching by utilizing a group portrait model to obtain a user portrait corresponding to the to-be-tested user; according to a motion normal form corresponding to the user demand, obtaining motion data models of a plurality of testers in the user portrait under the motion normal form; under the condition that the sensor data meets a first preset condition, determining that the to-be-tested user can reuse data in the motion data model, the first preset condition comprises that the data characteristics of the sensor data and the motion data model are consistent and / or the spatial position indicated by the sensor data meets the requirement and / or the time corresponding to the sensor data acquired by the plurality of acquisition sensors is consistent. According to the embodiment of the invention, data evaluation can be efficiently and accurately carried out, and model multiplexing migration is realized.

Description

Technical Field

[0001] This invention relates to the field of data service technology, and in particular to a data evaluation method, apparatus, equipment, medium, and program product. Background Technology

[0002] In existing technologies, wearable devices reuse the experience of different limb positions and different types of sensors to contribute weights to behavior recognition, which helps users to select different combinations of sensors for different limb positions as needed, reducing the number and dimensionality of sensors.

[0003] However, whether the detection results of the same limb position can be directly reused from the experimental results of the existing model depends on whether the data generated by the user's worn device is consistent with the data of the existing model. If they are consistent, the requirements for model reuse and migration can be met. However, the existing technology cannot accurately and efficiently assess data consistency, and therefore cannot achieve model reuse and migration. Summary of the Invention

[0004] The purpose of this invention is to provide a data evaluation method, apparatus, device, medium, and program product to address the shortcomings of existing technologies in accurately and efficiently evaluating data consistency, which in turn prevents model reuse and migration.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a data evaluation method, comprising:

[0006] Acquire static information, motion information, user needs, and sensor data from the sensors worn by the user under test;

[0007] Based on the static information and the motion information, a group profile model is used to perform profile matching to obtain a user profile corresponding to the user to be tested. The group profile model is pre-constructed based on the test information of multiple testers.

[0008] Based on the motion paradigm corresponding to the motion information, obtain motion data models of multiple testers in the user profile under the motion paradigm;

[0009] If the sensor data meets the first preset condition, it is determined that the user under test can reuse the data in the motion data model. The first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent; the spatial position indicated by the sensor data meets the requirements; and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0010] Optionally, based on the motion paradigm corresponding to the user's needs, obtain motion data models of multiple testers in the user profile under the motion paradigm, including:

[0011] Based on the user's needs, a movement paradigm is matched, and if the match is successful, the movement paradigm is obtained.

[0012] Based on the exercise paradigm, obtain the exercise test data of multiple testers in the user profile under the exercise paradigm, the types of exercise in the exercise paradigm, and the mean static correlation coefficient of the corresponding multiple testers;

[0013] Based on the exercise test data, the type of exercise, and the mean of the static correlation coefficient, intraclass correlation calculation is performed to obtain the first intraclass correlation coefficient of the exercise paradigm.

[0014] If the correlation coefficient within the first category is greater than the first threshold, obtain the motion data models of multiple testers in the user profile under the motion paradigm.

[0015] Optionally, the method further includes:

[0016] The sensor data is input into the converter model for feature encoding to obtain the first feature code output by the embedding layer of the converter model;

[0017] Obtain the second feature code corresponding to the data in the motion data model;

[0018] If the first feature code and the second feature code are consistent, it is determined that the sensor data satisfies the data features of the sensor data and the motion data model in the first preset condition.

[0019] Optionally, the method further includes:

[0020] When the user under test moves in the test scenario, the first motion data collected by the optical capture device worn by the user under test and the second motion data collected by the acquisition sensor worn by the user under test are acquired in real time. The first motion data includes the first spatial information of the user under test at each time step.

[0021] The second motion data is input into the converter model and feature-encoded to obtain a third feature code corresponding to the output of the converter model; and the second motion data is input into the converter model and position-encoded to obtain a first position code corresponding to the output of the converter model.

[0022] The location encoding and the third feature encoding are fused to obtain a first location vector, which includes the second spatial information of the user under test at each time step.

[0023] The first spatial information is compared with the second spatial information. If the spatial positions indicated by the first spatial information and the second spatial information are consistent, it is determined that the spatial position indicated by the sensor data meets the requirements of the first preset condition.

[0024] Optionally, the method further includes:

[0025] When the user under test moves in the test scenario according to the preset movement mode, the third movement data collected by the optical capture device worn by the user under test and the fourth movement data collected by the acquisition sensor worn by the user under test are acquired in real time. The preset movement mode includes repeating the preset movement a first preset number of times.

[0026] The fourth motion data is input into the converter model and feature-encoded to obtain the fourth feature code corresponding to the output of the converter model; the fourth motion data is input into the converter model and position-encoded to obtain the second position code corresponding to the output of the converter model; and the fourth motion data is input into the converter model and time-encoded to obtain the time code corresponding to the output of the converter model.

[0027] The time code, the second position code, and the fourth feature code are fused to obtain a time vector, which includes data collected by the multiple acquisition sensors at each time step.

