Interpolation method of sparse pressure sensor array for flexible insole

Abstract

The invention provides an interpolation method of a sparse pressure sensor array for a flexible insole, and the method comprises the steps: collecting sole pressure dense data, foot anatomical parameters and gait cycles of different people, and forming a training data set; constructing a core interpolation model based on a neural network, embedding a physiological rule mask in a network output layer of the core interpolation model, and training through the training data set; collecting pressure time sequence data of a user through a sparse pressure sensor array deployed on the flexible insole, and evaluating a risk value based on the pressure time sequence data; when the risk value is greater than a preset threshold value, processing the pressure time sequence data by a core interpolation model, and outputting a plantar pressure distribution map; according to the method, the plantar pressure distribution diagram is visually displayed, a personalized intervention scheme is generated, and after the physiological rule mask is embedded into the core interpolation model, accurate interpolation can be performed based on the foot physiological structure, so that subsequent reasonable health assessment and intervention are facilitated.

Description

Technical Field

[0001] This invention relates to the field of foot monitoring technology, and in particular to an interpolation method for a sparse pressure sensor array for flexible insoles. Background Technology

[0002] In the field of flexible smart wearable devices, plantar pressure distribution monitoring is of great significance for scenarios such as medical and health management and sports posture optimization. Flexible insoles have become the mainstream carrier due to their advantages of conforming to the foot and being comfortable to wear. However, due to the structural characteristics and manufacturing costs of flexible substrates, sensor arrays usually need to be deployed in a sparse manner. Traditional interpolation algorithms, such as bicubic splines and Kriging, are based on pure mathematical models and only focus on the numerical continuity fitting of discrete data points, without considering the physiological characteristics of plantar pressure. Plantar pressure distribution is affected by the anatomical structure of the foot and shows significant regional differences. The heel and metatarsal areas are high-pressure concentrated areas, while the pressure in the arch area is significantly lower. Moreover, the pressure transmission conforms to the mechanical laws of the gait cycle. Traditional algorithms are difficult to adapt to these physiological characteristics, resulting in non-physiologically reasonable pressure anomalies in the interpolation results. They cannot accurately restore the true pressure distribution and affect the accuracy of subsequent health assessments and interventions. Summary of the Invention

[0003] In view of this, the present invention proposes an interpolation method for a sparse pressure sensor array for flexible insoles, which can perform accurate interpolation based on the physiological structure of the foot, so as to facilitate subsequent reasonable health assessment and intervention.

[0004] The technical solution of this invention is implemented as follows: An interpolation method for a sparse pressure sensor array for flexible insoles includes the following steps: Step S1: Collect dense data on plantar pressure, foot anatomy parameters, and gait cycles from different population groups to form a training dataset; Step S2: Construct a core interpolation model based on a neural network, embed physiological rule masks into the network output layer of the core interpolation model, and train it using a training dataset; Step S3: Collect the user's pressure time-series data through a sparse pressure sensor array deployed on the flexible insole, and assess the risk value based on the pressure time-series data; Step S4: When the risk value is greater than the preset threshold, the core interpolation model processes the pressure time series data and outputs a plantar pressure distribution map. Step S5: Visualize the plantar pressure distribution map and generate a personalized intervention plan.

[0005] Preferably, step S1 includes the following steps: Step S11: Recruit volunteers through social networks and collect dense foot pressure data based on flexible insoles with dense pressure sensor arrays deployed inside the volunteers' shoes; Step S12: Obtain the arch height and metatarsal key point positions of the volunteers as foot anatomical parameters using a 3D foot scanner; Step S13: Collect gait cycles of volunteers, including heel strike, full foot support, and toe lift events, using an inertial measurement unit; Step S14: Combine the plantar pressure dense data, foot anatomical parameters, and gait cycle of a single volunteer into a training dataset, and label it according to the volunteer's foot structure.

[0006] Preferably, after obtaining the training dataset, the training dataset is subjected to data cleaning, normalization, and augmentation processing.

[0007] Preferably, step S2 includes the following specific steps: Step S21: Construct a core interpolation model based on a lightweight U-Net network with a multi-scale attention mechanism. The core interpolation model includes an encoding layer, a decoding layer, and a network output layer. Step S22: Set pressure constraints for each area of ​​the foot according to the anatomical parameters of the foot, and form a physiological rule mask; Step S23: In the network output layer of the core interpolation model, perform a dot product operation between the physiological rule mask and the original output; Step S24: Define a composite loss function that integrates interpolation error, pressure gradient continuity, and physiological rule matching degree, and train the core interpolation model using the training dataset.

