Joint motion estimation method

Abstract

This application relates to the field of human-computer interaction technology and provides a joint motion estimation method. The method includes: collecting training data; the training data includes electromyographic (EMG) signal data samples at N time points; decomposing the EMG signal data samples at the N time points into multiple initial fuzzy sets, obtaining the first membership degree of each EMG signal data sample to each initial fuzzy set; fusing the multiple initial fuzzy sets to obtain multiple fused fuzzy sets; the number of fused fuzzy sets is less than the number of initial fuzzy sets; projecting the data in each fused fuzzy set onto a feature space to obtain a relationship expression between joint angles and EMG signals; and using the relationship expression to calculate the joint angles of the EMG signal data to be estimated, obtaining the joint angle data corresponding to the EMG signal data to be estimated. This application can improve the accuracy of joint angle prediction.

Description

Technical Field

[0001] This application belongs to the field of human-computer interaction, and in particular relates to a joint motion estimation method. Background Technology

[0002] In recent years, active exoskeleton robots have been widely used in medical rehabilitation, assisted walking, and other fields. During human-computer interaction, continuous estimation of joint motion is a key technology for achieving human-machine collaboration, compliant control, and smooth, stable movement. Since electromyography (EMG) signals can precede human movement, it provides a foundation for continuous estimation of joint motion.

[0003] Currently, muscle models and machine learning methods are the two mainstream approaches for predicting continuous joint motion. The Hill model can explain muscle movement; however, it has several parameters that are difficult to measure, making parameter identification challenging. Machine learning methods directly establish a mapping model between electromyographic (EMG) signals and joint motion, commonly using linear models, higher-order polynomial models, neural network models, etc. This modeling approach is simple and direct. However, due to the ambiguity, variability, and strong nonlinearity of EMG signals, the model's predictions are prone to significant deviations, especially when muscle activity changes (e.g., muscle fatigue), which can easily lead to model failure. Summary of the Invention

[0004] This application provides a joint motion estimation method that can solve the problem of low accuracy in joint angle prediction.

[0005] This application provides a joint motion estimation method, including:

[0006] Collect training data; the training data includes electromyographic signal data samples at N time points;

[0007] The electromyography (EMG) signal data samples at N time points are decomposed into multiple initial fuzzy sets, and the first membership degree of the EMG signal data sample at each time point to each initial fuzzy set is obtained.

[0008] Multiple initial fuzzy sets are fused to obtain multiple fused fuzzy sets; the number of fused fuzzy sets is less than the number of initial fuzzy sets.

[0009] By projecting the data from each fused fuzzy set onto the feature space, the relationship expression between joint angles and electromyographic signals is obtained.

[0010] The joint angle data corresponding to the electromyographic signal data to be estimated is obtained by using relational expressions to calculate the joint angle data.

[0011] Optionally, the electromyography (EMG) signal data samples at N time points are decomposed into multiple initial fuzzy sets to obtain the first membership degree of the EMG signal data sample at each time point to each initial fuzzy set, including:

[0012] Based on the fuzzy clustering method, the electromyography signal data samples at N time points are decomposed into L initial fuzzy sets;

[0013] Calculate the cluster centers and variances of each initial fuzzy set;

[0014] Based on the cluster center and variance of each initial fuzzy set, calculate the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set.

[0015] Optionally, based on the cluster centers and variances of each initial fuzzy set, the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set is calculated, including:

[0016] Through calculation formula

[0017]

[0018] Obtain the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set;

[0019] in, x represents the electromyographic signal data sample at time i. i For the first membership degree of the k-th initial fuzzy set, μ k Let τ represent the cluster center of the k-th initial fuzzy set. k Let represent the variance of the k-th initial fuzzy set, k = 1, ..., L, i = 1, ..., N.