[0028] Based on the number of times the preset motion is repeated and the time vector, intra-class correlation calculation is performed to obtain the second intra-class correlation coefficient corresponding to the time vector;

[0029] If the correlation coefficient within the second category is greater than the second threshold, it is determined that the sensor data satisfies the first preset condition that the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0030] Optionally, the method further includes:

[0031] Acquire test information from multiple testers, including static test information, motion test information, and sensor test data corresponding to different motion test information from the acquisition sensors worn by the testers;

[0032] Based on digital twin technology, the sensor test data are subjected to correlation screening and cluster evaluation according to the static test information and motion test information, respectively, to construct the group profile model.

[0033] This invention also provides a data evaluation apparatus, comprising:

[0034] The first acquisition module is used to acquire static information, motion information, user needs, and sensor data of the acquisition sensors worn by the user under test.

[0035] The first matching module is used to perform profile matching using a group profile model based on the static information and the motion information to obtain a user profile corresponding to the user to be tested, wherein the group profile model is pre-constructed based on the test information of multiple testers;

[0036] The second acquisition module is used to acquire the motion data models of multiple testers in the user profile under the motion paradigm according to the motion paradigm corresponding to the user needs.

[0037] The first determining module is used to determine, when the sensor data meets a first preset condition, that the user to be tested can reuse the data in the motion data model, wherein the first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent; the spatial position indicated by the sensor data meets the requirements; and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0038] This invention also provides a network device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the data evaluation method as described in any of the preceding embodiments.

[0039] This invention also provides a readable storage medium, comprising: a program stored on the readable storage medium, wherein the program, when executed by a processor, implements the steps of the data evaluation method as described in any of the preceding claims.

[0040] This invention also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the data evaluation method as described in any of the preceding embodiments.

[0041] At least one of the above technical solutions of the present invention has the following beneficial effects:

[0042] The above scheme provides a data evaluation method. To evaluate the consistency between the sensor data of the user under test and the data of the model that the user needs to reuse and transfer, firstly, based on the static and motion information of the user under test, a user profile corresponding to the user under test is obtained by using a group profile model for profile matching, i.e., confirming the group profile that is consistent with the user's basic information (static and motion information); then, based on the user's needs, the motion paradigm that the user under test needs to detect is determined, and the motion data models of multiple testers in the user profile under this motion paradigm are obtained, i.e., the motion data model that the user under test needs to reuse and transfer is determined when the motion paradigm is consistent; finally, if the sensor data meets a first preset condition, i.e., the sensor data and the motion data model have consistent data characteristics, the spatial location indicated by the sensor data meets the requirements, and the time corresponding to the sensor data collected by multiple acquisition sensors is consistent, it is confirmed whether the user under test can reuse the motion data model data. This embodiment of the invention evaluates the data of the user under test from multiple aspects such as group, motion paradigm, data characteristics, space, and time, which can efficiently and accurately evaluate whether the sensor data and the motion data model data are consistent, thereby realizing the reuse and transfer of the motion data model. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the data evaluation method according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the motion paradigm of an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of the data evaluation device according to an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0049] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), and other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used not only in the systems and radio technologies mentioned above, but also in other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and NR terminology is used in most of the following description; however, these technologies can also be applied to applications beyond NR systems, such as 6th Generation (6G) communication systems.

[0050] like Figure 1 As shown, an embodiment of the present invention provides a data evaluation method, including:

[0051] Step S101: Obtain the static information, motion information, user needs, and sensor data of the acquisition sensors worn by the user to be tested;

[0052] In step S101, static information includes, but is not limited to, one or more of the following: age, gender, height, and weight; motion information includes, but is not limited to, walking, running, climbing stairs, and other motion states; user requirements include, but are not limited to, one or more of the following: the location of the wearable sensor, the type of motion to be detected, and the disease to be detected; the sensor is generally an inertial measurement unit (IMU), which includes 3-axis, 6-axis, 9-axis, and 10-axis IMUs. Specifically, a 3-axis IMU is used to detect acceleration data, a 6-axis IMU is used to detect acceleration data and gyroscope data, a 9-axis IMU is used to detect acceleration data, gyroscope data, and magnetometer data, and a 10-axis IMU is used to detect acceleration data, gyroscope data, magnetometer data, and barometer data. When applying the device, it must be worn according to the actual measurement needs of the user being tested.

[0053] Step S102: Based on the static information and the motion information, perform profile matching using a group profile model to obtain a user profile corresponding to the user to be tested. The group profile model is pre-constructed based on the test information of multiple testers.

[0054] In step S102, the test information of the tester includes static test information, motion test information and sensor test data of the acquisition sensors worn by the tester. The static test information includes, but is not limited to, one or more of the following: age, gender, height and weight. The motion test information includes, but is not limited to, the motion states such as walking, running, and going up and down stairs.

[0055] Input the static and motion information of the user to be tested into the cloud-based group profiling model to perform profile matching, obtain and download the user profile corresponding to the user to be tested, determine the group to which the user to be tested belongs, and make subsequent judgments based on the user profile corresponding to the group.

