Shoulder motion estimation method based on glenohumeral joint positioning and electromyography cooperation

Abstract

The application discloses a shoulder movement estimation method based on glenohumeral joint positioning and electromyography cooperation, and relates to the technical field of robot-assisted rehabilitation, and comprises the following steps: collecting surface electromyography signals and kinematics data; according to the surface electromyography signals and the kinematics data, solving a dynamic glenohumeral joint position in real time by minimizing distance residual based on anatomical constraints and step residual for ensuring continuity; determining muscle activation according to the surface electromyography signals; the muscle activation constitutes a muscle activation matrix; decomposing the muscle activation matrix into a muscle cooperation matrix and a command matrix; and estimating an upper arm angle according to elements in the command matrix. The method can realize high-precision, low-cost and long-term stable estimation of shoulder joint movement, and provides reliable technical support for robot-assisted rehabilitation.

Description

Technical Field

[0001] This application relates to the field of robot-assisted rehabilitation technology, and in particular to a method, device, equipment and storage medium for estimating shoulder motion based on glenohumeral joint localization and electromyography coordination. Background Technology

[0002] In the field of robot-assisted rehabilitation technology, accurately quantifying human movement intentions is crucial for achieving natural and efficient human-computer interaction. Surface electromyography (sEMG) signals, due to their rich neuromuscular activity information and non-invasive, low-cost characteristics, have become a major technical approach for long-term motion estimation and intention recognition. However, the development of existing sEMG decoding technologies has significant limitations in motion measurement accuracy, especially when applied to the shoulder joint complex—one of the most complex joints in the human body—where research results show a significant gap between research results and practical application requirements. The shoulder joint, composed of the scapula, clavicle, and humerus, has more than seven degrees of freedom, and its movement is controlled by multiple muscles in synergy. This anatomical complexity leads to extremely high measurement uncertainty. Currently, sEMG-based motion estimation methods largely rely on classic models such as the Hill muscle model, but these models are usually based on a simplified structure where a single muscle controls a single bone, making them difficult to directly apply to the complex scenario of multi-muscle synergistic control in the shoulder joint. In terms of posture quantification, existing technologies often use markers on the skin surface (such as the acromion) to represent the joint center position, which cannot be directly measured, in order to simplify calculations. Alternatively, they may directly regard the angle between adjacent body segments captured by an inertial measurement unit as the joint angle. Although optical motion capture systems can provide high-precision "gold standard" measurement results, their demanding measurement environment and high equipment cost make them difficult to popularize in clinical rehabilitation settings. A more fundamental problem is that for a non-"line-point-line" articulated structure like the shoulder joint, the position of the humerus is constantly changing due to the movement of the clavicle and scapula, and the center of rotation is not a fixed point. Existing methods usually assume that the trunk is fixed and simplify the shoulder joint angle to the angle between the upper arm and the trunk coordinate system. This simplification ignores the compensatory mechanism of scapular coordinated movement and cannot accurately reflect the true joint kinematics, which is the underlying reason for low positioning accuracy. In addition, existing sEMG driven models mostly use open-loop estimation strategies, and errors accumulate continuously during continuous recursive calculations, causing the estimation results to diverge over time, failing to meet the requirements of long-term, stable motion tracking. The aforementioned shortcomings in measurement accuracy, model applicability, environmental dependence, and estimation stability collectively restrict the progress of sEMG technology in shoulder joint rehabilitation robots in terms of precision and practical application.

[0003] Therefore, how to achieve high-precision shoulder motion estimation has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, equipment, and storage medium for shoulder motion estimation based on glenohumeral joint localization and electromyography coordination, which can achieve high-precision shoulder joint motion estimation and promote robot-assisted rehabilitation applications.

[0005] To address the aforementioned technical problems, this application provides a shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination, comprising:

[0006] Acquire surface electromyography signals and kinematic data;

[0007] Based on the surface electromyography signals and the kinematic data, the dynamic glenohumeral joint position is determined by minimizing the sum of squares of the distance residual and the step residual; the distance residual is based on anatomical constraints.

[0008] Muscle activation is determined based on surface electromyography signals; the muscle activation constitutes a muscle activation matrix.

[0009] The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix;

[0010] The upper arm angle is estimated based on the elements in the command matrix.

