Mass distribution model-based plantar impact force dynamic estimation method

Abstract

The invention discloses a plantar impact force dynamic estimation method based on a mass distribution model, and relates to the technical field of biomechanical measurement, and the plantar impact force dynamic estimation method comprises the following steps: collecting three-axis acceleration of a human leg through an inertial measurement unit; carrying out gait segmentation based on the three-axis acceleration, and recognizing a landing point; a preset number of acceleration sampling points before and after each landing point are extracted, and the average value of acceleration amplitudes is calculated to serve as impact acceleration; based on the current relative gait postures of the two feet, the supporting states of the feet on the side and the opposite side are determined, and the mass distribution proportion is calculated according to the supporting states; estimating the plantar impact force according to a formula F = ma in combination with the impact acceleration and the mass distribution proportion; and outputting the estimated value of the plantar impact force. According to the method, the landing point is obtained through gait segmentation, after the acceleration characteristics are extracted, biomechanical correction is conducted in combination with the leg mass distribution proportion, and higher-precision estimation of the plantar impact force is achieved.

Description

Technical Field

[0001] This invention relates to the field of biomechanical measurement technology, and more specifically, to a method for dynamic estimation of plantar impact force based on a mass distribution model. Background Technology

[0002] Gait analysis is an important tool for studying human movement patterns and is widely used in sports health, rehabilitation medicine, and smart wearable devices. With the development of inertial measurement units (IMUs), gait monitoring methods based on wearable devices have gradually become a research and application hotspot. IMUs can collect real-time triaxial acceleration and angular velocity data of the human leg, providing a rich information foundation for gait cycle recognition, motion event segmentation, and motion behavior understanding.

[0003] Plantar impact force is a crucial parameter for measuring gait stability and load status. Traditional methods often rely on force platforms or pressure sensor arrays to acquire impact force data, but these devices are typically fixed in laboratory settings, making them unsuitable for daily wear and dynamic monitoring. Therefore, how to achieve dynamic estimation of plantar impact force using non-contact, low-power devices such as IMUs has become a key research direction in recent years.

[0004] In impact force estimation, the mass distribution of the human body during walking is closely related to the current support state of both feet. The distribution of mass between the feet dynamically changes during different gait phases (such as the support phase, the lift-off phase, and the swing phase). Therefore, relying solely on the acceleration changes of a single foot for impact force estimation is insufficient to fully reflect the actual force distribution. Combining motion information acquired by IMU with techniques such as gait phase recognition, support state inference, and mass distribution modeling has become a key approach to improving the accuracy of impact force estimation. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a dynamic estimation method for plantar impact force based on a mass distribution model, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for dynamically estimating plantar impact force based on a mass distribution model includes the following steps:

[0008] The triaxial acceleration of the human leg is collected using an inertial measurement unit;

[0009] Gait segmentation is performed based on the triaxial acceleration to identify the landing point;

[0010] Extract a preset number of acceleration sampling points before and after each impact point, and calculate the average value of the acceleration amplitude as the impact acceleration.

[0011] Based on the current relative gait posture of both feet, determine the support state of the foot on this side and the foot on the opposite side, and calculate the mass distribution ratio accordingly.

[0012] Based on the impact acceleration and mass distribution ratio, the plantar impact force is estimated using the formula F=ma;

[0013] Output the estimated value of the plantar impact force.

[0014] In some embodiments, the gait segmentation includes the following steps:

[0015] The acceleration amplitude is calculated from the collected triaxial acceleration, and the acceleration amplitude is the composite amplitude of the acceleration in the three directions;

[0016] The acceleration amplitude is detected point by point using a sliding window. When the standard deviation of the acceleration amplitude is less than the rest threshold, the absolute value of the acceleration amplitude slope is less than the slope threshold, and the average value of the acceleration amplitude within the sliding window is close to zero, it is identified as the gait start point.

[0017] Within a preset time window after the gait initiation point, find the point with the maximum Z-axis acceleration and identify it as the departure point;

[0018] After leaving the location, when the standard deviation of the acceleration amplitude sequence within the sliding window is lower than the set rest threshold again, and the mean of the composite acceleration amplitude in the XY plane is lower than the preset amplitude threshold, it is identified as the gait end point;

[0019] Within a preset time window before the end of the gait, the point with the maximum Z-axis acceleration is identified as the landing point.

