Health monitoring method, system and equipment based on flexible plantar pressure array and medium

Abstract

The invention belongs to the technical field of health monitoring, and discloses a health monitoring method, system and device based on a flexible plantar pressure array and a medium, and the method comprises the steps: collecting the dynamic pressure data of each functional region of the plantar of a user based on the flexible plantar pressure array; extracting biomechanical characteristic parameters of the dynamic pressure data to obtain foot health assessment characteristics and weight prediction characteristics; based on the foot health assessment features, the foot arch type and uneven stress quantitative data are judged, and a foot health assessment result is obtained; inputting the body weight prediction features into a body weight prediction model for prediction, and outputting a body weight prediction result; wherein the body weight prediction model is constructed based on a random forest model and an XGBoost model; and performing health monitoring on the user based on the foot health assessment result and the weight prediction result. According to the technical scheme, objective assessment of foot health and daily noninvasive weight prediction can be carried out, and integrated monitoring is achieved.

Description

Technical Field

[0001] This invention belongs to the field of health monitoring technology, and in particular relates to a health monitoring method, system, device and medium based on a flexible plantar pressure array. Background Technology

[0002] Young people (especially college students) are in a period of rapid physical development, experiencing frequent weight fluctuations. Furthermore, due to sedentary lifestyles and insufficient exercise, the incidence of foot health problems (such as collapsed arches and uneven foot pressure) and obesity is significantly increased. Studies show that the overall incidence of foot pain among adult educators is as high as 74%, with approximately 40% suffering from moderate to severe pain. Without timely intervention, this can easily develop into chronic conditions such as flat feet and plantar fasciitis, and even trigger lower limb musculoskeletal problems.

[0003] Traditional foot health assessments rely on doctors' subjective observation or imaging examinations such as X-rays and CT scans, which have limitations such as large size, high cost, risk of ionizing radiation, and lack of real-time capability, making it difficult to meet the needs of daily non-invasive monitoring. Weight, as a core indicator of health management, is traditionally monitored using specialized scales, which are limited by the scenario and cannot achieve continuous monitoring; existing non-invasive weight prediction methods (such as bioelectrical impedance analysis and human morphological characteristics) are easily interfered with, cumbersome to operate, and have large errors (bioelectrical impedance analysis error ±8.81kg).

[0004] Furthermore, high-quality labeled datasets in the field of foot pressure are scarce. Existing publicly available data are mostly small-sample, static data, lacking dynamic gait sequences, health status annotations, and physiological ground truth values, which restricts algorithmic innovation and cross-disciplinary research comparisons. Existing technologies mostly focus on single tasks (foot health assessment only or weight prediction only), failing to form an integrated solution and making it difficult to meet users' comprehensive health monitoring needs.

[0005] To this end, this invention proposes a multi-task scheme based on a flexible plantar pressure array, which simultaneously achieves objective assessment of foot health and daily non-invasive weight prediction, filling a gap in the existing technology. Summary of the Invention

[0006] The purpose of this invention is to provide a health monitoring method, system, device, and medium based on a flexible plantar pressure array to solve the problems existing in the prior art.

[0007] To achieve the above objectives, the present invention provides a health monitoring method based on a flexible plantar pressure array, comprising: Dynamic pressure data of the user's foot is collected by a flexible plantar pressure array, which includes 48 sensing points per foot, covering four functional areas: heel, midfoot, forefoot, and toes. Foot health assessment features and weight prediction features are extracted based on the plantar dynamic pressure data. The foot health assessment features include COP trajectory area, X-direction offset standard deviation, Y-direction offset standard deviation, and foot arch sensor activation ratio. The weight prediction features include baseline pressure features, trajectory features, zone pressure features, pressure density features, activated sensor features, dynamic features, symmetry features, and maximum pressure features. A quantitative assessment model is constructed based on the foot health assessment characteristics, and the results of the arch type determination and the quantitative data of uneven force distribution are output. An RF-XGBoost fusion model is constructed based on the weight prediction features to output non-invasive weight prediction results.

