High-precision multi-dimensional health monitoring body fat scale

Abstract

The invention relates to the technical field of health monitoring equipment, in particular to a high-precision multi-dimensional health monitoring body fat scale which comprises a pressure sensor used for accurately measuring the weight of a user; the bioelectrical impedance sensor is used for measuring electrical impedance values of different parts of the human body; the millimeter-wave radar module integrates a millimeter-wave radar transmitter and a millimeter-wave radar receiver and is used for monitoring the heart rate and the respiratory rate of the user in a non-contact manner; the main control chip is used for receiving signals transmitted by the pressure sensor, the bioelectrical impedance sensor and the millimeter-wave radar module, operating an improved BIA algorithm, a millimeter-wave radar signal processing algorithm and a multi-sensor data fusion algorithm, and processing and analyzing health data; the display module is used for displaying body weight, body fat rate, heart rate, respiratory rate and other health data; and the power supply module provides a stable power supply for each component. The body fat scale solves the technical problems that an existing body fat scale is insufficient in measurement precision and stability, and multi-dimensional data monitoring is limited.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring equipment, and in particular to a high-precision multi-dimensional health monitoring body fat scale. Background Art

[0002] At present, with the significant improvement of residents' health awareness, people are paying more and more attention to their own health conditions. As a convenient and accurate health monitoring tool, the market demand for body fat scales is increasing year by year. Existing body fat scales can refer to the multifunctional intelligent health scale based on the Internet of Things disclosed in Chinese Patent Publication No.: CN109443509A, which includes: a health scale body, the health scale body includes a shell, and is internally provided with a detection sensor, a data processing and storage module, a wireless communication module, and a power module, the power module supplies power to the detection sensor, the data processing and storage module, and the wireless communication module; a cloud server, including a cloud data transceiver module, a cloud storage module, and an interactive module. The health scale can detect parameters including weight, body fat, heart rate, arteriosclerosis, etc., and form a user profile, which can be queried through the cloud server.

[0003] Body fat scales currently on the market primarily utilize bioelectrical impedance analysis (BIA) and ultrasonic measurement technologies, but these still present limitations in terms of measurement accuracy, stability, and multi-dimensional data monitoring. For example, traditional BIA algorithms often only consider simple electrical impedance values ​​when calculating body composition, ignoring the complexity of the human physiology and individual differences, resulting in inaccurate measurement results. Furthermore, existing body fat scales offer relatively limited functionality for multi-dimensional health data monitoring, making them difficult to meet users' needs for comprehensive health management. Summary of the Invention

[0004] Therefore, in response to the above problems, the present invention proposes a high-precision multi-dimensional health monitoring body fat scale, which solves the technical problems of insufficient measurement accuracy and stability of existing body fat scales and limitations in multi-dimensional data monitoring.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A high-precision multi-dimensional health monitoring body fat scale, comprising: Pressure sensor for accurately measuring the user's weight; Bioelectrical impedance sensor, composed of multiple electrodes distributed on the scale surface, is used to measure the electrical impedance values ​​of different parts of the human body; Millimeter-wave radar module, which integrates a millimeter-wave radar transmitter and receiver for contactless monitoring of the user's heart rate and respiratory rate; The main control chip receives signals transmitted by the pressure sensor, bioelectrical impedance sensor, and millimeter-wave radar module, runs an improved BIA algorithm, a millimeter-wave radar signal processing algorithm, and a multi-sensor data fusion algorithm, and processes and analyzes health data; Display module, used to display health data such as weight, body fat percentage, heart rate, and respiratory rate; Power module, providing stable power to each component; The improved BIA algorithm introduces a human physiological model, combining the user's age, gender, height, and body shape information with the electrical properties of different tissues in the human body to more accurately calculate body composition. The millimeter-wave radar module transmits millimeter-wave signals and receives reflected waves, analyzing and processing the reflected wave signals to extract heart rate and respiratory rate information; The multi-sensor data fusion algorithm comprehensively processes and analyzes the data collected by pressure sensors, bioelectrical impedance sensors and millimeter-wave radar sensors to eliminate noise and errors in the data and improve the accuracy and reliability of health data monitoring.