[0020] In some embodiments, the process of estimating the mass distribution ratio includes:

[0021] Gait segmentation was performed on both legs separately to identify the landing and takeoff points in their respective gait cycles, and to divide the corresponding support phase and takeoff phase.

[0022] When the lateral foot is at the landing point, determine the current gait phase of the contralateral foot and accordingly determine the contralateral foot's support posture coefficient 'a', where:

[0023] When the opposite side is completely off the ground, α = 0;

[0024] When the opposite side is in a fully supported state, α = 1;

[0025] When the contralateral foot is in a partially supported state, α is a proportional value between 0 and 1;

[0026] Calculate the mass distribution ratio ρ borne by the foot on this side when it lands, satisfying:

[0027]

[0028] Where ρ represents the percentage of mass borne by the foot on this side when it lands.

[0029] In some embodiments, the calculation process for the support attitude coefficient, which is between 0 and 1, includes:

[0030] After detecting the time point corresponding to the landing point of the foot on this side, determine the current gait phase of the opposite foot;

[0031] Obtain the start and end time intervals of the support phase within the current gait cycle of the contralateral foot;

[0032] Map the current time point to the position of the opposite foot within the current cycle as the percentage of support phase completion;

[0033] The percentage of support phase completion is used as the support posture coefficient of the opposite foot, representing the transition state of its load-bearing capacity between 0 and 1.

[0034] In some embodiments, the estimation of the plantar impact force F is based on the following calculation formula:

[0035] F = ρ·M·σ;

[0036] in:

[0037] ρ represents the mass distribution ratio of the foot on the landing side at the current point of impact;

[0038] M represents the total mass of the human body;

[0039] 'a' represents the impact acceleration at the point of impact.

[0040] In some embodiments, the method further includes a dynamic weight estimation mechanism for estimating the equivalent body weight based on IMU data in the absence of user weight input, specifically including:

[0041] Extract triaxial acceleration data of the human leg during multiple gait cycles;

[0042] Based on a pre-built supervised learning-based machine learning model, the body weight M is estimated using the triaxial acceleration data;

[0043] The estimated body weight is used for subsequent calculation of equivalent mass, thereby completing the estimation of plantar impact force without body weight input.

[0044] During the training process, the machine learning model collects triaxial acceleration data samples from multiple users with known weight labels in their natural gait.

[0045] A regression model training set with the user's real weight as the supervised label was constructed, and a machine learning model was used for fitting and training to obtain the nonlinear mapping relationship between acceleration features and weight.

[0046] The trained weight estimation model is used as the computation module of the dynamic weight estimation mechanism, and the acceleration sequence of unknown users is input during the inference phase to output the estimated weight value.

[0047] In some embodiments, the machine learning model is a multilayer perceptron, support vector regression, or gradient boosting tree.

[0048] In some embodiments, the method further includes:

[0049] In each complete bipedal gait cycle, the impact force ratio between the unilateral foot and the contralateral foot is calculated, and its time series is recorded. This data is used to identify the unbalanced trend in impact load distribution. The ratio is calculated as follows:

[0050] Extracting the estimated impact force F L With F R , where F L F represents the estimated impact force of the left foot. R This indicates the estimated impact force on the right foot;

[0051] Calculate the impact force ratio This ratio will be used as an indicator of the symmetry of the impact force in the current cycle;

[0052] The impact force symmetry index for multiple consecutive periods is constructed into a time series curve for analyzing user gait load.

[0053] In some embodiments, the process of analyzing user gait load includes:

[0054] Constructing the time series of impact force ratio {R t}, where R t This represents the ratio of the impact force of the left foot to that of the right foot during the t-th gait cycle;

[0055] The mean μ and standard deviation σ of the ratio sequence over N consecutive periods are calculated using the sliding window method.

[0056] If there are multiple consecutive periods R t Satisfy |R t If -μ|>h·σ, it is identified as a sharp offset event, where h is the offset sensitivity coefficient;

[0057] If the moving average μ shows a monotonically changing trend in multiple windows and the slope exceeds a set threshold, it is determined to be a continuous load offset.

[0058] The judgment results will be used as the trigger for abnormal warnings or rehabilitation guidance.

[0059] The advantages of this invention over the prior art are:

[0060] This invention combines triaxial acceleration data acquired by an inertial measurement unit (IMU), identifies key event points (such as landing points) through gait segmentation, and further calculates the ratio of impact acceleration to mass distribution. Based on Newton's second law (F=ma), it estimates the dynamic impact force. Compared to traditional estimation methods based on static body weight or pressure sensors, this method offers higher dynamic responsiveness and estimation accuracy.

