Automatic test data dynamic optimization method based on big data

Abstract

The invention belongs to the technical field of smart bracelets, and particularly relates to an automatic test data dynamic optimization method based on big data, which is applied to a smart bracelet and comprises the following steps: step 1, acquiring load data of a smart bracelet wearing user through the smart bracelet; 2, extracting load bearing characteristics of a user wearing the smart bracelet from load bearing data collected by the smart bracelet; 3, optimizing the calorie calculation model of the intelligent bracelet calculation wearing user according to the load bearing characteristics of the intelligent bracelet wearing user; 4, carrying out multi-scene load test on the calorie calculation model of the smart bracelet; the priority is dynamically adjusted according to a test result; and 5, constructing a personalized user portrait of the smart bracelet wearing user and continuously optimizing the personalized user portrait. According to the invention, deep coupling of load monitoring and energy consumption calculation is realized, and more accurate and reliable exercise energy consumption data is provided for users.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart bracelets, and in particular relates to a method for dynamic optimization of automated test data based on big data. Background Art

[0002] Accurately monitoring calorie consumption is a key function in smart wristband technology. However, existing technologies have significant shortcomings when it comes to handling weight-bearing exercise scenarios. Traditional calorie calculation models, such as the Mifflin-St Jeor equation, do not factor weight into the calculation, making it difficult to accurately reflect energy expenditure during weight-bearing conditions. Summary of the Invention

[0003] The purpose of the present invention is to provide an automated test data dynamic optimization method based on big data, which can achieve deep coupling of load monitoring and energy consumption calculation, and provide users with more accurate and reliable exercise energy consumption data.

[0004] The technical solutions adopted by the present invention are as follows:

[0005] A method for dynamic optimization of automated test data based on big data, applied to a smart bracelet, comprises the following steps:

[0006] Step 1: Collect the weight data of the user wearing the smart bracelet through the smart bracelet;

[0007] Step 2: Extract the weight characteristics of the user wearing the smart bracelet from the weight data collected by the smart bracelet;

[0008] Step 3: Optimize the calorie calculation model of the smart bracelet according to the weight-bearing characteristics of the user wearing the smart bracelet;

[0009] Step 4: Conduct multiple load tests on the calorie calculation model of the smart bracelet; and dynamically adjust the priority based on the test results;

[0010] Step 5: Build personalized user portraits for smart bracelet wearers and continuously optimize them.

[0011] Furthermore, the load data includes three-axis acceleration and angular velocity data when the user wearing the smart bracelet exercises with a load.

[0012] Furthermore, the step 1 includes the following steps:

[0013] Step 101: Using the built-in motion sensor of the smart bracelet to collect motion data of the user wearing the smart bracelet during weight-bearing exercise;

[0014] Step 102: Using the smart bracelet to synchronously monitor the heart rate of the user wearing the smart bracelet, and establishing a "weight-heart rate reserve" association model corresponding to the user;

[0015] Step 103: The user wearing the smart bracelet manually inputs his or her own weight information through the smart bracelet.

[0016] Furthermore, the motion sensor built into the smart bracelet in step 101 includes an accelerometer and a gyroscope, which are used to collect three-axis acceleration and angular velocity data when the user wearing the smart bracelet exercises with a weight.

[0017] Furthermore, the step 2 includes the following steps:

[0018] Step 201: extracting time domain features based on weight data;

[0019] Step 202: extracting frequency domain features based on the weight data;

[0020] Step 203: Fuse the time domain features, frequency domain features, heart rate data, and weight information input by the user to construct a weight feature vector including multiple dimensions.

[0021] Furthermore, the step 3 includes the following steps:

[0022] Step 301: Based on the Mifflin-St Jeor equation, introduce the load correction coefficient:

[0023] Calorie consumption = (A + B × weight) × exercise time × (1 + C × load characteristic value)

[0024] Among them, A represents basal metabolic equivalent, B=0.1 is the body weight coefficient, the exercise time unit is minutes, and C is the load sensitivity coefficient obtained through big data training;

[0025] Step 302: Train the calorie calculation model.

[0026] Furthermore, step 302 includes the following steps:

[0027] Step A: Collect exercise data from users wearing smart bracelets under different loads and exercise modes, mark the actual calorie consumption, and divide the data into training set, validation set, and test set according to the ratio of 7:2:1;

[0028] Step B: Train the calorie counting model with the Adam optimizer, with the goal of making the mean absolute error less than 8% of the actual consumption.