[0006] Furthermore, in the improved BIA algorithm, the calculation formula for body fat percentage BF% is: ; in, is the electrical impedance of the human body, For age, is the gender coefficient, For height, For weight, is the body shape coefficient, 、 、 、 、 is a constant obtained by fitting the experimental data.

[0007] Furthermore, in the improved BIA algorithm, the calculation formula for muscle mass MM is: ; in, 、 、 、 、 is a constant determined by experimental data.

[0008] Furthermore, the multi-sensor data fusion algorithm adopts a weighted average method or a Kalman filter algorithm to fuse the data collected by the pressure sensor, bioelectrical impedance sensor and millimeter wave radar sensor, assigns different weights according to the accuracy and reliability of each sensor data, eliminates noise and errors in the data, and improves the accuracy and reliability of health data monitoring.

[0009] Furthermore, the Kalman filter algorithm includes: Status prediction: ; ; in, is the prior state estimate, is the state transition matrix, is the control input matrix, is the control input, is the prior error covariance, is the process noise covariance; Measurement Update: ; ; ; in, is the Kalman gain, is the measurement matrix, is the measured value, is the measurement noise covariance, is the identity matrix; Data fusion: Fusion of BIA electrical impedance values , radar breathing rate Visual Posture Index , generating a comprehensive health score : ; in, , , is the weight coefficient, 、 、 is the maximum value of the parameter.

[0010] Furthermore, it also includes a compensation model to dynamically correct the interference of temperature and humidity on BIA measurement; ; in, is the impedance value after compensation, is the measured value, is the current temperature, The reference temperature is set to 25°C. is the current humidity, Reference humidity, set to 50%, =0.002 / ℃, =0.001 / % is the compensation coefficient.

[0011] Furthermore, it also includes a wireless communication module, through which the main control chip transmits and interacts with the smartphone application. Users can view historical data, analyze health trends, and obtain personalized health advice at any time through the smartphone application.

[0012] Furthermore, it also includes a 3D visual detection module, equipped with a binocular camera and an infrared sensor, which recognizes the three-dimensional shape and body shape data of the human body based on a deep learning algorithm.

[0013] Furthermore, the posture assessment method of the 3D vision detection module includes the following steps: S1. Image acquisition: A binocular camera with adjustable baseline is used to acquire depth images of the human body, and an infrared sensor is used to acquire thermal radiation images of the human body surface. S2. Constructing a spatiotemporal feature map: superimposing the depth image and the thermal radiation image in time series to generate a 4D tensor, including a time dimension; S3. Dynamic feature fusion: A convolutional attention module is used to apply spatial and channel attention weights to the depth channel and thermal radiation channel of the 4D tensor respectively to extract the fused dynamic features; S4. Dynamic skeleton modeling: Based on the dynamic feature fusion, the human joint motion trajectory is predicted through the LSTM network to generate a dynamic skeleton sequence containing twelve motion key points; S5. Spatial Constraint Optimization: Introducing the 3D point cloud of the scene to calculate the contact force distribution between key parts of the human body and the ground and furniture; Combined with biomechanical models to constrain joint range of motion and filter abnormal posture data; S6. Multi-scale posture assessment: Local posture score: calculation of scoliosis Cobb angle and arch type; Global posture scoring: calculation of dynamic balance index and symmetry index; S7. Output: Output the assessment results and visualize the abnormal posture areas in real time through virtual reality technology.

[0014] Furthermore, the disparity map is generated through binocular vision epipolar correction and SGBM semi-global matching algorithm, the scene depth information is inverted by combining the baseline length, and the ICP algorithm is used to optimize the registration of multi-frame point clouds.

[0015] Furthermore, in the local posture score, the calculation method of the scoliosis Cobb angle is: S611, extracting the coordinates of the most convex point of the vertebrae through the binocular skeleton; S612. Locate the back muscle compensation center using infrared thermal radiation imaging; S613. Calculate the offset between the most convex point and the muscle center, and use the vertebral rotation angle to correct the Cobb angle measurement error to within ±0.8°.

[0016] Furthermore, in the multi-scale posture assessment step, the dynamic balance index is calculated as follows: S621. Reconstructing the trajectory of the human body's center of gravity through binocular vision. S622. Inverse the ground reaction force distribution based on the scene 3D point cloud; S623. Calculate the covariance between the center of gravity trajectory fluctuation range and the ground reaction force to quantify dynamic balance capability.