[0061] This invention innovatively establishes a mechanism for calculating the mass distribution ratio under bipedal support conditions, considering the support state of the contralateral foot at the moment of landing, thereby dynamically adjusting the actual mass borne by the landing foot. This modeling method overcomes the estimation errors caused by the assumption of average distribution or fixed proportion of total mass in traditional methods, making the impact force estimation more consistent with the actual biomechanics of the human body.

[0062] This invention further proposes a dynamic weight estimation mechanism, which establishes a nonlinear mapping relationship between triaxial acceleration features and user weight by training a supervised machine learning model (such as a multilayer perceptron, support vector regression, or gradient boosting tree). Even when the user does not provide weight information, the equivalent weight can still be automatically calculated, improving the system's adaptability and user experience.

[0063] Based on the estimation of impact forces in the left and right feet, this invention constructs a time series of impact force ratios to identify the imbalance trend of gait load. It can be used to monitor gait symmetry, identify abnormal deviation events, assist in sports rehabilitation assessment or abnormal behavior early warning, and has important clinical and health management value.

[0064] The overall estimation process of this invention relies solely on acceleration data acquired by a wearable IMU sensor, avoiding the use of expensive pressure pads or force platforms. It has advantages such as low cost, convenient wearability, and suitability for home or outdoor environments, making it easy to promote on a large scale and for personalized health monitoring applications. Attached Figure Description

[0065] Figure 1 This is a general schematic diagram of the method of the present invention;

[0066] Figure 2 This is a flowchart of the process for obtaining the mass distribution ratio of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] This invention uses a wearable inertial measurement unit (IMU) to collect triaxial acceleration data of the human leg in real time. Combined with gait analysis and mass distribution model, it dynamically estimates the impact force on the sole of the foot. This not only provides quantitative data support for sports training and rehabilitation assessment, but also identifies potential abnormalities in gait by analyzing the impact force distribution of both feet.

[0069] like Figure 1 As shown, the method of the present invention includes the following steps:

[0070] The triaxial acceleration of the human leg is collected using an inertial measurement unit;

[0071] Gait segmentation is performed based on the triaxial acceleration to identify the landing point;

[0072] Extract a preset number of acceleration sampling points before and after each impact point, and calculate the average value of the acceleration amplitude as the impact acceleration.

[0073] Based on the current relative gait posture of both feet, determine the support state of the foot on this side and the foot on the opposite side, and calculate the mass distribution ratio accordingly.

[0074] Based on the impact acceleration and mass distribution ratio, the plantar impact force is estimated using the formula F=ma;

[0075] Output the estimated value of the plantar impact force.

[0076] To acquire dynamic information about the legs during movement, this invention first utilizes IMU sensors mounted on both legs for data acquisition. The sensors are fixed to the thighs or calves and can record the acceleration changes of the legs in three spatial directions at a high sampling frequency (e.g., 100Hz or higher). The data output by each IMU includes acceleration components in the x, y, and z directions.

[0077] Suppose that at time t, the IMU collects triaxial accelerations a, ... x (t), a y (t) and a z If (t), then the acceleration amplitude a(t) can be obtained by the following formula:

[0078]

[0079] Based on the landing point and the takeoff point, the gait cycle can be divided into two main phases: the time from the landing point to the takeoff point is called the support phase, which represents the time when the foot is in contact with the ground and bears the weight; the time from the takeoff point to the landing point of the next cycle is called the takeoff phase, which represents the time when the foot swings in the air.

[0080] After identifying the impact point, acceleration data from its vicinity needs to be extracted to calculate the impact acceleration. Specifically, a certain number of sampling points are selected before and after the impact point, for example, 5 points before and after (a total of 10 points). The average acceleration amplitude of these points is calculated as the impact acceleration. The number of sampling points can be adjusted according to the sampling frequency and gait speed. For example, at a sampling frequency of 100Hz, 10 points correspond to a 0.1-second time window, which can effectively cover the main part of the impact event. If the movement speed is fast (such as running), the number of sampling points can be appropriately increased to 15 or 20.