[0008] Preferably, in step S22, after setting pressure constraints for each area of ​​the foot, a binary mask matrix is ​​formed, wherein the effective foot contact area is 1 and the non-contact area is 0, and the binary mask matrix is ​​output as a physiological rule mask.

[0009] Preferably, step S3 includes the following specific steps: Step S31: Place a flexible insole with a sparse pressure sensor array inside the shoe of the test user, wherein the pressure sensor array is deployed at the position corresponding to the metatarsal head; Step S32: Collect time-series pressure data of the user during walking / movement using a sparse pressure sensor array; Step S33: After filtering, denoising and unit conversion of the pressure time series data, extract the peak pressure, pressure duration and pressure change rate as comparative features. Step S34: Compare the comparison features with the user's preset health baseline, and obtain a quantified risk value based on the comparison results.

[0010] Preferably, the specific steps of step S33 are as follows: compare the comparison features with the user's preset health baseline one by one, calculate the peak pressure risk score, pressure duration risk score and pressure change rate risk score respectively, and then sum the peak pressure risk score, pressure duration risk score and pressure change rate risk score after assigning weights to them respectively to obtain a quantified risk value.

[0011] Preferably, step S4 includes the following specific steps: Step S41: Compare the risk value with a preset threshold. If the risk value is greater than the preset threshold, trigger the interpolation process. Step S42: Convert the collected pressure time series data into a sparse feature map according to the position of the sparse sensor array; Step S43: Call the core interpolation model to perform interpolation calculations on the sparse feature map, and after filtering out non-physiologically reasonable values ​​through physiological rule masks, generate a plantar pressure distribution map.

[0012] Preferably, step S5 includes the following specific steps: Step S51: Render the plantar pressure distribution map using a heat map format and use different colors to indicate the pressure level; Step S52: Extract key feature parameters from the heat map, including the maximum pressure value, the area of ​​the high-pressure zone, and the pressure distribution asymmetry index; Step S53: Based on key feature parameters combined with the user's foot anatomy parameters and gait cycle, match personalized intervention plans from the preset intervention rule base; Step S54: Push the heat map and personalized intervention plan to the user's smart terminal.

[0013] Compared with the prior art, the beneficial effects of the present invention are: This invention discloses an interpolation method for a coefficient pressure sensor array used in flexible insoles. Unlike traditional mathematical interpolation methods, this method employs artificial intelligence interpolation. After forming a training dataset using dense plantar pressure data from different populations, foot anatomical parameters, and gait cycles, a core interpolation model is trained. Simultaneously, a physiological rule mask is embedded in the network output layer of the core interpolation model. When the core interpolation model performs interpolation, the physiological structure of the foot can be considered, and the output of non-contact areas of the foot can be forced to be 0, thereby achieving accurate interpolation. This facilitates obtaining an accurate plantar pressure distribution map, which in turn enables subsequent reasonable health assessments and interventions. In addition, the user's plantar pressure data does not need to be interpolated and supplemented at all times. After the sparse pressure sensor array collects the pressure time series data, the risk value is evaluated based on the pressure time series data. When the risk value is higher than the preset threshold, it is determined that there is a risk of gait abnormality or plantar ulcer. Only at this time will the interpolation process be triggered to supplement the pressure data for visualization and intervention plan generation. By using coarse screening + fine diagnosis, power consumption can be reduced. Interpolation is only performed when there is a risk, thus extending the monitoring time of the flexible insole and control system. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of an interpolation method for a sparse pressure sensor array for flexible insoles according to the present invention. Figure 2 This is a flowchart of step S1 of an interpolation method for a sparse pressure sensor array for a flexible insole according to the present invention. Figure 3 This is a flowchart of step S2 of an interpolation method for a sparse pressure sensor array for a flexible insole according to the present invention. Figure 4 This is a flowchart of step S3 of an interpolation method for a sparse pressure sensor array for a flexible insole according to the present invention. Figure 5 This is a flowchart of step S4 of an interpolation method for a sparse pressure sensor array for a flexible insole according to the present invention. Figure 6 This is a flowchart of step S5 of an interpolation method for a sparse pressure sensor array for flexible insoles according to the present invention. Detailed Implementation

[0016] To better understand the technical content of this invention, a specific embodiment is provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0017] See Figures 1 to 6 The present invention provides an interpolation method for a sparse pressure sensor array for flexible insoles, comprising the following steps: Step S1: Collect dense data on plantar pressure, foot anatomy parameters, and gait cycles from different population groups to form a training dataset; Step S2: Construct a core interpolation model based on a neural network, embed physiological rule masks into the network output layer of the core interpolation model, and train it using a training dataset; Step S3: Collect the user's pressure time-series data through a sparse pressure sensor array deployed on the flexible insole, and assess the risk value based on the pressure time-series data; Step S4: When the risk value is greater than the preset threshold, the core interpolation model processes the pressure time series data and outputs a plantar pressure distribution map. Step S5: Visualize the plantar pressure distribution map and generate a personalized intervention plan.

