A lower limb action correction method fusing angle and pressure data

Abstract

This invention relates to the field of motion correction technology, specifically to a method for lower limb motion correction that integrates angle and pressure data. The method includes acquiring real-time angle and pressure data during lower limb movement and establishing an angle-pressure time-series matrix; identifying the phase difference distribution pattern between angle and pressure feature points based on the matrix; comparing the phase difference distribution pattern with an individual's standard motion pattern; generating targeted motion correction commands when abnormal phase relationships are detected; and dynamically optimizing the individual's standard motion pattern based on the execution effect of the motion correction commands. This invention achieves cross-modal fusion of motion control signals and external reaction force signals. This fusion enhances the comprehensiveness of motion anomaly detection and improves the sensitivity of identifying subtle motion imbalances or phase misalignments, making it suitable for identifying problems that conventional models struggle to capture, such as gait deviations and incorrect force application sequences.

Description

Technical Field

[0001] This invention relates to the field of motion correction technology, and in particular to a method for lower limb motion correction that integrates angle and pressure data. Background Technology

[0002] In applications such as rehabilitation training, sports medicine, and intelligent training guidance, accurate assessment and real-time correction of lower limb movements have become key technologies for improving training effectiveness and preventing injuries. Traditional movement assessment methods mainly rely on visual observation, two-dimensional video analysis, or single-type sensors, which have significant limitations in terms of accuracy, stability, and real-time feedback capabilities, making it difficult to comprehensively reflect the coordination relationships and force sequence differences among multiple parts of the body during movement.

[0003] With the development of wearable technology and multimodal sensors, angular data and plantar pressure data have become two important indicators for studying lower limb movement. Angular data reflects the movement trajectory and rhythm of major joints such as the hip, knee, and ankle, while pressure data provides information on the force distribution and support dynamics in different areas of the foot. These two types of data characterize the internal control and external response features of lower limb movement from different dimensions. Their synergistic analysis is expected to reveal deeper patterns in movement coordination, such as the rationality of the force application sequence and the timeliness of support transitions.

[0004] However, current technologies still collect and analyze angle and pressure data separately, lacking a fusion matrix modeling method under a unified temporal benchmark. This makes it difficult to accurately capture the temporal coordination relationship between the two, limiting the depth and accuracy of motion anomaly detection. Furthermore, current motion analysis methods often focus on single-point angle deviations or pressure value mutations, neglecting the phase difference relationship between angle and pressure—that is, the temporal coupling mode between motion initiation and force response—and thus failing to effectively identify implicit errors caused by coordination mismatch. Existing correction strategies mostly use preset templates or fixed feedback, unable to dynamically update standard motion patterns based on individual training progress, nor can they implement differentiated interventions based on multi-dimensional anomaly types, resulting in a low level of intelligence. Summary of the Invention

[0005] This invention provides a method for lower limb movement correction that integrates angle and pressure data. This method can integrate multi-source information on angle and pressure, identify phase difference distribution patterns, implement personalized correction, and has adaptive optimization capabilities, thereby improving the accuracy of lower limb movement analysis, the real-time nature of feedback, and the individual adaptability of the system.

[0006] A method for correcting lower limb movements by integrating angle and pressure data includes the following steps:

[0007] S1: Collect real-time angle and pressure data during lower limb movement and establish an angle-pressure time series matrix;

[0008] S2: Based on the angle-pressure time series matrix, identify the phase difference distribution pattern between angle feature points and pressure feature points;

[0009] S3: Compare the phase difference distribution pattern with the individual standard action pattern, and generate a targeted action correction instruction when an abnormal phase relationship is detected;

[0010] S4: Based on the execution effect of the action correction instruction, dynamically optimize the individual standard action pattern.

[0011] Optionally, S1 collects angle data of the hip, knee and ankle joints through wearable inertial measurement units arranged in key parts of the lower limbs; and collects pressure data of multiple areas of the sole of the foot through a pressure sensor array embedded in the insole or standing platform.

[0012] Optionally, S1 further includes time synchronization processing of the collected angle data and pressure data to ensure that the data have a unified time reference; and organizing the synchronized angle data and pressure data into a matrix form according to the time series to form an angle-pressure time series matrix, wherein each row corresponds to a sampling time point and each column corresponds to a specific angle data or pressure data.

[0013] Optionally, based on the angle-pressure time series matrix, identifying the phase difference distribution pattern between angle feature points and pressure feature points specifically includes:

[0014] S21, extract the feature points of each joint angle during the motion cycle;

[0015] S22, extract pressure feature points of each region of the sole from the pressure data components of the angle-pressure time series matrix;

[0016] S23, Under a unified temporal framework, calculate the time delay between angular feature points and corresponding pressure feature points to form the original phase difference dataset;

[0017] S24, Perform statistical distribution analysis on the original phase difference dataset to construct a phase difference distribution pattern. The phase difference distribution pattern includes a central tendency index, a dispersion index, and distribution morphology characteristics of the phase difference.

[0018] Optionally, the feature points of each joint angle within the motion cycle include the angle maximum point, the angle minimum point, and the feature moment when the angle change rate exceeds a first preset threshold.

[0019] Optionally, the pressure characteristic points of each area of ​​the sole include pressure peak points, turning points of the pressure center trajectory, and characteristic moments when the pressure change rate exceeds a second preset threshold.