[0029] Furthermore, the step 4 includes the following steps:

[0030] Step 401: Perform a weight gradient test on the calorie calculation model of the smart bracelet;

[0031] Step 402: Perform boundary condition testing on the calorie calculation model of the smart bracelet;

[0032] Step 403: Perform a multi-factor combination test on the calorie calculation model of the smart bracelet;

[0033] Step 404: Based on historical test data, dynamically adjust the test case priority using a reinforcement learning algorithm.

[0034] Furthermore, the step 5 includes the following steps:

[0035] Step 501: Build a personalized weight-bearing calorie consumption model for the user wearing the smart bracelet based on the user's historical weight-bearing exercise data;

[0036] Step 502: Perform real-time error correction on the calorie consumption predicted by the smart bracelet, and regularly update the personalized weight-bearing calorie consumption model.

[0037] The technical effects achieved by the present invention are:

[0038] The present invention's method for dynamic optimization of automated test data based on big data improves the accuracy of calorie calculation in weight-bearing mode. In the process of monitoring user calorie consumption by a smart bracelet, the user's weight is deeply added as a core calculation factor. By fusing multi-source data such as motion sensors and heart rate to construct a multi-dimensional weight-bearing feature vector, and combining the improved Mifflin-St Jeor equation with the CNN-LSTM model, the weight-bearing state is converted into a quantifiable coefficient, so that the calorie consumption calculation error is strictly controlled within 8%. At the same time, the calculation logic is dynamically adjusted according to the type of weight, weight and exercise mode, realizing deep coupling of weight monitoring and energy consumption calculation, providing users with more accurate and reliable exercise energy consumption data. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is described in detail below with reference to the following examples. It should be understood that the following text is only used to describe one or more specific embodiments of the present invention and does not strictly limit the scope of protection of the present invention.

[0041] like Figure 1 As shown in the figure, a big data-based automated test data dynamic optimization method is applied to smart wearable devices such as smart bracelets and smart watches.

[0042] The automated test data dynamic optimization method includes the following steps:

[0043] Step 1: Collect the weight data of the user wearing the smart bracelet through the smart bracelet;

[0044] Specifically, step 1 includes the following steps:

[0045] Step 101: Using the built-in motion sensor of the smart bracelet to collect motion data of the user wearing the smart bracelet during weight-bearing exercise;

[0046] Here, in step 101, the motion sensor built into the smart bracelet includes at least one of an accelerometer and a gyroscope, and the accelerometer and gyroscope can be used to collect motion data such as three-axis acceleration and angular velocity data when the user wearing the smart bracelet is exercising with a weight. For example, in a weighted running scenario, the vertical acceleration peak and the angular velocity change of the arm swing movement are recorded. The weight can increase the user's vertical acceleration peak by 15%-20% and the root mean square value of the arm swing angular velocity by 10%-15%.

[0047] Step 102: Use the smart bracelet to synchronously monitor the heart rate of the user wearing the smart bracelet and establish a "load-heart rate reserve" correlation model corresponding to the user. When the user carries a load ≥ 5kg, at the same exercise intensity, the heart rate reserve (measured heart rate - resting heart rate / maximum heart rate - resting heart rate) will increase by 8%-12%, thereby assisting in determining the load status.

[0048] Step 103: The user wearing the smart bracelet manually inputs his or her own weight information through the smart bracelet, such as weight, weight type (such as backpack, handheld dumbbells), etc., as labeled data for algorithm training.

[0049] Step 2: Extract the weight characteristics of the user wearing the smart bracelet from the weight data collected by the smart bracelet;

[0050] Specifically, step 2 includes the following steps:

[0051] Step 201: Extract time-domain features based on the load data; for example, calculate the root mean square (RMS) value of the acceleration signal and the impact factor (vertical acceleration peak value / horizontal acceleration average value). As the load increases, the RMS value and impact factor increase significantly, and can serve as key features for load identification.

[0052] Step 202: extracting frequency domain features based on the weight data;

[0053] By analyzing the acceleration signal spectrum through fast Fourier transform (FFT), it is found that weight will cause the energy proportion of low-frequency components (0.5-2Hz) to increase by 10%-15%. The frequency domain features are automatically extracted using a convolutional neural network.