[0020] Optionally, multiple initial fuzzy sets can be fused to obtain multiple fused fuzzy sets, including:

[0021] Calculation formula:

[0022]

[0023] Obtain the distance D(A1, A2) between the initial fuzzy set A1 and the initial fuzzy set A2; the initial fuzzy set A1 and the initial fuzzy set A2 are two initial fuzzy sets from the L initial fuzzy sets;

[0024] Calculation formula:

[0025]

[0026] Obtain the similarity S(A1, A2) between the initial fuzzy sets A1 and A2;

[0027] Through formula Determine whether to fuse the initial fuzzy sets A1 and A2; θ represents the similarity threshold, and K(A1, A2) represents the fusion indicator;

[0028] If the fusion indicator K(A1, A2) = 1, then the initial fuzzy set A1 and the initial fuzzy set A2 are fused to obtain the fused fuzzy set.

[0029] Optionally, after fusing multiple initial fuzzy sets to obtain multiple fused fuzzy sets, the estimation method may further include:

[0030] Through calculation formula

[0031]

[0032] Obtain the second membership φ of the electromyography signal data samples at each time step to each fused fuzzy set. j (x i ); where s k A represents the weighting factor. k Let F represent the k-th initial fuzzy set. j Let j represent the j-th fused fuzzy set, j = 1, ..., R, where R represents the total number of fused fuzzy sets, and R << L;

[0033] Through calculation formula

[0034]

[0035] Obtain the cluster center m of the j-th fused fuzzy set j ;

[0036] Through calculation formula

[0037]

[0038] The variance σ of the j-th fused fuzzy set is obtained. j .

[0039] Optionally, the data from each fused fuzzy set is projected into the feature space to obtain the relationship expression between joint angles and electromyographic signals, including:

[0040] The data in each fused fuzzy set is projected onto the feature space to obtain the correlation between joint angles and electromyographic signals in each fused fuzzy set;

[0041] Based on the correlation between joint angles and electromyographic signals in each fused fuzzy set, the relationship between joint angles and electromyographic signals is obtained;

[0042] Solve for the parameters in the relation and use the obtained parameter values ​​to obtain the relationship expression between the joint angle and the electromyographic signal.

[0043] Optionally, the training data may also include N time-stamped joint angle data samples.

[0044] Optionally, the correlation between each fused fuzzy set joint angle and electromyographic signal is as follows:

[0045]

[0046] Among them, f j (x i ) represents the electromyographic signal data sample x in the j-th fused fuzzy set. i Corresponding joint angle data samples, Let p be the spatial projection function. g represents the sigmoid function, q represents the number of nodes in the spatial projection, and w p and t p Both represent spatial projection parameters, a j and b j Both represent consequent parameters in the fuzzy rule of the j-th fused fuzzy set, a jp Let a represent the space projection function corresponding to the p-th spatial projection function. j .

[0047] Optionally, the relationship between joint angle and electromyographic signal is as follows:

[0048]

[0049] Where, f(x) i ) represents electromyographic signal data sample x i The corresponding joint angle data sample.

[0050] Optionally, the parameters in the relation include the spatial projection parameter {w}. p , t p}, number of nodes in spatial projection q, weight factor s k And the consequent parameter {a j b j};

[0051] Optionally, the parameters in the relation can be solved, including:

[0052] Based on the GSA optimization algorithm, the relation is solved using training data and the first membership degree to obtain the spatial projection parameters {w}. p , t p}, weighting factor s k And the number of nodes q in the spatial projection;

[0053] According to the spatial projection parameter {w p , t p}, thus obtaining the spatial projection function

[0054] Based on the least squares method, using spatial projection functions Solving the relation with the first membership degree yields the consequent parameter {a}. j b j}

[0055] Optionally, the relationship between joint angle and electromyographic signal can be expressed as follows:

[0056]

[0057] Where f(x) represents the joint angle data corresponding to the electromyographic signal data to be estimated, and x represents the electromyographic signal data to be estimated.

[0058] Optionally, the joint angles of the electromyography (EMG) signal data to be estimated are calculated using a relational expression to obtain the joint angle data corresponding to the EMG signal data to be estimated, including:

[0059] Through calculation formula

[0060]

[0061] The second membership γ of the electromyography signal data to be estimated to each fused fuzzy set is obtained. j ; where γ j This represents the second membership degree of the electromyographic signal data to be estimated to the j-th fused fuzzy set, where j = 1, ..., R;

[0062] Optionally, let φ j (x)=γ j , The joint angle data f(x) corresponding to the electromyographic signal data to be estimated is calculated.