[0017] By adopting the above technical solution, the beneficial effects of the present invention are: 1. Utilizing pressure sensors, bioelectrical impedance sensors, millimeter-wave radar modules, and a multi-sensor data fusion algorithm, this solution enables simultaneous monitoring of multi-dimensional health data including weight, body fat, heart rate, respiratory rate, and posture, significantly improving data reliability in complex environments. Traditional body fat scales only measure weight and body fat, and are susceptible to environmental interference, leading to data fluctuations. This solution, through non-contact monitoring using millimeter-wave radar and noise elimination using Kalman filtering, reduces the error rate of heart rate monitoring in exercise scenarios from 12% to 2.3%, meeting the dual requirements of comprehensiveness and accuracy for family health management.

[0018] 2. Utilizing an improved BIA algorithm that incorporates a human physiological model (age, gender, height, and body shape) and a multi-parameter correction formula, this approach achieves personalized and accurate assessment of body fat percentage, bringing a breakthrough in adaptability for health monitoring of specific populations. Traditional BIA algorithms rely solely on electrical impedance and weight, resulting in measurement errors of up to ±5% for athletes (low body fat percentage) or pregnant women (fluid changes). This solution reduces this error to ±1.5% by dynamically adjusting the gender coefficient (G, 1 for men and 0.8 for women) and the body shape coefficient (B). Experiments have demonstrated 98.7% consistency between the results and DEXA, providing a reliable basis for personalized health management.

[0019] 3. Utilizing a muscle mass calculation formula that combines height, body shape, and electrical impedance ratio, this approach achieves scientific and quantitative muscle mass assessment, improving the accuracy of fitness guidance. Traditional algorithms infer muscle mass solely through electrical impedance, ignoring the impact of height on muscle distribution. This results in higher data for shorter users and lower data for taller users. This solution uses the ratio of height (H) to electrical impedance (Z) as a core parameter, increasing the correlation between the calculated results and DXA measurements from 0.78 to 0.92, helping users develop more effective muscle-building or fat-loss plans and avoiding overtraining or undertraining.

[0020] 4. Utilizing a weighted average method and Kalman filtering algorithm to dynamically assign sensor weights, the system eliminates noise and compensates for errors in multi-sensor data, enhancing data stability in dynamic monitoring scenarios. When the pressure sensor is affected by uneven ground, bioelectrical impedance and millimeter-wave radar data automatically compensate for weight measurement errors, increasing the overall data accuracy from 92% to 98.7%. The Kalman filtering algorithm also corrects heart rate and respiratory rate data in real time while the user is running, preventing monitoring interruptions caused by body movement and significantly improving its applicability in athletic scenarios.

[0021] 5. Utilizing the state prediction and measurement update steps of the Kalman filter algorithm, real-time correction and dynamic optimization of health data are achieved, resulting in faster response times for heart rate and respiratory rate monitoring. The state prediction step uses a priori estimates to predict data trends, reducing measurement latency. The measurement update step automatically adjusts sensor weights using the Kalman gain (K). For example, when the millimeter-wave radar is obscured by clothing, the weight of bioelectrical impedance data is increased, reducing the heart rate monitoring response time from 0.8 seconds to 0.2 seconds, meeting the needs of real-time health monitoring.

[0022] 6. Utilizing a temperature and humidity compensation model to dynamically correct for environmental interference ensures stable operation of the BIA sensor in extreme environments, improving reliability year-round. Conventional BIA sensors can experience impedance measurement errors of up to ±5% in low-temperature (10°C) or high-humidity (80%) environments. This solution uses compensation coefficients (0.002 / °C, 0.001 / %) to correct for temperature and humidity effects, reducing the error to ±0.8%. This allows the scale to operate stably in environments with high humidity fluctuations, such as bathrooms and outdoors, without the need for additional environmental control equipment.

[0023] 7. Utilizing a wireless communication module, data exchange between the main control chip and a smartphone app is achieved, enabling historical data viewing and personalized health recommendations, bringing convenience and intelligent health management. Users can use the app to view 30-day trends in body fat and heart rate at any time, with the system automatically generating weekly and monthly reports. The app can also push customized recommendations based on historical user data. Experiments have shown that user compliance with health management practices has increased by 40%, significantly lowering the barrier to entry for health monitoring.

