Body fat weighing, measuring and correcting system and method based on Internet of Things

Abstract

The invention discloses a body fat scale measurement correction system and method based on the Internet of Things, and particularly relates to the technical field of intelligent health monitoring equipment, and the system comprises a terminal sensing layer, an Internet of Things transmission layer, a cloud data processing layer and a user interaction layer. The Internet of Things transmission layer realizes data interaction between the terminal sensing layer and the cloud data processing layer and the user interaction layer; the cloud data processing layer is used for processing the received data, and the cloud data processing layer comprises a data preprocessing module, a feature extraction module, a dynamic correction model module and a model iteration updating module. Multi-path electrical impedance values are collected through an 8-electrode BIA module, feature extraction is optimized in combination with a WKPCA algorithm, dynamic correction is achieved through an AGA-BP model, superposition environment parameters are compensated in real time, the measurement error rate is obviously reduced compared with a traditional body fat scale, and the measurement precision of special crowds is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent health monitoring equipment technology, and more specifically, to an Internet of Things-based body fat scale measurement and correction system and method. Background Technology

[0002] With increasing health awareness, body fat scales have become a core device for home health monitoring, and their measurement accuracy directly affects users' health assessment results. Current body fat scales are mainly based on bioelectrical impedance analysis (BIA), calculating parameters such as body fat percentage and muscle mass by measuring the body's electrical impedance values. However, they have the following technical shortcomings: Sensitive to environmental interference: Traditional body fat scales do not consider the influence of environmental factors such as temperature, humidity, and ground flatness on electrical impedance measurement. Poor adaptability to individual differences: Existing models mostly use fixed formulas (such as the BodPod formula and the DEXA formula), and do not personalize them for users of different ages, genders, and body shapes. The measurement error will be further increased, especially for special groups (such as pregnant women, the elderly, and athletes). Static calibration model: The calibration parameters of traditional body fat scales are fixed at the factory and cannot be dynamically updated according to the data feedback during user use. After long-term use, the measurement accuracy will continue to decline due to equipment aging and changes in usage scenarios.

[0003] Therefore, to address the above problems, a body fat scale measurement and calibration system and method based on the Internet of Things are proposed. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, this application provides a body fat scale measurement and correction system and method based on the Internet of Things to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this application provides the following technical solution: a body fat scale measurement and calibration system based on the Internet of Things, comprising a terminal sensing layer, an Internet of Things transmission layer, a cloud data processing layer and a user interaction layer, wherein the terminal sensing layer is integrated into the body fat scale hardware device and includes a bioelectrical impedance measurement module (BIA measurement module), an environmental parameter sensing module and a user basic information collection module. The IoT transmission layer adopts a transmission mode combining Bluetooth 5.0 and NB-IoT network to realize data interaction between the terminal perception layer, the cloud data processing layer, and the user interaction layer. The IoT transmission layer enables short-range transmission: real-time data interaction between the body fat scale and the user's mobile terminal is achieved via Bluetooth 5.0, supporting the uploading of basic user information and feedback of calibration results; long-range transmission: multi-source data (impedance value, environmental parameters, user information) collected by the terminal is uploaded to the cloud server via the NB-IoT network, supporting low-power long-term online operation. The cloud data processing layer is used to process the received data, and includes a data preprocessing module, a feature extraction module, a dynamic correction model module, and a model iteration update module. The user interaction layer uses a mobile terminal APP to display measurement results, provide user feedback, and generate personalized health suggestions.

[0006] Furthermore, the bioelectrical impedance measurement module adopts an 8-electrode layout for collecting multipath impedance values ​​of the human body. ,in To measure the path, , This represents the total number of measurement paths. To measure the number of times, , This refers to the number of measurements taken in a single session. The environmental parameter sensing module uses multiple sensors to collect ambient temperature data. ,humidity with the angle of inclination of the body Furthermore, it incorporates a variety of sensors, including temperature sensors, humidity sensors, and tilt sensors. The user basic information collection module is used to obtain the user's age. ,gender ,height ,weight .

[0007] Furthermore, the data preprocessing module is used to preprocess the data, specifically by preprocessing the impedance value. The average value was calculated by removing outliers using the 3σ criterion. with standard deviation Remove Data; and through The data is normalized, among which The original data, , These are the historical maximum and minimum values ​​of this type of raw data. This is the normalized data.

