Method and system for analyzing a motion state based on plantar pressure

Abstract

The application relates to the technical field of pressure monitoring, and discloses a method and system for analyzing a motion state according to plantar pressure, the method comprising the following steps: receiving multi-modal biomechanics data related to a landing event; extracting corresponding motion feature vectors from the multi-modal biomechanics data; inputting the motion feature vectors into a motion state analysis model for identification to obtain a corresponding injury risk index, and outputting the corresponding injury risk index. The application can comprehensively capture biomechanics information in a motion process from multiple dimensions by receiving pressure distribution data detected by a plantar pressure sensing array, shear force data detected by a shear force calculation module, and kinematics data sequences detected by an inertial measurement unit. Compared with a single data collection mode, the fusion of multi-modal data can more accurately reflect the real state of human motion, and provide a richer and more accurate data basis for subsequent motion state analysis.

Description

Technical Field

[0001] This invention relates to the field of pressure monitoring technology, and specifically to a method and system for analyzing motion state based on plantar pressure. Background Technology

[0002] Currently, foot biomechanics monitoring mainly relies on two types of sensors: the first type is platform-type pressure sensors (such as the NOVELPedar system), but these need to be embedded in a ground platform, only support static monitoring, cannot acquire dynamic gait data, and are also bulky and susceptible to psychological interference; the second type is wearable sensors (such as the Tekscan F-scan insole), which can monitor dynamically, but the rigid materials cause discomfort when worn, and can only measure vertical pressure, unable to capture foot spatial movement and shear force. Newly emerging flexible sensing systems offer excellent wearability and comfort, can monitor foot movement in real time, and have high sensitivity and resolution. However, in current technologies, piezoelectric / piezoresistive flexible sensors, due to their single sensing mechanism, cannot simultaneously monitor dynamic and static pressure, and can only measure vertical pressure. Furthermore, plantar shear force, a key mechanical factor leading to foot strain and skin ulcers, currently lacks a mature wearable monitoring solution. Foot spatial motion plays a crucial role in monitoring foot health. Inertial measurement units (IMUs) are excellent measurement devices; however, when used independently, the lack of spatial anchoring based on plantar biomechanical data makes it difficult to establish a correlation between movement posture and foot load. Therefore, if vertical pressure, shear force, and spatial motion data are distributed across different systems, significant timing synchronization errors occur, making it impossible to construct a unified mechanical model. Consequently, designing a scheme capable of accurate motion state analysis has become a pressing technical problem for those skilled in the art. Summary of the Invention

[0003] To address the aforementioned shortcomings, this invention discloses a method for analyzing motion state based on plantar pressure, which enables accurate analysis of motion state.

[0004] The first aspect of this invention discloses a method for analyzing motion state based on plantar pressure, comprising:

[0005] Receive multimodal biomechanical data related to a single landing event, including pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and a sequence of kinematic data detected by an inertial measurement unit;

[0006] The corresponding motion feature vectors are extracted from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate;

[0007] The motion feature vector is input into the motion state analysis model for identification to obtain the corresponding damage risk index, and the corresponding damage risk index is output.

[0008] As an optional implementation, in the first aspect of the present invention, the method for motion state analysis further includes:

[0009] Receive low-resolution pressure data frames detected by the plantar pressure sensor array;

[0010] The low-resolution stress data frame is input into a trained image super-resolution reconstruction model; wherein, the image super-resolution reconstruction model is a deep learning model based on a convolutional neural network, which is trained by a large number of low-resolution stress matrices and corresponding high-resolution stress distribution ground value data, and the image super-resolution reconstruction model is used to learn the mapping relationship from low-resolution stress patterns to high-resolution stress patterns.

[0011] The image super-resolution reconstruction model outputs a corresponding high-resolution pressure distribution matrix, and a corresponding mechanical contour map is generated based on the high-resolution pressure distribution matrix.

[0012] The biomechanical cloud map is displayed on the user interface. It improves the accuracy of plantar pressure data through super-resolution reconstruction, while visualizing the results, further optimizing the practicality and accuracy of motion state analysis.

[0013] As an optional implementation, in the first aspect of the present invention, the motion state analysis model is a network model formed by combining a convolutional neural network and a long short-term memory network;

[0014] The step of inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding injury risk index includes:

[0015] The spatial pressure distribution image sequence detected by the plantar pressure sensing array is input into a convolutional neural network to extract spatial features;

[0016] The motion feature vectors of the time series are input into a long short-term memory network to extract the corresponding time series features;

[0017] The spatial features extracted by the convolutional neural network and the temporal series features extracted by the long short-term memory network are fused at the back end of the model and used together to calculate the injury risk index. This combined model of convolutional neural network and long short-term memory network simultaneously captures both spatial and temporal features of motion, significantly improving the accuracy and reliability of the injury risk index calculation.

[0018] As an optional implementation, in the first aspect of the present invention, after inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding damage risk index, the method further includes:

[0019] If the damage risk index exceeds the set risk threshold, targeted biomechanical optimization suggestions will be automatically generated.

[0020] The output of the corresponding damage risk index includes:

[0021] It outputs corresponding damage risk indices and targeted biomechanical optimization suggestions. Based on quantifying damage risk, it adds targeted biomechanical optimization suggestions, achieving a closed loop from risk assessment to optimization guidance, significantly improving the practicality and implementation value of the technical solution.

[0022] As an optional implementation, in a first aspect of the present invention, the automatic generation of targeted biomechanical optimization suggestions includes:

[0023] Identify one or more biomechanical features that contribute most to the damage risk index;

[0024] Based on one or more identified biomechanical features, select matching explanatory statements and training suggestions from a pre-defined knowledge base;

[0025] The explanatory statements and training suggestions are combined to form the targeted biomechanical optimization suggestions. By accurately identifying core risk features and matching them with a pre-defined knowledge base, these suggestions are generated, making the biomechanical optimization recommendations more targeted, persuasive, and actionable.