[0063] The above-mentioned solution in this application has the following beneficial effects:

[0064] In this embodiment, electromyography (EMG) signal data samples at N time points are decomposed into multiple initial fuzzy sets. The first membership degree of the EMG signal data sample at each time point to each initial fuzzy set is calculated. Then, the multiple initial fuzzy sets are fused to obtain multiple fused fuzzy sets. The data in each fused fuzzy set is then projected onto a feature space to obtain a relational expression between joint angles and EMG signals. Finally, the joint angles of the EMG signal data to be estimated are calculated using the relational expression to obtain the joint angle data corresponding to the EMG signal data to be estimated. By decomposing the EMG signal data samples into multiple initial fuzzy sets, the fuzziness and differences of EMG signals can be better handled. Fusing the original fuzzy sets into fused fuzzy sets reduces the number of fuzzy sets. Projecting the data from the fused fuzzy sets onto the feature space eliminates the strong nonlinearity of EMG signals, thereby significantly improving the accuracy of joint angle prediction in this application.

[0065] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of this application, 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 A flowchart illustrating a joint angle estimation method provided in an embodiment of this application;

[0068] Figure 2 This is a schematic diagram of the initial fuzzy set being fused into a fused fuzzy set in one embodiment of this application. Detailed Implementation

[0069] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0070] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0071] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0072] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0073] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0074] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0075] To address the low prediction accuracy of existing joint angle prediction methods, this application proposes a joint motion estimation method. This method decomposes electromyography (EMG) signal data samples at N time points into multiple initial fuzzy sets. The first membership degree of each EMG signal data sample to each initial fuzzy set is calculated. These initial fuzzy sets are then fused to obtain multiple fused fuzzy sets. The data in each fused fuzzy set is then projected onto a feature space to obtain a relationship expression between joint angles and EMG signals. Finally, this relationship expression is used to calculate the joint angles corresponding to the EMG signal data to be estimated. By decomposing the EMG signal data samples into multiple initial fuzzy sets, the fuzziness and variability of EMG signals can be better handled. The fusion of the original fuzzy sets into fused fuzzy sets reduces the number of fuzzy sets. Projecting the data from the fused fuzzy sets onto the feature space eliminates the strong nonlinearity of EMG signals, thus significantly improving the accuracy of joint angle prediction.

[0076] The joint motion estimation method provided in this application will be described exemplarily below.

[0077] like Figure 1 As shown, the joint motion estimation method provided in this application includes the following steps:

[0078] Step 11: Collect training data, which includes N electromyographic signal data samples at different times.

[0079] Electromyography (EMG) signals are the temporal and spatial superposition of motor unit action potentials (MUAPs) from numerous muscle fibers. Among them, surface electromyography (SEMG) can reflect neuromuscular activity to a certain extent, thus SEMG has important practical value in clinical medicine, ergonomics, rehabilitation medicine, and sports science.

[0080] Electromyographic (EMG) signals can be acquired using sampling electrodes, and then transmitted via wires to a data acquisition card or specific medical instruments for analysis and processing.

[0081] It should be noted that the collected training data also includes joint angle data samples at the aforementioned N time points.

[0082] Step 12: Decompose the electromyographic signal data samples at N time points into multiple initial fuzzy sets, and obtain the first membership degree of the electromyographic signal data sample at each time point to each initial fuzzy set.

[0083] Decomposing the electromyographic signal data samples at N time points into multiple initial fuzzy sets is to obtain the first membership degree of the electromyographic signal data sample at each time point to each initial fuzzy set. The first membership degree can realize the fuzzy representation of the electromyographic signal, thereby avoiding the deviation of joint motion estimation caused by the fuzziness and differences of the electromyographic signal.

[0084] Step 13: Fuse multiple initial fuzzy sets to obtain multiple fused fuzzy sets.

[0085] Because electromyographic signals have strong nonlinearity, multiple fuzzy sets are needed to accurately represent the characteristics of electromyographic signals. However, when the number of fuzzy sets increases, it is easy to cause the curse of dimensionality. Therefore, fusing the initial fuzzy sets can reduce the number of fuzzy sets and avoid the occurrence of the curse of dimensionality.