[0020] Optionally, S3 specifically includes:

[0021] S31, extract multiple statistical features of the phase difference distribution pattern;

[0022] S32, calculate the Mahalanobis distance between the statistical features and the corresponding baseline features in the individual's standard action pattern to obtain a multi-dimensional difference measure.

[0023] Optionally, the statistical features in S31 specifically include the mean, variance, and skewness of the phase difference.

[0024] Optionally, S4 specifically includes:

[0025] S41, within a predetermined time window after the execution of the action correction command, new angle data and pressure data are collected, and a new phase difference distribution pattern is calculated.

[0026] S42, assess the degree of improvement in coordination between the new phase difference distribution pattern and the current individual standard movement pattern;

[0027] S43, when the degree of coordination improvement exceeds the improvement threshold and the new phase difference distribution pattern remains stable in multiple consecutive motion cycles, the new phase difference distribution pattern is taken as the optimized individual standard action pattern.

[0028] S44, using a recursive update algorithm, the optimized individual standard action pattern is fused with the historical individual standard action pattern to generate an updated individual standard action pattern.

[0029] Optionally, the degree of coordination improvement in S43 is quantified by calculating the similarity between the new phase difference distribution pattern and the current individual standard movement pattern.

[0030] The beneficial effects of this invention are:

[0031] This invention, by constructing an angle-pressure time-series matrix, integrates three-dimensional angle information (from the hip, knee, and ankle joints) with plantar pressure data (from the insole or platform pressure array) in a time-synchronized manner for the first time, forming a unified data structure. Compared with traditional methods that rely solely on single-source angle or pressure data, this method achieves cross-modal fusion of motion control signals and external reaction force signals. This fusion enhances the comprehensiveness of motion anomaly detection and improves the sensitivity of identifying subtle motion imbalances or phase misalignments. It is suitable for identifying problems that are difficult to capture by conventional models, such as gait deviation and incorrect force application sequence.

[0032] This invention proposes a method for recognizing the phase difference distribution pattern of angle and pressure feature points, and combines Mahalanobis distance difference measurement and an adaptive threshold determination mechanism to achieve quantitative assessment of motion temporal coordination. This technology breaks through traditional threshold judgment or template matching methods, enabling difference analysis across different dimensions, including mean, variance, and skewness of phase difference. It determines the source of anomalies through dimensional combinations, and then dynamically matches personalized action correction strategies from a predefined correction rule base based on different combinations of anomaly dimensions. This achieves targeted and quantitative intelligent correction, improving the pertinence and effectiveness of correction commands.

[0033] This invention designs an optimization logic based on dual criteria of coordination improvement and movement stability, and updates individual standard movement patterns through a recursive fusion algorithm. It is applicable to various scenarios such as rehabilitation training, sports training, and professional movement standardization, and has long-term deployment value and continuous optimization capabilities. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a logical framework diagram of an embodiment of the present invention;

[0036] Figure 2 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0037] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0038] like Figure 1-2 As shown, a method for lower limb movement correction that integrates angle and pressure data includes the following steps:

[0039] S1: Collect real-time angle and pressure data during lower limb movement and establish an angle-pressure time series matrix;

[0040] S1 specifically includes:

[0041] S11, Multi-source sensor data acquisition:

[0042] Wearable inertial measurement units (IMUs) are deployed in key areas of the lower limbs, including the hip, knee, and ankle joints, to collect their three-dimensional angle data in real time.

[0043] Embed an array of pressure sensors in the insole or standing platform to collect real-time pressure data from multiple areas of the sole of the foot.

[0044] The S11 uses wearable inertial measurement units (IMUs) deployed at key points in the lower limbs to collect three-dimensional angle data of the hip, knee, and ankle joints. Simultaneously, an array of pressure sensors is embedded in the insole or standing platform to acquire real-time pressure distribution across multiple areas of the foot. These two types of sensors work together to collect data, providing raw input for subsequent motion state analysis and correction processing.

[0045] To comprehensively characterize the key biomechanical features of lower limb movement, it is necessary to simultaneously acquire information on both joint angle changes and plantar force states. Inertial Measurement Units (IMUs) offer advantages such as portability, real-time performance, and high resolution, making them suitable for continuously acquiring dynamic angle data of joints like the hip, knee, and ankle in three-dimensional space. Meanwhile, plantar pressure sensor arrays reflect force changes in various regions during the gait cycle, closely related to posture, balance, and other movement qualities. These two types of sensor data are significantly complementary in both time and space: angle data describes the trajectory and direction of movement, while pressure data reveals the support pattern and force structure. Their fusion significantly improves the accuracy of motion state perception and recognition. Furthermore, constructing a unified angle-pressure temporal matrix not only facilitates subsequent feature extraction and phase difference analysis but also provides a standardized data foundation for motion anomaly detection and correction strategy generation.

[0046] S12, perform time synchronization processing on the above angle data and pressure data to ensure that all sampled data have a unified timestamp. To construct a unified time-series data structure;

[0047] Lower limb movements are a highly dynamic and continuous process. Angle and pressure data must be on the same time reference to accurately correspond and reflect the true relationship between movement and force. Because inertial measurement units (IMUs) and pressure sensors may differ in sampling mechanisms, frequencies, or transmission paths, time misalignment will occur between data without time synchronization. This will prevent accurate pairing of angle and pressure states at the same moment, affecting subsequent phase difference analysis and movement recognition accuracy. Therefore, by unifying sampling timestamps or introducing system-level synchronization mechanisms, it is essential to ensure that all sensor data strictly correspond to the same point in time, constructing a consistent time-series data format.