[0018] This invention discloses an interpolation method for a sparse pressure sensor array in flexible insoles. Unlike traditional mathematical interpolation methods, this method uses a core interpolation model built on a neural network for intelligent interpolation, which improves interpolation efficiency. After building the core interpolation model using a neural network, it needs to be trained. The training dataset consists of a large amount of actually collected data, including dense plantar pressure data, foot anatomical parameters, and gait cycles from different populations. These populations are categorized based on foot structure, including people with normal feet, flat feet, high arches, and diabetic foot. The plantar pressure composition differs among these populations. For example, in people with high arches, the arch does not contact the insole, and the pressure is concentrated on both sides. Prolonged standing or high-intensity exercise can cause leg damage. The core interpolation model can be trained on different training datasets to adapt to various populations, thereby obtaining accurate interpolation results.

[0019] After the core interpolation model is trained, it can enter a ready-to-use state. For users who need monitoring, a sparse pressure sensor array can be deployed on their flexible insoles. The sparse pressure sensor array can collect time-series pressure data when the user is walking or exercising. By evaluating the pressure time-series data, a risk value can be assessed. Based on the risk value, it can be determined whether the user is at risk of gait abnormalities or plantar ulcers. When a risk is found, the interpolation process can be triggered, waking up the core interpolation model to process the pressure time-series data and output a plantar pressure distribution map. By visualizing the plantar pressure distribution map, users can easily understand the corresponding pressure data on their feet. At the same time, personalized intervention plans can be generated and pushed to users so that they can intervene in a timely manner to avoid aggravating foot risks. By using the risk value assessment + core interpolation model wake-up method, the system power consumption can be reduced and the service life can be extended to facilitate long-term monitoring of the user's feet.

[0020] Preferably, step S1 includes the following steps: Step S11: Recruit volunteers through social networks and collect dense foot pressure data based on flexible insoles with dense pressure sensor arrays deployed inside the volunteers' shoes; Step S12: Obtain the arch height and metatarsal key point positions of the volunteers as foot anatomical parameters using a 3D foot scanner; Step S13: Collect gait cycles of volunteers, including heel strike, full foot support, and toe lift events, using an inertial measurement unit; Step S14: Combine the plantar pressure dense data, foot anatomical parameters, and gait cycle of a single volunteer into a training dataset, and label it according to the volunteer's foot structure.

[0021] When constructing the training dataset, volunteers can be recruited through social networks or by partnering with medical institutions. The volunteers' insoles are then replaced with flexible insoles equipped with dense pressure sensor arrays. Commands to walk or move are issued to the volunteers. During their activities, the dense pressure sensor array collects corresponding plantar pressure data. Simultaneously, an inertial measurement unit (IMU) collects the entire gait cycle of the volunteer's walking or movement. After data collection, the volunteers remove their shoes, and a 3D foot scanner collects their foot anatomy parameters. Then, for each volunteer, the plantar pressure data, foot anatomy parameters, and gait cycle are combined to form a training dataset. The dataset is also labeled according to the volunteer's foot structure; for example, a training dataset labeled with flat feet (A) can be added to the training dataset. Training datasets with different foot structures can be stored in a structured manner.

[0022] Preferably, after obtaining the training dataset, the training dataset is subjected to data cleaning, normalization, and augmentation processing.

[0023] To ensure that the core interpolation model can be trained accurately, the training dataset needs to be preprocessed after it is obtained, including data cleaning, normalization and augmentation, to improve the size and quality of the dataset.

[0024] Preferably, step S2 includes the following specific steps: Step S21: Construct a core interpolation model based on a lightweight U-Net network with a multi-scale attention mechanism. The core interpolation model includes an encoding layer, a decoding layer, and a network output layer. Step S22: Set pressure constraints for each area of ​​the foot according to the anatomical parameters of the foot, and form a physiological rule mask; Step S23: In the network output layer of the core interpolation model, perform a dot product operation between the physiological rule mask and the original output; Step S24: Define a composite loss function that integrates interpolation error, pressure gradient continuity, and physiological rule matching degree, and train the core interpolation model using the training dataset.