[0054] When a user performs weight-bearing exercise, their movement patterns change, such as heavier steps and more strenuous arm swings. These changes are reflected in the acceleration signal, increasing the energy content of low-frequency components (0.5-2Hz) by 10%-15%. By analyzing this energy change in this interval and combining it with other characteristics, the smart bracelet can help determine whether the user is carrying weight and the approximate weight-bearing status, thereby optimizing calorie consumption calculations.

[0055] Step 203: Fusing the time domain features, frequency domain features, heart rate data, and weight information input by the user to construct a weight feature vector including multiple dimensions.

[0056] Step 3: Optimize the calorie calculation model of the smart bracelet wearer according to the weight-bearing characteristics of the smart bracelet wearer.

[0057] Specifically, step 3 includes the following steps:

[0058] Step 301: Based on the Mifflin-St Jeor equation, introduce the load correction coefficient:

[0059] Calorie consumption = (A + B × weight) × exercise time × (1 + C × load characteristic value)

[0060] Among them, A stands for basal metabolic equivalent, which is a benchmark value used to measure the energy required for the human body to maintain basic life activities (such as breathing and heartbeat).

[0061] B=0.1 is the body weight coefficient, which is used to reflect the weight of the impact of body weight on basic energy consumption. The unit of body weight is kilogram.

[0062] The unit of exercise time is minutes, which indicates the duration of the user's exercise.

[0063] C is the load sensitivity coefficient obtained through big data training, which varies according to different exercise modes.

[0064] For example, in the running scenario, C = 0.06; in the walking scenario, C = 0.04, which reflects the different degrees of influence of weight on calorie consumption in different exercise modes.

[0065] The weight characteristic value is a comprehensive indicator obtained by fusing multi-dimensional information such as motion sensor data and heart rate data, and undergoing a series of weight feature engineering processes such as time domain feature extraction and frequency domain feature analysis. It is used to quantitatively reflect the user's weight status.

[0066] Specifically, the load characteristic value can be calculated using a linear regression model:

[0067] The constructed multi-dimensional load feature vector is used as the independent variable, and the known actual load weight is used as the dependent variable. A large amount of historical data is used to train the linear regression model. After the training is completed, the new feature vector is input into the model. The model assigns corresponding weights to each feature dimension and calculates a value by summing them. This value is the load feature value. For example, if the feature vector is [x1,x2,...,xn] and the corresponding weight is [w1,w2,...,wn], then the load feature value is

[0068] =w1x1+w2x2+...+wn*xn.

[0069] Step 302: training a calorie calculation model;

[0070] Specifically, step 302 includes the following steps:

[0071] Step A: Collect exercise data from users wearing smart bracelets under different loads (0-20 kg, step length 2 kg) and exercise modes (running, walking, climbing), mark the actual calorie consumption (which can be measured by a professional metabolic bike), and divide the data into training set, validation set, and test set according to the ratio of 7:2:1.

[0072] Step B: Train the calorie counting model using the Adam optimizer, with the goal of making the mean absolute error (MAE) less than 8% of actual consumption.

[0073] Among them, the calorie calculation model adopts a lightweight CNN-LSTM network structure. CNN extracts the spatial features of the acceleration signal, and LSTM captures temporal features such as heart rate and exercise duration to output the predicted value of calorie consumption.

[0074] The optimized calorie calculation model must undergo multi-scenario testing to verify its reliability. The test results will in turn drive adjustments to the model parameters, forming a closed loop of 'data collection - model optimization - test iteration'."

[0075] Step 4: Conduct multiple load-bearing tests on the smart bracelet's calorie calculation model; and dynamically adjust the priority based on the test results;

[0076] Specifically, step 4 includes the following steps:

[0077] Step 401: Perform a weight gradient test on the calorie calculation model of the smart bracelet;

[0078] The load gradient test generates virtual test data with a load of 0-20kg (step length of 1kg), combined with different exercise modes (walking on flat ground, climbing, running), to test the calorie calculation accuracy of the smart bracelet when the load changes, detect whether the calculation logic of the model is correct under different load parameters, and whether the output results are reasonable. It also evaluates the responsiveness of the calorie calculation model to load changes and its stability in complex situations, to ensure that the calorie calculation model can accurately handle various load-related calculation tasks.

[0079] Step 402: Perform boundary condition testing on the calorie calculation model of the smart bracelet;

[0080] The boundary condition test injects extreme load data into the calorie calculation model of the smart bracelet, such as a load exceeding 25% of the user's weight, or a sudden increase / decrease of 5kg in the load, to verify the robustness of the bracelet's algorithm and ensure that there is no calculation crash or abnormal jump.

