A myoelectric estimation method and measurement device for elbow joint torque

Abstract

The present invention discloses an electromyographic estimation method and a measuring device for elbow joint torque, and relates to the fields of human biological signal processing and upper limb rehabilitation exoskeleton active control. The present invention comprises the following steps: collecting a training set, wherein the training set comprises an electromyographic signal, an elbow joint angle signal, and a joint torque signal; inputting the training set into a random forest regression algorithm model and a fruit fly optimization algorithm model respectively, wherein the fruit fly optimization algorithm model outputs optimal parameters, and the random forest regression algorithm model uses the optimal parameters to construct an elbow joint torque estimation model; and using the elbow joint torque estimation model to perform online estimation of the elbow joint torque. The present invention effectively improves the accuracy of electromyographic estimation.

Description

Technical Field

[0001] The present invention relates to the fields of human biological signal processing and upper limb rehabilitation exoskeleton active control, and more specifically to an electromyographic estimation method and a measuring device for elbow joint torque. Background Art

[0002] Skeletal muscles drive human joints to move, which is the main way for the nervous system to interact with the outside world. Muscle force estimation has important research significance in the fields of upper limb rehabilitation exoskeleton active control, human neuromuscular and prosthetic control.

[0003] The electromyographic signals and joint torque data during elbow joint movement are extracted, and a mapping between the electromyographic signals and joint torque data is established, thereby realizing the estimation of the upper limb elbow joint torque based on the electromyographic signals.

[0004] Application No. 202111437003.8 (A method for mapping joint torques directly driven by human electromyographic signals) proposes a method for mapping joint torques directly driven by human electromyographic signals based on the Hill model, but the calculation of this method is still relatively complex and requires parameter identification. Application No. 201910283322.4 (A method for intelligently predicting human joint torques based on artificial neural networks), Application No. 202010830177.X (An elbow contraction muscle force estimation method based on an adaptive long short-term memory network) and Application No. 201910091542.7 (A method for estimating knee joint torques based on surface electromyography and motion signals) all use neural networks to establish the relationship between electromyography and joint torques.

[0005] The estimation accuracy of existing technologies needs to be further improved. Summary of the invention

[0006] In view of this, the present invention proposes an electromyographic estimation method and a measuring device for elbow joint torque to solve the problems in the background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A myoelectric estimation method for elbow joint torque is proposed, and an improved Fruit Fly Optimization Algorithm Random Forest Regression (IFOA-RFR) method is proposed, which includes the following steps:

[0009] Collecting a training set, wherein the training set includes electromyographic signals, elbow joint angle signals, and joint torque signals;

[0010] The training set is input into the random forest regression algorithm model and the fruit fly optimization algorithm model respectively,

[0011] The fruit fly optimization algorithm model outputs optimal parameters, and the random forest regression algorithm model uses the optimal parameters to construct an elbow joint torque estimation model;

[0012] The elbow joint torque estimation model is used to perform online estimation of the elbow joint torque.

[0013] Optionally, the calculation steps of the random forest regression algorithm model are as follows:

[0014] Randomly extract m sample points from the training set to obtain a new sub-training set;

[0015] Use the sub-training set to train the decision tree. During the training process, the splitting rule for each node is to first randomly select K features from all features, and then select the optimal split point from these K features to divide the left and right subtrees;

[0016] Then generate many CART regression tree models;

[0017] The final prediction result of each decision tree is the mean of the leaf nodes reached by the sample point;

[0018] The final prediction result of the random forest is the mean of the prediction results of all decision trees.

[0019] Optionally, the calculation steps of the fruit fly optimization algorithm model are as follows:

[0020] Initialize the position of the fruit fly group;

[0021] Define the random distance and direction that fruit flies use their sense of smell to search for objects;

[0022] Estimate the distance between the fruit fly and the origin, and then calculate the taste concentration judgment value;

[0023] Calculate the taste concentration of the individual fruit fly position based on the taste concentration judgment value and the taste concentration judgment function;

[0024] Find the fruit fly with the highest concentration of flavor in the fruit fly population;

[0025] The highest flavor concentration value and the corresponding coordinates are retained, and the fruit fly group uses vision to fly to the corresponding coordinate position;

[0026] Repeat multiple times until the preset zone conditions are met.