[0008] Optionally, the step of collecting dynamic plantar pressure data of the user's foot through a flexible plantar pressure array specifically includes: The flexible plantar pressure array uses a polyimide flexible film substrate with a thickness of <0.3mm, a single sensing point pressure range of 0~50kg, and a response time of <20ms. Data is transmitted via a Bluetooth data acquisition device using the BLE 5.2 protocol. The maximum sampling frequency for a single foot is 50Hz, and the maximum sampling frequency for both feet is 20Hz. The device simultaneously records timestamps, sensor number, pressure values, subject IDs, and health status information.

[0009] Optionally, the step of extracting foot health assessment features based on the plantar dynamic pressure data specifically includes: The plantar dynamic pressure data were preprocessed using eight-connected region filtering, outlier removal, and signal normalization. Calculate the COP coordinates of the pressure center, and obtain the COP coordinates at each time point t using the pressure weighted average method; Four foot health assessment features were calculated based on the COP coordinates: The COP trajectory area reflects the stability of the foot arch's center of gravity, and is calculated by the area enclosed by the COP coordinate movement trajectory. The standard deviation of the X-direction offset reflects the lateral balance function of the arch of the foot. The standard deviation is calculated by the offset of the COP coordinate in the lateral (left-right) direction of the foot. The standard deviation of the Y-direction reflects the longitudinal cushioning and support synergy of the arch of the foot. The standard deviation is calculated by the offset of the COP coordinate in the longitudinal (anteroposterior) direction of the foot. The activation rate of sensors in the arch area directly reflects the support integrity of the arch. It is determined by the percentage of sensors with an effective pressure value > 0 in the arch area.

[0010] Optionally, the formula for calculating the COP coordinates of the pressure center is: ; ; In the formula, , Let be the coordinates of the i-th sensor. Let be the pressure value of the i-th sensor at time t.

[0011] Optionally, the step of constructing a quantitative assessment model based on the foot health assessment characteristics specifically includes: Four foot health assessment features are input into a preset threshold judgment model, and the arch type is judged into two categories: normal arch and collapsed arch. When the COP trajectory area exceeds the first threshold, the standard deviation of the X-direction offset exceeds the second threshold, the standard deviation of the Y-direction offset exceeds the third threshold, or the activation ratio of the foot arch sensor exceeds the fourth threshold, it is judged as a data anomaly. Two data points combined were deemed abnormal, indicating a risk of arch collapse. The three abnormal data points indicate foot arch collapse. The uniformity of plantar pressure distribution is analyzed by quantitative analysis of the force unevenness coefficient. The force unevenness coefficient is calculated by the ratio of the standard deviation of pressure in each region to the average pressure, and the quantitative error is ≤3.2%.

[0012] Optionally, the step of extracting weight prediction features based on the plantar dynamic pressure data specifically includes 15 features: Basic pressure characteristics: average total pressure, average total pressure of the right foot, average total pressure of the left foot, and pressure ratio between the left and right feet; Trajectory characteristics: mean COP distance, COP distance stability (standard deviation); Zone pressure characteristics: right heel pressure ratio; Pressure density characteristics: Pressure density of the right foot; Activated sensor characteristics: number of sensors activated by the right foot, number of sensors activated by the left foot; Dynamic characteristics: pressure fluctuations (standard deviation), mean pressure change rate; Symmetry characteristics: Symmetry of pressure on both feet; Maximum pressure characteristics: maximum pressure on the right foot, maximum pressure on the left foot.

[0013] Optionally, the construction of the RF-XGBoost fusion model based on the weight prediction features specifically includes: Construct a random forest regression model containing 20 decision trees, setting the maximum number of splits for each tree to 15 and the minimum number of leaf samples to 15, and output the first predicted value; Construct an XGBoost regression model with 40 weak learners and a learning rate of 0.1, and output the second predicted value; A dynamic weight fusion mechanism is designed based on the mean absolute error (MAE) of the validation set, and the calculation formula is as follows: ; ; ; In the formula, , These are the weights for the Random Forest model and the XGBoost model, respectively. The first predicted value, This is the second predicted value. This is the predicted final weight.