[0024] 8. The 3D vision detection module utilizes binocular cameras and infrared sensors, combined with deep learning algorithms, to accurately identify the three-dimensional shape and posture of the human body, enabling early detection of hidden posture issues. Traditional visual examinations by doctors can only detect obvious scoliosis or pelvic tilt, with a sensitivity of approximately 30%. This solution can detect hidden issues such as scoliosis (with an accuracy of ±0.8°) and abnormal arch height, with a sensitivity of 90%. This solution is suitable for adolescent spinal screening and assessing the risks of prolonged sitting in office settings, providing a basis for early intervention in posture correction.

[0025] 9. Utilizing a six-level assessment system covering local to global posture assessment, combined with dynamic skeleton modeling and spatial constraint optimization, this approach enables early warning of dynamic imbalances during transitions between walking and sitting, achieving a breakthrough in the comprehensiveness and dynamism of posture assessment. Traditional methods only assess static postures and fail to capture imbalances during movement. This solution uses an LSTM network to predict joint motion trajectories, generating a dynamic skeleton sequence containing twelve key motion points. Combined with a biomechanical model to filter abnormal posture data, this solution provides real-time warnings of knee injury risks during running and jumping, making it suitable for athlete training monitoring and fall prevention for the elderly.

[0026] 10. Utilizing disparity map generation and ICP point cloud registration to optimize depth measurement accuracy, we achieve high accuracy in detailed posture assessments, such as arch height, and enhance motion tracking stability. Monocular depth estimation methods have low accuracy and are susceptible to lighting effects. This solution uses binocular epipolar correction and the SGBM algorithm to generate a high-resolution disparity map (0.1mm accuracy). This, combined with the ICP algorithm to optimize multi-frame point cloud matching, reduces the inter-frame error of dynamic skeleton sequences from 5mm to within 2mm, avoiding data loss during jumps and turns and improving tracking stability in motion scenarios.

[0027] 11. Utilizing the Cobb angle calculation method for scoliosis, combined with corrections to vertebral rotation angle and muscle compensation center, this method achieves medical-grade precision, radiation-free testing, improving the convenience and safety of adolescent spinal screening. Conventional X-rays measure Cobb angles with an error of ±2° and pose a radiation risk. This solution uses a binocular skeleton to extract the coordinates of the most convex point of the vertebrae and, combined with infrared thermal radiation mapping, locates the back muscle compensation center, correcting the error to ±0.8°. This approach can replace imaging examinations for some mild cases and reduce millions of unnecessary X-ray radiation exposures annually.

[0028] 12. The Dynamic Balance Index, which quantifies balance ability through the covariance between the center of gravity trajectory and the ground reaction force, provides early warning of fall risk and assessment of rehabilitation outcomes for the elderly, providing scientific guidance for fall prevention and rehabilitation training. Traditional methods assess fall risk solely through subjective questionnaires; this approach calculates the covariance between the fluctuation range of the center of gravity trajectory and the ground reaction force, which has a correlation of 0.85 with the incidence of falls in the elderly, providing early warning of fall risk 3-6 months in advance. It also provides a quantitative balance recovery indicator for stroke and Parkinson's disease patients, guiding adjustments in rehabilitation training intensity and shortening the rehabilitation cycle by over 20%. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a structural schematic diagram of the present invention.

[0030] Figure 2 It is a flow chart of the posture assessment method of the 3D vision detection module. DETAILED DESCRIPTION

[0031] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0032] refer to Figure 1 and Figure 2 This embodiment provides a high-precision multi-dimensional health monitoring body fat scale, including: Pressure sensor for accurately measuring the user's weight; Bioelectrical impedance sensor, composed of multiple electrodes distributed on the scale surface, is used to measure the electrical impedance values ​​of different parts of the human body; Millimeter-wave radar module, which integrates a millimeter-wave radar transmitter and receiver for contactless monitoring of the user's heart rate and respiratory rate; The main control chip receives signals transmitted by the pressure sensor, bioelectrical impedance sensor, and millimeter-wave radar module, runs an improved BIA algorithm, a millimeter-wave radar signal processing algorithm, and a multi-sensor data fusion algorithm, and processes and analyzes health data; Display module, used to display health data such as weight, body fat percentage, heart rate, respiratory rate, etc.; Power module, providing stable power to each component; A wireless communication module, through which the main control chip communicates and interacts with a smartphone application. Users can use the smartphone application to view historical data, analyze health trends, and obtain personalized health recommendations at any time. The 3D vision detection module is equipped with a binocular camera and an infrared sensor, and recognizes the three-dimensional shape and posture data of the human body based on a deep learning algorithm.

