Foot biometric monitoring and evaluation system based on multi-modal sensor feedback

Abstract

The present application relates to the technical field of human biological feature measurement, and discloses a foot biological feature monitoring and evaluation system based on multi-modal sensing feedback, comprising: a signal acquisition module that collects plantar vertical pressure, horizontal shear force and contact surface deformation signals and fuses them into concurrent time sequence signals; a feature extraction module that generates a load query vector according to the concurrent time sequence signals, iterates a standard feature library to match a minimum Euclidean distance record to generate a state quantization result; and a state evaluation module that compares an initial reference model to calculate a deviation degree and generates a state drift parameter representing a physiological evolution trajectory, wherein the present application eliminates static data faults through a multi-source signal concurrent mechanism, converts high-dimensional time sequence analysis into discrete feature retrieval, avoids the power overload of equation solving, and outputs a low-delay and high-fidelity physiological state evaluation result.

Description

Technical Field

[0001] This invention relates to a foot biometric monitoring and evaluation system based on multimodal sensing feedback, belonging to the field of human biometric measurement technology. Background Technology

[0002] Currently, the monitoring of human foot biomechanical characteristics generally uses plantar pressure sensor arrays to collect discrete vertical load distribution signals as the basic parameters for gait stability assessment. As the monitoring conditions extend to continuous movement scenarios, the physical interaction between the sole and the support surface exhibits high-frequency fluctuations and time-varying characteristics. Hardware architectures that rely on a single vertical pressure acquisition cannot capture horizontal shear stress and dynamic deformation parameters of the arch structure, resulting in physical dimension truncation of the raw data during the acquisition process. Consequently, the system outputs a static judgment benchmark for deviation when analyzing gait compensation movements, and accumulates evaluation errors during continuous long-term monitoring.

[0003] Beyond the lack of data acquisition dimensions at the hardware level, existing monitoring and auxiliary protection systems generally suffer from lagging assessment and control logic, making it difficult to achieve real-time dynamic monitoring of physiological states. For example, Chinese utility model patent CN203898529U discloses a magnetic restraint belt. Although this solution utilizes magnetic induction locking to improve physical protection and safety, it is essentially still a static mechanical fixation device. Its control logic is limited to a binary switching level of locking and unlocking, completely failing to perceive and provide real-time data feedback on the biomechanical characteristics of the patient's foot in a medical environment. Consequently, when facing long-term monitoring conditions, this type of technology not only fails to obtain key mechanical characteristics through multimodal fusion, but also... Furthermore, there is a lack of intelligent algorithms that can automatically analyze physiological evolution trajectories from high-dimensional time-series signals. This situation of data acquisition gaps and dulled evaluation mechanisms means that existing technologies often output static judgment benchmarks for deviations when analyzing gait compensation movements, and accumulate evaluation errors during continuous monitoring. Simply increasing the spatial deployment density or time sampling frequency of pressure sensor arrays can only refine the data granularity of vertical loads but cannot restore the missing nonlinear spatiotemporal characteristics. Increasing the data flow acquired by single-modal hardware density allows the post-processing unit to use a global physical simulation model to deduce the missing dynamic mechanical state. The nonlinear mechanical transmission characteristics of human soft tissue generate high-dimensional matrix operations after being involved in the solution mechanism of the continuum equation, causing iterative calculation divergence and system computing power overload.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a concurrent sensing mechanism that covers multi-dimensional parameters of vertical load and shear deformation, reconstruct discrete temporal characteristics without the need for complex physical simulation, and establish a closed-loop correction path with adaptive feature alignment capability. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A foot biometric monitoring and evaluation system based on multimodal sensing feedback, comprising: The signal acquisition module includes a pressure sensing unit and a flexible strain sensing unit. It collects vertical pressure signals, horizontal shear force signals and contact surface deformation signals at the foot interface, and fuses the vertical pressure signals, horizontal shear force signals and contact surface deformation signals into concurrent timing signals. The feature extraction module, connected to the signal acquisition module, receives concurrent time-series signals, extracts peak distribution parameters from the concurrent time-series signals, defines a feature projection region based on the peak distribution parameters, converts the concurrent time-series signals within the feature projection region into peak pressure scalar and impulse integral values, concatenates the peak pressure scalar and impulse integral values ​​to generate a load query vector, traverses a preset standard feature library, calculates the Euclidean distance between the load query vector and each standard tolerance vector in the standard feature library, extracts the standard load record corresponding to the minimum value of the Euclidean distance, and generates a state quantization result based on the standard load record. The state assessment module, connected to the feature extraction module, receives the state quantization results, compares the state quantization results with the pre-stored initial reference model, calculates the degree of deviation between the state quantization results and the initial reference model, generates state drift parameters representing the physiological evolution trajectory based on the degree of deviation, and outputs the state drift parameters.