[0026] As an optional implementation, in a first aspect of the present invention, identifying one or more biomechanical features that contribute most to the damage risk index includes:

[0027] The contribution of each feature in the motion feature vector to the corresponding injury risk index is calculated using the SHAP method.

[0028] Obtain the weight outputs of the built-in attention layers in the CNN-LSTM combined network to determine the most critical data segments for model decisions in both the temporal and spatial dimensions;

[0029] The system ranks features based on their contribution and identifies the biomechanical feature with the highest contribution. This approach balances the quantitative accuracy of feature contributions with the decision-making relevance across time and space, making the identification of risk sources more precise and scientific.

[0030] As an optional implementation, in the first aspect of the present invention, the shear force data detected by the shear force calculation module includes:

[0031] The corresponding output charge is obtained through the piezoelectric unit matrix of the shear force calculation module, and the outer shear force is calculated in real time based on the adjacent unit strain difference algorithm; the adjacent unit strain difference algorithm includes a first shear force formula, a second shear force formula, and a shear force direction formula;

[0032] The first shear force formula is: ;

[0033] The second shear force formula is: ;

[0034] The formula for the direction of shear force is: ,in, , , , These are four piezoelectric capacitors connected in parallel for the i-th detection region. These four parallel piezoelectric capacitors together form a piezoelectric unit matrix. and These are the calibration coefficients in the x-direction and y-direction determined through material mechanics calibration experiments. and These are the shear forces in the x-direction and y-direction of the i-th detection region, respectively; and These are the total shear force outputs in the x-direction and y-direction of the foot, respectively; The direction of the resultant force is indicated. Through a specific algorithm combined with a piezoelectric element matrix, it achieves real-time, accurate, and multi-dimensional measurement of shear force, providing reliable shear force data support for motion state analysis and damage risk assessment.

[0035] As an optional implementation, in the first aspect of the present invention, the shear force calculation module includes multiple sensor clusters, which are distributed in a partitioned manner in multiple detection areas on the sole of the foot. Each sensor cluster includes a piezoelectric unit matrix, and the detection areas include the outer heel area, the inner heel area, the first and second metatarsal bone areas, and the fourth and fifth metatarsal bone areas.

[0036] The shear force data detected by the shear force calculation module includes:

[0037] It receives the shear force component output by each shear force sensing unit in the sensing cluster in real time;

[0038] Compare the output signals of each unit to identify outlier signals;

[0039] A dynamic confidence weight is assigned to the output of each shear force sensing unit, wherein units with high consistency with the outputs of other units are assigned higher weights, and units identified as outliers are assigned lower weights or excluded.

[0040] Based on the aforementioned confidence weights, the shear force components of all valid units within the cluster are weighted and fused to output the total shear force of the key biomechanical region. Through partitioned distributed arrangement and a dynamic weighted fusion algorithm, this approach achieves high reliability and regionalized accurate measurement of shear force data, further improving the quality of multimodal biomechanical data.

[0041] A second aspect of this invention discloses a system for analyzing motion state based on plantar pressure, comprising:

[0042] Receiving module: used to receive multimodal biomechanical data related to a single landing event, including pressure distribution data detected by the plantar pressure sensor array, shear force data detected by the shear force calculation module, and kinematic data sequence detected by the inertial measurement unit;

[0043] Feature extraction module: used to extract corresponding motion feature vectors from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate;

[0044] Identification module: used to input the motion feature vector into the motion state analysis model for identification to obtain the corresponding damage risk index, and output the corresponding damage risk index.

[0045] A third aspect of the present invention discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the method for analyzing motion state based on plantar pressure disclosed in the first aspect of the present invention.

[0046] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute the method for analyzing motion state based on plantar pressure disclosed in the first aspect of the present invention.

[0047] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0048] In this embodiment of the invention, by receiving pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and kinematic data sequences detected by an inertial measurement unit, biomechanical information during movement can be comprehensively captured from multiple dimensions. Compared to a single data acquisition method, the fusion of multimodal data can more accurately reflect the true state of the human body during movement, providing a richer and more accurate data foundation for subsequent movement state analysis. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying 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.

[0050] Figure 1 This is a flowchart illustrating the method for analyzing motion state based on plantar pressure disclosed in an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of the process for generating mechanical cloud maps disclosed in an embodiment of the present invention;

[0052] Figure 3 This is the specific identification and calculation process of the damage risk index disclosed in the embodiments of the present invention;

[0053] Figure 4 This is a flowchart illustrating the data acquisition and analysis logic disclosed in an embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of the sensor structure disclosed in an embodiment of the present invention;

[0055] Figure 6 This is a schematic diagram of a system for analyzing motion state based on plantar pressure, provided in an embodiment of the present invention.

[0056] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] It should be noted that the terms first, second, third, fourth, etc., in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order. The terms used in the embodiments of this invention include and have, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0059] Example 1

[0060] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for analyzing motion state based on plantar pressure, as disclosed in an embodiment of the present invention. The execution entity of the method described in this embodiment is an execution entity composed of software and / or hardware. This execution entity can receive relevant information via wired or / or wireless means and can send certain instructions. It may also have certain processing and storage functions. This execution entity can control multiple devices, such as remote physical servers or cloud servers and related software, or local hosts or servers and related software that perform related operations on devices located in a certain location. In some scenarios, multiple storage devices can also be controlled; these storage devices may be placed in the same location as the device or in different locations. Figures 1-5 As shown, the method for analyzing motion state based on plantar pressure includes the following steps:

[0061] S101: Receive multimodal biomechanical data related to a landing event, the multimodal biomechanical data including pressure distribution data detected by the plantar pressure sensor array, shear force data detected by the shear force calculation module, and kinematic data sequence detected by the inertial measurement unit;

[0062] S102: Extract the corresponding motion feature vectors from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate;

[0063] S103: Input the motion feature vector into the motion state analysis model for identification to obtain the corresponding damage risk index, and output the corresponding damage risk index.