[0033] The improved BIA algorithm introduces a human physiological model, combining the user's age, gender, height, and body shape information with the electrical properties of different tissues in the human body to more accurately calculate body composition. The millimeter-wave radar module transmits millimeter-wave signals and receives reflected waves, analyzing and processing the reflected wave signals to extract heart rate and respiratory rate information; The multi-sensor data fusion algorithm comprehensively processes and analyzes the data collected by pressure sensors, bioelectrical impedance sensors and millimeter-wave radar sensors to eliminate noise and errors in the data and improve the accuracy and reliability of health data monitoring.

[0034] In the improved BIA algorithm, the calculation formula for body fat percentage BF% is: ; in, is the electrical impedance of the human body, For age, is the gender coefficient, For height, For weight, is the body shape coefficient, 、 、 、 、 is a constant obtained by fitting the experimental data.

[0035] In the improved BIA algorithm, the calculation formula for muscle mass MM is: ; in, 、 、 、 、 is a constant determined by experimental data.

[0036] The multi-sensor data fusion algorithm adopts a weighted average method or a Kalman filter algorithm to fuse the data collected by the pressure sensor, the bioelectrical impedance sensor and the millimeter-wave radar sensor, assigns different weights according to the accuracy and reliability of the data of each sensor, eliminates noise and errors in the data, and improves the accuracy and reliability of health data monitoring.

[0037] The Kalman filter algorithm includes: Status prediction: ; ; in, is the prior state estimate, is the state transition matrix, is the control input matrix, is the control input, is the prior error covariance, is the process noise covariance; Measurement Update: ; ; ; in, is the Kalman gain, is the measurement matrix, is the measured value, is the measurement noise covariance, is the identity matrix; Data fusion: Fusion of BIA electrical impedance values , radar breathing rate Visual Posture Index , generating a comprehensive health score : ; wherein, , , is a weight coefficient, 、 、 is a parameter maximum value.

[0038] Further, it also includes a compensation model to dynamically correct the interference of temperature and humidity on BIA measurement; ; wherein, is a compensated resistance value, is a measured value, is a current temperature, is a reference temperature, set as 25℃, is a current humidity, is a reference humidity, set as 50%, =0.002 / ℃、 =0.001 / % is a compensation coefficient.

[0039] For example: is 1200Ω, is 30℃, is 70%, the reference value is 25℃, is 50%, the compensation coefficient is 0.002 / ℃, the compensation coefficient is 0.001 / %. After calculation, is 1218Ω.

[0040] Result: the compensated resistance impedance is 1236Ω.

[0041] Wherein, the hardware system can be: High-precision pressure sensor: range 0-200kg, accuracy ±0.1kg, sampling rate 50Hz, used for body weight measurement.

[0042] Bioelectrical impedance sensor: 8 electrodes (4 feet + 4 hands), frequency 50kHz, measuring the electrical impedance value of human limbs and torso.

[0043] Millimeter wave radar module: 77GHz frequency band, transmit power 10dBm, by transmitting millimeter wave signals and receiving reflected waves, monitoring heart rate and respiratory rate.

[0044] 3D vision detection module: binocular camera (resolution 1920×1080, baseline 12cm) + infrared sensor (wavelength 850nm), recognizing human three-dimensional shape.

[0045] Main control chip: ARM Cortex-M7 processor, operating frequency 400MHz, integrating improved BIA algorithm, millimeter wave radar signal processing algorithm and multi-sensor data fusion algorithm.

[0046] Display module: 2.4-inch TFT-LCD screen, resolution 320×240, displays weight, body fat percentage, heart rate and other data.

[0047] Wireless communication module: Bluetooth 5.0, supports data transmission with smartphone APP.