[0048] S13, Matrix Structured Representation: Organizing the synchronized data into an angle-pressure time series matrix according to the time series. , is represented as:

[0049] ;

[0050] matrix It is a two-dimensional data structure organized in chronological order to represent angle and pressure data at each time point during lower limb movement. Each row represents a sampling time point; each column represents a specific measurement channel, including angle and pressure channels; the left side contains three-dimensional angle data for multiple lower limb joints, and the right side contains pressure sensor data for various areas of the sole of the foot; among them, Indicates the total number of sampling frames. The total number of angle channels is determined by the number of joints and the spatial dimension of each joint, typically 9 (3 joints × 3 spatial axes). This indicates the total number of pressure channels, corresponding to the number of measurement areas in the pressure sensor array. Indicates the first The time point, the first Angle values ​​for each angle channel; Indicates the first The time point, the first The pressure value of each pressure channel. Represents the angle-pressure time series matrix. This represents a time point index, with values ​​ranging from a lower limb movement correction method that integrates angle and pressure data. , This represents the angle index, with a value range of 100. , This represents the pressure channel index, with a value range of [value range missing]. ;

[0051] The matrix structured representation in S13 aims to organize synchronized angle and pressure data into a unified two-dimensional matrix with a clear mathematical structure. This is based on the fact that each moment in the lower limb movement process contains posture information of multiple joints and force information of the sole of the foot, which must be expressed in a unified data format for normalization and subsequent analysis. By organizing each frame of data into a row of a matrix in chronological order, the entire movement process becomes a multi-dimensional time series structure, which is beneficial for subsequent operations such as feature extraction, statistical modeling, and phase analysis.

[0052] The structure:

[0053] (1) Unified data dimensions: Unify multi-channel and multi-type data within a mathematical framework to simplify the input form of the algorithm;

[0054] (2) Preservation of temporal features: Each row corresponds to a sampling time point, which can completely preserve the dynamic evolution process of the action;

[0055] (3) Facilitates comparative analysis: Both the standard action pattern and the actual action can be converted into the same type of matrix, which facilitates frame-by-frame comparison and similarity calculation.

[0056] Therefore, matrixing angle and pressure data is a fundamental step in achieving accurate identification and correction analysis, and it has sufficient technical rationality and practicality.

[0057] S14, for the matrix Outlier detection is performed on the data from each channel, and invalid data exceeding the preset threshold range is removed. Missing data points are filled in using methods such as linear interpolation and spline interpolation to ensure data continuity and integrity.

[0058] The core of S14 is to perform data quality control on the acquired angle-pressure time series matrix, specifically including outlier detection and missing data imputation. During actual sensor acquisition, factors such as hardware drift, motion interference, and signal loss often result in numerical anomalies, abrupt jumps, or missing samples in the raw data. Directly using this data for analysis can easily lead to misjudgments, model deviations, and even the generation of incorrect correction commands, affecting the system's stability and reliability. Therefore, firstly, by setting a reasonable physical range, outlier detection is performed on the data from each channel, removing invalid data that significantly exceeds the joint's range of motion or the normal pressure range. Subsequently, to fill the data gaps caused by outlier removal or missing samples, linear interpolation and spline interpolation methods are used to imput the data, ensuring the continuity and smoothness of the entire matrix in the time dimension.

[0059] S2: Based on the angle-pressure time series matrix, identify the phase difference distribution pattern between angle feature points and pressure feature points;

[0060] S2 specifically includes:

[0061] S21, Angular feature point extraction:

[0062] S211, extract key feature points of each joint within one or more motion cycles from the angle data portion of the angle-pressure time series matrix M. The set of extracted angle feature points is represented as:

[0063] The meaning of this expression is from a certain angle channel (the first... (from 1 channel) extracts representative feature time points throughout the entire motion process. These feature points include three types:

[0064] Maximum angle point: This indicates the local maximum angle reached by the joint at a certain moment in its motion trajectory, that is, the maximum swing or extension in a certain direction is completed.

[0065] Minimum angle point: This indicates that the joint reaches a local minimum angle at a certain moment, that is, it completes the maximum contraction or swing in a certain direction;

[0066] Points where the rate of change of angle exceeds the threshold: This indicates that the joint has undergone a drastic change in angle at a certain moment, which is an important moment of movement transition or force exertion;

[0067] in, Indicates the first The values ​​of each angle channel at time point t Indicates angle channel The set of characteristic time points, Indicates angle channel The rate of change; This represents the first preset threshold value indicating the rate of change of angle. The value is... For moderate-speed movements such as daily walking and climbing stairs, the joint angular velocity is mostly in the range of 40-90; for fast movements, including jumping, sudden stops and turns, it exceeds 100. Therefore, the range of 30-120 is selected as the dynamic adjustment range, which can capture effective features and avoid excessive noise triggering.