[0008] Furthermore, the feature extraction module employs an improved weighted kernel principal component analysis algorithm, WKPCA, to perform feature dimensionality reduction on the multipath impedance values, introducing sample confidence weights. Construct a weighted kernel matrix Before extraction One principal component is used as the input feature, where... The Gaussian kernel function; This represents the weighted kernel function value. and Sample data and Corresponding weights; Sample credibility weights Samples with high measurement stability (small standard deviation) are assigned higher weights to improve feature extraction accuracy. ,in For the first Standard deviation of electrical impedance measurement for each sample A cumulative contribution rate of ≥95% is considered a principal component.

[0009] Furthermore, the dynamic correction model module employs an adaptive genetic algorithm, specifically an AGA-optimized BP neural network (AGA-BP model), to calculate the final corrected body fat percentage using a linear compensation formula for environmental parameters. The BP neural network structure is as follows: Input layer - 1 hidden layer - Output layer, where the number of nodes in the input layer is... Equal to the number of principal components and the number of output layer nodes in WKPCA Number of hidden layer nodes pass Confirmed, and To adjust the parameters.

[0010] Furthermore, the specific process of optimizing the BP neural network using AGA includes: A1. Population Initialization: Encode the weights and thresholds of the BP neural network as real-number chromosomes, and determine the population size. Chromosome length ; A2. Adaptive Crossover and Mutation: Crossover Probability The calculation formula is: Probability of mutation The calculation formula is: ;in ; ; ; ; The maximum fitness of the population; The average fitness of the population; For individual fitness; A3. Fitness Calculation: Using... As a fitness function Predict body fat percentage for the model; Standard body fat percentage, measured using a DEXA bone densitometer; The number of samples used in the calculation; In the AGA-BP model, environmental parameters are compensated for to eliminate environmental interference using a linear compensation formula, which is: ,in This is the final corrected body fat percentage; , , These are standard environmental parameters; , , This is the compensation coefficient.

[0011] Furthermore, the model iteration update module uses an incremental learning algorithm to update the model parameters. Specifically, the strategy is as follows: model iteration is triggered when the cumulative amount of user measurement data on the cloud server reaches a threshold (e.g., 1000 data points / user). The incremental learning algorithm updates the weights and thresholds of the AGA-BP model, specifically: ,in Forgetting factor, For the old weight, For the updated weights calculated based on the new data, For the new weights.

[0012] Furthermore, the user interaction layer is implemented through a mobile terminal APP: Display function: Real-time display of corrected body fat percentage, muscle mass, water percentage and other parameters, as well as historical data trend charts; Feedback function: Users can mark the accuracy of measurement results (such as "accurate", "too high", "too low"), and the feedback data is uploaded to the cloud server for model optimization; Health advice feature: Based on calibrated measurement data and combined with the user's health goals (such as fat loss and muscle gain), it generates personalized health advice.

[0013] Furthermore, a body fat scale measurement and calibration method based on the Internet of Things is characterized by the following steps: S1. Data Acquisition; S11. Users input basic information via the body fat scale touchscreen or mobile app, including user age. ,gender ,height ,weight ; S12. Multipath impedance values ​​are collected using the 8-electrode BIA module of the body fat scale. Number of samplings Remove the first 5 and last 5 data points, and keep the middle 40 data points; S13. Environmental parameter sensing module collects current ambient temperature. ,humidity with the angle of inclination of the body The sampling frequency is 1 time / second, and the average value is taken after 5 consecutive samplings. S2. Data preprocessing; S21. Resistive impedance value Outliers were removed using the 3σ criterion. S22. Normalize the preprocessed impedance values, environmental parameters, and user basic information to obtain a normalized dataset. ,in This represents the normalized electrical impedance value; These are the normalized temperature data, humidity data, angle data, user age data, gender data, height data, and weight data, respectively. S3. Feature extraction; S31. From normalized datasets Extracting electrical impedance value Calculate the confidence weight for each sample ; S32. Construct the kernel matrix of WKPCA Extracting the pre-evaluation value through eigenvalue decomposition One principal component is used to obtain the input features. ; S4. Dynamic correction; S41. Input features Inputting the data into a pre-trained AGA-BP model yields a preliminary corrected body fat percentage. ; S42. Adopt Incorporating environmental parameter compensation, the final corrected body fat percentage is calculated. ; S43. Based on user weight and height ,pass Calculate the corrected muscle mass ,in This represents the rate of muscle mass decay with age. This is the coefficient of gender difference; S5. Data Feedback and Model Iteration; S51 cloud server will and Feedback is sent to the mobile terminal APP via IoT transport layer 2; S52. Users can mark the accuracy of measurement results via the APP and upload the feedback data to the cloud; S53. When the number of valid measurement data for a user in the cloud reaches 1,000, the AGA-BP model is updated using an incremental learning algorithm to generate a personalized correction model for that user. S54. Perform a global model optimization on all user data once a month to improve the system's adaptability to new users.