[0027] Optionally, a dynamic adaptive step size adjustment method is also included, as shown in the following formula:

[0028]

[0029] Where j is the current iteration number; i is the current fruit fly individual; X(i), Y(i) are the best fruit fly positions in the current iteration; randValue is a random variable; (X(i) j ,Y(i) j ) is the position of the i-th fruit fly individual in the j-th iteration; ω is the adaptive weight factor; ω0 is the initial adjustment weight factor; max is the maximum number of iterations; iteration is the current number of iterations; n is the weight order.

[0030] An elbow joint torque measuring device comprises a fixed frame, an angle adjustment frame and a force measuring component, wherein the angle adjustment frame is arranged on the fixed frame through an angle adjuster, the force measuring component is arranged on the fixed frame, and a wrist guard is arranged at the input end of the force measuring component.

[0031] Optionally, a fixing component is provided at the bottom of the fixed frame, and the fixing component is used to fix the fixed frame on the desktop.

[0032] Optionally, the angle adjuster is fixedly connected to the fixed frame, and the output end of the angle adjuster is connected to the angle adjustment frame.

[0033] Optionally, the angle adjuster is provided with a plurality of gears, and the difference between adjacent gears is 15°.

[0034] Optionally, the force measuring assembly includes a rigid rope fixer, a force meter and a rigid rope, the wrist guard is connected to the force meter via a first rigid rope, the force meter is connected to the rigid rope fixer via a second rigid rope, and the rigid rope fixer is fixedly connected to the angle adjustment frame.

[0035] Optionally, the angle adjustment frame is further provided with a fixed pulley, and the fixed pulley is used to cooperate with the first rigid rope to change the direction of the first rope.

[0036] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method and a device for myoelectric estimation of elbow joint torque, which have the following beneficial effects:

[0037] 1. A dynamic adaptive step-size adjustment strategy is proposed to improve the convergence speed of the fruit fly optimization algorithm and the efficiency of finding the global optimum.

[0038] 2. An elbow joint torque measurement device was designed to achieve simple and convenient measurement of elbow joint torque values.

[0039] 3. A myoelectric estimation method for elbow joint torque is proposed to achieve accurate elbow joint torque estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0041] Figure 1 It is a schematic flow chart of the improved fruit fly optimization-anytime forest regression estimation method of the present invention;

[0042] Figure 2 It is a schematic flow chart of the myoelectric estimation method of elbow joint torque of the present invention;

[0043] Figure 3 It is a schematic diagram of the elbow joint torque measuring device of the present invention;

[0044] Among them, 1. fixed frame, 2. angle adjustment frame, 3. angle adjuster, 4. wrist guard, 5. fixed pulley, 6. rigid rope, 7. force meter, 8. rigid rope fixer. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The embodiment of the invention discloses a myoelectric estimation method and a measuring device for elbow joint torque.

[0047] Specifically, the following steps are included: Figure 1-2 As shown, an offline elbow joint torque estimation model is established and the elbow joint torque is estimated online using the model.

[0048] Firstly, the sEMG signals of a pair of agonist and antagonist muscles, elbow angle and joint torque are collected during elbow flexion and extension. Then, the proposed IFOA-RFR method is used to establish an offline elbow torque estimation model.

[0049] Secondly, the sEMG signals of a pair of agonist and antagonist muscles during elbow flexion and extension are collected in real time, and the sEMG signals are input into the established offline model to output the estimated value, thus realizing the online estimation of the elbow joint torque.