[0011] In some embodiments, determining the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the surface electromyography signal and the kinematic data includes:

[0012] A system of linear equations is obtained by minimizing the sum of squares of the distance residual and the step residual;

[0013] The dynamic glenohumeral joint position is obtained by solving the linear equations based on the surface electromyography signal and the kinematic data.

[0014] In some embodiments, determining muscle activation based on surface electromyography signals includes:

[0015] Nerve activation is obtained by solving a second-order linear model based on surface electromyography signals.

[0016] Muscle activation is determined based on the described neural activation.

[0017] In some embodiments, decomposing the muscle activation matrix into a muscle coordination matrix and a command matrix includes:

[0018] The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix using nonnegative matrix factorization.

[0019] In some embodiments, before determining the dynamic glenohumeral joint position based on the surface electromyography signal and the kinematic data by minimizing the sum of squares of the distance residual and the step residual, the method further includes:

[0020] The surface electromyography (EMG) signals and the kinematic data are preprocessed to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the preprocessed EMG signals and kinematic data.

[0021] In some embodiments, preprocessing the surface electromyography signal includes:

[0022] The surface electromyography signal is noise-removed by a filter and normalized by maximal voluntary isometric contraction.

[0023] Preprocessing the kinematic data includes:

[0024] The kinematic data is smoothed and filtered before being transformed from the global coordinate system to the local coordinate system.

[0025] In some embodiments, it also includes:

[0026] The model predictions are compared with the actual measurements by using an extended Kalman filter, and the optimal correction is calculated based on the noise statistics.

[0027] The output is a shoulder motion state estimate corrected by the optimal correction amount; the shoulder motion state estimate includes the dynamic glenohumeral joint position and upper arm angle.

[0028] To address the aforementioned technical problems, this application also provides a shoulder motion estimation device based on glenohumeral joint localization and electromyography coordination, comprising:

[0029] The acquisition module is used to acquire surface electromyography signals and kinematic data;

[0030] A dynamic glenohumeral joint position measurement module is used to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the surface electromyography signal and the kinematic data; the distance residual is based on anatomical constraints.

[0031] A muscle activation determination module is used to determine muscle activation based on surface electromyography signals; the muscle activations constitute a muscle activation matrix.

[0032] The matrix decomposition module is used to decompose the muscle activation matrix into a muscle coordination matrix and a command matrix.

[0033] The upper arm angle estimation module is used to estimate the upper arm angle based on the elements in the command matrix.

[0034] To address the aforementioned technical problems, this application also provides an electronic device, comprising:

[0035] Memory, used to store computer programs;

[0036] A processor, used to execute the computer program to implement the steps of the shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination as described above.

[0037] To address the aforementioned technical problems, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the shoulder motion estimation steps described above.

[0038] The shoulder motion estimation method based on glenohumeral joint localization and electromyography synergy provided in this application includes: acquiring surface electromyography signals and kinematic data; determining the dynamic glenohumeral joint position by minimizing the sum of squares of distance residuals and step residuals based on the surface electromyography signals and the kinematic data; the distance residuals being based on anatomical constraints; determining muscle activation based on the surface electromyography signals; the muscle activations constituting a muscle activation matrix; decomposing the muscle activation matrix into a muscle synergy matrix and a command matrix; and estimating the upper arm angle based on the elements in the command matrix.

[0039] As can be seen, the shoulder motion estimation method based on glenohumeral joint localization and electromyography (EMG) provided in this application achieves high-precision tracking of the shoulder joint rotation center through dynamic glenohumeral joint position measurement based on anatomical constraints and residual minimization. Dynamic glenohumeral joint position measurement significantly reduces distance residuals. Furthermore, this method, based on anatomical structure, reduces the influence of individual differences and enhances adaptability. In addition, this method reduces dependence on optical systems, uses surface EMG signals as the primary input, and is cost-effective.

[0040] The shoulder motion estimation device, equipment, and computer-readable storage medium based on glenohumeral joint localization and electromyography provided in this application all have the aforementioned technical effects. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the prior art and embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A schematic flowchart illustrating a shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination provided in this application embodiment;

[0043] Figure 2 This is a schematic diagram of a skin marker point provided in an embodiment of this application;

[0044] Figure 3This is a schematic diagram of the upper arm position provided in an embodiment of this application;

[0045] Figure 4 This is a schematic diagram of an angle explosion provided in an embodiment of this application;

[0046] Figure 5 A schematic diagram of the overall process for estimating complex shoulder joint motion as provided in an embodiment of this application;

[0047] Figure 6 A schematic diagram illustrating an experimental result provided in an embodiment of this application;

[0048] Figure 7 This is a schematic diagram illustrating another experimental result provided in an embodiment of this application. Detailed Implementation

[0049] The core of this application is to provide a method, device, equipment, and storage medium for shoulder motion estimation based on glenohumeral joint localization and electromyography coordination, which can achieve high-precision shoulder joint motion estimation and promote robot-assisted rehabilitation applications.