[0081] Boundary condition testing can be performed for extreme load data. This involves injecting unconventional load data into the smart band's calorie calculation model and algorithm, such as setting the load to exceed 25% of the user's body weight. This extreme data verifies whether the band can accurately calculate calorie consumption under loads far exceeding those in daily use scenarios, avoiding calculation crashes, abnormal data jumps, or error prompts, thereby ensuring the reliability of the algorithm under extreme loads.

[0082] Boundary condition testing can be performed in a sudden load change scenario, simulating a sudden increase or decrease of 5kg during exercise. This test verifies the wristband's ability to accurately and promptly detect changes in load, quickly adjust its calorie calculation logic, and output reasonable results. This verifies the wristband's responsiveness and calculation accuracy under dynamic load changes, preventing significant deviations in calorie calculations caused by sudden changes in load.

[0083] Step 403: Perform a multi-factor combination test on the calorie calculation model of the smart bracelet;

[0084] The multi-factor combination test combines the weight with environmental factors (temperature, humidity, altitude) and exercise intensity to generate an orthogonal test matrix. For example, the calorie calculation error of the test user when running with a 10kg load in a high temperature (35°C) and high altitude (3000 meters) environment is tested. This is essentially a simulation of the user performing weight-bearing exercise in a real complex environment, testing the performance of the calorie calculation function of the smart bracelet in a scenario close to actual use, so as to discover possible problems that may arise in the actual application of the product. At the same time, the use of reinforcement learning algorithms based on historical test data to dynamically adjust the test case priority is also to more efficiently simulate the diverse usage scenarios of users and improve the practicality and effectiveness of the test.

[0085] Step 404: Based on historical test data, the reinforcement learning algorithm is used to dynamically adjust the test case priority. If the historical error rate of a load scenario (such as a 15kg hill climb) is high, the test frequency of this scenario is increased; conversely, the test weight of low-error scenarios is reduced to improve test efficiency.

[0086] Step 5: Build personalized user portraits for smart bracelet wearers and continuously optimize them.

[0087] Specifically, step 5 includes the following steps:

[0088] Step 501: Construct a personalized weight-bearing calorie consumption model for the user wearing the smart bracelet based on the user's historical weight-bearing exercise data.

[0089] The personalized weight-bearing calorie-burning model is based on the user's historical weight-bearing exercise data, including sensor data such as acceleration and heart rate during exercise, as well as user-entered weight information and exercise type. Through in-depth analysis of this data, the relationship between each user's unique movement patterns, body metabolic characteristics, and weight load is revealed. For example, a long-term fitness user has strong muscles. Under the same load, their movement deformation is minimal, and their acceleration characteristics do not change significantly, which is a significant difference from the average user. The personalized weight-bearing calorie-burning model learns from these individual differences and establishes unique calculation parameters and logic for each user.

[0090] In practice, when a user performs weight-bearing exercise, the smart bracelet collects real-time data and feeds it into the personalized weight-bearing calorie consumption model. This model then quickly and accurately calculates calorie consumption based on the user's learned individual characteristics. Compared to general calorie calculation models, this personalized model more accurately reflects the user's actual energy expenditure, providing users with more tailored exercise data feedback and health management recommendations.

[0091] Step 502: Perform real-time error correction on the calorie consumption predicted by the smart bracelet, and regularly update the personalized weight-bearing calorie consumption model.

[0092] Among them, the real-time error correction of the calorie consumption predicted by the smart bracelet is to compare the calorie consumption predicted by the smart bracelet with the actual value (obtained through manual input by the user or third-party equipment) after each exercise of the user, and use the Kalman filter algorithm to correct the model parameters in real time to reduce the prediction error of a single exercise.

[0093] The Kalman filter algorithm can fuse real-time sensor data with model predictions to dynamically correct errors caused by sudden changes in load or sensor noise. For example, when the load suddenly increases, the model weights are updated through recursive estimation.

[0094] Regularly update the personalized weight-bearing calorie consumption model to collect new user exercise data every week and retrain the personalized weight-bearing calorie consumption model; retrain the entire model every month. As the user's usage time increases and data accumulates, the accuracy of calorie calculation in weight-bearing scenarios will be continuously improved.