[0064] This invention, through receiving pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and kinematic data sequences detected by an inertial measurement unit, can comprehensively capture biomechanical information during movement from multiple dimensions. Compared to single data acquisition methods, the fusion of multimodal data can more accurately reflect the true state of the human body during movement, providing a richer and more accurate data foundation for subsequent movement state analysis.

[0065] Motion feature vectors, such as peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate, are extracted from multimodal biomechanical data. These feature vectors are key indicators reflecting motion status and potential injury risk. For example, an excessively high peak impact force may indicate a greater impact on the joint, increasing the likelihood of injury; an abnormal peak inversion angular velocity may be associated with ankle instability. By accurately extracting these feature vectors, abnormalities during motion can be identified more effectively.

[0066] By inputting motion feature vectors into a motion state analysis model for identification to obtain a corresponding injury risk index, this quantitative assessment method provides users with intuitive and clear injury risk information. Compared with traditional injury risk assessment methods based on experience or subjective judgment, data-driven models can more objectively and scientifically assess injury risk, helping to take preventive measures in advance and reduce the incidence of sports injuries.

[0067] Based on the output injury risk index, personalized exercise guidance can be provided to different individuals. For example, for athletes or sports enthusiasts with a high risk of injury, it may be recommended to adjust exercise methods, strengthen local muscle groups, or choose appropriate sports equipment to optimize athletic performance and reduce the risk of injury. In the field of rehabilitation, this method can be used to monitor the patient's movement status during rehabilitation, assess rehabilitation effects, and provide objective evidence for adjusting rehabilitation plans.

[0068] More preferably, such as Figure 2 As shown, the method for motion state analysis further includes:

[0069] S1011: Receive low-resolution pressure data frames detected by the plantar pressure sensor array;

[0070] S1012: Input the low-resolution stress data frame into the trained image super-resolution reconstruction model; wherein, the image super-resolution reconstruction model is a deep learning model based on a convolutional neural network, which is trained by a large number of low-resolution stress matrices and corresponding high-resolution stress distribution ground value data, and the image super-resolution reconstruction model is used to learn the mapping relationship from low-resolution stress patterns to high-resolution stress patterns.

[0071] S1013: The image super-resolution reconstruction model outputs a corresponding high-resolution pressure distribution matrix, and generates a corresponding mechanical cloud map based on the high-resolution pressure distribution matrix;

[0072] S1014: Display the mechanical cloud diagram on the user interface.

[0073] In practice, low-resolution pressure data frames are reconstructed using a trained convolutional neural network model, which can restore the high-resolution pressure distribution matrix. This process compensates for the loss of details in low-resolution data, accurately captures pressure differences in minute areas of the sole, and makes the subsequently extracted pressure-related features (such as peak impact force and pressure distribution uniformity) more consistent with the actual movement state, providing more reliable data support for injury risk assessment.

[0074] Based on a high-resolution pressure distribution matrix, a biomechanical cloud map is generated, transforming abstract pressure data into an intuitive visual form. Users, without specialized knowledge, can quickly perceive the distribution location and intensity differences of plantar pressure through the cloud map, clearly identifying areas of concentrated pressure or abnormal distribution patterns, thus lowering the barrier to data interpretation.

[0075] The user interface displays a biomechanical cloud map, enabling interactive visualization of the analysis results. Athletes, coaches, or rehabilitation therapists can directly observe the dynamic changes in plantar pressure during exercise through the interface. Combined with the previously output injury risk index, it is easier to pinpoint the source of risk (such as the correlation between pressure concentration points and high injury risk), providing intuitive evidence for personalized exercise adjustments, equipment selection, or the development of rehabilitation training programs.

[0076] Low-resolution plantar pressure sensor arrays typically offer advantages such as low cost and low power consumption. However, by using super-resolution model reconstruction, high-precision pressure data can be obtained without relying on high-cost, high-resolution sensors. This design balances device practicality and analytical accuracy, lowers the cost barrier to technology implementation, and facilitates large-scale application.

[0077] More preferably, such as Figure 3 As shown, the motion state analysis model is a network model formed by combining a convolutional neural network and a long short-term memory network;

[0078] The step of inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding injury risk index includes:

[0079] S1031: Input the spatial pressure distribution image sequence detected by the plantar pressure sensing array into the convolutional neural network to extract spatial features;

[0080] S1032: Input the temporal motion feature vector into the Long Short-Term Memory network to extract the corresponding time series features;

[0081] S1033: The spatial features extracted by the convolutional neural network and the time series features extracted by the long short-term memory network are fused at the back end of the model and used together to calculate the damage risk index.

[0082] Convolutional Neural Networks (CNNs) and Long Short-Term Memory Networks (LSTMs) are two key components. CNNs excel at processing spatially distributed data, accurately capturing spatial features such as the spatial morphology and regional differences of pressure distribution from spatial pressure distribution image sequences. LSTMs, on the other hand, focus on temporal data processing, effectively extracting the patterns of motion feature vector changes over time (such as the timing of impact peaks and the temporal trend of inversion angle velocity). The combination of these two technologies achieves comprehensive coverage of motion features, avoiding the limitation of a single network capturing only single-dimensional features.

[0083] Sports injuries are often associated with abnormal spatial pressure distribution and abrupt changes in motion parameters over time. The back-end fusion of spatial and temporal features allows the model to calculate the injury risk index based on both static spatial distribution and dynamic temporal changes—two key types of information. This fusion calculation method better reflects the formation mechanism of sports injuries, and compared to models with single feature inputs, the evaluation results are more scientific and can effectively reduce misjudgments or omissions.