[0006] Preferably, the signal acquisition module further includes a multi-channel synchronous sampling unit; the multi-channel synchronous sampling unit is connected to the pressure sensing unit and the flexible strain sensing unit respectively; the multi-channel synchronous sampling unit sends a global clock pulse signal with a unified timestamp to the pressure sensing unit and the flexible strain sensing unit; the pressure sensing unit and the flexible strain sensing unit perform synchronous sampling actions according to the global clock pulse signal, and add a unified timestamp identifier to the acquired vertical pressure signal, horizontal shear force signal and contact surface deformation signal to generate concurrent timing signals.

[0007] Preferably, the execution logic of the feature extraction module for extracting the peak distribution parameter in the concurrent time-series signal is as follows: the feature extraction module identifies the positive and negative zero-crossing points of the vertical pressure signal relative to the preset reference equilibrium surface in the concurrent time-series signal; the feature extraction module establishes the time span between the positive and negative zero-crossing points as a single basic gait cycle; the feature extraction module extracts the absolute maximum point of the vertical pressure signal within the single basic gait cycle as the peak distribution parameter.

[0008] Preferably, the standard feature library contains pre-set standard tolerance vectors classified by age group and foot morphology; when the feature extraction module calculates the Euclidean distance between the load query vector and the standard tolerance vector, it calculates the square of the numerical difference between each dimension component of the load query vector and the corresponding dimension component of the standard tolerance vector, and then performs a square root operation after summing the squares of all numerical differences to obtain the Euclidean distance.

[0009] Preferably, when extracting the standard load record corresponding to the minimum Euclidean distance, the feature extraction module calculates the product of the Euclidean distance and the preset environmental interference constant to obtain the comprehensive matching index. The feature extraction module determines the matching level of the standard load record based on the comprehensive matching index, and the determination rule follows the formula: M=D×K, where M is the comprehensive matching index, D is the calculated Euclidean distance, and K is the preset environmental interference constant characterizing the degree of nonlinear deformation interference of the foot contact surface. When the comprehensive matching index is less than the preset deviation threshold, the feature extraction module locks the corresponding standard load record as the target feature record.

[0010] Preferably, the state assessment module includes a historical trend tracking unit; the historical trend tracking unit has a storage sequence inside; the state assessment module stores the state drift parameters generated at multiple consecutive time points into the storage sequence in sequence, and calculates the time change rate of the state drift parameters in the storage sequence; when the time change rate of the state drift parameters continues to be greater than the preset change rate benchmark value, the state assessment module generates a biomechanical imbalance warning signal and outputs the biomechanical imbalance warning signal to the external display terminal.

[0011] Preferably, the system further includes a foot three-dimensional morphology mapping module; the foot three-dimensional morphology mapping module is connected to the state evaluation module; after the state evaluation module outputs state drift parameters, the foot three-dimensional morphology mapping module receives the state quantization result and acquires the pre-stored standardized plantar basic boundary mesh data; the foot three-dimensional morphology mapping module discretizes the three-dimensional envelope space defined by the standardized plantar basic boundary mesh data into a cubic voxel array with a side length of 2mm to 5mm to construct a foot physiological morphology reference space; the foot three-dimensional morphology mapping module adjusts the spatial coordinate distribution of specific voxel nodes in the cubic voxel array according to the state quantization result to generate foot three-dimensional physiological deformation mapping data.

[0012] Preferably, the foot three-dimensional morphology mapping module has a pre-set support density mapping rule; the foot three-dimensional morphology mapping module extracts the peak pressure scalar contained in the state quantization result and compares the peak pressure scalar with the preset compressive stress benchmark value; when the peak pressure scalar is greater than the preset compressive stress benchmark value, the foot three-dimensional morphology mapping module increases the node filling density of the corresponding local cubic voxel array according to the support density mapping rule, and outputs the density-adjusted foot plantar force distribution voxel model data.

[0013] Preferably, the system further includes a plantar shear force feature analysis unit; the plantar shear force feature analysis unit is connected to the foot three-dimensional morphology mapping module; the plantar shear force feature analysis unit acquires the horizontal shear force signal in the concurrent time-series signal and calculates the high-frequency fluctuation variance value of the horizontal shear force signal in a single gait cycle; the plantar shear force feature analysis unit generates a plantar local friction risk index based on the high-frequency fluctuation variance value, and merges the plantar local friction risk index with the plantar force distribution voxel model data for output.

[0014] Preferably, the signal acquisition module also includes a physical limiting unit and an electrostatic shielding unit; the pressure sensing unit and the flexible strain sensing unit are jointly configured in the physical limiting unit to maintain the consistency of the detection space; the electrostatic shielding unit is configured in the bottom space of the physical limiting unit, and receives the external static charge generated by the friction of the foot when the physical limiting unit deforms, and conducts the external static charge to the grounding terminal.