[0081] like Figure 2 As shown, the impact force on the sole of the foot is not only related to the impact acceleration but also affected by the support state of both feet. Therefore, it is necessary to calculate the mass distribution ratio borne by each foot upon landing based on the relative gait posture of both feet. This process requires gait segmentation of the data for both legs separately, identifying their respective landing and takeoff points, and dividing the phase into the support phase and the takeoff phase.

[0082] Assuming the current analysis focuses on the impact force of the left foot (the foot on its own side), when the left foot is at the landing point, it is necessary to determine the gait state of the right foot (the opposite foot) to ascertain its support posture coefficient α. The value of α reflects the weight-bearing capacity of the opposite foot, and the specific rules are as follows:

[0083] If the right foot is in the off-ground phase, that is, not in contact with the ground at all, then α = 0, indicating that the right foot does not bear any mass.

[0084] If the right foot is in the support phase, that is, fully in contact with the ground and providing stable support, then α = 1, indicating that the right foot is fully involved in bearing the weight.

[0085] If the right foot is in the transition phase of the support phase (partial support), then α is between 0 and 1, indicating that the load-bearing capacity is in an intermediate state.

[0086] For partial support scenarios, the calculation of α needs further refinement. The method involves determining the start and end times of the support phase within the current gait cycle of the right foot (e.g., from t...). s to t e Then, calculate the relative position of the current time point t within the support phase to obtain the completion percentage p:

[0087]

[0088] This percentage p is directly used as the value of α. For example, if the right foot has just entered 25% of the support phase, then α = 0.25; if it has completed 75%, then α = 0.75. This linear mapping method is simple and conforms to the progressive characteristics of gait.

[0089] With α, the mass distribution ratio ρ of the left foot can be calculated using the following formula:

[0090]

[0091] ρ represents the proportion of mass borne by the left foot upon landing. For example, if α = 0 (right foot completely off the ground), then ρ = 1, and the left foot bears the entire mass; if α = 1 (right foot fully supporting the weight), then ρ = 0.5, and each foot bears half the mass. This calculation method fully considers the dynamic interaction between the two feet, ensuring the rationality of mass distribution. After obtaining the impact acceleration 'a' and the mass distribution ratio 'ρ', the plantar impact force 'F' can be estimated using the following formula:

[0092] F = ρ·M·a;

[0093] Where M is the total mass of the human body, in kg; and a is the impact acceleration, in m / s². 2 ρ is a dimensionless proportionality constant. The calculated result F is in Newtons (N), representing the magnitude of the impact force on the sole of the foot.

[0094] The theoretical basis of this formula is Newton's second law, F=ma, which states that force equals mass multiplied by acceleration. During gait, the total mass M of the human body is dynamically distributed to each foot through both feet, and the role of ρ is to adjust the total mass to the equivalent mass ρ·M of the current foot.

[0095] For example, when one foot is supporting the weight, ρ = 1, and the impact force is M·a; when both feet are supporting the weight simultaneously, ρ = 0.5, and the force borne by each foot is halved. This design can realistically reflect the mechanical changes in gait.

[0096] If the user's weight M is known (e.g., through manual input), this value can be used directly. If it is unknown, it needs to be calculated through a subsequent dynamic weight estimation mechanism.

[0097] To address situations where no weight input is available, this invention introduces a dynamic weight estimation mechanism based on IMU data, utilizing a machine learning model to deduce the equivalent body weight M from acceleration data. The implementation process is as follows.

[0098] First, triaxial acceleration data of both legs are collected over multiple gait cycles (e.g., 10 cycles) to form a time-series dataset. Features are extracted from this dataset, including the mean, variance, maximum, and minimum values ​​of the acceleration amplitude, as well as frequency domain features (such as the dominant frequency component after Fourier transform). These features can reflect the intensity and regularity of the gait and have a potential correlation with body weight.

[0099] Next, these features are analyzed using a pre-trained machine learning model to output an estimated weight M. The model training process includes the following steps:

[0100] Acceleration data from multiple users with known weights is collected, such as natural walking data from 100 users, with at least 100 gait cycles provided for each user. Each sample is labeled with the user's actual weight as a supervision label.

[0101] Extract the above acceleration features for each sample to form a feature vector.