[0084] The spatial distribution and temporal variation patterns of motion features differ across various sports scenarios (e.g., the spatial distribution and temporal parameter changes of pressure during running and jumping are completely different). The combined architecture of convolutional neural networks and long short-term memory networks possesses stronger feature learning and adaptation capabilities, enabling it to handle the differences in motion data across different sports types and individuals. The model can flexibly capture key features across various scenarios, improving the universality and robustness of injury risk assessment.

[0085] More preferably, after inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding injury risk index, the method further includes:

[0086] If the damage risk index exceeds the set risk threshold, targeted biomechanical optimization suggestions will be automatically generated.

[0087] The output of the corresponding damage risk index includes:

[0088] Output the corresponding damage risk index and targeted biomechanical optimization suggestions.

[0089] In practice, simply outputting an injury risk index only informs users of the level of risk, while targeted biomechanical optimization suggestions directly provide solutions. These suggestions are generated based on motion feature vectors (such as excessively high peak impact force, abnormal inversion angular velocity, etc.), accurately identifying the source of risk and helping users understand how to make adjustments, thus preventing risk assessment from becoming a mere formality.

[0090] Biomechanical optimization recommendations are not general templates, but rather generated by combining specific features extracted from multimodal data. For example, if the risk stems from excessively high peak shear force, recommendations might focus on adjusting the landing angle or selecting equipment with better cushioning performance; if it's due to abnormal inversion velocity, recommendations could point to ankle stability training. This precise matching makes optimization measures easier to implement and increases users' willingness to follow through.

[0091] The above method eliminates the need for users to possess biomechanical expertise or consult with professionals. The system automatically completes the entire process from risk identification to cause correlation to suggestion generation. Users can directly adjust their exercise methods, intensity, or equipment based on the suggestions, making complex motion state analysis technology more relevant to the actual needs of ordinary users.

[0092] Previous damage risk indices only identified risks, while optimization recommendations enabled risk intervention. Combining the two forms a closed loop of detection, assessment, guidance, and adjustment, which not only provides early warnings of risks but also reduces the probability of injury through proactive intervention, truly realizing the damage prevention value of technical solutions.

[0093] More preferably, the automatic generation of targeted biomechanical optimization suggestions includes:

[0094] Identify one or more biomechanical features that contribute most to the damage risk index;

[0095] Based on one or more identified biomechanical features, select matching explanatory statements and training suggestions from a pre-defined knowledge base;

[0096] The explanatory statements and training suggestions are combined to form the targeted biomechanical optimization suggestions.

[0097] The solution in this invention identifies the core causes of risk by pinpointing the biomechanical characteristics that contribute most to the damage risk index (such as excessively high peak impact force or abnormal inversion velocity). The recommendations are no longer generalized guidelines, but are generated specifically for the source of the risk. For example, it may only provide suggestions for adjusting the landing angle if the peak shear force exceeds the limit, avoiding irrelevant suggestions and allowing users to focus on key issues for optimization.

[0098] The optimization suggestions include both explanatory statements and training recommendations. They inform users which biomechanical characteristic the risk stems from (e.g., excessively high peak inversion velocity, which can easily lead to uneven stress on the ankle joint) and provide specific solutions. This combination of cause and solution helps users understand the logical basis of the suggestions, reduces skepticism, and increases their willingness to implement them.

[0099] Matching statements and suggestions are selected from a pre-set knowledge base that integrates professional theories, clinical experience, and training protocols in the field of biomechanics. This ensures that the generated suggestions are scientifically based, avoids subjective and unprofessional guidance, and provides consistent professional advice to different users facing the same risk characteristics, thereby enhancing the credibility of the technical solution.

[0100] By employing a standardized process of identifying core features, matching them to a knowledge base, and combining suggestions, optimization recommendations can be generated quickly and automatically. This eliminates the need for complex real-time analysis and calculations, ensuring efficient recommendation generation while allowing users to receive targeted guidance immediately after obtaining a risk index. This shortens the time lag between risk awareness and action, improving the timeliness of damage prevention.

[0101] More preferably, the identification of one or more biomechanical features that contribute most to the damage risk index includes:

[0102] The contribution of each feature in the motion feature vector to the corresponding injury risk index is calculated using the SHAP method.

[0103] Obtain the weight outputs of the built-in attention layers in the CNN-LSTM combined network to determine the most critical data segments for model decisions in both the temporal and spatial dimensions;

[0104] The features were ranked based on their contribution, and the biomechanical feature with the highest contribution was determined.

[0105] In this embodiment of the invention, the SHAP method quantifies the specific contribution of each motion feature vector to the injury risk index from the perspective of model interpretability, and realizes the accurate decomposition of the influence of a single feature (such as clarifying the risk contribution ratio of the peak impact force and the peak inversion velocity), avoiding the deviation caused by subjective judgment.

[0106] The attention layer weight output focuses on the key spatiotemporal segments of the model's decision-making (such as the spatial distribution of pressure at the moment of landing, or the change of shear force in a certain time interval), revealing when and where features have the greatest impact on risk assessment, and supplementing the spatiotemporal correlation that cannot be reflected by the contribution value of a single feature.

[0107] The combination of these two approaches forms a multi-dimensional analysis of single-feature contribution and spatiotemporal key segments, making the identification of core risk characteristics more comprehensive and accurate, and avoiding the omission of key factors hidden in spatiotemporal changes.

[0108] Traditional black-box models often struggle to trace the root causes of risk assessments. However, this invention's approach, through SHAP values ​​and attention weights, makes the model's calculation logic for the damage risk index transparent. Users not only know which features are critical but also understand how these features influence the model's decisions (e.g., changes in inversion velocity over a certain time period are given high attention weights, directly increasing the risk index). This interpretability enhances user trust in the risk assessment results and provides a clear basis for the rationality of subsequent recommendations.