[0050] Among them, the processing means of the electromyographic signal includes: collecting the electromyographic signal of the relevant muscles of the upper limbs of the human body, preprocessing and extracting the characteristics of the electromyographic signal. This embodiment is mainly aimed at the elbow joint, so the muscles collected are the biceps and triceps. The electromyographic sensor can use the TrignoTM Wirless acquisition device of Delsys, USA. The preprocessing of the electromyographic signal includes: using a 4th-order Butterworth bandpass filter to perform a 20-500Hz bandpass filtering on the collected electromyographic signal; using a 50hz notch filter to process the electromyographic signal to remove the 50hz power frequency interference caused by the power supply. The characteristic of the extracted electromyographic signal is the root mean square (RMS).

[0051]

[0052] Where N is the number of sampling points, x i is the value of the i-th sample.

[0053] The Random Forest Regression (RFR) algorithm is composed of multiple regression trees, and there is no correlation between each decision tree in the forest. The final output of the model is jointly determined by each decision tree in the forest. The randomness of the random forest is reflected in two aspects: a. The randomness of the sample, a certain number of samples are randomly selected from the training set as the root node samples of each regression tree; b. The randomness of the feature, when establishing each regression tree, a certain number of candidate features are randomly selected, and the most suitable feature is selected as the split node.

[0054] The principle of random forest regression algorithm is as follows:

[0055] A. Randomly extract m sample points from the training sample set S to obtain a new sub-training set S1…Sm;

[0056] B. Use the sub-training set to train the CART regression tree (decision tree). During the training process, the splitting rule for each node is to first randomly select K features from all features, and then select the optimal split point from these K features to divide the left and right subtrees;

[0057] C. Generate many CART regression tree models through step B;

[0058] D. The final prediction result of each CART regression tree is the mean of the leaf nodes reached by the sample point;

[0059] E. The final prediction result of the random forest is the mean of the prediction results of all CART regression trees.

[0060] The number of regression trees generated in the random forest regression algorithm (ntree), the maximum depth of the decision tree (mdepth), and the number of candidate split attributes (mtry) will affect the performance of the model. Among them, ntree represents the number of subtrees established. When the number of subtrees increases, the stability and prediction accuracy of the model are better. However, when the number of ntrees reaches a certain threshold, increasing this value will not improve the accuracy of the model, but will lead to a decrease in model efficiency. mdepth is the maximum growth depth of the tree. A higher depth will be conducive to learning specific relationships of samples. mtry represents the maximum number of features selected by the subtree, which affects the randomness of the model. Therefore, determining the appropriate ntree, mdepth, and mtry has a significant impact on the prediction results of the random forest regression algorithm.

[0061] This embodiment uses the fruit fly optimization algorithm to find the best parameters. The fruit fly optimization algorithm (Fruit Fly Optimization Algorithm FOA) summarizes the foraging behavior of fruit flies into a random search and visual positioning process and proposes a swarm intelligence optimization algorithm. The standard FOA optimization algorithm steps are as follows:

[0062] A. Randomly initialize the position of the fruit fly group (X, Y);

[0063] B. Give fruit flies random distances and directions to search for objects using their sense of smell.

[0064] X(i) = X + R (1);

[0065] Y(i)=Y+R (2);

[0066] Among them, X(i), Y(i) represent the position of the i-th fruit fly, and R represents the step length;

[0067] C. Since the location of the food is unknown, the distance D from the origin is estimated first, and then the taste concentration judgment value S is calculated, where S is the reciprocal of D;

[0068]

[0069]

[0070] D. Substitute the smell concentration judgment value S into the smell concentration judgment function to find the smell concentration smell at the position of the fruit fly;

[0071] smell i =F(S i ) (5);

[0072] smell i represents the taste concentration value of the i-th fruit fly individual, and F represents the fitness function;

[0073] E. Find the fruit fly with the highest flavor concentration in the fruit fly population;

[0074] smell best =min(smell) (6);

[0075] F. Keep the best taste concentration value smell best and (X best ,Y best ) coordinates, at which point the fruit fly colony uses vision to fly to that location;

[0076]

[0077] G. Enter iterative optimization, repeat steps BE, and determine whether the flavor concentration is better than the flavor concentration of the previous iteration. If so, execute step F. The algorithm stops after the preset iteration condition is met.