[0102] Use a regression model to fit the relationship between features and weight. Optional models include multilayer perceptrons, support vector regression, or gradient boosting trees. Taking a multilayer perceptron as an example, its structure can be designed as an input layer (feature dimension, e.g., 10 features), two hidden layers (50 neurons each, using the ReLU activation function), and an output layer (1 neuron, outputting the weight value). During training, mean squared error is used as the loss function, and parameters are optimized using gradient descent. Model performance is evaluated using hold-out or cross-validation to ensure that the estimation error is controlled within a reasonable range (e.g., ±5 kg).

[0103] In practical applications, for new users, their acceleration data is input into a pre-trained model to obtain an estimated weight M, which is then used to calculate the impact force. This method requires no additional equipment, enhancing the system's portability and versatility.

[0104] To further evaluate gait balance and load distribution, this invention also includes a ratio analysis of impact forces on both feet. The specific implementation is as follows.

[0105] In each complete bipedal gait cycle, the estimated impact forces of the left and right feet are calculated separately and denoted as F. L and F R Then calculate the impact force ratio R:

[0106]

[0107] R, as a symmetry indicator, reflects the relative magnitude of the impact force from both feet. By continuously recording the R values ​​over multiple periods, a time series {R} is constructed. t}, where R t This represents the ratio in the t-th period.

[0108] Based on time series data, the sliding window method can be used for statistical analysis. The window size N is set (e.g., 10 periods), and R within the window is calculated. t Mean μ and standard deviation σ:

[0109]

[0110] Next, anomalies are detected using the following rules:

[0111] If R is in consecutive cycles (e.g., 3 cycles) t Satisfy |Rt If -μ|>h·σ, then a sharp shift is considered to have occurred. h is the shift sensitivity coefficient, typically taking a value of 2 or 3, representing the significance level of the deviation from the mean. For example, if h = 2, then a ratio deviating from twice the standard deviation is considered an anomaly.

[0112] Observe the trend of μ within multiple sliding windows. If μ monotonically increases or decreases, and the slope of change exceeds a set threshold (e.g., 0.05 per window), it is determined to be a continuous load offset. The slope can be calculated using linear regression.

[0113] These abnormal detection results can serve as early warning signals, alerting users to gait imbalances or potential sports injury risks.

[0114] Finally, the above method for gait segmentation specifically includes the following steps:

[0115] During the detection of gait cycle, the gait cycle length can be determined by analyzing the periodic characteristics of the acceleration amplitude sequence and identifying the time interval between the starting points of adjacent cycles.

[0116] At the beginning of the gait cycle, the feet are in a stationary state with their feet touching the ground. At this time, the changes in triaxial acceleration are weak, and the overall acceleration amplitude is stable and close to zero. Therefore, by detecting the standard deviation, slope and mean of the acceleration amplitude through a sliding window, this stationary initial state can be effectively identified.

[0117] During the foot lift phase, the foot begins to rise, and the Z-axis (vertical direction) acceleration increases significantly, exhibiting a distinct positive peak. Therefore, the maximum Z-axis acceleration after the gait initiation point is identified as having a high degree of physical consistency with the departure point. Similarly, upon foot landing, the strong impact between the sole and the ground causes the Z-axis acceleration to reach another local peak, allowing the landing point to be located within the window before the gait ends.

[0118] As the gait gradually ends, the feet stabilize on the ground again, acceleration fluctuations weaken, the overall amplitude converges to a low level, and the disturbances on the XY plane also decrease significantly. At this point, the end point of the gait can be accurately identified by judging the standard deviation and amplitude threshold.

[0119] Based on the above principles, this invention establishes dynamic parameters that depend on event detection. These parameters need to be adjusted in real time according to the characteristics of the current gait cycle. Specifically, the dynamic parameters include, but are not limited to, the following:

[0120] static threshold θ st The standard deviation threshold of acceleration used to determine the start and end points of gait can range from 0.05 to 0.2g (g is the acceleration due to gravity, approximately 9.8 m / s²). 2 ).

[0121] Slope threshold (θ) sl ): The acceleration slope threshold used to detect the gait initiation point, with a value range of 0.1 to 0.5 g / s.

[0122] Amplitude threshold (θ) a ): The threshold value of the average magnitude of the composite acceleration in the XY plane used to determine the end point of gait. The value range can be from 0.1 to 0.3g.

[0123] Time window: The range used to find the take-off and landing points near the start and end points of the gait cycle. It is recommended to set it to 0.2 to 0.5 seconds.

[0124] The update process for these parameters is based on the characteristics of the current gait cycle. For example, if a gait cycle length of 1.2 seconds and an average acceleration amplitude of 1.5g are detected, the historical parameters can be updated using a weighted average.