[0109] The core biomechanical features identified by ranking feature contributions are the direct targets for generating optimization suggestions. For example, if the SHAP value shows that the impact load rate contributes the most, and the attention layer indicates that the pressure spatial distribution 0.1-0.3 seconds after landing is a critical segment, then the suggestion can precisely focus on adjusting the landing cushioning rhythm to reduce the impact load rate, and provide specific movement correction plans based on the abnormal pressure distribution during this period. This kind of suggestion based on precise targets is more likely to directly address the core problem than generalized guidance, improving the effectiveness of motion optimization and injury prevention.

[0110] More preferably, the shear force data detected by the shear force calculation module includes:

[0111] The corresponding output charge is obtained through the piezoelectric unit matrix of the shear force calculation module, and the outer shear force is calculated in real time based on the adjacent unit strain difference algorithm; the adjacent unit strain difference algorithm includes a first shear force formula, a second shear force formula, and a shear force direction formula;

[0112] The first shear force formula is: ;

[0113] The second shear force formula is: ;

[0114] The formula for the direction of shear force is: ,in, , , , These are four piezoelectric capacitors connected in parallel for the i-th detection region. These four parallel piezoelectric capacitors together form a piezoelectric unit matrix. and These are the calibration coefficients in the x-direction and y-direction determined through material mechanics calibration experiments. and These are the shear forces in the x-direction and y-direction of the i-th detection region, respectively; and These are the total shear force outputs in the x-direction and y-direction of the foot, respectively; The direction of the resultant force.

[0115] In this embodiment of the invention, the shear force is directly calculated based on the adjacent element strain differential algorithm combined with the output charge of the piezoelectric element matrix. The algorithm logic is clear and highly targeted, reducing errors in the data conversion process. By introducing calibration coefficients in the x and y directions, which are determined through material mechanics calibration experiments, the accuracy of shear force calculation in different directions is ensured, closely matching the physical characteristics of the actual testing scenario.

[0116] It can not only measure the shear force in the x and y directions of a single detection area, but also summarize the total shear force on the plantar surface, and obtain the direction of the resultant shear force through a directional formula. This multi-dimensional data comprehensively reflects the magnitude, distribution, and directional characteristics of the plantar shear force, overcoming the limitations of single-dimensional measurement and providing complete data for extracting key motion features such as peak shear force and directional angle.

[0117] The structural design of the piezoelectric unit matrix and the efficiency of the differential algorithm support real-time calculation of shear force data, enabling the capture of instantaneous changes in shear force during movement (such as landing and pushing off). Shear force is calculated separately for different detection areas on the sole of the foot, accurately locating regions of concentrated shear force and providing fine-grained data support for analyzing risk points of uneven force distribution on the sole during movement. As a core component of multimodal biomechanical data, the accuracy of shear force data directly affects the quality of motion feature vector extraction.

[0118] The standardized, high-precision shear force data output by this calculation method can be efficiently integrated with plantar pressure data and kinematic data, providing high-quality input for the CNN-LSTM model and indirectly improving the accuracy of injury risk index assessment.

[0119] More preferably, the shear force calculation module includes multiple sensor clusters, which are distributed in multiple detection areas on the sole of the foot. Each sensor cluster includes a piezoelectric unit matrix, and the detection areas include the outer heel area, the inner heel area, the first and second metatarsal bone areas, and the fourth and fifth metatarsal bone areas.

[0120] The shear force data detected by the shear force calculation module includes:

[0121] It receives the shear force component output by each shear force sensing unit in the sensing cluster in real time;

[0122] Compare the output signals of each unit to identify outlier signals;

[0123] A dynamic confidence weight is assigned to the output of each shear force sensing unit, wherein units with high consistency with the outputs of other units are assigned higher weights, and units identified as outliers are assigned lower weights or excluded.

[0124] Based on the aforementioned confidence weights, the shear force components of all effective units within the cluster are weighted and fused to output the total shear force of the biomechanical key region.

[0125] In this embodiment of the invention, the sensor clusters are arranged in zones in key biomechanical regions such as the lateral and medial sides of the heel, the first and second metatarsals, and the fourth and fifth metatarsals. These regions are where shear forces are concentrated during movement and are highly correlated with injury risk. Zoned measurement can accurately capture the differences in shear forces in different regions, avoiding the ambiguity of key region data caused by overall measurement, and providing fine-grained data for extracting regionalized motion features.

[0126] By identifying outlier signals and assigning dynamic confidence weights, abnormal data (such as erroneous outputs caused by sensor unit failures or transient interference) is effectively filtered out. Units with high consistency receive high weights, while outlier units are downweighted or excluded, reducing the impact of individual unit errors on the overall data and ensuring that the output shear force data more closely matches the actual stress state, thus enhancing its anti-interference capability.

[0127] The weighted fusion calculation of effective units within a cluster combines the output information of multiple sensing units, rather than relying on data from a single unit, reducing the randomness of single-point measurements. The dynamic weight allocation mechanism can adapt to the working state of sensing units under different motion scenarios in real time, ensuring that the data output remains stable and providing high-quality, highly consistent shear force data support for subsequent feature extraction and risk assessment.

[0128] The combination of partitioned distributed design and weighted fusion algorithm can quickly respond to the dynamic changes in shear force in different areas of the foot during movement (such as the heel bearing the force first upon landing, and the metatarsal area bearing the increased force upon pushing off the ground). It receives and processes data from each unit in real time, ensuring the real-time nature of shear force data, and perfectly adapts to the detection needs of dynamic sports scenarios such as running and jumping.