[0056] Step S103: Based on the motion paradigm corresponding to the user's needs, obtain motion data models of multiple testers in the user profile under the motion paradigm;

[0057] In step S103, a motion paradigm is a simple description of a motion or action. Its core characteristic is the selection of appropriate features as the measurement or judgment criterion for the motion. Motion paradigms can be categorized into temporal, spatial, and spatiotemporal joint types based on different features. For example: Figure 2As shown, the actions in the motion paradigm are linear movement and 180-degree turns through walking. When the motion paradigm is temporal, the focus is mainly on the trajectory and speed. When the motion paradigm is spatial, the focus is mainly on the spatial position and offset of the foot. When the motion paradigm is spatiotemporally combined, both temporal and spatial aspects are considered.

[0058] Based on the user needs of the test users, the required motion paradigms can be determined, and then motion data models of multiple testers in the user profile under these paradigms can be obtained. It should be noted that, to ensure the accuracy of the data in the subsequent migration and reuse process, certain constraints need to be set when obtaining the motion data models, only acquiring motion data models that meet the conditions, thus ensuring the reusability of the motion data models for the test users.

[0059] Step S104: If the sensor data meets the first preset condition, determine that the user to be tested can reuse the data in the motion data model. The first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent; the spatial position indicated by the sensor data meets the requirements; and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0060] In step S104, the consistency of sensor data and motion data model characteristics in the first preset condition is to ensure that data can be successfully, accurately, and quickly received when the hardware (acquisition sensor) transmits data with the motion data model. The spatial position indicated by the sensor data in the first preset condition meets the requirements to ensure the accuracy of the position of the acquisition sensor worn by the user under test, and the accuracy of the spatial information of the position detected by the user indicated by the sensor data, making the data transmission between the acquisition sensor and the motion data model more precise. The consistency of the time corresponding to the sensor data collected by the multiple acquisition sensors in the first preset condition is to ensure that the time corresponding to the sensor data collected by the multiple acquisition sensors is synchronized when the user under test wears multiple acquisition sensors, thereby making the data transmission between the acquisition sensor and the motion data model more precise.

[0061] If the sensor data meets the first preset condition, it is determined that the user under test can reuse the data in the motion data model. By collecting a small amount of data from the user under test, more relevant data and / or detection results of the user under test can be obtained based on the motion data model.

[0062] This invention provides a data evaluation method. To assess the consistency between sensor data from a user under test and the data of a model that the user needs to reuse and transfer, firstly, based on the static and motion information of the user under test, a user profile is obtained by using a group profile model for profile matching, i.e., confirming a group profile consistent with the user's basic information (static and motion information). Then, based on the user's needs, the required motion paradigm for the user under test is determined, and motion data models of multiple testers in the user profile under this motion paradigm are obtained. That is, under the condition of consistent motion paradigm, the motion data model that the user under test needs to reuse and transfer is determined. Finally, if the sensor data meets a first preset condition, i.e., the sensor data and the motion data model have consistent data characteristics, the spatial location indicated by the sensor data meets requirements, and the time corresponding to the sensor data collected by multiple acquisition sensors is consistent, it is confirmed whether the user under test can reuse the motion data model data. This invention evaluates the data of the user under test from multiple aspects, including group, motion paradigm, data characteristics, space, and time, enabling efficient and accurate assessment of the consistency between sensor data and motion data model data, thereby achieving the reuse and transfer of motion data models. In practical applications of the reuse migration function, by collecting a small amount of data from the user under test, more relevant data and / or detection results of the user under test can be obtained based on the motion data model, making the acquisition of information about the user under test more efficient and convenient.

[0063] Optionally, based on the motion paradigm corresponding to the user's needs, obtain motion data models of multiple testers in the user profile under the motion paradigm, including:

[0064] Based on the user's needs, a movement paradigm is matched, and if the match is successful, the movement paradigm is obtained.

[0065] Based on the exercise paradigm, obtain the exercise test data of multiple testers in the user profile under the exercise paradigm, the types of exercise in the exercise paradigm, and the mean static correlation coefficient of the corresponding multiple testers;

[0066] Based on the exercise test data, the type of exercise, and the mean of the static correlation coefficient, intraclass correlation calculation is performed to obtain the first intraclass correlation coefficient of the exercise paradigm.

[0067] If the correlation coefficient within the first category is greater than the first threshold, obtain the motion data models of multiple testers in the user profile under the motion paradigm.

[0068] In this embodiment of the invention, firstly, motion paradigm matching is performed according to user needs to obtain the actions and types (temporal, spatial, and spatiotemporal combinations) of the successfully matched motion paradigm; then, the motion test data of multiple testers in the user profile under the motion paradigm, the types of motion in the motion paradigm, and the average static correlation coefficient of the corresponding multiple testers are obtained.