[0095] In summary, this technical solution focuses on improving the accuracy of calorie calculation in weight-bearing mode. When a smart bracelet monitors a user's calorie consumption, it deeply adds the user's weight as a core calculation factor. By fusing multi-source data such as motion sensors and heart rate to construct a multi-dimensional weight-bearing feature vector, and combining the improved Mifflin-St Jeor equation with the CNN-LSTM model, the weight-bearing state is converted into a quantifiable coefficient, strictly controlling the calorie consumption calculation error within 8%. At the same time, the calculation logic is dynamically adjusted according to the type of weight, weight, and exercise mode, achieving a deep coupling of weight monitoring and energy consumption calculation, providing users with more accurate and reliable exercise energy consumption data.

[0096] The foregoing is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained herein shall, unless otherwise specified or limited, be implemented in accordance with conventional means in the art.

Claims

1. A method for dynamic optimization of automated test data based on big data, applied to smart bracelets, characterized by: The following steps are involved: Step 1: Collect the weight data of the user wearing the smart bracelet through the smart bracelet; Step 2: Extract the weight characteristics of the user wearing the smart bracelet from the weight data collected by the smart bracelet; Step 3: Optimize the calorie calculation model of the smart bracelet according to the weight-bearing characteristics of the user wearing the smart bracelet; Step 4: Conduct multiple load tests on the calorie calculation model of the smart bracelet; and dynamically adjust the priority based on the test results; Step 5: Build personalized user portraits for smart bracelet wearers and continuously optimize them.

2. The method for dynamic optimization of automated test data based on big data according to claim 1, characterized in that: The load data includes the three-axis acceleration and angular velocity data of the user wearing the smart bracelet when exercising with a load.

3. The method for dynamic optimization of automated test data based on big data according to claim 1, characterized in that: The step 1 comprises the following steps: Step 101: Using the built-in motion sensor of the smart bracelet to collect motion data of the user wearing the smart bracelet during weight-bearing exercise; Step 102: Using the smart bracelet to synchronously monitor the heart rate of the user wearing the smart bracelet, and establishing a "weight-heart rate reserve" association model corresponding to the user; Step 103: The user wearing the smart bracelet manually inputs his or her own weight information through the smart bracelet.

4. The method for dynamic optimization of automated test data based on big data according to claim 3, characterized in that: The motion sensor built into the smart bracelet in step 101 includes an accelerometer and a gyroscope, which are used to collect three-axis acceleration and angular velocity data when the user wearing the smart bracelet is exercising with a load.

5. The method for dynamic optimization of automated test data based on big data according to claim 4, characterized in that: The step 2 comprises the following steps: Step 201: extracting time domain features based on weight data; Step 202: extracting frequency domain features based on the weight data; Step 203: Fuse the time domain features, frequency domain features, heart rate data, and weight information input by the user to construct a weight feature vector including multiple dimensions.

6. The method for dynamic optimization of automated test data based on big data according to claim 1, characterized in that: The step 3 comprises the following steps: Step 301: Based on the Mifflin-St Jeor equation, introduce the load correction coefficient: Calorie consumption = (A + B × weight) × exercise time × (1 + C × load characteristic value) Among them, A represents basal metabolic equivalent, B=0.1 is the body weight coefficient, the exercise time unit is minutes, and C is the load sensitivity coefficient obtained through big data training; Step 302: Train the calorie calculation model.

7. The method for dynamic optimization of automated test data based on big data according to claim 6, characterized in that: The step 302 includes the following steps: Step A: Collect exercise data from users wearing smart bracelets under different loads and exercise modes, mark the actual calorie consumption, and divide the data into training set, validation set, and test set according to the ratio of 7:2:1; Step B: Train the calorie counting model with the Adam optimizer, with the goal of making the mean absolute error less than 8% of the actual consumption.

8. The method for dynamic optimization of automated test data based on big data according to claim 1, characterized in that: The step 4 comprises the following steps: Step 401: Perform a weight gradient test on the calorie calculation model of the smart bracelet; Step 402: Perform boundary condition testing on the calorie calculation model of the smart bracelet; Step 403: Perform a multi-factor combination test on the calorie calculation model of the smart bracelet; Step 404: Based on historical test data, dynamically adjust the test case priority using a reinforcement learning algorithm.

9. The method for dynamic optimization of automated test data based on big data according to claim 1, characterized in that: The step 5 comprises the following steps: Step 501: Build a personalized weight-bearing calorie consumption model for the user wearing the smart bracelet based on the user's historical weight-bearing exercise data; Step 502: Perform real-time error correction on the calorie consumption predicted by the smart bracelet, and regularly update the personalized weight-bearing calorie consumption model.

Images