[0086] It should be noted that the number of fused fuzzy sets is less than the number of initial fuzzy sets.

[0087] Step 14: Project the data from each fused fuzzy set onto the feature space to obtain the relationship expression between joint angles and electromyographic signals.

[0088] While step 13 can reduce the number of fuzzy sets and avoid the curse of dimensionality, it also leads to enhanced nonlinearity of the data in the fuzzy sets. The data in the fuzzy sets cannot be linearly represented, which is not conducive to establishing the relationship between joint angles and electromyographic signals. However, by projecting the data in each fused fuzzy set onto the feature space, the data in each fused fuzzy set can present a linear relationship, eliminating the drawbacks caused by the strong nonlinearity of electromyographic signals.

[0089] Step 15: Calculate the joint angles using the relational expression to obtain the joint angle data corresponding to the electromyography signal data to be estimated.

[0090] The ability to calculate joint angles by collecting electromyographic signal data to be estimated demonstrates the real-time nature and simplicity of this method.

[0091] It is worth mentioning that, since the electromyography signal data samples are decomposed into multiple initial fuzzy sets, the fuzziness and differences of the electromyography signals can be handled better. The original fuzzy sets are fused into a fused fuzzy set, reducing the number of fuzzy sets. Then, the data in the fused fuzzy set is projected onto the feature space, which can eliminate the strong nonlinearity of the electromyography signals. As a result, the joint angle estimation method of this application can greatly improve the accuracy of joint angle prediction.

[0092] The specific steps of step 12 are illustrated below.

[0093] Step 12.1: Based on the fuzzy clustering method, decompose the electromyography signal data samples at N time points into L initial fuzzy sets.

[0094] The aforementioned fuzzy clustering method can be the fuzzy C-means clustering method, which decomposes the electromyographic signal data samples collected in step 11 at N time points into L initial fuzzy sets. The electromyographic signal data in each initial fuzzy set has the highest similarity, while the electromyographic signal data in different initial fuzzy sets have the lowest similarity.

[0095] It is worth mentioning that decomposing the electromyographic signal data samples into an initial fuzzy set can eliminate the negative impact of the variability of electromyographic signals on joint angle estimation.

[0096] Step 12.2: Calculate the cluster center and variance of each initial fuzzy set.

[0097] By continuously optimizing the center of the initial fuzzy set and the distance from the electromyography signal data sample to the center of the initial fuzzy set, the cluster center μ of each initial fuzzy set is obtained. k The variance τ of the initial fuzzy set k .

[0098] It should be noted that the cluster center μ of each initial fuzzy set is calculated. k The variance τ of the initial fuzzy set k The method is common knowledge and will not be elaborated here.

[0099] Step 12.3: Calculate the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set based on the cluster center and variance of each initial fuzzy set.

[0100] Through calculation formula

[0101]

[0102] Obtain the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set; where, x represents the electromyographic signal data sample at time i. i For the first membership degree of the k-th initial fuzzy set, μ k Let τ represent the cluster center of the k-th initial fuzzy set. k Let represent the variance of the k-th initial fuzzy set, k = 1, ..., L, i = 1, ..., N.

[0103] It is worth mentioning that using membership to represent electromyographic signals can achieve fuzzy representation of electromyographic signals and eliminate the negative impact of the fuzziness of electromyographic signals on joint angle estimation.

[0104] The specific steps of step 13 are illustrated below.

[0105] The first step is to use the calculation formula:

[0106]

[0107] Obtain the distance D(A1, A2) between the initial fuzzy set A1 and the initial fuzzy set A2; the initial fuzzy set A1 and the initial fuzzy set A2 are two initial fuzzy sets in the L initial fuzzy sets.

[0108] The second step is to use the calculation formula:

[0109]

[0110] Obtain the similarity S(A1, A2) between the initial fuzzy set A1 and the initial fuzzy set A2.

[0111] It is understandable that the shorter the distance between fuzzy sets, the higher the similarity between them.

[0112] The third step is to use the formula. Determine whether to fuse the initial fuzzy sets A1 and A2; θ represents the similarity threshold, which can be preset according to the actual situation, and K(A1, A2) represents the fusion indicator.