[0014] The technical effects and advantages of this application are as follows: Compared with existing technologies, this IoT-based body fat scale measurement and calibration system and method significantly improves accuracy: It acquires multi-path impedance values ​​through an 8-electrode BIA module, optimizes feature extraction using the WKPCA algorithm, achieves dynamic calibration using the AGA-BP model, and superimposes real-time environmental parameter compensation, resulting in a significantly lower measurement error rate compared to traditional body fat scales, and improved measurement accuracy for specific populations. It also exhibits strong environmental adaptability, incorporating compensation mechanisms for environmental parameters such as temperature, humidity, and tilt angle, effectively controlling the error rate even in extreme environments (low temperature and low humidity, high temperature and high humidity), thus improving environmental adaptability compared to traditional body fat scales. Personalized adaptation: Based on user basic information, historical measurement data and feedback data, a personalized correction model is generated through incremental learning. At the same time, scenario-based compensation factors are added for different scenarios to meet diverse usage needs. Data linkage capability: Relying on IoT technology, it enables real-time data interaction between terminals, cloud, and user terminals, and can be connected with systems such as health records and fitness records to break down data silos and provide precise data support for health management, medical intervention, and fitness guidance. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system flow of this application; Figure 2 A schematic diagram illustrating the specific process of optimizing the BP neural network using AGA in the system of this application; Figure 3 This is a schematic diagram of the method flow of this application.

[0016] The attached diagram is labeled as follows: 1. Terminal sensing layer; 101. Bioelectrical impedance measurement module; 102. Environmental parameter sensing module; 103. User basic information collection module; 2. Internet of Things transmission layer; 3. Cloud data processing layer; 301. Data preprocessing module; 302. Feature extraction module; 303. Dynamic correction model module; 304. Model iteration update module; 4. User interaction layer. Detailed Implementation

[0017] 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, and 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.

[0018] Example 1 like Figure 1The body fat scale measurement and calibration system based on the Internet of Things (IoT) shown includes a terminal sensing layer 1, an IoT transmission layer 2, a cloud data processing layer 3, and a user interaction layer 4. The terminal sensing layer 1 is integrated into the body fat scale hardware device and includes a bioelectrical impedance measurement module 101 (BIA measurement module), an environmental parameter sensing module 102, and a user basic information collection module 103. The bioelectrical impedance measurement module 101 adopts an 8-electrode layout, which is different from the traditional 4-electrode layout. Two electrodes are positioned in each of the left front, left rear, right front, and right rear areas of the weighing surface, allowing simultaneous acquisition of multi-path impedance values ​​such as hand-to-foot, left foot-to-right foot, and hand-to-hand measurements. It is used for acquiring multi-path impedance values ​​of the human body. ,in To measure the path, , This represents the total number of measurement paths. To measure the number of times, , This refers to the number of measurements taken in a single session. The environmental parameter sensing module 102 is based on multiple sensors for collecting ambient temperature data. ,humidity with the angle of inclination of the body Furthermore, it incorporates a variety of sensors, including temperature sensors, humidity sensors, and tilt sensors. The user basic information collection module 103 is used to obtain the user's age. ,gender ,height ,weight ; The IoT transmission layer 2 adopts a transmission mode combining Bluetooth 5.0 and NB-IoT network to realize data interaction between the terminal perception layer 1, the cloud data processing layer 3, and the user interaction layer 4. The IoT transmission layer 2 enables short-range transmission: real-time data interaction between the body fat scale and the user's mobile terminal via Bluetooth 5.0, supporting the uploading of basic user information and feedback of calibration results; long-range transmission: uploading multi-source data (electrical impedance value, environmental parameters, user information) collected by the terminal to the cloud server via the NB-IoT network, supporting low-power long-term online operation. The cloud data processing layer 3 is used to process the received data, and the cloud data processing layer 3 includes a data preprocessing module 301, a feature extraction module 302, a dynamic correction model module 303, and a model iteration update module 304. The data preprocessing module 301 is used to preprocess the data. The specific strategy is to preprocess the impedance value. The average value was calculated by removing outliers using the 3σ criterion. with standard deviation Remove Data; and through The data is normalized, among which The original data, , These are the historical maximum and minimum values ​​of this type of raw data. This is the normalized data.