[0014] Optionally, data preprocessing procedures may also be included: Remove unsteady-state data during the weighing and unweighing phases, and remove 30 frames from the beginning and end of each phase. A sliding window of 30 frames was used to detect steady-state data segments, and 30 consecutive frames of data whose total pressure was closest to the average of three measurements were selected. Outliers are identified through a consistency test. If the relative difference between a measurement and the mean of the other two measurements exceeds 30%, it is considered an outlier and removed. A robust average is performed on the normal data to generate a representative data file containing 30 frames of steady-state pressure sequences.

[0015] The present invention also provides a health monitoring system based on a flexible plantar pressure array, comprising: The data acquisition module is used to collect dynamic pressure data of various functional areas of the user's sole based on the flexible plantar pressure array; The feature extraction module is used to extract the biomechanical feature parameters of the dynamic pressure data to obtain foot health assessment features and weight prediction features; The health monitoring module is used to determine the arch type and quantify the uneven force distribution based on the foot health assessment features to obtain the foot health assessment results; input the weight prediction features into the weight prediction model for prediction and output the weight prediction results; wherein, the weight prediction model is constructed based on the random forest model and the XGBoost model; and perform health monitoring on the user based on the foot health assessment results and the weight prediction results.

[0016] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the health monitoring method based on a flexible plantar pressure array.

[0017] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned health monitoring method based on a flexible plantar pressure array.

[0018] The technical effects of this invention are as follows: This invention is the first to achieve the synergy between foot health assessment and weight prediction. No additional equipment is required; dual-task monitoring can be completed simply by embedding a flexible pressure insole in the shoe, meeting the needs of daily use and being imperceptible, and realizing integrated monitoring.

[0019] This invention uses a flexible and lightweight sensor (thickness < 0.3mm), which fits the sole of the foot without any foreign body sensation, supports daily wear, requires no professional operation for data collection, and is suitable for use in multiple scenarios; The accuracy rate of this scheme in determining foot arch type is 93.1% (100% accuracy rate in determining foot arch collapse); the weight prediction MAE is 0.59kg and RMSE is 0.81kg, which is significantly better than existing single models.

[0020] This invention collects data using a flexible plantar pressure array, and combines multi-dimensional feature extraction and fusion modeling to achieve integrated and high-precision monitoring of foot health and weight. It solves many pain points of traditional technologies and has broad application prospects and promotional value in the fields of youth health management, smart wearable devices, and rehabilitation medicine. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is an overall flowchart of an embodiment of the present invention; Figure 2 This is a time series comparison of the X and Y coordinates of the COP of the left and right feet of the subjects with collapsed arches in this embodiment of the invention. Figure 3 The image shows a thermogram of the sole of a subject with collapsed arches in this embodiment of the invention. Figure 4 This is a thermogram of the sole of a subject with normal arches in an embodiment of the present invention; Figure 5 This is a scatter plot comparing the weight prediction results of each model in the embodiments of the present invention; Figure 6 This is a radar chart showing the model performance in an embodiment of the present invention; Figure 7 This is a flowchart illustrating the implementation of an embodiment of the present invention. Detailed Implementation

[0023] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as a limitation of the present invention, but rather as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0024] It should be understood that the terminology used in this invention is merely for describing particular embodiments and is not intended to limit the invention. Furthermore, with respect to numerical ranges in this invention, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Every smaller range between any stated value or intermediate value within a stated range, and any other stated value or intermediate value within said range, is also included in this invention. The upper and lower limits of these smaller ranges may be independently included or excluded from the range.

[0025] Various modifications and variations can be made to the specific embodiments described in this specification without departing from the scope or spirit of the invention, as will be apparent to those skilled in the art. Other embodiments derived from this specification will also be obvious to those skilled in the art. This application specification and embodiments are merely exemplary.