[0113] If the fusion indicator K(A1, A2) = 1, then the initial fuzzy set A1 and the initial fuzzy set A2 are fused to obtain the fused fuzzy set; otherwise, it means that the initial fuzzy set A1 and the initial fuzzy set A2 have low similarity and do not need to be fused.

[0114] After the above steps, all similar initial fuzzy set pairs are merged into a larger fused fuzzy set. If multiple initial fuzzy sets are similar, they are merged together into a larger fused fuzzy set. For example... Figure 2 As shown, the initial fuzzy set A k (k = 1, 2, ..., L) are reconstituted into a fusion fuzzy set F. j (j = 1, 2, ..., R), if the fuzzy set F is fused j From the initial fuzzy set A1~A k It is formed by fusion.

[0115] It is worth mentioning that by merging initial fuzzy sets with high similarity into a fused fuzzy set, the number of fuzzy sets can be reduced, thus avoiding the curse of dimensionality caused by an excessive number of fuzzy sets.

[0116] It should be noted that after fusing multiple initial fuzzy sets to obtain multiple fused fuzzy sets, the following operations are still required:

[0117] Through calculation formula

[0118]

[0119] Obtain the second membership φ of the electromyography signal data samples at each time step to each fused fuzzy set. j (x i Where sk represents the weighting factor, and A k Let F represent the k-th initial fuzzy set. j Let J represent the j-th fused fuzzy set, where j = 1, ..., R, and R represents the total number of fused fuzzy sets, where R << L. Figure 2 middle, This represents the calculation of the second membership degree φ1 of the electromyographic signal data sample to the fused fuzzy set F1. This indicates the computation of electromyographic signal data samples to fuse the fuzzy set F. RThe second membership degree φ R s1, s2, ..., s l1 , ..., s lq+1 , ..., s L All of these represent weighting factors.

[0120] Through calculation formula

[0121]

[0122] Obtain the cluster center m of the j-th fused fuzzy set j .

[0123] Through calculation formula

[0124]

[0125] The variance σ of the j-th fused fuzzy set is obtained. j .

[0126] It is worth mentioning that the above calculation is performed to calculate the second membership degree of the actual electromyographic signal data to be estimated to each fused fuzzy set, which facilitates the subsequent calculation of joint angles.

[0127] The specific steps of step 14 are illustrated below.

[0128] Step 14.1: Project the data of each fused fuzzy set onto the feature space to obtain the correlation between joint angles and electromyographic signals in each fused fuzzy set.

[0129] The correlation between the joint angles and electromyographic signals obtained for each fused fuzzy set is as follows:

[0130]

[0131] Among them, f j (x i ) represents the electromyographic signal data sample x in the j-th fused fuzzy set. i Corresponding joint angle data samples, Let p be the spatial projection function. g represents the sigmoid function, q represents the number of nodes in the spatial projection, and w p and t p Both represent spatial projection parameters, a j and b j Both represent consequent parameters in the fuzzy rule of the j-th fused fuzzy set, a jp Let a represent the space projection function corresponding to the p-th spatial projection function. j .

[0132] Step 14.2: Based on the correlation between joint angles and electromyographic signals in each fused fuzzy set, obtain the relationship between joint angles and electromyographic signals.

[0133] The relationship between the joint angle and the electromyographic signal is as follows:

[0134]

[0135] Where, f(x) i ) represents electromyographic signal data sample x i The corresponding joint angle data sample.

[0136] Step 14.3: Solve for the parameters in the relation and use the obtained parameter values ​​to obtain the relational expression between the joint angle and the electromyographic signal.

[0137] The parameters in the above relation include the spatial projection parameter {w} p , t p}, number of nodes in spatial projection q, weight factor s k And the consequent parameter {a j b j}

[0138] Step 14.3.1, based on the Gravity Search Algorithm (GSA optimization algorithm), utilize the training data (electromyography signal data samples at N time points and joint angle data samples at N time points) and the first membership degree. ) on relational expressions Solving for the spatial projection parameters {w} yields the results. p , t p}, weighting factor s k And the number of nodes q in the spatial projection.

[0139] It should be noted that the GSA optimization algorithm is common knowledge and will not be elaborated upon here.