[0125] θ n =β·θ o +(1-β)·θ c

[0126] Where β is the update weight (ranging from 0.7 to 0.9), θ c The parameter value θ is calculated for the current period. o These are the parameter values ​​before the update. This update mechanism allows the system to adapt to changes in the gait habits or movement states of different users.

[0127] Key events in the gait cycle include the start point, departure point, landing point, and end point, and their identification process is based on dynamic parameters and sliding window analysis.

[0128] To identify the gait initiation point, a sliding window is used to detect acceleration data point by point. Assuming a window size of w (e.g., 20 sampling points), at each time point t, the following features are calculated:

[0129] Acceleration standard deviation in This represents the mean value within the window.

[0130] acceleration slope The sampling interval is denoted as .

[0131] Mean acceleration amplitude

[0132] The time point is determined as the gait start point when the following conditions are met:

[0133] σ(t)<θ st (For example, 0.1g);

[0134] |s(t)|<θsl (For example, 0.3 g / s);

[0135] (For example, 0.2g);

[0136] These conditions reflect the body's relatively still state before gait begins. For example, when a user begins to walk from a standstill, there will be slight changes in the magnitude and slope of acceleration, but the overall fluctuation is small.

[0137] After detecting the starting point, within a preset time window (e.g., 0.3 seconds), the maximum value of the Z-axis acceleration is found, i.e.:

[0138]

[0139] This point in time is identified as the lift-off point because the Z-axis acceleration typically reaches its peak when the foot leaves the ground. For example, during a normal walk, the lift-off point might correspond to the sudden increase in acceleration at the instant the heel leaves the ground.

[0140] The identification of gait termination points is also based on sliding window analysis. When the standard deviation σ(t) of the acceleration sequence within the window is below the rest threshold (e.g., 0.1g), and the mean of the composite acceleration amplitude in the XY plane is... When the amplitude drops below a threshold (e.g., 0.2g), that point in time is considered the end of the gait. This reflects the state at which the body returns to stillness at the end of the gait.

[0141] After detecting the endpoint, within a preset time window (e.g., 0.3 seconds) before it, find the point of maximum Z-axis acceleration, i.e.:

[0142]

[0143] This point in time is identified as the landing point because the Z-axis acceleration reaches its peak again when the foot lands. For example, when a user takes a step and the toes touch the ground, the Z-axis acceleration will increase significantly.

[0144] The starting point, departure point, landing point, and ending point identified by the above method are arranged in chronological order to form a complete gait event sequence. The departure point and landing point can be used to set the departure phase and landing phase, and the impact force can be measured using the method of this invention.

[0145] In summary, this invention achieves real-time estimation of plantar impact force by acquiring acceleration data using an IMU and combining it with a mass distribution model. Compared to traditional methods, it requires no complex mechanical platform, is low-cost, and highly portable. The dynamic weight estimation mechanism further expands its applicability, while the impact force ratio analysis provides a new perspective for gait health assessment. In the future, the accuracy can be further improved by optimizing the model with more samples or integrating other physiological signals such as heart rate.

[0146] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic estimation of plantar impact force based on a mass distribution model, characterized in that, Includes the following steps: The triaxial acceleration of the human leg is collected using an inertial measurement unit; Gait segmentation is performed based on the triaxial acceleration to identify the landing point; Extract a preset number of acceleration sampling points before and after each impact point, and calculate the average value of the acceleration amplitude as the impact acceleration. Based on the current relative gait posture of both feet, determine the support state of the foot on this side and the foot on the opposite side, and calculate the mass distribution ratio accordingly. Based on the impact acceleration and mass distribution ratio, the plantar impact force is estimated using the formula F=ma; Output the estimated value of the plantar impact force.

2. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 1, characterized in that, The gait segmentation includes the following steps: The acceleration amplitude is calculated from the collected triaxial acceleration, and the acceleration amplitude is the composite amplitude of the acceleration in the three directions; The acceleration amplitude is detected point by point using a sliding window. When the standard deviation of the acceleration amplitude is less than the rest threshold, the absolute value of the acceleration amplitude slope is less than the slope threshold, and the average value of the acceleration amplitude within the sliding window is close to zero, it is identified as the gait start point. Within a preset time window after the gait initiation point, find the point with the maximum Z-axis acceleration and identify it as the departure point; After leaving the location, when the standard deviation of the acceleration amplitude sequence within the sliding window is lower than the set rest threshold again, and the mean of the composite acceleration amplitude in the XY plane is lower than the preset amplitude threshold, it is identified as the gait end point; Within a preset time window before the end of the gait, find the point with the maximum Z-axis acceleration and identify it as the landing point; The time interval between the landing point and the departure point is defined as the support phase, and the time interval between the departure point and the landing point of the next cycle is defined as the departure phase.

3. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 2, characterized in that, The process of estimating the mass distribution ratio includes: Gait segmentation was performed on both legs separately to identify the landing and takeoff points in their respective gait cycles, and to divide the corresponding support phase and takeoff phase. When the lateral foot is at the landing point, determine the current gait phase of the contralateral foot and accordingly determine the contralateral foot's support posture coefficient 'a', where: When the opposite side is completely off the ground, α = 0; When the opposite side is in a fully supported state, α = 1; When the contralateral foot is in a partially supported state, 'a' is a proportional value between 0 and 1. Calculate the mass distribution ratio ρ borne by the foot on this side when it lands, satisfying: Where ρ represents the percentage of mass borne by the foot on this side when it lands.

4. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 3, characterized in that, The calculation process for the support attitude coefficient, which is between 0 and 1, includes: After detecting the time point corresponding to the landing point of the foot on this side, determine the current gait phase of the opposite foot; Obtain the start and end time intervals of the support phase within the current gait cycle of the contralateral foot; Map the current time point to the position of the opposite foot within the current cycle as the percentage of support phase completion; The percentage of support phase completion is used as the support posture coefficient of the opposite foot, representing the transition state of its load-bearing capacity between 0 and 1.

5. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 4, characterized in that, The estimation of the plantar impact force F is based on the following calculation formula: F = ρ·M·a; in: ρ represents the mass distribution ratio of the foot on the landing side at the current point of impact; M represents the total mass of the human body; 'a' represents the impact acceleration at the point of impact.

6. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 1, characterized in that, The method further includes a dynamic weight estimation mechanism for estimating the equivalent body weight based on IMU data in the absence of user weight input, specifically including: Extract triaxial acceleration data of the human leg during multiple gait cycles; Based on a pre-built supervised learning-based machine learning model, the body weight M is estimated using the triaxial acceleration data; The estimated body weight is used for subsequent calculation of equivalent mass, thereby completing the estimation of plantar impact force without body weight input.

7. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 6, characterized in that, During the training process, the machine learning model collects triaxial acceleration data samples from multiple users with known weight labels in their natural gait. A regression model training set with the user's real weight as the supervised label was constructed, and a machine learning model was used for fitting and training to obtain the nonlinear mapping relationship between acceleration features and weight. The trained weight estimation model is used as the computation module of the dynamic weight estimation mechanism, and the acceleration sequence of unknown users is input during the inference phase to output the estimated weight value.

8. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 7, characterized in that, The machine learning model is a multilayer perceptron, support vector regression, or gradient boosting tree.

9. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 1, characterized in that, The method further includes: In each complete bipedal gait cycle, the impact force ratio between the unilateral foot and the contralateral foot is calculated, and its time series is recorded. This data is used to identify the unbalanced trend in impact load distribution. The ratio is calculated as follows: Extracting the estimated impact force F L With F R F L F represents the estimated impact force of the left foot. R This indicates the estimated impact force on the right foot; Calculate the impact force ratio This ratio will be used as an indicator of the symmetry of the impact force in the current cycle; The impact force symmetry index for multiple consecutive periods is constructed into a time series curve for analyzing user gait load.

10. The method for dynamic estimation of plantar impact force based on a mass distribution model according to claim 9, characterized in that, The process of analyzing user gait load includes: Constructing the time series of impact force ratio {R t }, where R t This represents the ratio of the impact force of the left foot to that of the right foot during the t-th gait cycle; The mean μ and standard deviation σ of the ratio sequence over N consecutive periods are calculated using the sliding window method. If there are multiple consecutive periods R t Satisfy |R t If -μ|>h·σ, it is identified as a sharp offset event, where h is the offset sensitivity coefficient; If the moving average μ shows a monotonically changing trend in multiple windows and the slope exceeds a set threshold, it is determined to be a continuous load offset. The judgment results will be used as the trigger for abnormal warnings or rehabilitation guidance.

Images