[0129] The total shear force output in this region is not dependent on a single module, but is obtained by comprehensively processing the output signals of four modules within the cluster through an advanced data fusion algorithm. The system compares the output values ​​of the four modules in real time. Ideally, they should be highly correlated. By setting a dynamic threshold, the system can identify outlier signals (p < 0.05) caused by local damage, drift, or interference. Simultaneously, the system assigns a confidence weight to the output of each module; modules with stable signals and high consistency with other modules have higher weights. The weights of modules identified as outliers are automatically reduced or even reset to zero. The specific formula is as follows:

[0130]

[0131]

[0132] Even if one or more modules within the cluster fail completely, the system can still use the data from the remaining three valid modules to provide accurate shear force measurements for that region, achieving functional degradation under fault conditions and greatly improving the robustness and reliability of the entire system. By decoupling the shear force vector through spatial multi-point measurement, tangential force measurement is achieved.

[0133] This invention embeds a commercially available six-axis IMU (Inertial Measurement Unit, typically comprising a three-axis accelerometer and a three-axis gyroscope) into a substrate in the arch of the foot. Its accelerometer range is typically set to ±16g, and the gyroscope range to ±2000° / s, to accommodate the extremely high accelerations and angular velocities that may occur during human movement. The sampling rate is set to above 200Hz to satisfy the Nyquist sampling theorem. Additionally, a magnetometer with an azimuth error of <1° is integrated for auxiliary calibration to reduce the cumulative azimuth error.

[0134] Hardware synchronization: A low-power MCU (microcontroller) is used as the main controller, which integrates a high-precision clock source. The MCU's hardware timer triggers all ADCs (analog-to-digital converters) to synchronously sample the pressure and shear force signals, and IMU data is synchronously read through the SPI interface. This hardware synchronization mechanism ensures the timing consistency of multi-source data and keeps the synchronization error at an extremely low level (theoretically less than 5ms).

[0135] Software Algorithm: A mobile app was developed to receive data via Bluetooth 5.2. A built-in convolutional neural network-long short-term memory (LSTM) combination network was used to process the data. The convolutional neural network was used to extract spatial features, while the LSM combination network was used to analyze time-series features, achieving simultaneous capture of changes in plantar multidimensional forces and spatial pose data across both spatial and temporal dimensions. Real-time rendering of mechanical cloud maps, motion-mechanical time-series alignment, and the marking and alarming of abnormal mechanical events (such as shear force exceeding thresholds and sudden changes in angular velocity) based on machine learning models such as support vector machines were achieved through image super-resolution reconstruction.

[0136] The model outputs a Damage Risk Index (DRI) of 0-1 in the embodiments of the present invention, with a DRI > 0.75 being considered high risk. If the system identifies a high-risk landing (e.g., DRI > 0.75 and mainly contributed by lateral shear force and inward angular velocity), it automatically generates targeted optimization suggestions.

[0137] By developing a mobile app and embedding a convolutional neuron-long short-term memory (LSTM) network, the system simultaneously captures changes in multidimensional plantar force and spatial pose data across both spatial and temporal dimensions. Combined with image super-resolution reconstruction technology, it enables real-time rendering of mechanical cloud maps, motion-mechanical time-series alignment, and abnormal mechanical event labeling and alarms based on machine learning models, thereby improving the accuracy and efficiency of data analysis.

[0138] The sensor structure in the embodiments of the present invention is as follows: Figure 5 As shown, a sandwich structure flexible sensing layer is used:

[0139] a) Vertical pressure monitoring module:

[0140] Static pressure region: a micro-convex piezoresistive unit made of carbon nanotubes (CNTs) and polydimethylsiloxane (PDMS).

[0141] Dynamic impact zone: Employs a laser-thinned lead zirconate titanate (PZT) piezoelectric ceramic unit array.

[0142] b) Shear force calculation module: It adopts a partitioned distributed design, and shear force sensing clusters are arranged in key areas of the sole of the foot. Each sensing cluster consists of four 2×2 matrix sensing modules based on PVDF piezoelectric film.

[0143] Motion attitude anchoring layer: Embedded with a commercial six-axis IMU (Inertial Measurement Unit).

[0144] 3. Multi-source data fusion terminal:

[0145] Hardware synchronization: A low-power MCU is used as the main controller, and a hardware timer is used to trigger all ADCs to synchronously sample the pressure and shear force signals.

[0146] Software Algorithm: Develop a mobile app that receives data via Bluetooth 5.2 and uses a built-in convolutional neuron-long short-term memory combined network for data processing.

[0147] Multiple shear force sensing units are arranged in a closely spaced square array to form a redundant sensing group; each shear force sensing unit includes: a piezoelectric film having a common lower electrode and multiple independent upper electrodes to form multiple piezoelectric capacitors; a force transmission structure precisely mounted above the center position of the multiple independent upper electrodes, with its bottom edge aligned with the corner of the multiple independent upper electrodes; wherein each shear force sensing unit is configured to: calculate the two axial shear force components acting on the unit through differential calculation based on the charge signals output by the multiple independent upper electrodes.

[0148] The system automatically generates personalized feedback, including high-risk alerts and specific suggestions, such as strengthening specific muscle groups and adjusting landing posture.

[0149] The specific working principle of this invention embodiment:

[0150] In the analysis of basketball players' jump and landing, the system can synchronously collect the following data through multimodal sensing: using PZT piezoelectric ceramic units to synchronously capture the dynamic impact force in the forefoot area at the moment of landing, and measuring the peak impact force in the forefoot area at the moment of landing; using a 2×2 piezoelectric unit matrix (unit spacing 5mm), the lateral shear force is calculated in real time based on the strain difference algorithm between adjacent units. When the shear force direction angle θ is detected to be continuously greater than 25° and the magnitude exceeds 15% of the body weight, it is determined to be a significant lateral shear force; the embedded IMU (MPU-9250, sampling rate 200Hz) measures the inversion velocity of the foot. When the peak angular velocity exceeds 150° / s, it is recorded as an extremely high inversion velocity.