[0025] The core interpolation model is built on the U-Net network, which mainly includes an encoding layer, a decoding layer, and a network output layer. The network output layer needs to embed a physiological rule mask. When setting pressure constraints for different areas of the foot based on foot anatomy parameters, corresponding settings are made according to different foot structures. For example, the upper limit of pressure in the arch area is dynamically adjusted with the arch height. After setting the pressure constraints for each area of ​​the foot, a binary mask matrix is ​​formed, where the effective foot contact area is 1 and the non-contact area is 0. The binary mask matrix is ​​output as a physiological rule mask. Then, in the network output layer, the physiological rule mask is multiplied by the original output to force the predicted values ​​of non-physiological areas to be zero. Finally, after defining the composite loss function, the core interpolation model can be trained using the training dataset. By embedding the physiological rule mask, the core interpolation model can take into account the foot structure when performing interpolation, thereby performing more accurate interpolation.

[0026] Preferably, step S3 includes the following specific steps: Step S31: Place a flexible insole with a sparse pressure sensor array inside the shoe of the test user, wherein the pressure sensor array is deployed at the position corresponding to the metatarsal head; Step S32: Collect time-series pressure data of the user during walking / movement using a sparse pressure sensor array; Step S33: After filtering, denoising and unit conversion of the pressure time series data, extract the peak pressure, pressure duration and pressure change rate as comparative features. Step S34: Compare the comparison features with the user's preset health baseline, and obtain a quantified risk value based on the comparison results.

[0027] After the core interpolation model is trained, it enters the ready-to-use stage. Then, foot monitoring can be performed on the test users. A sparse pressure sensor array is deployed on the flexible insole of the test users. The sparse pressure sensor array can collect the pressure time series data of the user when walking / exercising. The metatarsal head is a high-risk area, and the pressure sensor array is deployed at the corresponding position of the metatarsal head, while other positions adopt a uniform deployment strategy. The sparse pressure sensor array can collect sparse pressure time series data of the foot. Then, the pressure time series data is preprocessed and feature extracted to obtain comparative features. Finally, the comparative features are compared with the user's preset health baseline to quantify the risk value.

[0028] Preferably, the specific steps of step S33 are as follows: compare the comparison features with the user's preset health baseline one by one, calculate the peak pressure risk score, pressure duration risk score and pressure change rate risk score respectively, and then sum the peak pressure risk score, pressure duration risk score and pressure change rate risk score after assigning weights to them respectively to obtain a quantified risk value.

[0029] Using a quantitative approach to assess risk values ​​facilitates comparison with subsequent preset thresholds. During quantification, each comparative feature is compared with a preset health baseline, and the difference is calculated. The ratio of the difference to the health baseline is then used as the risk score. After obtaining the peak pressure risk score, pressure duration risk score, and pressure change rate risk score, a weighted summation method can be used to obtain the quantified risk value.

[0030] Preferably, step S4 includes the following specific steps: Step S41: Compare the risk value with a preset threshold. If the risk value is greater than the preset threshold, trigger the interpolation process. Step S42: Convert the collected pressure time series data into a sparse feature map according to the position of the sparse sensor array; Step S43: Call the core interpolation model to perform interpolation calculations on the sparse feature map, and after filtering out non-physiologically reasonable values ​​through physiological rule masks, generate a plantar pressure distribution map.

[0031] When the risk value exceeds a preset threshold, it is determined that the user is at risk of gait abnormalities or plantar ulcers. At this point, the interpolation process can be initiated. First, the collected pressure time-series data is converted into a sparse feature map based on the position of the sparse sensor array. Then, the sparse feature map is input into the core interpolation model, which can then perform interpolation calculations. At the network output layer, after filtering out non-physiologically reasonable values ​​through a physiological rule mask, a corresponding plantar pressure distribution map can be generated. By embedding the physiological rule mask, it can be ensured that the interpolation calculation conforms to the foot structure, thereby improving the interpolation accuracy.

[0032] Preferably, step S5 includes the following specific steps: Step S51: Render the plantar pressure distribution map using a heat map format and use different colors to indicate the pressure level; Step S52: Extract key feature parameters from the heat map, including the maximum pressure value, the area of ​​the high-pressure zone, and the pressure distribution asymmetry index; Step S53: Based on key feature parameters combined with the user's foot anatomy parameters and gait cycle, match personalized intervention plans from the preset intervention rule base; Step S54: Push the heat map and personalized intervention plan to the user's smart terminal.

[0033] After obtaining the plantar pressure distribution map, it can be converted into a heat map. In the heat map, different colors are used to mark the pressure level so that users can intuitively see the pressure distribution of the foot. Then, key feature parameters are extracted from the pressure map. These key feature parameters are combined with the user's foot anatomy parameters and gait cycle, and matched with a preset intervention rule base to find a personalized intervention plan that matches the user. Finally, the heat map and personalized intervention plan are pushed to the user's smart terminal so that the user can view the pressure distribution and intervene in a timely manner to avoid aggravating foot risks.