[0069] The motion test data is displayed in tabular form, with each type of motion corresponding to a separate table. Each row contains the values ​​of each axis of the acquisition sensor collected in chronological order, and each column of the table is the time sequence of the values ​​of each axis in chronological order.

[0070] The types of movement in the movement paradigm can be one or more, depending on actual needs; this invention does not impose any limitations.

[0071] The group profile model corresponding to the tester includes various static test information, motion information and sensor test data of the tester, as well as the Pearson correlation coefficient between each static test information and the sensor test data under each motion state. The mean of the static correlation coefficient is the mean of the above multiple Pearson correlation coefficients.

[0072] Secondly, obtain the intra-class correlation coefficient (ICC(m,k)) for multiple testers in the user profile under the movement paradigm. This is because whether sensor data from the same limb range can be transferred depends not only on static population characteristics but also on the dynamic movement type, which may affect the performance of different sensors. For example, a pressure sensor might be effective at detecting walking but not running. Therefore, the intra-class correlation coefficient is needed, and the formula is as follows: Among them, MS R MS is the row mean square of this table. E MS is the mean square error of this table. E Let n be the column mean square of the table, n be the number of testers in the user profile, m be the number of exercise categories participated in, and k be the mean static correlation coefficient. If the correlation coefficient within the first category is greater than a first threshold (generally 0.5), it indicates that the exercise corresponding to this exercise paradigm is consistent, and the corresponding exercise data may transfer between each other. Therefore, we obtain the exercise data model of multiple testers in the user profile under this exercise paradigm.

[0073] Optionally, the method further includes:

[0074] The sensor data is input into the converter model for feature encoding to obtain the first feature code output by the embedding layer of the converter model;

[0075] Obtain the second feature code corresponding to the data in the motion data model;

[0076] If the first feature code and the second feature code are consistent, it is determined that the sensor data satisfies the data features of the sensor data and the motion data model in the first preset condition.

[0077] In this embodiment of the invention, if the user to be tested can migrate and copy data from the motion data model, during application, the sensor data of the user to be tested needs to be input into the motion data model through a body area network (BNB) data center. The BNB data center is the most powerful mobile smart node within the BNB, typically a mobile phone. To make the output results of the motion data model more accurate, the data characteristics of the sensor data need to be consistent with the data characteristics in the motion data model. These data characteristics include, but are not limited to, data type, format, and frequency. To make the comparison of data characteristics more efficient and accurate, the sensor data is feature-encoded through the embedding layer of a Transformer model to obtain the first feature code.

[0078] For example, the original multidimensional time-series IMU data S is transformed into a high-dimensional embedding vector E using a Transformer model. In this embodiment, the input IMU data S mainly consists of nine-axis sensor data, therefore its dimension is T×9, where T represents the time step and 9 represents the data dimension in each time step, including three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer data. This embedding process is implemented by a combination of a linear mapping and a nonlinear activation function, specifically expressed as follows:

[0079] E = ReLU(SW) e +b e )

[0080] Among them, W e It is a learnable weight matrix of dimension 9×D′, used to transform the IMU data at each time step into a higher-dimensional space. D′ is the dimension of the target embedding vector, determined by model design and specific application requirements. e It is a learnable bias vector with the same dimensions as D′, used to add the necessary offset during the transformation process.

[0081] Finally, the first feature code is compared with the second feature code corresponding to the data in the motion data model. If they match, it is determined that the sensor data meets the first preset condition that the sensor data and the motion data model have the same data features.

[0082] Optionally, the method further includes:

[0083] When the user under test moves in the test scenario, the first motion data collected by the optical capture device worn by the user under test and the second motion data collected by the acquisition sensor worn by the user under test are acquired in real time. The first motion data includes the first spatial information of the user under test at each time step.

[0084] The second motion data is input into the converter model and feature-encoded to obtain a third feature code corresponding to the output of the converter model; and the second motion data is input into the converter model and position-encoded to obtain a first position code corresponding to the output of the converter model.

[0085] The location encoding and the third feature encoding are fused to obtain a first location vector, which includes the second spatial information of the user under test at each time step.

[0086] The first spatial information is compared with the second spatial information. If the spatial positions indicated by the first spatial information and the second spatial information are consistent, it is determined that the spatial position indicated by the sensor data meets the requirements of the first preset condition.

[0087] In this embodiment of the invention, to ensure more accurate data transmission between the acquisition sensor and the motion data model, it is necessary to ensure the accuracy of the spatial information indicating the location detected by the test user as shown by the sensor data. The test user is placed in a test scenario, typically a professional measurement laboratory. First, the body area network data center simultaneously connects to the optical capture device (e.g., the Vicon optical motion capture system) worn by the test user and the acquisition sensor. The first motion data acquired by the optical capture device is more accurate, including the gait trajectory data TR = {tr} of the test user for each test. k}:tr k ={(q xi,k ,q yi,k ,q zi,k ,q wi,k ,x i,k ,y i,k ,z i,k The system records the precise position and rotation of the foot at each point in time in three-dimensional space, where x, y, and z represent the spatial coordinates of the foot, q represents the quaternion of the foot in three-dimensional space (representing rotation), and k represents the gait trajectory data of the kth test.