[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0051] Given the following technical shortcomings of existing technologies: 1. Low shoulder joint measurement accuracy: Traditional methods simplify joint positions to surface points (such as the acromion), ignoring dynamic GH displacement, resulting in large motion estimation errors. 2. Limitations of HMM models: Only applicable to single muscle and single bone control, unable to handle multi-muscle coordination problems in the shoulder joint. 3. Strong environmental dependence: High requirements for optical methods, unsuitable for long-term, low-cost applications. 4. Open-loop estimation error accumulation: sEMG decoding models lack feedback, and angle estimation is prone to divergence. This application provides a shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination, aiming to improve the accuracy of shoulder joint motion quantification, extend HMM to three-dimensional space, and achieve multi-muscle control estimation based on muscle coordination and angle decomposition. A closed-loop motion estimation model is constructed using extended Kalman filtering (EKF) to reduce error accumulation. This achieves low-cost, high-precision shoulder joint motion estimation, promoting robot-assisted rehabilitation applications. This method combines dynamically changing glenohumeral joint center position measurement, muscle-coordination-based Hill model extension, and extended Kalman filter closed-loop estimation, thus systematically solving the problems of accuracy, model applicability, and stability in shoulder joint motion estimation.

[0052] Please refer to Figure 1 , Figure 1 A flowchart illustrating a shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination, provided in this application embodiment, is shown below. Figure 1 As shown, the method includes:

[0053] S101: Acquire surface electromyography signals and kinematic data.

[0054] sEMG (Surface Electromyography) signals can be acquired from seven muscles (trapezius TRA, anterior deltoid DA, medial deltoid DM, posterior deltoid DP, pectoralis major PM, triceps brachii long head TRI-LO, and biceps brachii BIC). The sampling frequency can be up to 1000 Hz.

[0055] Kinematic data represents the three-dimensional positions of pre-defined skin markers. These positions are captured using an optical system. The pre-defined skin markers include key anatomical points such as the acromion and distal humerus. For example, ... Figure 2 As shown, the three-dimensional positions of 15 skin markers are captured using an optical system. Figure 2 The numbers 1 to 15 represent 15 skin markers.

[0056] S102: Based on the surface electromyography signal and the kinematic data, the dynamic glenohumeral joint position is determined by minimizing the sum of squares of the distance residual and the step residual; the distance residual is based on anatomical constraints.

[0057] Based on the surface electromyography signals and the kinematic data, the dynamic glenohumeral joint position is determined by minimizing the sum of squares of the distance residual and the step residual, including:

[0058] A system of linear equations is obtained by minimizing the sum of squares of the distance residual and the step residual;

[0059] The dynamic glenohumeral joint position is obtained by solving the linear equations based on the surface electromyography signal and the kinematic data.

[0060] In some embodiments, before determining the dynamic glenohumeral joint position based on the surface electromyography signal and the kinematic data by minimizing the sum of squares of the distance residual and the step residual, the method further includes:

[0061] The surface electromyography (EMG) signals and the kinematic data are preprocessed to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the preprocessed EMG signals and kinematic data.

[0062] The preprocessing of the surface electromyography signal includes:

[0063] The surface electromyography signal is noise-removed by a filter and normalized by maximal voluntary isometric contraction.

[0064] Preprocessing the kinematic data includes:

[0065] The kinematic data is smoothed and filtered before being transformed from the global coordinate system to the local coordinate system.

[0066] For surface electromyography signals, noise can be removed using a bandpass filter (4-200Hz) and a limiting filter (50Hz), and then normalized using methods such as maximum voluntary isometric contraction.

[0067] For kinematic data, after smoothing and filtering, the data is transformed from the global coordinate system to the local coordinate system (with the acromion as the reference).