[0015] Compared to existing technologies, the advantages of this invention are as follows: In multimodal sensing feedback foot biometric monitoring, a multimodal biomechanical sensing network is constructed. By setting up pressure film sensors and flexible strain sensor arrays in parallel, continuous deformation signals of vertical pressure, horizontal shear force, and local contact surfaces of the foot are collected concurrently throughout the entire time period. The processing unit receives the above-mentioned multidimensional concurrent signals, performs gait period division and temporal fusion calculations, and directly generates dynamic biometric parameters based on the extracted peak distribution. This multi-source signal concurrent extraction mechanism eliminates the data discontinuity phenomenon caused by single-modal static sampling, objectively maps the complex dynamic interaction and evolution process between the sole and the support interface, and improves the physical dimension integrity and original data fidelity of the biomechanical feature acquisition process. The system employs a data dimensionality reduction and reconstruction architecture based on discretization mapping in its intelligent processing module. The processing unit converts dynamic sensing data of a specific projection area into peak pressure scalar and impulse integral values, and sequentially splices them to generate a one-dimensional spatiotemporal load query vector. The system then traverses a pre-set standard feature database based on this query vector, calculates the Euclidean distance between the input vector and the standard load tolerance vector, and extracts the data record corresponding to the minimum distance value. This processing link transforms the complex analytical requirements of high-dimensional nonlinear time-series signals into distance measurement and discrete data retrieval actions of the underlying feature vector, bypassing the extreme computational power consumption and convergence / divergence risks caused by traditional continuous medium physical equation simulation solutions, and outputs low-latency and highly deterministic biomechanical state quantification results. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system functional modules and feature matching principle of the multimodal sensing of the present invention; Figure 2 This is a timing diagram of the entire process of concurrent acquisition of multi-source signals and assessment of physiological state in this invention.

[0017] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] A foot biometric monitoring and assessment system based on multimodal sensing feedback, comprising: The signal acquisition module includes a pressure sensing unit and a flexible strain sensing unit. It collects vertical pressure signals, horizontal shear force signals and contact surface deformation signals at the foot interface, and fuses the vertical pressure signals, horizontal shear force signals and contact surface deformation signals into concurrent timing signals. The feature extraction module, connected to the signal acquisition module, receives concurrent time-series signals, extracts peak distribution parameters from the concurrent time-series signals, defines a feature projection region based on the peak distribution parameters, converts the concurrent time-series signals within the feature projection region into peak pressure scalar and impulse integral values, concatenates the peak pressure scalar and impulse integral values ​​to generate a load query vector, traverses a preset standard feature library, calculates the Euclidean distance between the load query vector and each standard tolerance vector in the standard feature library, extracts the standard load record corresponding to the minimum value of the Euclidean distance, and generates a state quantization result based on the standard load record. The state assessment module, connected to the feature extraction module, receives the state quantization results, compares the state quantization results with the pre-stored initial reference model, calculates the degree of deviation between the state quantization results and the initial reference model, generates state drift parameters representing the physiological evolution trajectory based on the degree of deviation, and outputs the state drift parameters.

[0020] Preferably, the signal acquisition module further includes a multi-channel synchronous sampling unit; the multi-channel synchronous sampling unit is connected to the pressure sensing unit and the flexible strain sensing unit respectively; the multi-channel synchronous sampling unit sends a global clock pulse signal with a unified timestamp to the pressure sensing unit and the flexible strain sensing unit; the pressure sensing unit and the flexible strain sensing unit perform synchronous sampling actions according to the global clock pulse signal, and add a unified timestamp identifier to the acquired vertical pressure signal, horizontal shear force signal and contact surface deformation signal to generate concurrent timing signals.

[0021] Preferably, the execution logic of the feature extraction module for extracting the peak distribution parameter in the concurrent time-series signal is as follows: the feature extraction module identifies the positive and negative zero-crossing points of the vertical pressure signal relative to the preset reference equilibrium surface in the concurrent time-series signal; the feature extraction module establishes the time span between the positive and negative zero-crossing points as a single basic gait cycle; the feature extraction module extracts the absolute maximum point of the vertical pressure signal within the single basic gait cycle as the peak distribution parameter.

[0022] Preferably, the standard feature library contains pre-set standard tolerance vectors classified by age group and foot morphology; when the feature extraction module calculates the Euclidean distance between the load query vector and the standard tolerance vector, it calculates the square of the numerical difference between each dimension component of the load query vector and the corresponding dimension component of the standard tolerance vector, and then performs a square root operation after summing the squares of all numerical differences to obtain the Euclidean distance.