[0088] Then, the second motion data is input into the converter model and subjected to feature encoding and position encoding respectively to obtain the third feature encoding E and the first position encoding P. The two are then fused to obtain the first position vector. The third feature encoding has the same format as the first feature encoding E in the previous embodiment, so it is also represented by E.

[0089] The introduction of the first position code P is crucial for realizing temporal signal processing. It provides key information about the temporal sequence, which is essential for accurate prediction of gait trajectories. The first position code P is a fixed set of vectors designed to enable the motion data model to distinguish inputs at different time steps. For each time step t and each dimension i, P is generated according to the following sine and cosine functions:

[0090] P (t,2i) =sin(t / 10000) 2i / D′ ); P (t,2i+1) =cos(t / 10000) 2i / D′ );

[0091] Here, t represents the time step position, i represents the dimensional position in the embedding vector, and D′ is the dimension of the embedding vector. This use of sine and cosine waves allows the model to learn positional information, thus better understanding and processing time-series data. The process of combining P and E can be represented as:

[0092] E′=E+P

[0093] E′ is the position vector, specifically an embedded vector that incorporates position information. It retains the original features of the input IMU data, enabling the model to effectively process time-series data and capture subtle changes in gait trajectories.

[0094] The first spatial information indicated by the first motion data is compared with the second spatial information indicated by the position vector. If they match, it is determined that the spatial position indicated by the sensor data meets the requirement of the first preset condition. It should be noted that the consistency of the spatial positions indicated by the first spatial information and the second spatial information does not mean that the data are completely consistent. It is sufficient that they are consistent within a certain error range. The error range is set according to the actual situation, and this invention does not impose any limitations.

[0095] Optionally, the method further includes:

[0096] When the user under test moves in the test scenario according to the preset movement mode, the third movement data collected by the optical capture device worn by the user under test and the fourth movement data collected by the acquisition sensor worn by the user under test are acquired in real time. The preset movement mode includes repeating the preset movement a first preset number of times.

[0097] The fourth motion data is input into the converter model and feature-encoded to obtain the fourth feature code corresponding to the output of the converter model; the fourth motion data is input into the converter model and position-encoded to obtain the second position code corresponding to the output of the converter model; and the fourth motion data is input into the converter model and time-encoded to obtain the time code corresponding to the output of the converter model.

[0098] The time code, the second position code, and the fourth feature code are fused to obtain a time vector, which includes data collected by the multiple acquisition sensors at each time step.

[0099] Based on the number of times the preset motion is repeated and the time vector, intra-class correlation calculation is performed to obtain the second intra-class correlation coefficient corresponding to the time vector;

[0100] If the correlation coefficient within the second category is greater than the second threshold, it is determined that the sensor data satisfies the first preset condition that the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0101] In this embodiment of the invention, to ensure more accurate data transmission between the sensors and the motion data model, when the user under test wears multiple sensors, it is necessary to ensure that the time corresponding to the sensor data collected by the multiple sensors is consistent. First, as in the previous embodiment, the user under test is placed in a test scenario. The body area network data center simultaneously connects to the optical capture device and the sensors worn by the user under test. The user under test moves in the test scenario according to a preset motion pattern (e.g., repeating a preset motion). Figure 2 During the exercise (20 repetitions of the exercise paradigm), third and fourth motion data were acquired. For example, the test user first performed a static test, standing still with feet 15 cm apart. In the walking test, the test user performed the following... Figure 2 The task is described as "go straight - turn - go straight - turn - go straight", and is repeated 20 times at a comfortable pace. The test user must make a 180° turn during the walk.

[0102] The third type of motion data includes the gait trajectory data of the user being tested for each test, TR = {tr k}:tr k ={(q xi,k ,q yi,k ,q zi,k ,q wi,k ,x i,k ,y i,k ,z i,kThe system records the precise position and rotation of the foot at each point in time in three-dimensional space, where x, y, and z represent the spatial coordinates of the foot, q represents the quaternion of the foot in three-dimensional space (representing rotation), and k represents the gait trajectory data of the kth test.

[0103] The fourth motion data is obtained through the dataset. The form is represented as S = {s}, which contains raw IMU sensor data from the subject's foot. k}, where k represents the k-th data segment, and also includes the gait index data g for each test. k g k This includes, but is not limited to, several common metrics such as stride length and stride speed. Therefore... It can be represented as

[0104] Each data segment s k Composed of continuous sensor readings, for ease of representation, the fourth motion data is input to the converter model and subjected to feature encoding, position encoding, and time encoding respectively, to obtain the fourth feature code, second position code, and time code output by the converter model. The time code is fused with the second position code and the fourth feature code to obtain a time vector. The time vector, as the column vector of the time sequence signal, provides key information on the time sequence, which is crucial for the accurate prediction of gait trajectory. The fourth feature code has the same format as the first feature code mentioned above, and the second position code has the same format as the first position code mentioned above, which will not be elaborated here.