[0034] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An interpolation method for a sparse pressure sensor array used in flexible insoles, characterized in that, Includes the following steps: Step S1: Collect dense data on plantar pressure, foot anatomy parameters, and gait cycles from different population groups to form a training dataset; Step S2: Construct a core interpolation model based on a neural network, embed physiological rule masks into the network output layer of the core interpolation model, and train it using a training dataset; Step S3: Collect the user's pressure time-series data through a sparse pressure sensor array deployed on the flexible insole, and assess the risk value based on the pressure time-series data; Step S4: When the risk value is greater than the preset threshold, the core interpolation model processes the pressure time series data and outputs a plantar pressure distribution map. Step S5: Visualize the plantar pressure distribution map and generate a personalized intervention plan.

2. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 1, characterized in that, The specific steps of step S1 include: Step S11: Recruit volunteers through social networks and collect dense foot pressure data based on flexible insoles with dense pressure sensor arrays deployed inside the volunteers' shoes; Step S12: Obtain the arch height and metatarsal key point positions of the volunteers as foot anatomical parameters using a 3D foot scanner; Step S13: Collect gait cycles of volunteers, including heel strike, full foot support, and toe lift events, using an inertial measurement unit; Step S14: Combine the plantar pressure dense data, foot anatomical parameters, and gait cycle of a single volunteer into a training dataset, and label it according to the volunteer's foot structure.

3. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 2, characterized in that, After obtaining the training dataset, the training dataset is cleaned, normalized, and augmented.

4. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 1, characterized in that, The specific steps of step S2 include: Step S21: Construct a core interpolation model based on a lightweight U-Net network with a multi-scale attention mechanism. The core interpolation model includes an encoding layer, a decoding layer, and a network output layer. Step S22: Set pressure constraints for each area of ​​the foot according to the anatomical parameters of the foot, and form a physiological rule mask; Step S23: In the network output layer of the core interpolation model, perform a dot product operation between the physiological rule mask and the original output; Step S24: Define a composite loss function that integrates interpolation error, pressure gradient continuity, and physiological rule matching degree, and train the core interpolation model using the training dataset.

5. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 4, characterized in that, In step S22, after setting pressure constraints for each area of ​​the foot, a binary mask matrix is ​​formed, where the effective foot contact area is 1 and the non-contact area is 0. The binary mask matrix is ​​then output as a physiological rule mask.

6. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 1, characterized in that, The specific steps of step S3 include: Step S31: Place a flexible insole with a sparse pressure sensor array inside the shoe of the test user, wherein the pressure sensor array is deployed at the position corresponding to the metatarsal head; Step S32: Collect time-series pressure data of the user during walking / movement using a sparse pressure sensor array; Step S33: After filtering, denoising and unit conversion of the pressure time series data, extract the peak pressure, pressure duration and pressure change rate as comparative features. Step S34: Compare the comparison features with the user's preset health baseline, and obtain a quantified risk value based on the comparison results.

7. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 6, characterized in that, The specific steps of step S33 are as follows: compare the comparison features with the user's preset health baseline one by one, calculate the peak pressure risk score, pressure duration risk score and pressure change rate risk score respectively, and then sum the peak pressure risk score, pressure duration risk score and pressure change rate risk score after assigning weights to them respectively to obtain the quantified risk value.

8. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 1, characterized in that, The specific steps of step S4 include: Step S41: Compare the risk value with a preset threshold. If the risk value is greater than the preset threshold, trigger the interpolation process. Step S42: Convert the collected pressure time series data into a sparse feature map according to the position of the sparse sensor array; Step S43: Call the core interpolation model to perform interpolation calculations on the sparse feature map, and after filtering out non-physiologically reasonable values ​​through physiological rule masks, generate a plantar pressure distribution map.

9. The interpolation method for a sparse pressure sensor array for flexible insoles according to claim 1, characterized in that, The specific steps of step S5 include: Step S51: Render the plantar pressure distribution map using a heat map format and use different colors to indicate the pressure level; Step S52: Extract key feature parameters from the heat map, including the maximum pressure value, the area of ​​the high-pressure zone, and the pressure distribution asymmetry index; Step S53: Based on key feature parameters combined with the user's foot anatomy parameters and gait cycle, match personalized intervention plans from the preset intervention rule base; Step S54: Push the heat map and personalized intervention plan to the user's smart terminal.

Images