[0026] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] like Figure 1 - Figure 7As shown, this embodiment provides a health monitoring method based on a flexible plantar pressure array, solving the problems of subjective foot health assessment, limited weight monitoring scenarios, and insufficient accuracy in traditional technologies. It achieves integrated, convenient, and high-precision health monitoring, including: collecting dynamic plantar pressure data of the user's feet using a flexible plantar pressure array; extracting foot health assessment features and weight prediction features based on the dynamic plantar pressure data. The foot health assessment features and force unevenness parameters include COP trajectory area, X-direction offset standard deviation, Y-direction offset standard deviation, and sensor activation ratio in the arch area. The weight prediction features include 15 features such as basic pressure features, trajectory features, and zoned pressure features; constructing a quantitative assessment model based on the foot health assessment features to achieve arch type determination and quantitative analysis of force unevenness; and constructing an RF-XGBoost fusion model based on the weight prediction features to achieve non-invasive and accurate weight prediction. This invention achieves integrated and convenient foot health assessment and weight monitoring, improving the accuracy and reliability of assessment and prediction, and is applicable to scenarios such as youth health management and smart wearable device development.

[0029] Through the implementation of the above technical solution, this embodiment achieves the first-ever synergistic monitoring of foot health assessment and weight prediction. No additional equipment is required; dual-task monitoring can be completed simply by embedding a flexible pressure insole in the shoe, meeting the needs of daily, seamless monitoring. Furthermore, the accuracy rate for determining arch type in this solution reaches 93.1% (100% accuracy rate for determining arch collapse); the weight prediction MAE is 0.59 kg, and RMSE is 0.81 kg, significantly better than existing single models. This embodiment also constructs an open-source, high-quality foot pressure dataset, including dynamic pressure sequences, health annotations, and physiological ground truth, providing standardized support for research in this field.

[0030] This embodiment collects data through a flexible plantar pressure array, and combines multi-dimensional feature extraction and fusion modeling to achieve integrated and high-precision monitoring of foot health and weight. It solves many pain points of traditional technologies and has broad application prospects and promotional value in the fields of youth health management, smart wearable devices, and rehabilitation medicine.

[0031] In terms of hardware system design and data acquisition, this embodiment uses a flexible pressure array sensor with 48 sensing points in 12 rows × 4 columns per foot. The sensing points are distributed to cover the heel area (numbers 1-3, 13-15, 25-27, 37-39), the midfoot area (numbers 4-7, 16-18, 28-31, 40-43, which are related to the arch), the forefoot area (numbers 6-11, 20-23, 32-35, 44-47), and the toe area (numbers 12, 24, 36, 48). This sensor uses a flexible polyimide film substrate with a thickness of <0.3mm, which can adaptively conform to the shape of the foot. In terms of performance, the pressure range is 0~50kg, the static resistance is >1MΩ, the hysteresis is <6%, the repeatability is ±8%, the response time is <20ms, the operating temperature range is -30~+60℃, and the data transmission uses a BLE5.2 Bluetooth acquisition device, which supports dual-foot synchronous sampling. The maximum sampling frequency for a single foot is 50Hz (sampling interval 20ms), and the maximum sampling frequency for both feet is 20Hz (sampling interval 50ms).

[0032] This embodiment uses a flexible and thin sensor (thickness < 0.3mm), which fits the sole of the foot without any foreign body sensation, supports daily wear, and data collection does not require professional operation, making it suitable for use in multiple scenarios; The data acquisition process begins with preprocessing and baseline calibration. Subjects remove their shoes and socks and sit quietly in the laboratory for 10 minutes to eliminate interference from movement and temperature changes on plantar pressure. An electronic scale (Omron HEM-7130, ESH certified) with an accuracy of ±0.1 kg is used to record their actual weight, and a height meter with an accuracy of ±0.5 cm is used to record their height as the true baseline. Simultaneously, the pressure insoles are checked; they are laid flat on a flat table and a standard pressure of 10 kg is applied to confirm that all 48 sensing points can output signals normally. Next, dynamic pressure data is acquired. Subjects stand with their feet shoulder-width apart in the center of the scale (with the pressure insoles placed on the surface). After the scale reading stabilizes, data is collected for 20 seconds. After collection, subjects slowly walk off the scale and rest for 30 seconds to restore blood circulation in their feet. This "stand-collect-rest" process is repeated three times to obtain three sets of valid data. For everyday wear scenarios, the pressure insoles are embedded in the subjects' daily commuting shoes, ensuring a snug fit without folds or obstructions, and the above calibration and acquisition process is repeated.