[0105] Based on the time vector, s k Represented as s k ={(a xi,k ,a yi,k ,a zi,k ,ω xi,k ,ω yi,k ,ω zi,k )}, where a represents acceleration, ω represents angular velocity, and i indexes the data points in each time series.

[0106] Finally, regarding the dataset Perform a consistency check to obtain the intra-class correlation coefficient. Among them, MS R The rows represent mean squared values, with each row containing the data from the sensors for each axis acquired in chronological order, in MS. W The mean square of the residual variance column is given, where each column represents the time series of values ​​for each axis in chronological order, and k is the number of repetitions. If the correlation coefficient within the second category is greater than the second threshold (generally 0.5), it indicates that the sensor data collected by the multiple acquisition sensors correspond to the same time period.

[0107] Optionally, the method further includes:

[0108] Acquire test information from multiple testers, including static test information, motion test information, and sensor test data corresponding to different motion test information from the acquisition sensors worn by the testers;

[0109] Based on digital twin technology, the sensor test data are subjected to correlation screening and cluster evaluation according to the static test information and motion test information, respectively, to construct the group profile model.

[0110] In this embodiment of the invention, static test information includes, but is not limited to, one or more of the following: age, gender, height, and weight; and exercise test information includes, but is not limited to, exercise states such as walking, running, and going up and down stairs.

[0111] When constructing a group profile model, it is necessary to perform correlation screening and cluster evaluation on the sensor test data based on static test information and motion test information, respectively. The specific methods are as follows:

[0112] Iterate through the static test information and sensor test data of each tester, calculate the correlation between them, and the calculation formula is as follows: Among them, X i,k Y represents the i-th data item in the static test information of the k-th user. j,k This represents the test data of the j-th sensor for the k-th user. This represents the average value of the i-th data item in the static test information (e.g., the i-th data item is age). (This represents the average age of all test takers.) This represents the average value of the j-th data point in the sensor test data, where N is the number of testers, and r is the mean value. i,j This represents the Pearson correlation coefficient between the i-th data item in the static test information and the j-th sensor test data. Then, the sensor test data are filtered for correlation based on the Pearson correlation coefficient.

[0113] Then, the test subjects are classified according to the K-Means gait profile clustering model and the cluster boundaries, and individual twin static profiles are calculated; the mean of the sensor test data of this group of people within the 95% confidence interval is calculated, and the calculation result is used as the boundary of the individual twin static profile of the cluster category; the test subjects are divided into individual twin static profiles according to the cluster boundaries, and the group profile model is constructed. This step is common knowledge in the field of digital twin technology and will not be elaborated here.

[0114] This invention also provides specific embodiments for the steps of acquiring motion data models and the evaluation of the effectiveness of motion data models.

[0115] Example 1: Based on the movement paradigm corresponding to user needs, obtain the movement data model of multiple testers in the user profile under the movement paradigm;

[0116] Assume the test subject in the user profile wears three 9-axis IMUs, which detect acceleration data, gyroscope data, and magnetometer data. Each type of data is collected in the x, y, and z directions, with a total of 9 channels outputting data. The datasets for all motion paradigms corresponding to the test subject are shown in Table 1: the subject performs the actions in the order shown in the table, and the labels during the transition periods between different actions are all 0.

[0117]

[0118]

[0119] Table 1

[0120] Based on the user's desired exercise paradigm, exercise test data of the test subject under the exercise paradigm is obtained. This embodiment focuses on gait-related information, therefore, six gait-related exercise paradigms are selected: standing, walking, running, cycling, Nordic walking, and rope skipping. Taking this as an example, exercise test data of the test subject in the above activity types are obtained, and the first intra-class correlation coefficient for each exercise is calculated. If the first intra-class correlation coefficient is greater than a first threshold, exercise data models of multiple test subjects in the user profile under the exercise paradigm are obtained. In addition, the effect of the exercise data model corresponding to each exercise paradigm can also be obtained, as shown in Table 2, which can be referenced in the subsequent migration and reuse process.

[0121]

[0122] Table 2

[0123] Example 2: After the sensor data meets the requirements of the spatial location indicated by the sensor data in the first preset condition, the sensor data captured by the acquisition sensors at different locations within the same limb valve is transmitted to the body area network data center, for example: different locations on the sole, instep, ankle and calf.

[0124] By training a gait recognition model, the contribution weights of the acquisition sensors at different locations are trained. The calculation results of the contribution weights at different locations under different motion paradigms and the overall contribution weight are shown in Table 3, which facilitates the selection of the placement position of the acquisition sensors according to user needs.

[0125]

[0126] Table 3

[0127] like Figure 3 As shown, embodiments of the present invention also provide a data evaluation apparatus, comprising:

[0128] The first acquisition module 301 is used to acquire static information, motion information and sensor data of the acquisition sensor worn by the user under test;

[0129] The first matching module 302 is used to perform profile matching using a group profile model based on the static information and the motion information to obtain a user profile corresponding to the user to be tested, wherein the group profile model is pre-constructed based on the test information of multiple testers;

[0130] The second acquisition module 303 is used to acquire the motion data models of multiple testers in the user profile under the motion paradigm according to the motion paradigm corresponding to the user needs.