[0068] The dynamic glenohumeral joint position was determined based on the preprocessed surface electromyography signals and preprocessed kinematic data.

[0069] The GH (Glenohumeral Joint) is the center of rotation of the shoulder joint. This application provides a method for measuring the dynamic GH position. The static GH (stGH) position is calculated in a standing posture. The static GH position can be calculated based on kinematic data of the geometric properties of the center of rotation, as shown in the following formula:

[0070] .

[0071] .

[0072] Indicates the upper arm marker point i at time t k The coordinates, x i(k) y i(k) z i(k) This represents the i-th marker point on the upper arm at time t. k R represents the three-dimensional coordinates along the x, y, and z axes of a Cartesian coordinate system. i It is the fixed distance from the marked point to stGH. T represents the transpose of the matrix.

[0073] Dynamic GH (dyGH) position r x(k) r y(k) r z(k) Indicates that the dynamic GH at time t k At that time, the three-dimensional coordinate values ​​along the x, y, and z axes of the spatial rectangular coordinate system.

[0074] The dynamic GH position is determined by solving a system of linear equations. The system of linear equations is as follows:

[0075] (1)

[0076] Where Q is a weighting constant. Q can take a value of 0.1. The value of Q can be adjusted to balance the residual weights.

[0077] Larm i (t k ) is the distance sequence from marker i to the acromion, Linn i (t k ) is the distance from stGH to the acromion, Lfunc i (t k ) is the distance from stGH to the marked point i.

[0078] Distance residual Based on anatomical constraints (humerus rigidity, acromion-humerus distance approximately fixed), the formula is:

[0079] (2)

[0080] Step residual To ensure the continuity of dyGH, the formula is:

[0081] (3)

[0082] Minimizing the sum of squared distance residuals and step residuals yields the linear equation system (1).

[0083] S103: Determine muscle activation based on surface electromyography signals; the muscle activation constitutes a muscle activation matrix;

[0084] The HMM model is extended to three dimensions for estimating upper arm angles. Step S103 aims to perform muscle activation calculations.

[0085] In some embodiments, determining muscle activation based on surface electromyography signals includes:

[0086] Nerve activation is obtained by solving a second-order linear model based on surface electromyography signals.

[0087] Muscle activation is determined based on the described neural activation.

[0088] After full-wave rectification of the surface electromyography signal, neural activation was obtained by solving a second-order linear model. .

[0089] (4)

[0090] Among them, e i(k) is the filtered surface electromyography signal. The values ​​of parameters d, α, β1, and β2 can be: d = 100ms, α = 0.95, β1 = -0.05, and β2 = 0.0006.

[0091] Muscle activation Segmented calculation:

[0092] (5)

[0093] The values ​​of parameters u0, δ, n1, n2, and c can be: u0=0.25, δ=10.48, n1=0.29, n2=0.84, and c=0.16.

[0094] S104: Decompose the muscle activation matrix into a muscle coordination matrix and a command matrix;

[0095] The muscle activation matrix Vm×n (m=7, n=time point) is decomposed into a muscle coordination matrix and a low-dimensional command matrix (i.e., the command matrix described above).

[0096] The muscle activation matrix is ​​decomposed into: Vm×n≈W m×j H j×n Where Wm×j is the muscle synergy matrix, j is the number of muscle synergies (e.g., j=2), H j×n This is a low-dimensional command matrix, which is recorded and used for real-time computation of muscle coordination units. The dimensionality reduction rate needs to exceed 90%.

[0097] In some embodiments, decomposing the muscle activation matrix into a muscle coordination matrix and a command matrix includes:

[0098] The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix using nonnegative matrix factorization.

[0099] This embodiment uses nonnegative matrix factorization (NNMF) to reduce the dimensionality of the muscle activation matrix into a muscle coordination matrix and a low-dimensional command matrix.

[0100] In addition, other dimensionality reduction methods can be used to replace nonnegative matrix factorization, such as using PCA (Principal Component Analysis) instead of NNMF.

[0101] S105: Estimate the upper arm angle based on the elements in the command matrix.

[0102] refer to Figure 3 and Figure 4As shown, the upper arm angle is decomposed into θy and θz. The position of the upper arm is defined by a vector from the acromion to the intercondylar point (DH) of the humerus, where DH is the midpoint between the lateral epicondyle (LEH) and the medial epicondyle (MEH) of the humerus. θy is the angle of the upper arm projected onto the XOY plane relative to the Y-axis, and θz is the angle of the upper arm relative to the XOY plane. Combining θy and θz provides a complete description of the upper arm angle.