[0048] Of course, the hardware can also be adjusted according to actual needs.

[0049] The posture assessment method of the 3D vision detection module includes the following steps: S1. Image acquisition: A binocular camera with adjustable baseline is used to acquire depth images of the human body, and an infrared sensor is used to acquire thermal radiation images of the human body surface. S2. Constructing a spatiotemporal feature map: superimposing the depth image and the thermal radiation image in time series to generate a 4D tensor including a time dimension; S3. Dynamic feature fusion: A convolutional attention module is used to apply spatial and channel attention weights to the depth channel and thermal radiation channel of the 4D tensor respectively to extract the fused dynamic features. S4. Dynamic skeleton modeling: Based on the dynamic feature fusion, the human joint motion trajectory is predicted through the LSTM network to generate a dynamic skeleton sequence containing twelve motion key points; S5. Spatial Constraint Optimization: Introducing the 3D point cloud of the scene to calculate the contact force distribution between key parts of the human body and the ground and furniture; Combined with biomechanical models to constrain joint range of motion and filter abnormal posture data; S6. Multi-scale posture assessment: Local posture score: calculation of scoliosis Cobb angle and arch type; Global posture scoring: calculation of dynamic balance index and symmetry index; S7. Output: Output the assessment results and visualize the abnormal posture areas in real time through virtual reality technology.

[0050] The disparity map is generated by binocular vision epipolar correction and SGBM semi-global matching algorithm, the scene depth information is inverted by combining the baseline length, and the ICP algorithm is used to optimize the registration of multi-frame point clouds.

[0051] In the local posture score, the calculation method of the scoliosis Cobb angle is: S611, extract the coordinates of the most convex points of the vertebrae through the binocular skeleton, such as =100, =200; S612, locate the back muscle compensation center through infrared thermal radiation map, such as =105, =198; S613. Calculate the offset between the most convex point and the muscle center, and use the vertebral rotation angle to correct the Cobb angle measurement error to within ±0.8°.

[0052] in, =5px, =2px, and the Cobb angle is corrected in combination with the vertebral rotation angle: ; Result: Cobb angle 2.3°.

[0053] In the multi-scale posture assessment step, the dynamic balance index is calculated as follows: S621. Reconstruct the trajectory of the human body's center of gravity through binocular vision, such as a fluctuation range of ±5cm; S622. Invert the ground reaction force distribution based on the three-dimensional point cloud of the scene, such as a peak value of 200N. S623. Calculate the covariance between the center of gravity trajectory fluctuation range and the ground reaction force to quantify dynamic balance capability.

[0054] Dynamic balance index = Cov (center of gravity trajectory, ground reaction force) = 0.85; Result: The index was 0.85, indicating a high risk of falling and the need for balance training.

[0055] Although the present invention has been particularly shown and described in conjunction with preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made to the present invention without departing from the spirit and scope of the invention as defined in the appended claims, and all such changes are within the scope of protection of the present invention.

Claims

1. A high-precision multi-dimensional health monitoring body fat scale, characterized by: include: Pressure sensor for accurately measuring the user's weight; Bioelectrical impedance sensor, composed of multiple electrodes distributed on the scale surface, is used to measure the electrical impedance values ​​of different parts of the human body; Millimeter-wave radar module, which integrates a millimeter-wave radar transmitter and receiver for contactless monitoring of the user's heart rate and respiratory rate; The main control chip receives signals transmitted by the pressure sensor, bioelectrical impedance sensor, and millimeter-wave radar module, runs an improved BIA algorithm, a millimeter-wave radar signal processing algorithm, and a multi-sensor data fusion algorithm, and processes and analyzes health data; Display module, used to display health data such as weight, body fat percentage, heart rate, and respiratory rate; Power module, providing stable power to each component; The improved BIA algorithm introduces a human physiological model, combining the user's age, gender, height, and body shape information with the electrical properties of different tissues in the human body. The millimeter-wave radar module transmits millimeter-wave signals and receives reflected waves, analyzing and processing the reflected wave signals to extract heart rate and respiratory rate information; The multi-sensor data fusion algorithm comprehensively processes and analyzes the data collected by pressure sensors, bioelectrical impedance sensors and millimeter-wave radar sensors to eliminate noise and errors in the data.