[0131] The first determining module 304 is used to determine, when the sensor data meets the first preset condition, that the user to be tested can reuse the data in the motion data model, wherein the first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent, the spatial position indicated by the sensor data meets the requirements, and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

[0132] Optionally, the second acquisition module 303 includes

[0133] The first acquisition unit is used to perform motion paradigm matching according to the user's needs, and acquire the motion paradigm if the matching is successful.

[0134] The second acquisition unit is used to acquire, according to the exercise paradigm, the exercise test data of multiple testers in the user profile under the exercise paradigm, the types of exercise in the exercise paradigm, and the average static correlation coefficient of the corresponding multiple testers;

[0135] The first calculation unit is used to perform intra-class correlation calculation based on the exercise test data, the type of exercise and the mean of the static correlation coefficient, to obtain the first intra-class correlation coefficient of the exercise paradigm.

[0136] The third acquisition unit is used to acquire the motion data models of multiple testers in the user profile under the motion paradigm when the correlation coefficient within the first category is greater than the first threshold.

[0137] Optionally, the device further includes:

[0138] The first encoding module is used to input the sensor data into the converter model for feature encoding to obtain the first feature code output by the embedding layer of the converter model;

[0139] The third acquisition module is used to obtain the second feature code corresponding to the data in the motion data model;

[0140] The second determining module is used to determine, when the first feature code and the second feature code are consistent, that the sensor data satisfies the data features of the sensor data and the motion data model in the first preset condition.

[0141] Optionally, the device further includes:

[0142] The fourth acquisition module is used to acquire, in real time, the first motion data collected by the optical capture device worn by the user under test and the second motion data collected by the acquisition sensor worn by the user under test when the user under test is moving in the test scenario, wherein the first motion data includes the first spatial information of the user under test at each time step;

[0143] The second encoding module is used to perform feature encoding on the second motion data input converter model to obtain a third feature code corresponding to the output of the converter model; and to perform position encoding on the second motion data input converter model to obtain a first position code corresponding to the output of the converter model.

[0144] The first fusion module is used to fuse the location code with the third feature code to obtain a first location vector, wherein the first location vector includes the second spatial information of the user under test at each time step;

[0145] The third determining module is used to compare the first spatial information with the second spatial information, and if the spatial positions indicated by the first spatial information and the second spatial information are consistent, determine that the spatial position indicated by the sensor data meets the requirements of the first preset condition.

[0146] Optionally, the device further includes:

[0147] The fifth acquisition module is used to acquire, in real time, the third motion data collected by the optical capture device worn by the user under test and the fourth motion data collected by the acquisition sensor worn by the user under test when the user under test moves in the test scenario according to the preset motion mode. The preset motion mode includes repeating the preset motion a first preset number of times.

[0148] The third encoding module is used to perform feature encoding on the fourth motion data input converter model to obtain the fourth feature code corresponding to the output of the converter model; and to perform position encoding on the fourth motion data input converter model to obtain the second position code corresponding to the output of the converter model; and to perform time encoding on the fourth motion data input converter model to obtain the time code corresponding to the output of the converter model.

[0149] The second fusion module is used to fuse the time code, the second position code and the fourth feature code to obtain a time vector, the time vector including data collected by the multiple acquisition sensors at each time step;

[0150] The first calculation module is used to perform intra-class correlation calculation based on the number of times the preset motion is repeated and the time vector to obtain the second intra-class correlation coefficient corresponding to the time vector;

[0151] The fourth determining module is used to determine that, when the correlation coefficient within the second category is greater than the second threshold, the time corresponding to the sensor data collected by the multiple acquisition sensors in the first preset condition is consistent.

[0152] Optionally, the device further includes:

[0153] The sixth acquisition module is used to acquire the test information of multiple testers, including static test information, motion test information, and sensor test data of the acquisition sensors worn by the testers corresponding to different motion test information;

[0154] The first construction module is used to construct the group profile model by performing correlation screening and cluster evaluation on the sensor test data based on the static test information and motion test information, respectively, using digital twin technology.

[0155] It should be noted that the embodiments of this device are devices corresponding to the embodiments of the above methods. All implementations in the embodiments of the above methods are applicable to the embodiments of this device and can achieve the same technical effect.