[0103] As shown in the following equation, the HMM model and forward dynamics are used for continuous motion estimation based on muscle activation:

[0104] (6)

[0105] , , These are the joint angle, angular velocity, and angular acceleration, respectively. S0 to S5 are constants, Ts is the sampling time, and a(k) is the muscle activation at time k.

[0106] Replacing a(k) with Ha(k) in the above equation, the HMM equation is extended to:

[0107] (7)

[0108] Ha(k) is a cooperative command. Ha(k) is controlled by θy and θz. or Replacement.

[0109] For example, H i (k) represents H j×n The element in the i-th row and k-th column. The constant is S. The constant matrix S is obtained through nonlinear fitting.

[0110] This application demonstrates how HMM models are applied to multi-muscle control scenarios through muscle coordination and angular decomposition.

[0111] The upper arm angle can be estimated based on the above formulas (6) and (7).

[0112] In some embodiments, it also includes:

[0113] The model predictions are compared with the actual measurements by using an extended Kalman filter, and the optimal correction is calculated based on the noise statistics.

[0114] The output is a shoulder motion state estimate corrected by the optimal correction amount; the shoulder motion state estimate includes the dynamic glenohumeral joint position and upper arm angle.

[0115] This embodiment aims to achieve closed-loop motion estimation. It uses EKF to fuse GH measurements with upper arm angle estimation, avoiding error accumulation.

[0116] Measurement equation: Regression relationship between GH position and angle:

[0117] .

[0118] Where A is the regression matrix and l is the input corresponding to the intercept term coefficient.

[0119] dyGH is normalized according to the following formula:

[0120] GH(k) = dyGH(k) - stGH.

[0121] The Hammerstein model is used to quickly estimate dyGH from Ha(k) as an input to the EKF.

[0122] EKF linearization: state matrix The HMM and measurement equations are linearized using Taylor expansion.

[0123] Receive low-dimensional collaborative commands Using the dynamic model f(·) and the Hammerstein model Hammer(·), the future motion state (θ) and the current observation (dyGH) are predicted, respectively. An extended Kalman filter (EKF) is used to continuously compare the model predictions with the actual measurements, and the optimal correction is calculated based on the noise statistics. The optimally corrected, more accurate estimate of the shoulder motion state (including dyGH and θ) is output, thus achieving closed-loop feedback.

[0124] This application's embodiments combine EKF and regression models to reduce sEMG decoding errors. The closed-loop design avoids error divergence and is suitable for long-term applications.

[0125] The overall process for estimating shoulder joint compound motion based on sEMG signals and optical marker data, through dynamic GH position measurement, muscle coordination analysis, and closed-loop estimation, provided in this application embodiment, can be referred to as follows: Figure 5 As shown. Figure 5 middle Representation matrix W m×n The pseudo-inverse, pinv() is the standard function for calculating the pseudo-inverse of a matrix, W m×n This represents the original muscle synergy matrix, with dimensions of m rows (number of muscles) and n columns (synergy pattern dimension or time sample). Through... The high-dimensional muscle activation matrix V can be projected onto a low-dimensional cooperative command space, thereby extracting the essential control signals driving the motion. .

[0126] This represents the extracted muscle co-activation command matrix, which is the core output of the entire motion decoding process and serves as the low-dimensional instruction driving the subsequent motion estimation model. j represents the number of co-activation modes (e.g., j=2), and n represents the length of the time series. Therefore, It is a matrix with j rows and n columns, where each row corresponds to the activation intensity of a cooperative mode as it changes over time.

[0127] This is the pseudoinverse of the muscle coordination matrix. Position in the data stream: Input is the preprocessed muscle activation matrix. Processing: Decoding is performed using the offline-learned muscle coordination matrix and its pseudoinverse. Output: Low-dimensional coordination commands. or This output will directly serve as the driving signal for the Hammerstein model and dynamics model in subsequent application modules, thereby achieving accurate and real-time estimation of shoulder movement.

[0128] This is the core of the application module and represents the mathematical model for closed-loop motion estimation (such as extended Kalman filtering). This set of equations defines the state-space model for real-time, optimal estimation of shoulder motion and is the foundation of the extended Kalman filter algorithm.