[0033] Finally, a preliminary screening of data quality is conducted. If there are ≥10 abnormal points in a certain data set with no signal (pressure value = 0) or exceeding the range (pressure value > 55kg), it is determined to be an invalid sample and re-collected. In the end, 3 sets of valid data are retained for each subject and stored in the SQLite database in association with the true value of height / weight.

[0034] In the data preprocessing and feature extraction stage, the data preprocessing first involves filtering and noise reduction. An eight-connected region filter is used to process the pressure data to remove environmental noise and sensor interference. Then, the 3σ criterion is used to remove outlier pressure values ​​that exceed the mean ± 3 times the standard deviation. Finally, Min-Max normalization is used to map the data to the [0,1] interval. The normalization formula is: ; In the formula, This is the original data. , These represent the minimum and maximum values ​​of the data, respectively.

[0035] Finally, steady-state data is extracted, and non-steady-state data from the first and last 30 frames of the weighing phase are removed. A sliding window of 30 frames is used to calculate the pressure change gradient between adjacent frames within the window. The 30 consecutive frames of data whose total pressure is closest to the average of the three measurements are selected as the steady-state segment. At the same time, abnormal measurements are identified through consistency checks. If the relative difference between a measurement and the average of the other two measurements exceeds 30%, the abnormal data is removed, and the normal data is subjected to robust averaging.

[0036] Foot health assessment feature extraction first uses the pressure-weighted average method to calculate the COP coordinates at each time point t, using the following formula: ; ; In the formula, , Let be the coordinates of the i-th sensor. Let be the pressure value of the i-th sensor at time t.

[0037] Then, based on the COP coordinates, four core features are calculated: COP trajectory area (based on the COP coordinate sequence, the area enclosed by the trajectory is calculated using the polygon area formula; the area of ​​the normal arch trajectory is small, and the area increases significantly when the arch collapses), X-direction offset standard deviation (calculate the standard deviation of the COP_x sequence, reflecting the degree of fluctuation of the center of gravity in the lateral (left-right) direction of the foot; the fluctuation intensifies when the arch collapses), Y-direction offset standard deviation (calculate the standard deviation of the COP_y sequence, reflecting the degree of fluctuation of the center of gravity in the longitudinal (front-back) direction of the foot, reflecting the synergy between arch cushioning and support), and arch area sensor activation ratio (statistically, the proportion of sensors with effective pressure values ​​> 0 in the midfoot area out of the total number of sensors in that area; the activation ratio of the normal arch is < 30%, and > 60% when the arch collapses).

[0038] COP trajectory area calculation process: ; The calculation process of the standard deviation of the X-direction offset: ; The calculation process of the standard deviation of the Y-direction offset: ; The calculation process for the activation ratio of sensors in the arch area: ; Figure 3 The images show the plantar thermograms of subjects with collapsed arches. The left side of the image shows the plantar thermogram of a 21-year-old male, and the right side shows the plantar thermogram of a 20-year-old female. Figure 4 The images show plantar thermograms of subjects with normal arches. The upper sub-image represents the plantar thermogram of females aged 18-21, while the lower sub-image represents the plantar thermogram of males aged 18-21.