[0151] All sensor data is synchronized by the MCU hardware clock (error <5ms). Feature vectors (peak impact force, peak shear force, shear force direction angle, peak inversion velocity, impact load rate) are extracted for each landing event. These feature vectors are then input into a convolutional neural network-long short-term memory (CNN-LSTM) trained on a large amount of biomechanical data. The CNN branch processes spatial pressure distribution images from piezoelectric and piezoresistive arrays, while the LSTM branch processes time-series data from the IMU. The model outputs a 0-1 Damage Risk Index (DRI), with a DRI > 0.75 considered high-risk. If the system identifies a high-risk landing (e.g., DRI > 0.75 and primarily contributed by lateral shear force and inversion velocity), it automatically generates targeted optimization suggestions. For example, if excessive inversion velocity is detected upon landing, it suggests strengthening the gluteus medius and attempting to align the knee with the second metatarsal upon landing; if the impact load rate is too high, it suggests increasing the knee flexion angle to prolong the cushioning time, with a target cushioning time > 50ms, etc.

[0152] This embodiment details the application of the system in basketball injury prevention. A professional basketball player wore this flexible insole to perform a standard vertical jump landing test. The data acquisition process is as follows: The system synchronously acquired multimodal data at the moment of landing at a sampling rate of 1kHz. Among them, the PZT piezoelectric unit array measured the peak dynamic impact force in the forefoot region to be 3.5 times the body weight (approximately 2450N, assuming a body weight of 70kg), and the impact load rate was as high as 85kN / s. At almost the same time point (hardware synchronization error <5ms), the piezoelectric shear force matrix arranged in the forefoot calculated a significant outward shear force with a direction angle θ=42° and a magnitude of approximately 18% of the body weight (approximately 123N) in the subtotal region of the third metatarsal bone according to the differential algorithm. At the same time, the six-axis IMU (MPU-9250) embedded in the arch of the foot measured the peak inversion velocity of the foot to be 195° / s. Data Analysis and Decision-Making Process: All data is transmitted to the terminal APP via Bluetooth 5.2. The built-in CNN-LSTM fusion model immediately processes the spatiotemporally aligned multidimensional feature vectors (impact force, shear force, angular velocity), outputting an Injury Risk Index (DRI) of 0.88 (>0.75, a high-risk threshold). The system automatically generates personalized feedback: Alarm: High-risk landing posture detected. Main causes: Excessive lateral shear force (123N) combined with high-speed foot inversion (195° / s). Recommendations: ① Strengthen the gluteus medius and peroneus muscles; ② Intentionally control the knee to align with the second toe upon landing, increasing the knee flexion angle to >30° to prolong the cushioning time. Based on this, the coach develops a targeted biomechanical training program for the athlete.

[0153] In this embodiment of the invention, by receiving pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and kinematic data sequences detected by an inertial measurement unit, biomechanical information during movement can be comprehensively captured from multiple dimensions. Compared to a single data acquisition method, the fusion of multimodal data can more accurately reflect the true state of the human body during movement, providing a richer and more accurate data foundation for subsequent movement state analysis.

[0154] Example 2

[0155] Please see Figure 6 , Figure 6 This is a schematic diagram of the system for analyzing motion state based on plantar pressure, as disclosed in an embodiment of the present invention. Figure 6 As shown, the system for analyzing motion state based on plantar pressure may include:

[0156] Receiving module 21: used to receive multimodal biomechanical data related to a single landing event, the multimodal biomechanical data including pressure distribution data detected by the plantar pressure sensor array, shear force data detected by the shear force calculation module, and kinematic data sequence detected by the inertial measurement unit;

[0157] Feature extraction module 22: used to extract corresponding motion feature vectors from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate;

[0158] Identification module 23: used to input the motion feature vector into the motion state analysis model for identification to obtain the corresponding damage risk index, and output the corresponding damage risk index.

[0159] In this embodiment of the invention, by receiving pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and kinematic data sequences detected by an inertial measurement unit, biomechanical information during movement can be comprehensively captured from multiple dimensions. Compared to a single data acquisition method, the fusion of multimodal data can more accurately reflect the true state of the human body during movement, providing a richer and more accurate data foundation for subsequent movement state analysis.

[0160] Example 3

[0161] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. The electronic device can be a computer, a server, etc. Of course, in certain cases, it can also be a mobile phone, tablet computer, monitoring terminal, or other smart device, as well as an image acquisition device with processing capabilities. Figure 7 As shown, the electronic device may include:

[0162] Memory 510 storing executable program code;

[0163] Processor 520 coupled to memory 510;

[0164] The processor 520 calls the executable program code stored in the memory 510 to execute some or all of the steps in the method for analyzing motion state based on plantar pressure in Embodiment 1.

[0165] This invention discloses a computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps in the method for analyzing motion state based on plantar pressure in Embodiment 1.

[0166] This invention also discloses a computer program product, wherein when the computer program product is run on a computer, the computer performs some or all of the steps in the method for analyzing motion state based on plantar pressure in Embodiment 1.

[0167] This invention also discloses an application publishing platform, which is used to publish computer program products. When the computer program products are run on a computer, the computer performs some or all of the steps in the method for analyzing motion state based on plantar pressure in Embodiment 1.

[0168] In various embodiments of the present invention, it should be understood that the sequence number of each process does not necessarily imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0169] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they can be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0171] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several requests to cause a computer device (which can be a personal computer, server, or network device, specifically a processor in the computer device) to execute some or all of the steps of the methods described in the various embodiments of the present invention.