[0019] Feature extraction module 302 employs an improved weighted kernel principal component analysis algorithm, WKPCA, to perform feature dimensionality reduction on multipath impedance values, introducing sample confidence weights. Construct a weighted kernel matrix Before extraction One principal component is used as the input feature, where... The Gaussian kernel function; This represents the weighted kernel function value. and Sample data and Corresponding weights; Sample credibility weights Samples with high measurement stability (small standard deviation) are assigned higher weights to improve feature extraction accuracy. ,in For the first Standard deviation of electrical impedance measurement for each sample A cumulative contribution rate of ≥95% is considered a principal component.

[0020] The dynamic correction model module 303 employs an adaptive genetic algorithm, specifically an AGA-optimized BP neural network (AGA-BP model), to calculate the final corrected body fat percentage using a linear compensation formula for environmental parameters. The BP neural network structure is as follows: Input layer - 1 hidden layer - Output layer, where the number of nodes in the input layer is... Equal to the number of principal components and the number of output layer nodes in WKPCA Number of hidden layer nodes pass Confirmed, and To adjust the parameters.

[0021] The specific process of AGA optimizing the BP neural network includes: A1. Population Initialization: Encode the weights and thresholds of the BP neural network as real-number chromosomes, and determine the population size. Chromosome length ; A2. Adaptive Crossover and Mutation: Crossover Probability The calculation formula is: Probability of mutation The calculation formula is: ;in ; ; ; ; The maximum fitness of the population; The average fitness of the population; For individual fitness; A3. Fitness Calculation: Using... As a fitness function Predict body fat percentage for the model; Standard body fat percentage, measured using a DEXA bone densitometer; The number of samples used in the calculation; In the AGA-BP model, environmental parameters are compensated for to eliminate environmental interference using a linear compensation formula, which is: ,in This is the final corrected body fat percentage; , , These are standard environmental parameters; , , This is the compensation coefficient.

[0022] The model iteration update module 304 uses an incremental learning algorithm to update the model parameters. Specifically, the strategy is as follows: model iteration is triggered when the cumulative amount of user measurement data on the cloud server reaches a threshold (e.g., 1000 data points per user). The incremental learning algorithm updates the weights and thresholds of the AGA-BP model, specifically: ,in Forgetting factor, For the old weight, For the updated weights calculated based on the new data, For the new weights; The user interaction layer 4 uses a mobile terminal APP to display measurement results, provide user feedback, and generate personalized health suggestions.

[0023] User interaction layer 4 is implemented through a mobile terminal APP: Display function: Real-time display of corrected body fat percentage, muscle mass, water percentage and other parameters, as well as historical data trend charts; Feedback function: Users can mark the accuracy of measurement results (such as "accurate", "too high", "too low"), and the feedback data is uploaded to the cloud server for model optimization; Health advice feature: Based on calibrated measurement data and combined with the user's health goals (such as fat loss and muscle gain), it generates personalized health advice.