[0039] Weight prediction feature extraction involves extracting 15 multi-dimensional features, covering eight categories: baseline pressure, trajectory, zoned pressure, pressure density, activated sensors, dynamics, symmetry, and maximum pressure. Specifically, these include average total pressure (arithmetic mean of pressure values ​​from all sensors), average total pressure of the right foot (arithmetic mean of pressure values ​​from 48 sensors on the right foot), average total pressure of the left foot (arithmetic mean of pressure values ​​from 48 sensors on the left foot), left-right foot pressure ratio (average total pressure of the right foot / average total pressure of the left foot), and average COP distance (mean Euclidean distance between the COP coordinates of the left and right feet). The following parameters are considered: COP distance stability (standard deviation of the COP distance sequence between the left and right feet), right heel pressure ratio (total pressure of sensors in the right heel area ÷ total pressure of the right foot), right foot pressure density (total pressure of the right foot ÷ area of ​​the sensor distribution area on the right foot, preset to 12cm×4cm), number of activated sensors on the right foot (number of sensors with effective pressure value > 0 on the right foot), number of activated sensors on the left foot (number of sensors with effective pressure value > 0 on the left foot), pressure fluctuation (standard deviation of the pressure value sequence of all sensors), average pressure change rate (arithmetic mean of the absolute difference of total pressure between adjacent frames), bipedal pressure symmetry (Pearson correlation coefficient of pressure values ​​of corresponding sensors on the left and right feet), maximum pressure on the right foot (maximum pressure value among all sensor pressure values ​​on the right foot), and maximum pressure on the left foot (maximum pressure value among all sensor pressure values ​​on the left foot).

[0040] Table 1. Calculation process and corresponding physiological significance of weight prediction characteristics. ; ; in, To ignore the total number of stress frames with all zeros, For the right foot, ignore all zeros in the pressure frame count. For the right foot, the non-zero pressure is measured by the i-th sensor in the t-th frame. Total number of frames To ignore the number of stress frames with all zeros on the left, For the left foot, the non-zero pressure is measured by the j-th sensor in frame t. Let the COP coordinate of the right foot be the coordinate of the right foot in frame t. Let COP be the coordinate of the left foot in frame t. The number of frames where neither the left nor right foot is all zero. Let be the Euclidean distance between the left and right foot COPs in frame t. For the right heel sensor index set (satisfying) , for One sensor coordinate), This represents the area of ​​the sensor distribution region on the right foot. The t-th pressure is the sum of the non-zero pressures of the left and right feet. The effective frame average pressure of the k-th sensor on the right foot.

[0041] During the model construction and training process, the foot health quantitative assessment model first set feature thresholds based on the clinical data and foot health annotations of 29 young volunteers (aged 17-25). The first threshold for COP trajectory area was 1500 mm², the second threshold for X-direction offset standard deviation was 8 mm, the third threshold for Y-direction offset standard deviation was 12 mm, and the fourth threshold for the activation ratio of the arch sensor was 50%.

[0042] Next, the arch type is determined. If at least 3 of the 4 characteristics exceed the corresponding threshold, it is determined to be an arch collapse; if 2 exceed the threshold, it is determined to be a risk of arch collapse; and if 1 or less exceeds the threshold, it is determined to be a normal arch.

[0043] The RF-XGBoost fusion weight prediction model first configured the dataset with 29 young volunteers, each with 3 sets of valid data, for a total of 870 samples. The dataset was divided into a training set (731 samples) and a test set (139 samples) at a ratio of 8.4:1.6. The feature input consisted of 15 weight prediction features, and the label was the subject's true weight (accuracy ±0.1kg).

[0044] Next, the model is trained. The random forest model constructs 20 decision trees, forming a random forest regression model. The maximum number of splits per tree is set to 15, and the minimum number of leaf samples is set to 15. The training process is monitored using out-of-bag error to avoid overfitting. The XGBoost model has 40 weak learners with a learning rate of 0.1. Each weak learner has a maximum number of splits of 5 and a minimum number of leaf samples of 10, and the mean squared error (MSE) is used as the loss function.

[0045] An adaptive weight allocation strategy is designed based on the validation set performance, and the final prediction value is a weighted sum of the prediction values ​​of the two models: ; The weighting coefficients are dynamically calculated based on the mean absolute error (MAE) of each model on the validation set: ; This weighting strategy allows models with higher prediction accuracy to receive greater weight in the fusion process, achieving organic complementarity of model advantages. In practical implementation, a grid search method is used to optimize the weights within the weight range [0, 1] with a step size of 0.1, and the final fusion ratio is determined by minimizing the root mean square error (RMSE) of the test set.