[0068] S212, Pressure Feature Point Extraction:

[0069] Key pressure feature points are extracted from various regions of the sole from the pressure data, and defined as follows:

[0070] ;

[0071] The meaning of this expression is from the first Extract the set of representative pressure feature time points during the motion from each pressure channel, denoted as . These feature points include the following three categories:

[0072] 1. Peak pressure point: This indicates the local maximum pressure experienced by a certain area of ​​the sole at a certain moment, usually corresponding to key action phases such as standing, supporting, and pushing off the ground;

[0073] 2. Turning point of the pressure center trajectory: This indicates the point in time when the direction of the plantar pressure center changes along the movement path, reflecting gait transition or posture adjustment;

[0074] 3. Points where the rate of change of pressure exceeds the second preset threshold: This indicates that the force state of a certain area of ​​the sole of the foot changes rapidly at a certain moment, which may represent dynamic behaviors such as landing, jumping, and speed change.

[0075] in, Indicates pressure channel The set of characteristic time points, Indicates the first Each pressure channel at a specific time point The value, Indicates pressure channel The rate of change; The second preset threshold represents the rate of pressure change, and its value is the pressure change rate threshold. , Specifically, it includes:

[0076] 1. The rate of change of plantar pressure differs significantly between static standing and dynamic walking;

[0077] 2. The rate of change of single-point pressure during normal gait is approximately 10-25;

[0078] 3. Considering the sensor sampling frequency and noise level, setting it in the range of 5-30 is more suitable for most situations;

[0079] 4. It can also be flexibly set according to the usage scenario, such as rehabilitation training vs. competition monitoring.

[0080] The core of S21 lies in extracting key feature points corresponding to angle data from the angle-pressure time series matrix to capture typical movement moments of lower limb joints during movement. Specifically, it identifies the maximum and minimum points in each angle channel, as well as the moments when the angle change rate exceeds a certain preset threshold, to form a representative set of feature time points. These feature points reflect key nodes in the joint's movement cycle, such as the moments when the joint completes maximum flexion and extension, transitions between movements, or exerts force rapidly.

[0081] S22, Phase difference calculation:

[0082] S221, On a unified time axis, calculate the time difference between each pair of corresponding angular feature points and pressure feature points to construct the original phase difference dataset:

[0083] ;

[0084] The meaning of this expression is that, under a unified timeline, the first... In the first angle channel The feature point at the angular angle, and the ... The first pressure channel The time difference between each pressure characteristic point is denoted as the phase difference. This time difference reflects the relative temporal relationship between joint movements (leg lift, knee flexion) and changes in plantar force (stepping on the ground, lifting off the ground) within the same movement cycle. It is used to calculate the phase difference between characteristic points of joint angles and characteristic points of plantar pressure, aiming to quantify the temporal coordination between movement (angle change) and force (pressure change) in lower limb movements. In actual movement, this phase difference can reflect the coordination of motor control. For example, ideally, leg lift should occur when plantar pressure decreases; if there is a continuous lag or premature movement, it may indicate gait abnormalities, postural control problems, or abnormal neural responses.

[0085] in, Indicates angle channel The Time for each feature point Indicates pressure channel The Time for each feature point This represents the time phase difference between corresponding feature point pairs.

[0086] S222, forming the original phase difference dataset: This term describes the typical temporal relative position between a characteristic point of angular change at a specific joint (such as the knee joint) and a characteristic point of pressure on a certain area of ​​the foot (such as the forefoot) within a certain type of movement. It reveals whether the joint movement usually precedes, lags, or is synchronized with the force changes in that area of ​​the foot. This average phase difference can be regarded as a temporal offset or a temporal regularity.

[0087] The core objective of S22 is to calculate the time difference between angular and pressure feature points on a unified time axis, thereby constructing an original phase difference dataset for subsequent motion coordination analysis and correction. Specifically, it matches feature points (maximums, minimums, or abrupt changes in angle) in each angular channel with feature points (pressure peaks, inflection points, or abrupt changes) in the pressure channel, calculating their difference on the time axis to obtain a numerical value called the phase difference. This phase difference represents the relative temporal sequence between two action signals, such as whether a joint movement occurs before or after a change in foot pressure.

[0088] S23, Phase difference distribution pattern construction:

[0089] For the original phase difference dataset Statistical analysis was performed to obtain the phase difference distribution pattern, which includes the following indicators:

[0090] S231, Indicator of Central Tendency (Mean):

[0091] ;

[0092] This expression represents a specific angular channel. With pressure channel Between all matched angular and pressure feature point pairs, the average value of the phase difference (i.e., the time difference) is calculated and denoted as . This value reflects the typical time relationship between the joint and the foot region throughout the movement.

[0093] S232, Dispersion index (standard deviation):

[0094] ;

[0095] Indicates the first The first angle channel and the first The standard deviation is the dispersion of the phase difference values ​​of all feature point pairs between pressure channels relative to their average phase difference. Here, L represents the total number of selected angle-pressure feature point pairs. Indicates the first The time phase difference between each angle-pressure feature point This represents the average phase difference corresponding to that channel. The standard deviation of the phase difference represents the channel's distribution and reflects its dispersion. Specifically, it includes three aspects: first, indicators of central tendency, such as the mean of the phase difference, which reflects the typical temporal relationship between the two signals; second, indicators of dispersion, such as the standard deviation, used to measure the stability of action coordination; and third, distribution morphology characteristics, such as skewness, kurtosis, or frequency histograms, which can further reveal whether the phase difference exhibits complex characteristics such as symmetry, skewness, or multiple peaks.

[0096] S233, Distribution morphology characteristics: including statistical descriptions reflecting the overall shape of the phase difference, such as skewness, kurtosis, and frequency histogram.