[0140] Step 14.3.2, based on the spatial projection parameters {w p , t p}, thus obtaining the spatial projection function

[0141] Step 14.3.3, based on the least squares method, using the spatial projection function and first membership degree Solve the relation to obtain the consequent parameter {a}. j b j}

[0142] Specifically, the spatial projection function and first membership degree Substitute these values ​​into the relation and solve for the consequent parameter {a} using the least squares method. j b j The equation can be transformed into ∑χ=η.

[0143] in,

[0144]

[0145]

[0146] Ω=(H1H2...H R ) N×(R×q)

[0147]

[0148] A = [a 11 ...a 1q ...a Rq ] T

[0149] Solving for x, we get χ = ​​(∑ T ∑) -1 ∑ T η.

[0150] First, the optimal parameters for each particle are obtained, and the average error is calculated based on these parameters. Then, the particle positions are updated until the model error converges. Finally, the globally optimal parameters are substituted to obtain the optimal parameters, which are the consequent parameters {a}. j b j}

[0151] Step 14.3.4: Substitute the obtained parameter values ​​into the relational expression to obtain the relationship expression between the joint angle and the electromyographic signal.

[0152] The specific steps of step 15 are illustrated below.

[0153] The relationship between the joint angle and the electromyographic signal obtained from the above steps is expressed as follows:

[0154]

[0155] Where f(x) represents the joint angle data corresponding to the electromyographic signal data to be estimated, and x represents the electromyographic signal data to be estimated.

[0156] Step 15.1, using the calculation formula

[0157]

[0158] The second membership γ of the electromyography signal data to be estimated to each fused fuzzy set is obtained. j ; where γj Let represent the second membership degree of the electromyographic signal data to be estimated to the j-th fused fuzzy set, where j = 1, ..., R.

[0159] Step 15.2, let φ j (x)=γ j , The joint angle data f(x) corresponding to the electromyographic signal data to be estimated is calculated.

[0160] It should be noted that when calculating the joint angle data corresponding to the electromyographic signal data to be estimated using the relationship expression between joint angle and electromyographic signal, since the parameters in the relationship expression have already been solved, the joint angle data f(x) corresponding to the electromyographic signal data to be estimated can be obtained by inputting the membership degree and spatial projection function corresponding to the electromyographic signal data to be estimated into the relationship expression. To better understand the technical solution provided in this application, the following illustrative examples of specific embodiments will be provided.

[0161] Collect training data Where x i Let y represent the electromyographic signal data sample at time i. i Let i represent the joint angle data sample at time i, where i = 1, ..., N.

[0162] Initialize the original rule L (the number of initial fuzzy sets) and the similarity threshold θ.

[0163] The FCM clustering algorithm was used to decompose the collected electromyographic signal data samples at N time points into L initial fuzzy sets, and then the cluster center μ of each initial fuzzy set was calculated. k and variance τ k Finally, based on the cluster center μ of each initial fuzzy set... k and variance τ k Calculate the first membership degree of the electromyography signal data sample at each time step to each initial fuzzy set. k = 1, ..., L.

[0164] Based on training data The first membership degree of each time step electromyography signal data sample to each initial fuzzy set. Calculate spatial projection parameters {w p , t p}, weighting factor s k And the number of nodes q in the spatial projection. The spatial projection parameters {w} p , t p Substituting the number of nodes q in the spatial projection into the spatial projection function yields the spatial projection function.

[0165] Randomly initialize M particles, and optimize the parameters as needed (possibly the consequent parameter {a}). j b j Define the dimension of each particle.

[0166] Based on the similarity criterion and fuzzy set fusion strategy (decomposing the EMG signal data samples at N time points into multiple initial fuzzy sets, obtaining the first membership degree of the EMG signal data sample at each time point to each initial fuzzy set), the original membership degree of the samples (the first membership degree of the EMG signal data samples at N time points to the initial fuzzy sets) is used to... ) and the particle's parameter s k Substitute into the formula ( This yields the second membership φ of the electromyography signal data samples at N time points to each fused fuzzy set. j (x) and the cluster center m and variance σ of the fused fuzzy set.