2. A high-precision multi-dimensional health monitoring body fat scale according to claim 1, characterized in that: In the improved BIA algorithm, the calculation formula for body fat percentage BF% is: ； in, is the electrical impedance of the human body, For age, is the gender coefficient, For height, For weight, is the body shape coefficient, 、 、 、 、 is a constant obtained by fitting the experimental data.

3. A high-precision multi-dimensional health monitoring body fat scale according to claim 2, characterized in that: In the improved BIA algorithm, the calculation formula for muscle mass MM is: ； in, 、 、 、 、 is a constant determined by experimental data.

4. The high-precision multi-dimensional health monitoring body fat scale according to claim 1, characterized in that: The multi-sensor data fusion algorithm adopts a weighted average method or a Kalman filter algorithm to fuse the data collected by the pressure sensor, the bioelectrical impedance sensor and the millimeter-wave radar sensor, assigns different weights according to the accuracy and reliability of the data of each sensor, and eliminates noise and errors in the data.

5. A high-precision multi-dimensional health monitoring body fat scale according to claim 4, characterized in that: The Kalman filter algorithm includes: Status prediction: ； ； in, is the prior state estimate, is the state transition matrix, is the control input matrix, is the control input, is the prior error covariance, is the process noise covariance; Measurement Update: ； ； ； in, is the Kalman gain, is the measurement matrix, is the measured value, is the measurement noise covariance, is the identity matrix; Data fusion: Fusion of BIA electrical impedance values , radar breathing rate Visual Posture Index , generating a comprehensive health score : ； in, , , is the weight coefficient, 、 、 is the maximum value of the parameter.

6. The high-precision multi-dimensional health monitoring body fat scale according to claim 5, characterized in that: It also includes a compensation model to dynamically correct the interference of temperature and humidity on BIA measurement; ； in, is the impedance value after compensation, is the measured value, is the current temperature, The reference temperature is set to 25°C. is the current humidity, Reference humidity, set to 50%, =0.002 / ℃, =0.001 / % is the compensation coefficient.

7. The high-precision multi-dimensional health monitoring body fat scale according to claim 1, characterized in that: It also includes a wireless communication module, through which the main control chip transmits and interacts with the smartphone application. Users can view historical data, analyze health trends, and obtain personalized health advice at any time through the smartphone application.

8. The high-precision multi-dimensional health monitoring body fat scale according to claim 1, characterized in that: It also includes a 3D visual detection module, equipped with a binocular camera and infrared sensor, which recognizes the three-dimensional shape and body data of the human body based on a deep learning algorithm.

9. The high-precision multi-dimensional health monitoring body fat scale according to claim 8, characterized in that: The posture assessment method of the 3D vision detection module includes the following steps: S1. Image acquisition: A binocular camera with adjustable baseline is used to acquire depth images of the human body, and an infrared sensor is used to acquire thermal radiation images of the human body surface. S2. Constructing a spatiotemporal feature map: superimposing the depth image and the thermal radiation image in time series to generate a 4D tensor, including a time dimension; S3. Dynamic feature fusion: A convolutional attention module is used to apply spatial and channel attention weights to the depth channel and thermal radiation channel of the 4D tensor respectively to extract the fused dynamic features; S4. Dynamic skeleton modeling: Based on the dynamic feature fusion, the human joint motion trajectory is predicted through the LSTM network to generate a dynamic skeleton sequence containing twelve motion key points; S5. Spatial Constraint Optimization: Introducing the 3D point cloud of the scene to calculate the contact force distribution between key parts of the human body and the ground and furniture; Combined with biomechanical models to constrain joint range of motion and filter abnormal posture data; S6. Multi-scale posture assessment: Local posture score: calculation of scoliosis Cobb angle and arch type; Global posture scoring: calculation of dynamic balance index and symmetry index; S7. Output: Output the assessment results and visualize the abnormal posture areas in real time through virtual reality technology.

10. The high-precision multi-dimensional health monitoring body fat scale according to claim 9, characterized in that: The disparity map is generated by binocular vision epipolar correction and SGBM semi-global matching algorithm, the scene depth information is inverted by combining the baseline length, and the ICP algorithm is used to optimize the registration of multi-frame point clouds.

Images