[0097] The core of S23 lies in the statistical analysis of the phase difference dataset between the extracted angle and pressure feature points, thereby constructing a phase difference distribution pattern that reflects the temporal relationship of the movement. This distribution pattern not only reflects the average time difference between joint movement and plantar force, but also describes whether this time difference is stable and its overall distribution pattern.

[0098] S3: Compare the phase difference distribution pattern with the individual standard movement pattern, and generate targeted movement correction instructions when an abnormal phase relationship is detected;

[0099] S3 specifically includes:

[0100] S31, Phase difference statistical feature extraction: Statistical features are extracted from the phase difference dataset constructed during the current motion process, mainly including:

[0101] S311, Mean (Central Tendency):

[0102] ;

[0103] This expression represents: at a specific angular channel With pressure channel Between, the phase difference values ​​extracted from all samples The arithmetic mean of the values, i.e., the average phase difference under this channel combination, is denoted as... This average value reflects the typical time difference between the angle feature point and the pressure feature point over multiple motion cycles or multiple samplings. A positive value indicates that the angle action occurs after the pressure event; a negative value indicates that the angle action occurs before the pressure event; and a value close to zero indicates that the two are approximately synchronized.

[0104] S312, Variance (Dispersion):

[0105] ;

[0106] This expression represents the first... The first angle channel and the first With the combination of pressure channels, the phase difference of all samples Relative to its average value The variance, i.e. the degree of fluctuation of the phase difference of the channel pair, measures the degree of dispersion of the phase difference between the angle and the pressure around its average value in multiple movements or cycles. The larger the value, the more unstable and volatile the timing coordination between the angle and the pressure is in the movement; the smaller the value, the more stable the timing of the movement and the better the coordination.

[0107] S313, Skewness (Distribution Symmetry):

[0108] ;

[0109] This expression represents the first... The first angle channel and the first With the combination of pressure channels, the phase difference of all samples The skewness is used to measure the symmetry or degree of skewness in the phase difference distribution; if This indicates a right-skewed distribution, meaning that most phase differences are less than the average, with a small number of larger phase differences; if This indicates a left-skewed distribution, meaning that most phase differences are greater than the average, with a small number of smaller phase differences; if ≈0 indicates that the distribution is roughly symmetrical and has no significant skewness. Among them, Indicates the first The phase difference value of each sample; L represents the number of samples. This represents the mean phase difference. Indicates the phase difference variance. Phase difference deviation, This indicates a combination of an angle channel and a pressure channel;

[0110] The core of S31 lies in its multi-dimensional comparison of the phase difference statistical features extracted from the current action with the individual's standard action model to determine whether there are abnormal phase relationships, and based on this, to generate targeted action correction instructions. Specifically, it first extracts multiple statistical feature indicators of the current phase difference, such as mean, variance, and skewness, to comprehensively describe the temporal coordination characteristics between angle and pressure during the action. Then, it uses Mahalanobis distance to measure the difference between the current features and the baseline features in the individual's standard model, accurately assessing the overall deviation between the current state and the standard while considering the correlation of various dimensions.

[0111] To avoid misjudgments due to fluctuations in a single indicator, the solution incorporates a multi-dimensional simultaneous threshold judgment mechanism: only when the differences in at least two statistical dimensions simultaneously exceed a set threshold is an abnormal phase relationship identified, thus improving the robustness of the judgment and reducing the false alarm rate. Finally, based on the specific combination of dimensions exceeding the threshold, a matching correction strategy is selected from a pre-set correction rule library to automatically generate motion correction instructions. These instructions are used to adjust the force application timing or movement sequence of key joints, improving the coordination and stability of the movement.

[0112] S32, Mahalanobis distance calculation:

[0113] S321, the statistical feature vector of the current action:

[0114] ;

[0115] This expression represents: [the function of] a channel at a certain angle. With a certain pressure channel The phase difference distribution pattern is represented by a three-dimensional vector, which consists of three statistical features:

[0116] The average phase difference of this channel reflects the typical temporal sequence between action and force.

[0117] The phase difference variance of this channel pair measures the degree of fluctuation of this time relationship over multiple periods;

[0118] The phase difference skewness of this channel pair reflects the symmetry or skewness trend of the phase difference distribution;

[0119] This vector is for the angle channel. With pressure channel This is a statistical feature description of the phase difference relationship between two points, compressing the original phase difference data into three dimensions with physical and statistical significance. This structured representation allows for more effective standard action comparison, anomaly detection, and subsequent correction processing.

[0120] S322, the corresponding feature vector in the individual standard movement model:

[0121] ;

[0122] This expression represents: within an individual's standard action pattern, for a specific angle channel. With a certain pressure channel The reference phase difference eigenvector established by the combination of these parameters consists of the following three statistics:

[0123] The average phase difference of this channel represents the typical time relationship between an individual's actions and forces under standard conditions.