[0024] Example 2 Please see Figure 3 A body fat scale measurement calibration method based on the Internet of Things, the method comprising the following steps: S1. Data Acquisition; S11. Users input basic information via the body fat scale touchscreen or mobile app, including user age. ,gender ,height ,weight ; S12. Multipath impedance values ​​are collected using the 8-electrode BIA module of the body fat scale. Number of samplings Remove the first 5 and last 5 data points, and keep the middle 40 data points; S13. Environmental parameter sensing module collects current ambient temperature. ,humidity with the angle of inclination of the body The sampling frequency is 1 time / second, and the average value is taken after 5 consecutive samplings. S2. Data preprocessing; S21. Resistive impedance value Outliers were removed using the 3σ criterion. S22. Normalize the preprocessed impedance values, environmental parameters, and user basic information to obtain a normalized dataset. ,in This represents the normalized electrical impedance value; These are the normalized temperature data, humidity data, angle data, user age data, gender data, height data, and weight data, respectively. S3. Feature extraction; S31. From normalized datasets Extracting electrical impedance value Calculate the confidence weight for each sample ; S32. Construct the kernel matrix of WKPCA Extracting the pre-evaluation value through eigenvalue decomposition One principal component is used to obtain the input features. ; S4. Dynamic correction; S41. Input features Inputting the data into a pre-trained AGA-BP model yields a preliminary corrected body fat percentage. ; S42. Adopt Incorporating environmental parameter compensation, the final corrected body fat percentage is calculated. ; S43. Based on user weight and height ,pass Calculate the corrected muscle mass ,in This represents the rate of muscle mass decay with age. This is the coefficient of gender difference; S5. Data Feedback and Model Iteration; S51 cloud server will and Feedback is sent to the mobile terminal APP via IoT transport layer 2; S52. Users can mark the accuracy of measurement results via the APP and upload the feedback data to the cloud; S53. When the number of valid measurement data for a user in the cloud reaches 1,000, the AGA-BP model is updated using an incremental learning algorithm to generate a personalized correction model for that user. S54. Perform a global model optimization on all user data once a month to improve the system's adaptability to new users.

Claims

1. A body fat scale measurement correction system based on Internet of Things, comprising a terminal perception layer (1), an Internet of Things transmission layer (2), a cloud data processing layer (3) and a user interaction layer (4), characterized in that: The terminal perception layer (1) is integrated in a body fat scale hardware device, and includes a bioelectrical impedance measurement module (101), an environmental parameter sensing module (102), and a user basic information acquisition module (103); The Internet of Things transmission layer (2) adopts a transmission mode combining Bluetooth 5.0 and NB-IoT networks to realize data interaction between the terminal perception layer (1), the cloud data processing layer (3), and the user interaction layer (4); The cloud data processing layer (3) is configured to process received data, and includes a data preprocessing module (301), a feature extraction module (302), a dynamic correction model module (303), and a model iteration updating module (304); The user interaction layer (4) realizes measurement result display, user feedback, and personalized health suggestion generation through a mobile terminal APP.

2. The body fat scale measurement correction system based on the Internet of Things according to claim 1, characterized in that: The bioelectrical impedance measurement module (101) adopts an 8-electrode layout for collecting multi-path electrical impedance values of a human body wherein is the total number of measurement paths; , is the total number of measurement paths; is the number of measurements, , is the number of single measurements; The environmental parameter sensing module (102) is based on multiple sensors for collecting environmental temperature , humidity and the inclination of the scale body , and the multiple sensors include a temperature sensor, a humidity sensor and an inclination sensor. The user basic information collection module (103) is used for acquiring user age , gender , height , weight .

3. The body fat scale measurement correction system based on the Internet of Things according to claim 2, characterized in that: The data preprocessing module (301) is used for preprocessing data, and the specific strategy is: preprocessing the electrical impedance value Adopting 3σ criterion to eliminate abnormal value to calculate average value And standard deviation , eliminate Data; and through Normalization processing is carried out on the data, wherein The original data, , The historical maximum value and the minimum value of the original data of this kind respectively, Normalized data.

4. The body fat scale measurement correction system based on the Internet of Things according to claim 3, characterized in that: The feature extraction module (302) uses an improved weighted kernel principal component analysis algorithm, namely WKPCA, to perform feature dimension reduction on the multi-path resistance impedance values, and introduces sample reliability weights A weighted kernel matrix is constructed Before extraction The first principal component is extracted as an input feature, wherein, The Gaussian kernel function is used; The kernel function value after the weighting process is represented by, And The sample data And The corresponding weights.