[0156] This invention also provides a network device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the data evaluation method as described in any of the preceding claims and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0157] This invention also provides a readable storage medium, comprising: a program stored on the readable storage medium, wherein when the program is executed by a processor, it implements the steps of the data evaluation method described in any of the preceding claims and achieves the same technical effect; to avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0158] This invention also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the steps of the data evaluation method described in any of the preceding claims and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0159] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0160] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A data evaluation method, characterized in that, include: Acquire static information, motion information, user needs, and sensor data from the sensors worn by the user under test; Based on the static information and the motion information, a group profile model is used to perform profile matching to obtain a user profile corresponding to the user to be tested. The group profile model is pre-constructed based on the test information of multiple testers. Based on the movement paradigm corresponding to the user's needs, obtain the movement data models of multiple testers in the user profile under the movement paradigm; If the sensor data meets the first preset condition, it is determined that the user under test can reuse the data in the motion data model. The first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent; the spatial position indicated by the sensor data meets the requirements; and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

2. The data evaluation method according to claim 1, characterized in that, Based on the movement paradigm corresponding to the user's needs, obtain movement data models of multiple testers in the user profile under the movement paradigm, including: Based on the user's needs, a movement paradigm is matched, and if the match is successful, the movement paradigm is obtained. Based on the exercise paradigm, obtain the exercise test data of multiple testers in the user profile under the exercise paradigm, the types of exercise in the exercise paradigm, and the mean static correlation coefficient of the corresponding multiple testers; Based on the exercise test data, the type of exercise, and the mean of the static correlation coefficient, intraclass correlation calculation is performed to obtain the first intraclass correlation coefficient of the exercise paradigm. If the correlation coefficient within the first category is greater than the first threshold, obtain the motion data models of multiple testers in the user profile under the motion paradigm.

3. The data evaluation method according to claim 1, characterized in that, The method further includes: The sensor data is input into the converter model for feature encoding to obtain the first feature code output by the embedding layer of the converter model; Obtain the second feature code corresponding to the data in the motion data model; If the first feature code and the second feature code are consistent, it is determined that the sensor data satisfies the data features of the sensor data and the motion data model in the first preset condition.

4. The data evaluation method according to claim 1, characterized in that, The method further includes: When the user under test moves in the test scenario, the first motion data collected by the optical capture device worn by the user under test and the second motion data collected by the acquisition sensor worn by the user under test are acquired in real time. The first motion data includes the first spatial information of the user under test at each time step. The second motion data is input into the converter model and feature-encoded to obtain a third feature code corresponding to the output of the converter model; and the second motion data is input into the converter model and position-encoded to obtain a first position code corresponding to the output of the converter model. The location encoding and the third feature encoding are fused to obtain a first location vector, which includes the second spatial information of the user under test at each time step. The first spatial information is compared with the second spatial information. If the spatial positions indicated by the first spatial information and the second spatial information are consistent, it is determined that the spatial position indicated by the sensor data meets the requirements of the first preset condition.

5. The data evaluation method according to claim 1, characterized in that, The method further includes: When the user under test moves in the test scenario according to the preset movement mode, the third movement data collected by the optical capture device worn by the user under test and the fourth movement data collected by the acquisition sensor worn by the user under test are acquired in real time. The preset movement mode includes repeating the preset movement a first preset number of times. The fourth motion data is input into the converter model and feature-encoded to obtain the fourth feature code corresponding to the output of the converter model; the fourth motion data is input into the converter model and position-encoded to obtain the second position code corresponding to the output of the converter model; and the fourth motion data is input into the converter model and time-encoded to obtain the time code corresponding to the output of the converter model. The time code, the second position code, and the fourth feature code are fused to obtain a time vector, which includes data collected by the multiple acquisition sensors at each time step. Based on the number of times the preset motion is repeated and the time vector, intra-class correlation calculation is performed to obtain the second intra-class correlation coefficient corresponding to the time vector; If the correlation coefficient within the second category is greater than the second threshold, it is determined that the sensor data satisfies the first preset condition that the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

6. The data evaluation method according to claim 1, characterized in that, The method further includes: Acquire test information from multiple testers, including static test information, motion test information, and sensor test data corresponding to different motion test information from the acquisition sensors worn by the testers; Based on digital twin technology, the sensor test data are subjected to correlation screening and cluster evaluation according to the static test information and motion test information, respectively, to construct the group profile model.

7. A data evaluation device, characterized in that, include: The first acquisition module is used to acquire static information, motion information, user needs, and sensor data of the acquisition sensors worn by the user under test. The first matching module is used to perform profile matching using a group profile model based on the static information and the motion information to obtain a user profile corresponding to the user to be tested, wherein the group profile model is pre-constructed based on the test information of multiple testers; The second acquisition module is used to acquire the motion data models of multiple testers in the user profile under the motion paradigm according to the motion paradigm corresponding to the user needs. The first determining module is used to determine, when the sensor data meets a first preset condition, that the user to be tested can reuse the data in the motion data model, wherein the first preset condition includes one or more of the following: the data characteristics of the sensor data and the motion data model are consistent; the spatial position indicated by the sensor data meets the requirements; and the time corresponding to the sensor data collected by the multiple acquisition sensors is consistent.

8. A network device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, the program implementing the data evaluation method as described in any one of claims 1 to 6 when executed by the processor.

9. A readable storage medium, characterized in that, include: The readable storage medium stores a program that, when executed by a processor, implements the steps of the data evaluation method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the data evaluation method as described in any one of claims 1 to 6.

Images