[0124] The phase difference variance reference value of this channel pair represents the stability of this time relationship in standard motion;

[0125] The phase difference skewness reference value of this channel pair represents the morphological characteristics of the time difference distribution in standard motion;

[0126] This expression models the temporal relationship between a certain joint (angle) and a certain foot area (pressure) in a person's standard / ideal movement state, and extracts three statistical features as a benchmark for subsequent comparative analysis with real-time movements;

[0127] S323, Perform a difference measurement and calculate the Mahalanobis distance:

[0128] ;

[0129] This expression is used to measure the multidimensional distance difference between the current action (such as the statistical characteristics of the phase difference between angle and pressure in a certain motion cycle) and an individual's standard action, where:

[0130] vector The characteristics representing the current action include three dimensions: mean, variance, and skewness;

[0131] vector This represents the user's standard reference characteristics;

[0132] S is the covariance matrix of the standard model samples, used to consider the correlation between the various feature dimensions;

[0133] Overall Results Mahalanobis distance is a weighted distance metric that measures the degree to which the current state deviates from the standard state. Unlike simple Euclidean distance, Mahalanobis distance considers the covariance between multiple feature dimensions; that is, if two features are highly correlated, their joint deviation should not be counted twice in the total distance. The covariance matrix represents the statistical characteristics of the standard action model samples. This represents the Mahalanobis distance between the current action and the standard action in the phase feature space.

[0134] S32 incorporates Mahalanobis distance calculation to perform multi-dimensional, weighted difference analysis on the phase difference statistical characteristics of the current action and the individual's standard action pattern. Its core idea is to organize the phase difference characteristics of the currently acquired angle and pressure channel combination, including mean, variance, and skewness, into a vector, and then compare it with a pre-established feature vector for the individual under standard conditions. Unlike simple Euclidean distance, Mahalanobis distance introduces the covariance matrix of standard action characteristics, allowing the correlation between features to be considered, thus more accurately reflecting the overall degree of deviation in a statistical sense.

[0135] S33, Determination of Abnormal Phase Relationships:

[0136] Setting multidimensional difference thresholds The actual deviation in each dimension is compared with its corresponding threshold. If the following conditions are met:

[0137] ;

[0138] If the difference in at least two dimensions exceeds the threshold simultaneously, then an abnormal phase relationship is determined to exist.

[0139] The expression means: Calculate the current value in each of the multiple statistical feature dimensions of the current action. The difference between the value and the standard value is used to determine whether the difference exceeds the corresponding preset threshold. If the number of dimensions exceeding the threshold reaches two or more, the action is considered to have a significant phase anomaly, requiring the triggering of a correction mechanism. Indicates the current action is in the 1st position. Feature values ​​(such as mean, variance, or skewness) on a statistical dimension. This indicates that in the standard action model of this individual, the first... Benchmark values ​​in each dimension express Adaptive judgment thresholds across multiple dimensions; This indicates that the number of dimensions that meet the "exceeds the threshold" condition is greater than or equal to 2.

[0140] Multidimensional difference threshold Specifically, it includes:

[0141] Mean deviation threshold : Values ∈[50, 150], representing the maximum allowable offset time between the mean phase difference of the current action and the standard value;

[0142] Variance bias threshold The value is ∈[30, 100]; indicates the allowable deviation of phase difference fluctuation from the standard action;

[0143] Skewness deviation threshold Values ∈[0.3,0.8]; indicates the allowable difference between the skewness of the distribution and the standard value;

[0144] S33 establishes a multi-dimensional phase difference anomaly judgment mechanism. By setting deviation thresholds for each feature dimension (mean, variance, skewness), it determines whether the current action state deviates from the individual's standard action pattern. Specifically, the current phase difference feature is compared with the standard feature dimension by dimension. If the difference in a certain dimension exceeds its corresponding preset threshold, it is considered that the dimension is abnormal. When two or more of the three dimensions exceed the threshold simultaneously, it can be determined that there is a significant phase anomaly, and an action correction command needs to be triggered.

[0145] S34, Motion correction instruction generation:

[0146] Based on the specific combination of dimensions that exceeds the threshold, including mean + variance, mean + skewness, etc., the system searches for matching items from the predefined correction rule library and selects the corresponding action correction instruction.

[0147] Each set of dimensions A corresponding correction strategy This strategy is used to adjust the timing of force application initiation, duration of force application, or sequence of coordination at certain key joints. The final output correction command can be expressed as:

[0148] ;

[0149] The meaning of this expression is that when the system identifies a specific phase difference anomaly type, it determines the anomaly based on the angle channel where the anomaly occurred. With pressure channel and the corresponding abnormal dimension combinations It selects matching action correction instructions from a predefined correction rule base; among which, This indicates the generated correction instructions; This represents the correction rule corresponding to a specific combination of phase anomaly dimensions; This indicates the corresponding combination of angle and pressure channel; This indicates the current combination of feature dimensions that exceeds the threshold, such as mean + skewness or variance + skewness, etc. This indicates which combination of angle and pressure channels is being analyzed, ensuring that the correction command is located at the specific joint and foot area.

[0150] When the phase difference characteristics of a certain angle-pressure channel pair are identified to deviate from the standard pattern simultaneously in multiple dimensions (such as mean, variance, and skewness), it is necessary not only to determine whether it is abnormal, but also to further clarify how to correct it. To this end, a correction rule base was designed, which pre-defines action correction strategies corresponding to different dimensional combinations. Each abnormal pattern (such as mean + variance abnormality or mean + skewness abnormality) is considered an independent identification type, corresponding to a specific action correction scheme. When the system detects a phase abnormality between a certain angle channel and pressure channel in the current action, and clarifies the abnormal dimensional combination... It will then search the rule base for correction rules that match the combination of that dimension. And combined with the current analysis of the channels Generate a specific action correction instruction This correction instruction will guide the user to adjust the starting time, rhythm, duration, or coordination sequence of a joint with other joints, thereby improving the timing coordination of the overall movement.