[0129] The measurement equation is dyGH(k). dyGH(k) represents the dynamic glenohumeral joint position observed or rapidly estimated at time k. This indicates that the Hammerstein model is based on the current muscle coordination command. Directly predict joint position. This indicates the measurement noise in the prediction.

[0130] Here is the state prediction equation. θz(k+1) represents the predicted angle of the upper arm around the Z-axis at time k+1. f(·) represents the forward dynamic function, i.e., the previously extracted extended Hill muscle model. This represents the coordinated muscle commands driving the motion. v(k) represents the process noise of the dynamic model.

[0131] The output prediction equation is defined as follows: dyGH(k+1) represents the predicted joint position at time k+1. g(·) represents the regression function, which establishes the geometric mapping relationship between the joint angle and the joint center position. This represents the output noise of this mapping relationship.

[0132] The above set of equations is the core of the high-precision, interference-resistant real-time estimation in the embodiments of this application, and its operating logic is as follows:

[0133] 1. Input: Receive low-dimensional collaborative commands from the "Computation Module". .

[0134] 2. Prediction: Using the dynamic model f(·) and the Hammerstein model Hammer(·), the future motion state (θ) and the current observation value (dyGH) are predicted respectively.

[0135] 3. Correction: The Extended Kalman Filter (EKF) continuously compares the model's predicted values ​​with actual measurements from other sensors (such as imaging systems) and adjusts them based on noise statistics. Calculate the optimal correction amount.

[0136] 4. Output: Outputs optimized and more accurate estimates of shoulder motion state (including dyGH and θ) to achieve closed-loop feedback.

[0137] refer to Figure 6 As shown, the experimental data results are as follows:

[0138] GH position measurement: The maximum offset of the dynamic GH trajectory is 56.20 mm. The average distance residual measured by the method provided in this application embodiment is only 3.20 mm. Compared with the traditional method mentioned in the background art that uses the acromion as the marker point (residual of about 12.95 mm), the accuracy is improved by about 75%, and the measurement error is significantly reduced, by 75.11% compared with the acromion. The repeatability test error is 2.25 mm.

[0139] refer to Figure 7 As shown, the experimental data results are as follows:

[0140] Angle estimation: The upper arm angle error θy is 0.65±0.19 rad, and θz is 0.14±0.08 rad. The open-loop model diverges after 30-40 seconds, while the closed-loop model remains convergent. The fact that the open-loop model begins to diverge after 30-40 seconds verifies the necessity of introducing EKF closed-loop correction for achieving long-term, stable motion estimation.

[0141] In addition, the experimental data results are as follows:

[0142] DH position estimation: error is 33.90±19.21mm.

[0143] This verifies the effectiveness of the shoulder motion estimation method based on glenohumeral joint localization and electromyography provided in the embodiments of this application.

[0144] In summary, the shoulder motion estimation method based on glenohumeral joint localization and electromyography (EMG) provided in this application achieves high-precision tracking of the shoulder joint rotation center through dynamic glenohumeral joint position measurement based on anatomical constraints and residual minimization. Dynamic glenohumeral joint position measurement significantly reduces distance residuals. Furthermore, this method, based on anatomical structure, reduces the influence of individual differences and enhances adaptability. Additionally, this method reduces reliance on optical systems, using surface EMG signals as the primary input, and is cost-effective.

[0145] This application also provides a shoulder motion estimation device based on glenohumeral joint localization and electromyography coordination, which can be referred to in conjunction with the method described above. The device includes:

[0146] The acquisition module is used to acquire surface electromyography signals and kinematic data;

[0147] A dynamic glenohumeral joint position measurement module is used to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the surface electromyography signal and the kinematic data; the distance residual is based on anatomical constraints.

[0148] A muscle activation determination module is used to determine muscle activation based on surface electromyography signals; the muscle activations constitute a muscle activation matrix.

[0149] The matrix decomposition module is used to decompose the muscle activation matrix into a muscle coordination matrix and a command matrix.

[0150] The upper arm angle estimation module is used to estimate the upper arm angle based on the elements in the command matrix.

[0151] In some embodiments, the dynamic glenohumeral joint position measurement module includes:

[0152] A linear equation system determination unit is used to obtain a linear equation system by minimizing the sum of squares of the distance residual and the step residual;

[0153] The dynamic glenohumeral joint position determination unit is used to solve the linear equations based on the surface electromyography signal and the kinematic data to obtain the dynamic glenohumeral joint position.