[0167] The spatial projection parameter {w} of the particle p , t p Substitute into the formula Obtain the number of spatial projection layers

[0168] Will and Substitute into the following formula

[0169]

[0170] Transform the equation into ∑χ=η.

[0171] in,

[0172]

[0173]

[0174] Ω=(H1H2...H R ) N×(R×q)

[0175]

[0176] A = [a 11 ...a 1q ...a Rq ] T

[0177] Solving for x, we get χ = ​​(∑ T ∑) -1 ∑ T η.

[0178] First, the optimal parameters for each particle are obtained, and the average error is calculated based on these parameters. Then, the particle positions are updated until the model error converges. Finally, the globally optimal parameters are substituted to obtain the optimal parameters, which are the consequent parameters {a}. j b j}

[0179] Substituting the obtained parameter values ​​into the relational formula, we obtain the expression relating the joint angle and the electromyographic signal:

[0180]

[0181] Next, calculate the second membership γ of the electromyographic signal data to be estimated to each fused fuzzy set. j Specifically, through calculation formula

[0182]

[0183] The second membership γ of the electromyography signal data to be estimated to each fused fuzzy set is obtained. j ; where γ j This represents the second membership degree of the electromyographic signal data to be estimated to the j-th fused fuzzy set, j = 1, ..., R.

[0184] Then use γ j The value of the parameter φ in the relational expression is replaced. j The value of (x) can be obtained by calculating the relational expression. Finally, the joint angle data corresponding to the electromyographic signal data to be estimated can be obtained.

[0185] This application is mainly used for the estimation of joint motion and has the following advantages:

[0186] (1) Based on fuzzy clustering, high-dimensional time-series electromyography data is decomposed into multiple fuzzy sets, and the center of fuzzy rules is used to replace the original high-dimensional feature data, which reduces the complexity of the model and can better handle the fuzziness and uncertainty of electromyography signals.

[0187] (2) A low-order adaptive fuzzy neural network modeling method was designed. This model further fuses rules based on the similarity of fuzzy sets, while optimizing the membership degree of the new fuzzy sets to obtain a new expression method for the fuzzy sets. An activation function is used to map the fused nonlinear data to the feature space, making it exhibit a linear relationship. Finally, a linear model is established for each fuzzy set, and real-time prediction of joint angles is achieved through defuzzification operations.

[0188] (3) A hybrid learning method was developed to identify the parameters, and the optimal parameters of the model were obtained. The above description is a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for estimating joint motion, characterized in that, include: Collect training data; The training data includes electromyographic signal data samples at N time points; The electromyography (EMG) signal data samples at N time points are decomposed into multiple initial fuzzy sets, and the first membership degree of the EMG signal data sample at each time point to each initial fuzzy set is obtained. Multiple initial fuzzy sets are fused to obtain multiple fused fuzzy sets; The number of fused fuzzy sets is less than the number of initial fuzzy sets; By projecting the data from each fused fuzzy set onto the feature space, the relationship expression between joint angles and electromyographic signals is obtained. The joint angle data corresponding to the electromyographic signal data to be estimated is obtained by using the relational expression to calculate the joint angle data of the electromyographic signal data to be estimated. The step of projecting the data from each fused fuzzy set onto the feature space to obtain the relationship expression between joint angles and electromyographic signals includes: The data in each fused fuzzy set is projected onto the feature space to obtain the correlation between joint angles and electromyographic signals in each fused fuzzy set; Based on the correlation between joint angles and electromyographic signals in each fused fuzzy set, the relationship between joint angles and electromyographic signals is obtained; Solve for the parameters in the relation, and use the obtained parameter values ​​to obtain the relationship expression between the joint angle and the electromyographic signal; The training data also includes joint angle data samples at the N time points, and the correlation between the joint angles in each fused fuzzy set and the electromyographic signals is as follows: Among them, f j (x i ) represents the electromyographic signal data sample x in the j-th fused fuzzy set. i Corresponding joint angle data samples, Let p be the spatial projection function. g represents the sigmoid function, q represents the number of nodes in the spatial projection, and w p and t p Both represent spatial projection parameters, a j and b j Both represent consequent parameters in the fuzzy rule of the j-th fused fuzzy set, a jp Let a represent the space projection function corresponding to the p-th spatial projection function. j .