[0023] Preferably, when extracting the standard load record corresponding to the minimum Euclidean distance, the feature extraction module calculates the product of the Euclidean distance and the preset environmental interference constant to obtain the comprehensive matching index. The feature extraction module determines the matching level of the standard load record based on the comprehensive matching index, and the determination rule follows the formula: M=D×K, where M is the comprehensive matching index, D is the calculated Euclidean distance, and K is the preset environmental interference constant characterizing the degree of nonlinear deformation interference of the foot contact surface. When the comprehensive matching index is less than the preset deviation threshold, the feature extraction module locks the corresponding standard load record as the target feature record.

[0024] Preferably, the state assessment module includes a historical trend tracking unit; the historical trend tracking unit has a storage sequence inside; the state assessment module stores the state drift parameters generated at multiple consecutive time points into the storage sequence in sequence, and calculates the time change rate of the state drift parameters in the storage sequence; when the time change rate of the state drift parameters continues to be greater than the preset change rate benchmark value, the state assessment module generates a biomechanical imbalance warning signal and outputs the biomechanical imbalance warning signal to the external display terminal.

[0025] Preferably, the system further includes a foot three-dimensional morphology mapping module; the foot three-dimensional morphology mapping module is connected to the state evaluation module; after the state evaluation module outputs state drift parameters, the foot three-dimensional morphology mapping module receives the state quantization result and acquires the pre-stored standardized plantar basic boundary mesh data; the foot three-dimensional morphology mapping module discretizes the three-dimensional envelope space defined by the standardized plantar basic boundary mesh data into a cubic voxel array with a side length of 2mm to 5mm to construct a foot physiological morphology reference space; the foot three-dimensional morphology mapping module adjusts the spatial coordinate distribution of specific voxel nodes in the cubic voxel array according to the state quantization result to generate foot three-dimensional physiological deformation mapping data.

[0026] Preferably, the foot three-dimensional morphology mapping module has a pre-set support density mapping rule; the foot three-dimensional morphology mapping module extracts the peak pressure scalar contained in the state quantization result and compares the peak pressure scalar with the preset compressive stress benchmark value; when the peak pressure scalar is greater than the preset compressive stress benchmark value, the foot three-dimensional morphology mapping module increases the node filling density of the corresponding local cubic voxel array according to the support density mapping rule, and outputs the density-adjusted foot plantar force distribution voxel model data.

[0027] Preferably, the system further includes a plantar shear force feature analysis unit; the plantar shear force feature analysis unit is connected to the foot three-dimensional morphology mapping module; the plantar shear force feature analysis unit acquires the horizontal shear force signal in the concurrent time-series signal and calculates the high-frequency fluctuation variance value of the horizontal shear force signal in a single gait cycle; the plantar shear force feature analysis unit generates a plantar local friction risk index based on the high-frequency fluctuation variance value, and merges the plantar local friction risk index with the plantar force distribution voxel model data for output.

[0028] Preferably, the signal acquisition module also includes a physical limiting unit and an electrostatic shielding unit; the pressure sensing unit and the flexible strain sensing unit are jointly configured in the physical limiting unit to maintain the consistency of the detection space; the electrostatic shielding unit is configured in the bottom space of the physical limiting unit, and receives the external static charge generated by the friction of the foot when the physical limiting unit deforms, and conducts the external static charge to the grounding terminal.

[0029] Example 1: Under long-cycle, high-frequency dynamic foot load monitoring conditions, a highly nonlinear dynamic interaction continuously occurs between the sole of the foot and the support interface. Traditional architectures relying on a single vertical pressure acquisition cannot capture horizontal shear stress and dynamic deformation parameters of the arch structure. Inputting high-dimensional time-series biomechanical data into a global physical simulation model leads to computational convergence difficulties, causing the system to output a deviated static judgment benchmark when analyzing gait compensation movements, and accumulating evaluation errors during continuous monitoring. The foot biometric monitoring and evaluation system based on multimodal sensor feedback addresses the above physical constraints by relying on the built-in signal acquisition module. It uses a multi-channel synchronous sampling unit to send a global clock pulse signal with a unified timestamp to the pressure sensing unit and the flexible strain sensing unit, enabling the pressure sensing unit and the flexible strain sensing unit to adjust according to the global clock pulse signal. The clock pulse signal synchronously acquires the vertical pressure signal, horizontal shear force signal, and contact surface deformation signal at the foot interface. The multi-channel synchronous sampling unit fuses the vertical pressure signal, horizontal shear force signal, and contact surface deformation signal into a concurrent timing signal. The feature extraction module receives the concurrent timing signal and identifies the positive and negative zero crossing points of the vertical pressure signal relative to the preset reference equilibrium surface. The time span between the positive and negative zero crossing points is established as a single basic gait cycle. Within this single basic gait cycle, the absolute maximum point of the vertical pressure signal is extracted as a peak distribution parameter to define the feature projection area. Finally, the concurrent timing signal within the feature projection area is converted into a peak pressure scalar and an impulse integral value, and the peak pressure scalar and impulse integral value are concatenated to generate a load query vector.