[0154] In some embodiments, the muscle activation determination module includes:

[0155] The neural activation determination unit is used to obtain neural activation by solving a second-order linear model based on surface electromyography signals.

[0156] A muscle activation determination unit is used to determine muscle activation based on the neural activation.

[0157] In some embodiments, the matrix factorization module is used for:

[0158] The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix using nonnegative matrix factorization.

[0159] In some embodiments, it also includes:

[0160] The preprocessing module is used to preprocess the surface electromyography signal and the kinematic data to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the preprocessed surface electromyography signal and kinematic data.

[0161] In some embodiments, the preprocessing module includes:

[0162] The first preprocessing unit is used to remove noise from the surface electromyography signal through a filter and normalize it through maximum voluntary isometric contraction.

[0163] The second preprocessing unit is used to smooth and filter the kinematic data and then transform it from the global coordinate system to the local coordinate system.

[0164] In some embodiments, it also includes:

[0165] A closed-loop motion estimation unit is used to fuse glenohumeral joint measurements and upper arm angle estimations using an extended Kalman filter.

[0166] The shoulder motion estimation device based on glenohumeral joint localization and electromyography (EMG) provided in this application achieves high-precision tracking of the shoulder joint rotation center through dynamic glenohumeral joint position measurement based on anatomical constraints and residual minimization. Dynamic glenohumeral joint position measurement significantly reduces distance residuals. Furthermore, this method, based on anatomical structure, reduces the influence of individual differences and enhances adaptability. Additionally, this method reduces reliance on optical systems, using surface EMG signals as the primary input, and is cost-effective.

[0167] This application also provides an electronic device that includes a memory and a processor.

[0168] Memory, used to store computer programs;

[0169] A processor is used to execute computer programs to perform the following steps:

[0170] Acquire surface electromyography (EMG) signals and kinematic data; determine the dynamic glenohumeral joint position by minimizing the sum of squares of distance residuals and step residuals based on the EMG signals and kinematic data; the distance residuals are based on anatomical constraints; determine muscle activation based on the EMG signals; the muscle activations constitute a muscle activation matrix; decompose the muscle activation matrix into a muscle coordination matrix and a command matrix; estimate the upper arm angle based on the elements in the command matrix.

[0171] For a description of the equipment provided in this application, please refer to the above method embodiments; further details will not be provided here.

[0172] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps:

[0173] Acquire surface electromyography (EMG) signals and kinematic data; determine the dynamic glenohumeral joint position by minimizing the sum of squares of distance residuals and step residuals based on the EMG signals and kinematic data; the distance residuals are based on anatomical constraints; determine muscle activation based on the EMG signals; the muscle activations constitute a muscle activation matrix; decompose the muscle activation matrix into a muscle coordination matrix and a command matrix; estimate the upper arm angle based on the elements in the command matrix.

[0174] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0175] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0176] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatuses, devices, and computer-readable storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant details can be found in the method section.

[0177] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0178] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0179] The foregoing has provided a detailed description of the shoulder motion estimation method, apparatus, device, and storage medium based on glenohumeral joint localization and electromyography coordination provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for estimating shoulder motion based on glenohumeral joint localization and electromyography coordination, characterized in that, include: Acquire surface electromyography signals and kinematic data; The kinematic data are the three-dimensional positions of preset skin markers; Based on the surface electromyography signals and the kinematic data, the dynamic glenohumeral joint position is determined by minimizing the sum of squares of the distance residual and the step residual; the distance residual is based on anatomical constraints; the distance residual... ; residual ;Larm i (t k ) is the distance sequence from marker i to the acromion, Linn i (t k () is the distance from stGH to the acromion, where stGH is the static glenohumeral joint position, and Lfunc is the distance from stGH to the acromion. i (t k () is the distance from stGH to the marked point i. Indicates the upper arm marker point i at time t k The coordinates, dyGH(t) k ) indicates at time t k Dynamic glenohumeral joint position, r x(k) r y(k) r z(k) Indicates the dynamic glenohumeral joint at time t k At that time, the three-dimensional coordinate values ​​along the x, y, and z axes of the spatial rectangular coordinate system, r x(k-1) r y(k-1) r z(k-1) Indicates the dynamic glenohumeral joint at time t k-1 When, the three-dimensional coordinate values ​​along the x, y, and z axes of the spatial rectangular coordinate system, Q is a weighting constant; Muscle activation is determined based on surface electromyography signals; the muscle activation constitutes a muscle activation matrix. The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix; Estimate the upper arm angle based on the elements in the command matrix; The model predictions are compared with the actual measurements by using an extended Kalman filter, and the optimal correction is calculated based on the noise statistics. The output is a shoulder motion state estimate corrected by the optimal correction amount; the shoulder motion state estimate includes the dynamic glenohumeral joint position and upper arm angle.