[0172] In the embodiments provided by this invention, it should be understood that B corresponding to A means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0173] Those skilled in the art will understand that some or all of the steps in the various methods of the embodiments described can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0174] The foregoing has provided a detailed description of the method, system, electronic device, and storage medium for analyzing motion state based on plantar pressure disclosed in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for analyzing motion state based on plantar pressure, characterized in that, include: Receive multimodal biomechanical data related to a single landing event, including pressure distribution data detected by a plantar pressure sensor array, shear force data detected by a shear force calculation module, and a sequence of kinematic data detected by an inertial measurement unit; The corresponding motion feature vectors are extracted from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate; The motion feature vector is input into the motion state analysis model for identification to obtain the corresponding damage risk index, and the corresponding damage risk index is output.

2. The method for analyzing motion state based on plantar pressure as described in claim 1, characterized in that, The method for motion state analysis further includes: Receive low-resolution pressure data frames detected by the plantar pressure sensor array; The low-resolution stress data frame is input into a trained image super-resolution reconstruction model; wherein, the image super-resolution reconstruction model is a deep learning model based on a convolutional neural network, which is trained by a large number of low-resolution stress matrices and corresponding high-resolution stress distribution ground value data, and the image super-resolution reconstruction model is used to learn the mapping relationship from low-resolution stress patterns to high-resolution stress patterns. The image super-resolution reconstruction model outputs a corresponding high-resolution pressure distribution matrix, and a corresponding mechanical contour map is generated based on the high-resolution pressure distribution matrix. The mechanical cloud diagram is displayed on the user interface.

3. The method for analyzing motion state based on plantar pressure as described in claim 2, characterized in that, The motion state analysis model is a network model formed by combining a convolutional neural network and a long short-term memory network. The step of inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding injury risk index includes: The spatial pressure distribution image sequence detected by the plantar pressure sensing array is input into a convolutional neural network to extract spatial features; The motion feature vectors of the time series are input into a long short-term memory network to extract the corresponding time series features; The spatial features extracted by the convolutional neural network and the time-series features extracted by the long short-term memory network are fused at the back end of the model and used together to calculate the damage risk index.

4. The method for analyzing motion state based on plantar pressure as described in claim 3, characterized in that, After inputting the motion feature vector into the motion state analysis model for identification to obtain the corresponding injury risk index, the method further includes: If the damage risk index exceeds the set risk threshold, targeted biomechanical optimization suggestions will be automatically generated. The output of the corresponding damage risk index includes: Output the corresponding damage risk index and targeted biomechanical optimization suggestions.

5. The method for analyzing motion state based on plantar pressure as described in claim 4, characterized in that, The automatically generated targeted biomechanical optimization suggestions include: Identify one or more biomechanical features that contribute most to the damage risk index; Based on one or more identified biomechanical features, select matching explanatory statements and training suggestions from a pre-defined knowledge base; The explanatory statements and training suggestions are combined to form the targeted biomechanical optimization suggestions.

6. The method for analyzing motion state based on plantar pressure as described in claim 5, characterized in that, The identification of one or more biomechanical features that contribute most to the damage risk index includes: The contribution of each feature in the motion feature vector to the corresponding injury risk index is calculated using the SHAP method. Obtain the weight outputs of the built-in attention layers in the CNN-LSTM combined network to determine the most critical data segments for model decisions in both the temporal and spatial dimensions; The features were ranked based on their contribution, and the biomechanical feature with the highest contribution was determined.

7. The method for analyzing motion state based on plantar pressure as described in claim 1, characterized in that, The shear force data detected by the shear force calculation module includes: The corresponding output charge is obtained through the piezoelectric unit matrix of the shear force calculation module, and the outer shear force is calculated in real time based on the adjacent unit strain difference algorithm; the adjacent unit strain difference algorithm includes a first shear force formula, a second shear force formula, and a shear force direction formula; The first shear force formula is: ； The second shear force formula is: ； The formula for the direction of shear force is: ,in, , , , These are four piezoelectric capacitors connected in parallel for the i-th detection region. These four parallel piezoelectric capacitors together form a piezoelectric unit matrix. and These are the calibration coefficients in the x-direction and y-direction determined through material mechanics calibration experiments. and These are the shear forces in the x-direction and y-direction of the i-th detection region, respectively; and These are the total shear force outputs in the x-direction and y-direction of the foot, respectively; The direction of the resultant force.

8. The method for analyzing motion state based on plantar pressure as described in claim 7, characterized in that, The shear force calculation module includes multiple sensor clusters, which are distributed in multiple detection areas on the sole of the foot. Each sensor cluster includes a piezoelectric unit matrix. The detection areas include the outer heel area, the inner heel area, the first and second metatarsal bone areas, and the fourth and fifth metatarsal bone areas. The shear force data detected by the shear force calculation module includes: It receives the shear force component output by each shear force sensing unit in the sensing cluster in real time; Compare the output signals of each unit to identify outlier signals; A dynamic confidence weight is assigned to the output of each shear force sensing unit, wherein units with high consistency with the outputs of other units are assigned higher weights, and units identified as outliers are assigned lower weights or excluded. Based on the aforementioned confidence weights, the shear force components of all effective units within the cluster are weighted and fused to output the total shear force of the biomechanical key region.

9. A system for analyzing motion state based on plantar pressure, characterized in that, include: Receiving module: used to receive multimodal biomechanical data related to a single landing event, including pressure distribution data detected by the plantar pressure sensor array, shear force data detected by the shear force calculation module, and kinematic data sequence detected by the inertial measurement unit; Feature extraction module: used to extract corresponding motion feature vectors from the multimodal biomechanical data, wherein the motion feature vectors include peak impact force, peak shear force, shear force direction angle, peak inversion angular velocity, and impact load rate; Identification module: used to input the motion feature vector into the motion state analysis model for identification to obtain the corresponding damage risk index, and output the corresponding damage risk index.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program causes a computer to perform the method for analyzing motion state based on plantar pressure as described in any one of claims 1 to 7.

Images