[0151] S4: Dynamically optimize individual standard action patterns based on the execution effect of action correction instructions;

[0152] S4 specifically includes:

[0153] S41, New Data Acquisition and Phase Difference Distribution Pattern Reconstruction:

[0154] The predetermined time window after the user executes the correction command Inside, new angle and pressure data are collected, processed, and an updated phase difference distribution pattern is obtained, represented as:

[0155] ;

[0156] This expression indicates that this three-dimensional vector is used to describe a pair of angular channels. With pressure channel The motion performance after the execution of the correction command was quantified from three statistical dimensions: the first dimension (mean) measures the average time difference between the current angle and the pressure feature point, reflecting the overall temporal relationship between the action and the force; the second dimension (variance) reflects the degree of fluctuation of this temporal relationship over multiple cycles, demonstrating the stability of motion control; and the third dimension (skewness) indicates whether the phase difference distribution is symmetrical, revealing whether there is a continuous advance or lag in the action. This represents the new mean phase difference. This represents the new phase difference variance; This indicates the new phase difference deviation. Indicates the angle-pressure channel pair index;

[0157] The purpose of S41 is to evaluate and model the motion effect after executing the motion correction command, in order to determine whether the synergistic relationship between angle and pressure has indeed been improved. Specifically, within a predetermined time window after the correction is executed, including several seconds or several motion cycles, angle and pressure data are re-acquired, and the phase difference distribution characteristics are recalculated based on the acquired results to obtain a new three-dimensional feature vector. This vector contains: the mean phase difference, representing the typical temporal relationship between the new motion and the applied force; the phase difference variance, indicating whether this temporal relationship is stable; and the phase difference skewness, reflecting whether there is a persistent trend of advance or lag.

[0158] S42, Calculation of the degree of improvement in coordination:

[0159] The degree of coordination improvement is quantified by calculating the similarity between the current standard movement pattern of an individual and the new phase difference distribution pattern. A similarity index is defined as follows:

[0160] ;

[0161] This expression represents the angle channel. With pressure channel Under the combination of these factors, the degree of improvement in the similarity between the current action and the standard action model after correction is called the degree of coordination improvement. This involves comparing the changes in the user's action features (feature vectors with phase difference as the core) before and after correction with the individual's standard action pattern. If the Mahalanobis distance after correction... Significantly smaller than the distance before correction This indicates that the current motion state is closer to the standard motion, signifying a good correction effect; the right side of the expression is a ratio, representing the proportion of the corrected distance to the original distance, which is obtained by subtracting this ratio from 1. This refers to the proportion of similarity improvement or the degree of coordination improvement, with a value range of [0,1]. This indicates no improvement; when This indicates some improvement; when A value of 0 indicates significant improvement; a negative value (which is extremely rare) may indicate a deterioration in the condition; among them, This represents the Mahalanobis distance between the new phase eigenvector and the standard eigenvector. The Mahalanobis distance between the phase eigenvector before correction and the standard eigenvector; This indicates the degree of improvement in coordination, with a value range of [0,1]. The larger the value, the more significant the improvement.

[0162] The core of S42 lies in quantifying the improvement in motor coordination by calculating the changes in the differences between the movement characteristics before and after correction and the individual's standard movement pattern. Specifically, it calculates the Mahalanobis distance between the phase feature vectors before and after correction and the standard model. A smaller Mahalanobis distance indicates that the current movement is closer to the standard movement. Then, it calculates the ratio of the two distances and subtracts this ratio from 1 to obtain the coordination improvement index η, which ranges from 0 to 1. A larger value indicates that the movement is closer to the standard and the correction effect is better.

[0163] S43, Determine stability and optimization conditions:

[0164] The current action mode is considered to be optimized and effective when the following two conditions are met:

[0165] 1. The degree of improvement in coordination exceeds the preset threshold. ,Right now ;

[0166] 2. New phase difference characteristics in continuous The fluctuation within each motion cycle is less than the set tolerance. That is, the stability satisfies:

[0167] ;

[0168] This expression represents: in the current angle channel With pressure channel Under the combination of these, the system is in continuous Within each motion cycle, the phase difference feature vector is extracted for each cycle. and compare it with the current overall average eigenvector. Perform difference calculation. If the maximum difference (i.e., fluctuation amplitude) across all periods is less than the set stability threshold... If so, it is assumed that the current feature remains stable over multiple periods.

[0169] The purpose of S43 is to establish a set of criteria for judging whether movement correction is effective and can be incorporated into the standard model. The core idea is to ensure, through two criteria—whether coordination has improved and whether characteristics have stabilized—that only truly improved and repeatable movement results will be used to update an individual's standard movement pattern. Specifically, the first condition requires that the corrected movement shows an improvement in temporal coordination compared to before the correction; that is, the degree of improvement is judged by calculating coordination improvement indicators (such as the relative reduction in Mahalanobis distance) to determine whether it exceeds a set threshold. This step is to confirm that the corrective action is effective and not just a random fluctuation or an ineffective adjustment. The second condition requires that this improvement result be continuous. The phase difference feature vector extracted in each cycle remains relatively stable, meaning that the deviation between the phase difference feature vector extracted in each cycle and the overall average value cannot exceed the set tolerance. This step is to verify whether the current state is stable and reproducible, and not a one-off or short-term change. Only when both conditions are met simultaneously is the current action pattern considered to have reached the optimization standard and can be used to update the individual standard action model; where: Indicates the first Phase difference eigenvectors within each period; This represents the stability tolerance threshold; This indicates the required number of consecutive cycles.