2. The shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination according to claim 1, characterized in that, Based on the surface electromyography signals and the kinematic data, the dynamic glenohumeral joint position is determined by minimizing the sum of squares of the distance residual and the step residual, including: A system of linear equations is obtained by minimizing the sum of squares of the distance residual and the step residual; The dynamic glenohumeral joint position is obtained by solving the linear equations based on the surface electromyography signal and the kinematic data.

3. The shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination according to claim 1, characterized in that, Determining muscle activation based on surface electromyography signals includes: Nerve activation is obtained by solving a second-order linear model based on surface electromyography signals. Muscle activation is determined based on the described neural activation.

4. The shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination according to claim 1, characterized in that, The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix, including: The muscle activation matrix is ​​decomposed into a muscle coordination matrix and a command matrix using nonnegative matrix factorization.

5. The shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination according to claim 1, characterized in that, Based on the surface electromyography signals and the kinematic data, the determination of the dynamic glenohumeral joint position, prior to minimizing the sum of squares of the distance residual and the step residual, further includes: The surface electromyography (EMG) signals and the kinematic data are preprocessed to determine the dynamic glenohumeral joint position by minimizing the sum of squares of the distance residual and the step residual based on the preprocessed EMG signals and kinematic data.

6. The shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination according to claim 5, characterized in that, Preprocessing of the surface electromyography signal includes: The surface electromyography signal is noise-removed by a filter and normalized by maximal voluntary isometric contraction. Preprocessing the kinematic data includes: The kinematic data is smoothed and filtered before being transformed from the global coordinate system to the local coordinate system.

7. A shoulder motion estimation device based on glenohumeral joint localization and electromyography coordination, characterized in that, include: The acquisition module is used to acquire surface electromyography signals and kinematic data; The kinematic data are the three-dimensional positions of preset skin markers; A dynamic glenohumeral joint position measurement module is used to determine the dynamic glenohumeral joint position based on the surface electromyography signal and the kinematic data by minimizing the sum of squares of the distance residual and the step residual; the distance residual is based on anatomical constraints; the distance residual... ; residual ;Larm i (t k ) is the distance sequence from marker i to the acromion, Linn i (t k () is the distance from stGH to the acromion, where stGH is the static glenohumeral joint position, and Lfunc is the distance from stGH to the acromion. i (t k () is the distance from stGH to the marked point i. Indicates the upper arm marker point i at time t k The coordinates, dyGH(t) k ) indicates at time t k Dynamic glenohumeral joint position, r x(k) r y(k) r z(k) Indicates the dynamic glenohumeral joint at time t k At that time, the three-dimensional coordinate values ​​along the x, y, and z axes of the spatial rectangular coordinate system, r x(k-1) r y(k-1) r z(k-1) Indicates the dynamic glenohumeral joint at time t k-1 When, the three-dimensional coordinate values ​​along the x, y, and z axes of the spatial rectangular coordinate system, Q is a weighting constant; A muscle activation determination module is used to determine muscle activation based on surface electromyography signals; the muscle activations constitute a muscle activation matrix. The matrix decomposition module is used to decompose the muscle activation matrix into a muscle coordination matrix and a command matrix. The upper arm angle estimation module is used to estimate the upper arm angle based on the elements in the command matrix. The shoulder motion estimation device based on glenohumeral joint localization and electromyography also compares the model prediction value with the actual measurement value through an extended Kalman filter, and calculates the optimal correction amount based on the noise statistical characteristics; it outputs the estimated shoulder motion state value after correction based on the optimal correction amount; the estimated shoulder motion state value includes the dynamic glenohumeral joint position and upper arm angle.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the shoulder motion estimation method based on glenohumeral joint localization and electromyography coordination as described in any one of claims 1 to 6.

Images