[0078] Since the search step length R of step B in the standard FOA algorithm is fixed, its convergence efficiency is low and it is easy to fall into the local optimum. In view of the above shortcomings, this embodiment proposes a dynamic adaptive step length adjustment method. As shown in the following formula:

[0079]

[0080] Where j is the current iteration number; i is the current fruit fly individual; X(i), Y(i) are the best fruit fly positions in the current iteration; randValue is a random variable; (X(i) j ,Y(i) j ) is the position of the i-th fruit fly individual in the j-th iteration; ω is the adaptive weight factor; ω0 is the initial adjustment weight factor; max is the maximum number of iterations; iteration is the current number of iterations; n is the weight order, which ranges from 0 to 10.

[0081] The value of n is related to the degree of change in the step length. The larger the value of n is, the more drastic the change in the step length is, and the more drastic the change in the random distance of the fruit fly is, and it is easier to find the global optimal value. Through testing, n in this embodiment is set to 2.

[0082] The initial adjustment weight factor of ω0 is 1.5, and randValue is a random number between 1 and 5.

[0083] The ω adaptive weight factor is related to the number of iterations. As the number of iterations increases, the step size changes more steadily, avoiding missing the optimal value. In the early stage, the number of iterations is small and ω is large, which increases the randomness of the fruit fly population and makes all the fruit flies dispersed. Through this dynamic adaptive step size adjustment method, the convergence speed and the performance of finding the global optimal solution are improved.

[0084] The improved IFOA-RFR algorithm replaces Formula 1 and Formula 2 with Formula 8 to improve the fruit fly optimization algorithm. The improved fruit fly optimization algorithm is used to find the optimal ntree, mdepth and mtry parameters, that is, to find the best RFR algorithm.

[0085] Before using an electromyographic estimation method for elbow joint torque to estimate the joint torque according to the electromyographic signal, an offline elbow joint torque estimation model needs to be established, that is, the IFOA-RFR network needs to be trained.

[0086] Depending on the conditions of muscle contraction, the physiological characteristics of the muscle may change. The types of muscle contraction can be divided into three categories: isometric contraction, isotonic contraction and isokinetic contraction (dynamic contraction). Among them, isometric contraction is when the muscle tension changes while the length of the muscle is inconvenient to maintain. Isotonic contraction is when the length of the muscle is stretched and contracted while the muscle strength remains unchanged. Isokinetic contraction is when the length of the muscle is contracted while the speed remains unchanged. Among them, isokinetic contraction is closer to the actual situation. The training data of this embodiment mainly comes from isometric contraction.

[0087] There is a great relationship between joint torque and joint angle, so this embodiment sets five conditions of elbow joint angles, namely 15 degrees, 30 degrees, 45 degrees, 60 degrees, and 75 degrees. A device for measuring elbow joint torque is designed, such as Figure 3 As shown. The device includes: 1. a fixed frame, 2. an angle adjustment frame, 3. an angle adjuster, 4. a wrist guard, 5. a fixed pulley, 6. a rigid rope, 7. a dynamometer, 8. a rigid rope fixer. The entire elbow joint torque measuring device is fixed on a table. The fixed frame carries all structures. The entire frame is built with 4040 aluminum profiles. The wrist passes through the wrist guard 4, and then the elbow joint is placed on the table. The angle between the angle adjustment frame 2 and the fixed frame 1 is adjusted by the angle adjuster 3. The fixed pulley 5 changes the direction of the rigid rope 6. The dynamometer 7 is connected to the wrist guard 4 through the rigid rope 6. The dynamometer 7 is connected to the rigid rope fixer 8 through the rigid rope 6. The rigid rope fixer 8 is connected to the angle adjustment frame 2.