[0046] Then, dynamic weights were calculated based on the validation set MAE, and the optimal weight combination was determined by grid search (step size 0.1). The fusion prediction results were output, and finally, the performance was evaluated. The model performance was evaluated using four indicators: MAE, RMSE, R², and MBE. The fusion model showed MAE=0.59kg, RMSE=0.81kg, R²=0.9960, and MBE=0.16kg, which was significantly better than models such as multiple linear regression, SVM, and single random forest.

[0047] ; ; ; ; Experimental results and verification show that, in terms of foot health assessment, out of 50 samples with normal arches, 46 were correctly identified, with an accuracy rate of 92.0%; out of 8 samples with collapsed arches, 8 were correctly identified, with an accuracy rate of 100.0%; and out of a total of 58 samples, 54 were correctly identified, with an accuracy rate of 93.1%. This can accurately reflect the health status of the soles of the feet.

[0048] In terms of weight prediction, the RF-XGBoost fusion model outperformed multiple linear regression (MLR), support vector machine (SVM), backpropagation neural network (BPNN), random forest (RF), and XGBoost single model in all four metrics: MAE, MBE, RMSE, and R². The predicted values ​​were 0.59 kg, MBE was 0.16 kg, RMSE was 0.81 kg, and R² was 0.9960. The predicted values ​​were highly consistent with the actual weight and there was no obvious systematic bias.

[0049] Table 2 Comparison results of each model

[0050] In summary, this embodiment achieves non-invasive and continuous acquisition of dynamic plantar pressure data through a flexible plantar pressure array. Combined with multi-dimensional feature extraction and dual-model construction, it simultaneously completes foot health assessment and weight prediction. Foot health assessment eliminates the subjective dependence of traditional methods, achieving high accuracy in determining arch type and precise quantification of uneven force distribution. Weight prediction requires no specialized equipment, integrating into daily wearable scenarios for seamless monitoring. The fusion model improves prediction accuracy and robustness. This embodiment also open-sources a high-quality foot pressure dataset, providing support for research in this field and possessing significant application value in areas such as youth health management and the development of smart wearable devices.

[0051] Implementable, this embodiment also provides a health monitoring system based on a flexible plantar pressure array, including: The data acquisition module is used to collect dynamic pressure data of various functional areas of the user's sole based on the flexible plantar pressure array; The feature extraction module is used to extract the biomechanical feature parameters of the dynamic pressure data to obtain foot health assessment features and weight prediction features; The health monitoring module is used to determine the arch type and quantify the uneven force distribution based on the foot health assessment features to obtain the foot health assessment results; input the weight prediction features into the weight prediction model for prediction and output the weight prediction results; wherein, the weight prediction model is constructed based on the random forest model and the XGBoost model; and perform health monitoring on the user based on the foot health assessment results and the weight prediction results.

[0052] In practice, this embodiment also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the health monitoring method based on a flexible plantar pressure array.

[0053] In practice, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned health monitoring method based on a flexible plantar pressure array.

[0054] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited to the above-described specific embodiments. All equivalent modifications, substitutions, or extensions based on the technical solution of this application fall within the scope of protection of this invention. For example, the number of sensor sensing points can be adjusted according to actual needs (e.g., 64 points, 96 points), weight prediction features can be added or deleted based on data optimization, and other adaptive strategies can be adopted for the weight calculation method of the fusion model. As long as the core idea is based on the synergy of foot health assessment and weight prediction using a flexible plantar pressure array, it falls within the scope of protection of this application.

Claims

1. A health monitoring method based on a flexible plantar pressure array, characterized in that, include: Dynamic pressure data of various functional areas of the user's foot are collected based on a flexible plantar pressure array. Biomechanical characteristic parameters of the dynamic pressure data are extracted to obtain foot health assessment features and weight prediction features; Based on the foot health assessment characteristics, the arch type and the quantitative data of uneven force distribution are determined to obtain the foot health assessment results; The weight prediction features are input into the weight prediction model for prediction, and the weight prediction result is output; wherein, the weight prediction model is constructed based on the random forest model and the XGBoost model; The user's health is monitored based on the foot health assessment results and the weight prediction results.