2. The estimation method according to claim 1, characterized in that, The process of decomposing the electromyographic signal data samples at N time points into multiple initial fuzzy sets, and obtaining the first membership degree of the electromyographic signal data sample at each time point to each initial fuzzy set, includes: Based on the fuzzy clustering method, the electromyography signal data samples at N time points are decomposed into L initial fuzzy sets; Calculate the cluster center and variance of each of the initial fuzzy sets; Based on the cluster center and variance of each initial fuzzy set, calculate the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set.

3. The estimation method according to claim 2, characterized in that, The step of calculating the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set based on the cluster center and variance of each initial fuzzy set includes: Through calculation formula Obtain the first membership degree of the electromyographic signal data sample at each time step to each initial fuzzy set; in, x represents the electromyographic signal data sample at time i. i For the first membership degree of the k-th initial fuzzy set, μ k Let τ represent the cluster center of the k-th initial fuzzy set. k Let represent the variance of the k-th initial fuzzy set, k = 1, ..., L, i = 1, ..., N.

4. The estimation method according to claim 3, characterized in that, The process of fusing multiple initial fuzzy sets to obtain multiple fused fuzzy sets includes: Calculation formula: Obtain the distance D(A1,A2) between the initial fuzzy set A1 and the initial fuzzy set A2; the initial fuzzy set A1 and the initial fuzzy set A2 are two initial fuzzy sets from the L initial fuzzy sets; Calculation formula: Obtain the similarity S(A1,A2) between the initial fuzzy set A1 and the initial fuzzy set A2; Through formula Determine whether to fuse the initial fuzzy sets A1 and A2; θ represents the similarity threshold, and K(A1,A2) represents the fusion indicator; If the fusion indicator K(A1,A2) = 1, then the initial fuzzy set A1 and the initial fuzzy set A2 are fused to obtain a fused fuzzy set.

5. The estimation method according to claim 4, characterized in that, After fusing the multiple initial fuzzy sets to obtain multiple fused fuzzy sets, the estimation method further includes: Through calculation formula Obtain the second membership φ of the electromyography signal data samples at each time step to each fused fuzzy set. j (x i ); where s k A represents the weighting factor. k Let F represent the k-th initial fuzzy set. j Let represent the j-th fused fuzzy set, j = 1, ..., R, where R represents the total number of fused fuzzy sets, R < 1 / R. <L; Through calculation formula Obtain the cluster center m of the j-th fused fuzzy set j ; Through calculation formula The variance σ of the j-th fused fuzzy set is obtained. j .

6. The estimation method according to claim 1, characterized in that, The relationship between the joint angle and the electromyographic signal is as follows: Where, f(x) i ) represents electromyographic signal data sample x i The corresponding joint angle data sample.

7. The estimation method according to claim 6, characterized in that, The parameters in the relation include the spatial projection parameter {w} p ,t p The number of nodes in the spatial projection q and the weighting factor s k And the consequent parameter {a j ,b j }; Solving for the parameters in the relation includes: Based on the GSA optimization algorithm, the relation is solved using the training data and the first membership degree to obtain the spatial projection parameter {w}. p ,t p }, weighting factor s k And the number of nodes q in the spatial projection; According to the spatial projection parameter {w p ,t p }, thus obtaining the spatial projection function Based on the least squares method, using the spatial projection function Solving the relation with the first membership degree yields the consequent parameter {a}. j ,b j } 8. The estimation method according to claim 7, characterized in that, The relationship between the joint angle and the electromyographic signal is expressed as follows: Where f(x) represents the joint angle data corresponding to the electromyographic signal data to be estimated, and x represents the electromyographic signal data to be estimated; The step of calculating joint angles using the relational expression to obtain joint angle data corresponding to the electromyographic signal data to be estimated includes: Through calculation formula The second membership degree γ of the electromyographic signal data to be estimated to each fused fuzzy set is obtained. j ; where γ j This represents the second membership degree of the electromyographic signal data to be estimated to the j-th fused fuzzy set, where j = 1,...,R; Let φ j (x)=γ j , The joint angle data f(x) corresponding to the electromyographic signal data to be estimated is calculated.

Images