[0030] The feature extraction module iterates through a standard feature library containing standard tolerance vectors categorized by age group and foot morphology using the load query vector. It calculates the square of the numerical difference between each dimension component of the load query vector and the corresponding dimension component of the standard tolerance vector. The square root of the sum of all squared differences is used to obtain the Euclidean distance. The system then calculates the product of this Euclidean distance and a preset environmental interference constant according to a decision rule, yielding a comprehensive matching index. This calculation satisfies the formula M = D × K, where M is the comprehensive matching index, D is the calculated Euclidean distance, and K is a preset environmental interference constant characterizing the degree of nonlinear deformation interference on the plantar contact surface. When the comprehensive matching index is less than a preset deviation threshold, the feature extraction module locks the corresponding standard load record as the target feature record and generates a state quantization result based on this standard load record. This processing link utilizes discrete data retrieval. Distance calculation with the underlying feature vector replaces global physical equation simulation, avoiding computational overload and the risk of divergence in nonlinear matrix solutions. The state assessment module receives the state quantization result and compares it with the pre-stored initial reference model, calculating the deviation between the state quantization result and the initial reference model. Based on the deviation, it generates state drift parameters representing the physiological evolution trajectory. The historical trend tracking unit sequentially stores the state drift parameters generated at multiple consecutive time points into an internal storage sequence, calculates the time change rate of the state drift parameters in the storage sequence, and when the time change rate of the state drift parameters continuously exceeds the preset change rate benchmark value, the state assessment module generates a biomechanical imbalance warning signal and outputs it to the external display terminal, maintaining the objectivity and consistency of physiological state evaluation. The foot three-dimensional morphology mapping module receives the state quantization result and extracts the peak pressure scalar. With impulse integral value The coordinate offset Δz of the voxel node in the stress area relative to the foot physiological morphology reference space is calculated. The value of Δz is determined by the preset pressure deformation mapping coefficient. The spatial distribution of adjacent voxel nodes is corrected by the Laplace smoothing operator, and the load scalar is converted into three-dimensional physiological deformation mapping data of the foot.

[0031] Example 2: This example addresses the engineering problem of nonlinear coupling between minute sliding at the support interface and arch collapse masking true biomechanical characteristics in long-term dynamic foot contact state monitoring scenarios. A dynamic foot load simulation test platform is constructed. The platform integrates a multi-axis force stage with a measurement resolution of 0.1N and a flexible dielectric elastomer strain sensor array with a strain resolution of 0.1%. Gaussian white noise with a signal-to-noise ratio of 15dB is injected into the underlying driver of the test platform, while simulated power frequency interference harmonics at a frequency of 50Hz are superimposed to create a test environment containing physical disturbances. The sampling frequency setting of the global clock pulse signal sent by the multi-channel synchronous sampling unit needs to strike a balance between real-time feature capture and backend data throughput load. Based on the built-in control logic, as the main lobe width of the contact surface deformation signal increases, the sampling frequency approaches the upper limit of the value range to avoid signal aliasing. Under this mechanism, the sampling frequency of the global clock pulse signal is set to 500Hz for normal fast-walking conditions. The test group uses concurrent timing signals including vertical pressure signal, horizontal shear force signal and contact surface deformation signal to extract features. The first control group only collects vertical pressure signal and horizontal shear force signal. The second control group sets the preset environmental interference constant to the limit value beyond the reasonable range. The test platform outputs the original sensing electrical signal containing the aforementioned 15dB noise. The multi-channel synchronous sampling unit of the test group fuses the signals of each channel according to the sampling frequency of 500Hz. The feature extraction module extracts the positive zero crossing point and the negative zero crossing point of the vertical pressure signal relative to the preset reference equilibrium surface in the concurrent timing signal. Due to the lack of contact surface deformation signal compensation, the time span of the positive zero crossing point in the original data of the first control group fluctuates drastically, with a time range of 112.5ms.