[0170] S43, Recursive fusion update of individual standard action patterns:

[0171] A recursive update algorithm (such as exponentially weighted average) is used to fuse historical standard motion feature vectors with new feature vectors to generate updated individual standard motion patterns.

[0172] ;

[0173] The above formula represents the weighted fusion of the newly acquired phase difference features, i.e., the new features formed after correction and stabilization, with the original individual standard action features to generate a new individual standard pattern; where, This represents the fusion coefficient, which controls the update speed; This represents the feature vector of the current standard action; This represents the updated standard action feature vector.

[0174] The core objective of S43 is to dynamically optimize individual standard movement patterns, allowing them to gradually adjust as the user's motor performance improves, thereby achieving true personalization and adaptability. Specifically, after confirming that the movement correction effect is significant and stable, the newly acquired phase difference features are weighted and fused with the existing standard movement model to generate an updated standard feature.

[0175] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0176] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for correcting lower limb movements by integrating angle and pressure data, characterized in that, Includes the following steps: S1: Collect real-time angle and pressure data during lower limb movement and establish an angle-pressure time series matrix; S2: Based on the angle-pressure time series matrix, identify the phase difference distribution pattern between angle feature points and pressure feature points; S3: Compare the phase difference distribution pattern with the individual standard action pattern, and generate a targeted action correction instruction when an abnormal phase relationship is detected; S4: Based on the execution effect of the action correction instruction, dynamically optimize the individual standard action pattern.

2. The method for lower limb movement correction by fusing angle and pressure data according to claim 1, characterized in that, The S1 collects angle data of the hip, knee and ankle joints through wearable inertial measurement units arranged in key parts of the lower limbs; and collects pressure data of multiple areas of the sole of the foot through pressure sensor arrays embedded in insoles or standing platforms.

3. The method for lower limb movement correction by fusing angle and pressure data according to claim 1, characterized in that, S1 further includes time synchronization processing of the collected angle data and pressure data to ensure that the data have a unified time reference; the synchronized angle data and pressure data are organized into a matrix form according to the time series to form an angle-pressure time series matrix, where each row corresponds to a sampling time point and each column corresponds to a specific angle data or pressure data.

4. The method for lower limb movement correction by fusing angle and pressure data according to claim 1, characterized in that, Based on the angle-pressure time series matrix, identifying the phase difference distribution pattern between angle feature points and pressure feature points specifically includes: S21, extract the feature points of each joint angle during the motion cycle; S22, extract pressure feature points of each region of the sole from the pressure data components of the angle-pressure time series matrix; S23, Under a unified temporal framework, calculate the time delay between angular feature points and corresponding pressure feature points to form the original phase difference dataset; S24, Perform statistical distribution analysis on the original phase difference dataset to construct a phase difference distribution pattern. The phase difference distribution pattern includes a central tendency index, a dispersion index, and distribution morphology characteristics of the phase difference.

5. The method for lower limb movement correction by fusing angle and pressure data according to claim 4, characterized in that, The characteristic points of each joint angle within the motion cycle include the angle maximum point, the angle minimum point, and the characteristic moment when the angle change rate exceeds the first preset threshold.

6. The method for lower limb movement correction by fusing angle and pressure data according to claim 4, characterized in that, The pressure characteristic points of each area of ​​the sole include pressure peak points, turning points of the pressure center trajectory, and characteristic moments when the pressure change rate exceeds a second preset threshold.

7. The method for lower limb movement correction by fusing angle and pressure data according to claim 1, characterized in that, S3 specifically includes: S31, extract multiple statistical features of the phase difference distribution pattern; S32, calculate the Mahalanobis distance between the statistical features and the corresponding baseline features in the individual standard action pattern to obtain a multi-dimensional difference measure; S33, compare the multi-dimensional difference measure with a preset adaptive threshold. When the difference measure of at least two dimensions exceeds the preset adaptive threshold at the same time, it is determined that there is an abnormal phase relationship. S34, based on specific dimension combinations that exceed the preset adaptive threshold, selects the corresponding action correction instruction from the predefined correction rule library, where different dimension combinations correspond to different force timing correction strategies.

8. The method for lower limb movement correction by fusing angle and pressure data according to claim 7, characterized in that, The statistical features in S31 specifically include the mean, variance, and skewness of the phase difference.

9. The method for lower limb movement correction by fusing angle and pressure data according to claim 1, characterized in that, S4 specifically includes: S41, within a predetermined time window after the execution of the action correction command, new angle data and pressure data are collected, and a new phase difference distribution pattern is calculated. S42, assess the degree of improvement in coordination between the new phase difference distribution pattern and the current individual standard movement pattern; S43, when the degree of coordination improvement exceeds the improvement threshold and the new phase difference distribution pattern remains stable in multiple consecutive motion cycles, the new phase difference distribution pattern is taken as the optimized individual standard action pattern. S44, using a recursive update algorithm, the optimized individual standard action pattern is fused with the historical individual standard action pattern to generate an updated individual standard action pattern.

10. A method for lower limb movement correction that integrates angle and pressure data according to claim 9, characterized in that, In S43, the degree of coordination improvement is quantified by calculating the similarity between the new phase difference distribution pattern and the current individual standard movement pattern.

Images