[0088] The angle adjuster was used to set five conditions: 15 degrees, 30 degrees, 45 degrees, 60 degrees, and 75 degrees. In each condition, the wrist passed through the 4 wrist guards and was subjected to isometric contraction (from the smallest force to the largest force). The data was collected three times in total. There was a 3-minute rest interval between each collection.

[0089] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0090] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A myoelectric estimation method for elbow joint torque, characterized in that: The following steps are involved: Collecting a training set, wherein the training set includes electromyographic signals, elbow joint angle signals, and joint torque signals; Inputting the training set into a random forest regression algorithm model and a fruit fly optimization algorithm model respectively; The fruit fly optimization algorithm model outputs optimal parameters, and the random forest regression algorithm model uses the optimal parameters to construct an elbow joint torque estimation model; Using the elbow joint torque estimation model to estimate the elbow joint torque online; The calculation steps of the fruit fly optimization algorithm model are as follows: Initialize the position of the fruit fly group; Define the random distance and direction that fruit flies use their sense of smell to search for objects; Estimate the distance between the fruit fly and the origin, and then calculate the taste concentration judgment value; Calculate the taste concentration of the individual position of the fruit fly based on the taste concentration judgment function according to the taste concentration judgment value; Find the fruit fly with the highest concentration of flavor in the fruit fly population; The highest flavor concentration value and the corresponding coordinates are retained, and the fruit fly group uses vision to fly to the corresponding coordinate position; Iterate multiple times until the preset iteration conditions are met; It also includes a dynamic adaptive step size adjustment method, as shown in the following formula: Among them, j is the current iteration number; i is the current fruit fly individual; X(i), Y(i) are the best fruit fly positions in the current iteration; randValue is a random variable; (X(i) j ,Y(i) j ) is the position of the i-th fruit fly individual in the j-th iteration; ω is the adaptive weight factor; ω0 is the initial adjustment weight factor; max is the maximum number of iterations; iteration is the current number of iterations; n is the weight order.

2. The method for myoelectric estimation of elbow joint torque according to claim 1, characterized in that: The calculation steps of the random forest regression algorithm model are as follows: Randomly extract m sample points from the training set to obtain a new sub-training set; Use the sub-training set to train the decision tree. During the training process, the splitting rule for each node is to first randomly select K features from all features, and then select the optimal split point from these K features to divide the left and right subtrees; Then generate many CART regression tree models; The final prediction result of each decision tree is the mean of the leaf nodes reached by the sample point; The final prediction result of the random forest is the mean of the prediction results of all decision trees.

3. An elbow joint torque measurement device, using the myoelectric estimation method of elbow joint torque according to any one of claims 1-2, characterized in that: It comprises a fixed frame, an angle adjustment frame and a force measuring component. The angle adjustment frame is arranged on the fixed frame through an angle adjuster. The force measuring component is arranged on the fixed frame. A wrist guard is arranged at the input end of the force measuring component.

4. The elbow joint torque measuring device according to claim 3, characterized in that: A fixing component is arranged at the bottom of the fixed frame, and the fixing component is used to fix the fixed frame on the desktop.

5. The elbow joint torque measuring device according to claim 3, characterized in that: The angle adjuster is fixedly connected to the fixed frame, and the output end of the angle adjuster is connected to the angle adjustment frame.

6. The elbow joint torque measuring device according to claim 3, characterized in that: The angle adjuster is provided with a plurality of gears, and the difference between adjacent gears is 15°.

7. The elbow joint torque measuring device according to claim 3, characterized in that: The force measuring assembly comprises a rigid rope fixer, a force measuring device and a rigid rope, the wrist guard is connected to the force measuring device via a first rigid rope, the force measuring device is connected to the rigid rope fixer via a second rigid rope, and the rigid rope fixer is fixedly connected to the angle adjustment frame.

8. The elbow joint torque measuring device according to claim 3, characterized in that: The angle adjustment frame is also provided with a fixed pulley, and the fixed pulley is used to cooperate with the first rigid rope to change the direction of the first rope.

Images