2. The method according to claim 1, characterized in that, The flexible plantar pressure array is a flexible pressure array sensor that includes several sensing points, which are distributed to cover the heel area, midfoot area, forefoot area, and toe area.

3. The method according to claim 1, characterized in that, The biomechanical characteristic parameters extracted from the dynamic pressure data specifically include: The dynamic pressure data is sequentially processed by filtering and noise reduction, abnormal pressure value removal, and normalization. After normalization, steady-state data is extracted to obtain preprocessed dynamic pressure data. The specific process includes: removing non-steady-state data of the first and last preset frame lengths; calculating the pressure change gradient between adjacent frames within the window using a sliding window of preset frame length; selecting the continuous data whose total pressure is closest to the average of three measurements as the steady-state segment; if the relative difference between a measurement and the average of the other two measurements exceeds 30%, it is identified as abnormal data and removed; and the normal data is subjected to robust averaging. Foot health assessment features and weight prediction features were extracted from preprocessed dynamic pressure data.

4. The method according to claim 3, characterized in that, The extraction process of the foot health assessment features specifically includes: The COP coordinates at each time point t are obtained based on the pressure-weighted average method; Foot health assessment features are calculated based on the COP coordinates. These features include the COP trajectory area, the standard deviation of the X-direction offset, the standard deviation of the Y-direction offset, and the activation ratio of sensors in the arch area.

5. The method according to claim 3, characterized in that, The weight prediction features include average total pressure, average total pressure of the right foot, average total pressure of the left foot, pressure ratio of the left and right feet, average COP distance, COP distance stability, pressure ratio of the right heel, pressure density of the right foot, number of activated sensors in the right foot, number of activated sensors in the left foot, pressure fluctuation, average pressure change rate, bipedal pressure symmetry, maximum pressure of the right foot, and maximum pressure of the left foot.

6. The method according to claim 4, characterized in that, The process of obtaining the foot health assessment results specifically includes: Foot arch type determination: When the COP trajectory area exceeds the first preset threshold, the standard deviation of X-direction offset exceeds the second preset threshold, the standard deviation of Y-direction offset exceeds the third preset threshold, and the activation ratio of foot arch area sensors exceeds the fourth preset threshold, the current data is determined to be abnormal. If the combined results of both data points are abnormal, it indicates a risk of arch collapse. When the cumulative results of the three data points are abnormal, it indicates foot arch collapse; Quantitative analysis of uneven stress distribution: The uniformity of plantar pressure distribution is quantitatively analyzed by the uneven stress coefficient, which is calculated by the ratio of the standard deviation of pressure in each region to the average pressure.

7. The method according to claim 5, characterized in that, The prediction process for the weight prediction result specifically includes: The weight prediction features are input into the random forest model and the XGBoost regression model respectively to predict weight, and the first predicted value and the second predicted value are output. The final weight prediction result is calculated based on the first predicted value and the second predicted value. The specific calculation process is as follows: ； In the formula, , These are the weights for the random forest model and the XGBoost regression model, respectively. The first predicted value, This is the second predicted value. This is the final weight prediction result.

8. A health monitoring system based on a flexible plantar pressure array, characterized in that, include: The data acquisition module is used to collect dynamic pressure data of various functional areas of the user's sole based on the flexible plantar pressure array; The feature extraction module is used to extract the biomechanical feature parameters of the dynamic pressure data to obtain foot health assessment features and weight prediction features; The health monitoring module is used to determine the arch type and quantify the uneven force distribution based on the foot health assessment features to obtain the foot health assessment results; input the weight prediction features into the weight prediction model for prediction and output the weight prediction results; wherein, the weight prediction model is constructed based on the random forest model and the XGBoost model; and perform health monitoring on the user based on the foot health assessment results and the weight prediction results.

9. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to enable the electronic device to perform a health monitoring method based on a flexible plantar pressure array according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements a health monitoring method based on a flexible plantar pressure array as described in any one of claims 1-7.

Images