[0032] The experimental group combined the contact surface deformation signal to correct the horizontal shear force deviation, so that the time span between the positive and negative zero crossings converged within a single basic gait period of 1.15s to 1.18s. When extracting peak distribution parameters, changing the sampling frequency induced a nonlinear evolution of feature recognition accuracy. When the sampling frequency was below 250Hz, the peak pressure scalar extraction deviation rate reached 14.5%. As the sampling frequency increased to 400Hz, the extraction deviation rate decreased and remained in the saturation plateau region of 1.2% to 1.5%. Further increasing the sampling frequency to above 600Hz, the incorporation of high-frequency noise caused the impulse integral value to diverge, and the extraction deviation rate deteriorated to 6.8%. Experimental data confirmed that a sampling frequency of 500Hz constituted the optimal parameter range that balanced feature recognition accuracy and noise resistance. The feature extraction module converted the concurrent time-series signal in the feature projection region into a peak pressure scalar of 42.5kPa and an impulse integral value of 12.8N·s, and concatenated the above parameters to generate a load query vector. The feature extraction module calculated the load. The system calculates the comprehensive matching index when extracting the standard load record corresponding to the minimum value of the Euclidean distance between the query vector and each standard tolerance vector in the standard feature library. This calculation satisfies the formula M=D×K, where M is the comprehensive matching index, D is the calculated Euclidean distance, and K is a preset environmental interference constant characterizing the degree of nonlinear deformation interference of the plantar contact surface. When the preset environmental interference constant is set to 0.85, the comprehensive matching index calculated by the experimental group is less than the preset deviation threshold for three consecutive single basic gait cycles. The system then locks the corresponding standard load record as the target feature record. The second control group forcibly sets the preset environmental interference constant to an out-of-range value of 2.50, which drastically increases the comprehensive matching index and reduces the target feature record matching accuracy to 41.5%. Due to the lack of the contact surface deformation signal dimension, the minimum Euclidean distance corresponding to the load query vector of the first control group is 314% larger than that of the experimental group, causing the time change rate of the state drift parameter generated by the state assessment module to abnormally exceed the preset change rate benchmark value.

[0033] Example 3: Under long-term dynamic foot monitoring conditions, the physical tension difference of the plantar fascia and the time-varying baseline drift of the flexible sensing interface will cause feature mapping distortion due to a fixed environmental interference constant. Before continuous monitoring starts, the signal acquisition module activates the parameter adaptive calibration link. The multi-channel synchronous sampling unit sends calibration trigger pulses to the pressure sensing unit and the flexible strain sensing unit, enabling the pressure sensing unit and the flexible strain sensing unit to acquire the zero-point reference level signal under no-load conditions and the static deformation signal under standard static load conditions. The feature extraction module receives the static deformation signal and extracts a steady-state monitoring window with a time span of 10 seconds. It calculates the discrete variance of the contact surface deformation signal within this time window to quantify the nonlinear elastic fluctuation of the plantar contact surface. Based on the quantified nonlinear elastic fluctuation, the feature extraction module establishes a preset environmental interference constant. The numerical calculation of this preset environmental interference constant satisfies the formula... , where K is the preset environmental interference constant, μ is the material damping coefficient of the flexible dielectric elastomer strain sensing array, and σ is the discrete variance of the calculated contact surface deformation signal.

[0034] The system acquires concurrent time-series signals within 50 consecutive single basic gait cycles in the initial stage and converts them into corresponding load query vector sets. The feature extraction module calculates the cluster center coordinates of this load query vector set in a multi-dimensional feature space and locks its corresponding underlying feature topology as the initial reference model. The state assessment module receives the real-time generated state quantization results in the subsequent monitoring stage. The state assessment module calls the initial reference model composed of cluster center coordinates to calculate the deviation between the state quantization results and the initial reference model, and generates state drift parameters based on the deviation. The historical trend tracking unit sequentially stores the state drift parameters at continuous time nodes into an internal storage sequence and calculates the time change rate of the state drift parameters in the storage sequence. The parameter calibration and model initialization link associates the variables in the comprehensive matching index with the physical biomechanical boundary, eliminating the static baseline offset caused by the general threshold and sensor aging residuals. The signal acquisition module acquires the zero-point reference level in the no-load state and calculates the discrete variance σ of the contact surface deformation signal within the steady-state monitoring window to characterize the nonlinear elastic fluctuation of the contact surface, according to the formula... A preset environmental interference constant K is determined, where μ is the damping coefficient of the sensing unit material. The temperature sensor inside the physical limiting unit is used to monitor the ambient temperature fluctuation, and the sensitivity gain compensation of the flexible sensing unit is performed according to the temperature deviation from the calibration value. This eliminates the deviation of the comprehensive matching index M caused by sensor temperature drift or material hysteresis, so that the time change rate of the state drift parameter reflects the physiological evolution of biomechanical characteristics.