5. The body fat scale measurement correction system based on the Internet of Things according to claim 4, characterized in that: The dynamic correction model module (303) adopts an adaptive genetic algorithm (AGA) optimized BP neural network (AGA-BP model) to calculate the final corrected body fat rate in combination with an environmental parameter linear compensation formula, and the BP neural network structure is as follows: an input layer-1 hidden layer-output layer, wherein the number of input layer nodes is equal to the number of principal components in WKPCA, the number of output layer nodes is equal to the number of principal components in WKPCA, the number of hidden layer nodes is determined by and is an adjustment parameter. input layer-1 hidden layer- output layer, wherein the number of input layer nodes is equal to the number of principal components in WKPCA, the number of output layer nodes , the number of hidden layer nodes is determined by , and is an adjustment parameter.

6. The body fat scale measurement correction system based on the Internet of Things according to claim 5, characterized in that: The specific process of optimizing the BP neural network by the AGA includes: A1. Population initialization: encode the weights and thresholds of the BP neural network as real-valued chromosomes, population size , chromosome length ; A2. Adaptive crossover and mutation: crossover probability The calculation formula of the crossover probability is: The calculation formula of the mutation probability is: ; wherein ; ; ; ; is the maximum fitness of the population; is the average fitness of the population; is the individual fitness;​ A3. Fitness calculation: Use as fitness function, for model predicted body fat percentage; for standard body fat percentage, for number of samples involved in the calculation; The environmental parameters in the AGA-BP model are compensated and eliminated from environmental interference by a linear compensation formula, which is: wherein is the final corrected body fat rate; , , is a standard environmental parameter; , , is a compensation coefficient.

7. The body fat scale measurement correction system based on the Internet of Things according to claim 6, characterized in that: The model iterative updating module (304) updates the model parameters by using an incremental learning algorithm, and the specific strategy is: the model iteration is started when the cumulative user measurement data amount of the cloud server reaches a threshold; the incremental learning algorithm updates the weights and thresholds of the AGA-BP model, and the specific method is: wherein is a forgetting factor, is an old weight, is an updated weight calculated based on new data, is a new weight. 8.The body fat scale measurement correction system based on the Internet of Things according to claim 7, wherein: The user interaction layer (4) realizes the following functions through a mobile terminal APP: Display function: real-time display of corrected body fat rate, muscle mass, water content, and historical data trend chart; Feedback function: the user can mark the accuracy of the measurement result, and the feedback data is uploaded to the cloud server for model optimization; Health suggestion function: based on the corrected measurement data, personalized health suggestions are generated in combination with user health goals (such as weight loss and muscle gain). 9.A body fat scale measurement correction method based on Internet of Things, characterized in that: The method includes the following steps: S1. Data acquisition; S11. The user inputs basic information through the body fat scale touch screen or the mobile terminal APP, and the basic information includes user age , gender , height , weight ; S12. Collect multi-path electrical impedance values by 8-electrode BIA module of body fat scale , sampling times , remove the first 5 and last 5 data, and keep the middle 40 data; S13. The environmental parameter sensing module collects the current environmental temperature , humidity , and the inclination angle of the scale The collection frequency is 1 time / second, and the average value is taken after continuous collection for 5 times. S2. Data preprocessing; S21. The electrical impedance value Anomalies were removed using the 3σ criterion; S22. Normalize the pre-processed electrical impedance values, environmental parameters, and user basic information to obtain a normalized data set wherein represents the normalized electrical impedance values; respectively, are the normalized temperature data, humidity data, angle data, user age data, gender data, height data, and weight data. S3. Feature extraction; S31. Extracting electrical impedance values from the normalized dataset Calculating a confidence weight for each sample ;​ S32. Constructing the kernel matrix of WKPCA , extracting the first principal components by eigenvalue decomposition, obtaining input features ; S4. Dynamic correction; S41. inputting the input features into the pre-trained AGA-BP model to obtain a preliminary corrected body fat rate ;​ S42. Adopting Incorporating environmental parameter compensation, calculating final corrected body fat rate ; S43. Based on the user's weight and height , the corrected muscle mass is calculated by where is the age decay coefficient for muscle mass, is the gender difference coefficient;​ S5. Data feedback and model iteration; S51. The cloud server will and Through the Internet of Things transmission layer 2 feedback to the mobile terminal APP; S52. The user marks the accuracy of the measurement result through the APP, and the feedback data is uploaded to the cloud; S53. When the cloud accumulates 1000 valid measurement data of a certain user, the AGA-BP model is updated using an incremental learning algorithm to generate a personalized correction model for the user; S54. The global model is optimized once a month based on all user data.

Images