[0035] Example 4: In the case of constructing a benchmark for large-scale heterogeneous foot states, the signal acquisition module continuously collects prior sample data covering different foot morphologies. The system inputs the initial physical measurement values ​​into the data preprocessing unit to perform outlier removal and time-series alignment. The feature extraction module extracts the peak distribution parameters of each sample under the basic gait and maps them to a high-dimensional feature space. The density clustering algorithm is used to perform unsupervised partitioning of the sample set in the high-dimensional feature space. For each partitioned sub-cluster, the statistical mean and standard deviation of the multi-dimensional feature components are calculated. The system defines the boundary envelope under different classification dimensions based on the above statistical parameters. The calculation of this boundary envelope satisfies the formula... Where E is the boundary envelope, The calculated statistical mean, The calculated standard deviation is given by λ, which is a preset envelope expansion coefficient. The system converts the boundary envelope into a standard tolerance vector and stores it into the standard feature library in sequence.

[0036] When the system faces the initial docking condition of a newly deployed object, the multi-channel synchronous sampling unit sends a start listening command to the pressure sensing unit and the flexible strain sensing unit for a duration of a preset calibration period. This enables the pressure sensing unit and the flexible strain sensing unit to acquire tentative vertical pressure signals and tentative contact surface deformation signals during the standard calibration action of the object. The feature extraction module converts the received tentative signals into initial tentative vectors and calculates the reference Euclidean distance between the initial tentative vectors and each standard tolerance vector in the standard feature library. Based on the minimum Euclidean distance, the system binds the currently deployed object to a specific foot morphology classification node, thereby establishing a personalized monitoring session and converging the retrieval range of the continuous monitoring phase to the exclusive feature subset under that classification node. The offline data filling and front-end node binding procedure establishes the physical benchmark for matching concurrent timing signals.

[0037] Example 5: Under conditions of complex terrain transient impacts and long-term sensor baseline drift interference, hard-coded evaluation thresholds can lead to fixed triggering conditions for biomechanical imbalance warning signals and false alarms in state analysis. Before performing routine monitoring, the state assessment module initiates a pre-dynamic threshold calibration procedure, controlling the multi-channel synchronous sampling unit to continuously collect and transmit time-series signals within 100 consecutive basic gait cycles of a specific test subject performing standard uniform linear walking. The feature extraction module calculates the absolute difference of state drift parameters between adjacent cycles within the aforementioned basic gait cycles, extracting the maximum value of the absolute difference as the individual baseline fluctuation. The state assessment module establishes a preset rate of change benchmark value based on this individual baseline fluctuation. The configuration logic of this parameter satisfies the formula... Where R is the preset rate of change benchmark value, η is the extracted maximum value, η is the environmental adjustment coefficient to adapt to different support surface material hardness, and W is the physiological safety tolerance constant preset based on large-scale sampling.

[0038] During the online monitoring phase, the signal acquisition module continuously monitors the instantaneous step slope of the contact surface deformation signal. When it exceeds the quantization boundary corresponding to the physical response limit of the sensor array, the feature extraction module triggers the abnormal data bypass elimination mechanism, marks the load query vector of the current gait cycle as an invalid matrix, and calls the mean vector of the previous three historical cycles to perform time-series interpolation replacement. This closed-loop control path, which integrates pre-quantization of background fluctuations and real-time compensation for instantaneous overload, anchors the abnormal warning boundary within the effective range formed by physical measurement characteristics and individual physiological baselines, filtering out signal analysis deviations introduced by nonlinear physical impacts.

[0039] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A foot biometric monitoring and assessment system based on multimodal sensing feedback, characterized in that, include: The signal acquisition module includes a pressure sensing unit and a flexible strain sensing unit. It collects vertical pressure signals, horizontal shear force signals and contact surface deformation signals at the foot interface, and fuses the vertical pressure signals, horizontal shear force signals and contact surface deformation signals into concurrent timing signals. The feature extraction module, connected to the signal acquisition module, receives concurrent time-series signals, extracts peak distribution parameters from the concurrent time-series signals, defines a feature projection region based on the peak distribution parameters, converts the concurrent time-series signals within the feature projection region into peak pressure scalar and impulse integral values, concatenates the peak pressure scalar and impulse integral values ​​to generate a load query vector, traverses a preset standard feature library, calculates the Euclidean distance between the load query vector and each standard tolerance vector in the standard feature library, extracts the standard load record corresponding to the minimum value of the Euclidean distance, and generates a state quantization result based on the standard load record. The state assessment module, connected to the feature extraction module, receives the state quantization results, compares the state quantization results with the pre-stored initial reference model, calculates the degree of deviation between the state quantization results and the initial reference model, generates state drift parameters representing the physiological evolution trajectory based on the degree of deviation, and outputs the state drift parameters.

2. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 1, characterized in that, The signal acquisition module also includes a multi-channel synchronous sampling unit; the multi-channel synchronous sampling unit is connected to the pressure sensing unit and the flexible strain sensing unit respectively; the multi-channel synchronous sampling unit sends a global clock pulse signal with a unified timestamp to the pressure sensing unit and the flexible strain sensing unit; the pressure sensing unit and the flexible strain sensing unit perform synchronous sampling actions according to the global clock pulse signal, and add a unified timestamp to the acquired vertical pressure signal, horizontal shear force signal and contact surface deformation signal to generate concurrent timing signals.

3. The foot biometric monitoring and evaluation system based on multimodal sensor feedback according to claim 1, characterized in that, The execution logic of the feature extraction module to extract the peak distribution parameter in the concurrent time series signal is as follows: the feature extraction module identifies the positive and negative zero crossing points of the vertical pressure signal relative to the preset reference equilibrium surface in the concurrent time series signal; the feature extraction module establishes the time span between the positive and negative zero crossing points as a single basic gait cycle; the feature extraction module extracts the absolute maximum point of the vertical pressure signal as the peak distribution parameter within the single basic gait cycle.

4. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 1, characterized in that, The standard feature library contains pre-set standard tolerance vectors classified by age group and foot morphology. When the feature extraction module calculates the Euclidean distance between the load query vector and the standard tolerance vector, it calculates the square of the numerical difference between each dimension component of the load query vector and the corresponding dimension component of the standard tolerance vector, and then performs a square root operation after summing the squares of all numerical differences to obtain the Euclidean distance.

5. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 4, characterized in that, When extracting the standard load record corresponding to the minimum Euclidean distance, the feature extraction module calculates the product of the Euclidean distance and the preset environmental disturbance constant to obtain the comprehensive matching index; features The extraction module determines the matching level of the standard load record based on the comprehensive matching index. The determination rule follows the formula: M=D×K, where M is the comprehensive matching index, D is the calculated Euclidean distance, and K is a preset environmental interference constant that characterizes the degree of nonlinear deformation interference of the foot contact surface. When the comprehensive matching index is less than the preset deviation threshold, the feature extraction module locks the corresponding standard load record as the target feature record.

6. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 1, characterized in that, The state assessment module includes a historical trend tracking unit; the historical trend tracking unit has a storage sequence inside; the state assessment module stores the state drift parameters generated at multiple consecutive time points into the storage sequence in sequence, and calculates the time change rate of the state drift parameters in the storage sequence; when the time change rate of the state drift parameters continues to be greater than the preset change rate benchmark value, the state assessment module generates a biomechanical imbalance warning signal and outputs the biomechanical imbalance warning signal to the external display terminal.

7. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 1, characterized in that, The system also includes a foot 3D morphology mapping module; the foot 3D morphology mapping module is connected to the state evaluation module; after the state evaluation module outputs the state drift parameters, the foot 3D morphology mapping module receives the state quantization results and obtains the pre-stored standardized foot base boundary mesh data; The foot 3D morphology mapping module discretizes the 3D envelope space defined by the standardized foot basic boundary mesh data into a cubic voxel array with a side length of 2mm to 5mm to construct the foot physiological morphology reference space. The foot 3D morphology mapping module adjusts the spatial coordinate distribution of specific voxel nodes in the cubic voxel array based on the state quantization results to generate 3D physiological deformation mapping data of the sole.

8. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 7, characterized in that, The foot 3D morphology mapping module has a pre-set support density mapping rule; the foot 3D morphology mapping module extracts the peak pressure scalar contained in the state quantization result and compares the peak pressure scalar with the preset compressive stress benchmark value; when the peak pressure scalar is greater than the preset compressive stress benchmark value, the foot 3D morphology mapping module increases the node filling density of the corresponding local cubic voxel array according to the support density mapping rule, and outputs the density-adjusted foot plantar force distribution voxel model data.

9. A foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 8, characterized in that, The system also includes a plantar shear force feature analysis unit; the plantar shear force feature analysis unit is connected to the foot three-dimensional morphology mapping module; the plantar shear force feature analysis unit acquires the horizontal shear force signal in the concurrent time-series signal and calculates the high-frequency fluctuation variance value of the horizontal shear force signal in a single gait cycle; the plantar shear force feature analysis unit generates a local plantar friction risk index based on the high-frequency fluctuation variance value, and merges the local plantar friction risk index with the voxel model data of the plantar force distribution and outputs it.

10. The foot biometric monitoring and evaluation system based on multimodal sensing feedback according to claim 1, characterized in that, The signal acquisition module also includes a physical limiting unit and an electrostatic shielding unit; the pressure sensing unit and the flexible strain sensing unit are jointly configured in the physical limiting unit to maintain the consistency of the detection space; the electrostatic shielding unit is configured in the bottom space of the physical limiting unit and receives the external static charge generated by the friction of the foot when the physical limiting unit deforms, and conducts the external static charge to the grounding terminal.

Images