Accurate recognition method for lower limb actions based on surface electromyogram signal decomposition and optimization processing

Abstract

The invention discloses a lower limb action accurate recognition method based on surface electromyogram signal decomposition and optimization processing, which comprises the following steps: acquiring a surface electromyogram signal, and decomposing the surface electromyogram signal to obtain sub-signals; carrying out optimization processing on the sub-signals through a two-stage modal selection method, carrying out feature extraction and fusion on the sub-signals after optimization processing to obtain a feature vector, and carrying out dimension reduction on the feature vector; wherein the two-stage modal selection method comprises an energy contribution selection method and a modal correlation selection method; and recognizing the dimensionality-reduced feature vector through a classification algorithm to obtain a lower limb action recognition result.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and in particular relates to a method for accurately identifying lower limb movements based on surface electromyography signal decomposition and optimization processing. Background Art

[0002] With the rapid development of biomechanics, sports medicine and human-computer interaction technology, lower limb motion recognition technology based on surface electromyography signals is increasingly widely used in rehabilitation medicine, prosthetic control, sports training and other fields. As a non-invasive physiological signal, surface electromyography signals can reflect the state of muscle activity and thus reflect the dynamic process of human movement. For different lower limb motion recognition tasks, how to extract accurate and effective features from surface electromyography signals has become a key research issue.

[0003] Traditional surface electromyography signal analysis methods mainly rely on the time domain, frequency domain or time-frequency domain characteristics of the signal. However, since surface electromyography signals are affected by factors such as noise and motion artifacts, directly processing the original surface electromyography signals may result in lower recognition accuracy. In addition, surface electromyography signals usually have highly nonlinear and time-varying characteristics, so it is particularly important to extract features that comprehensively and accurately characterize the complexity of surface electromyography signals.

[0004] At present, although there are some lower limb movement recognition methods based on surface electromyography signals, there are still problems such as low signal quality and incomplete feature extraction. Therefore, how to improve the processing accuracy of surface electromyography signals and build an efficient and accurate lower limb movement recognition system is still a technical challenge that needs to be solved urgently. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a method for accurately identifying lower limb movements based on surface electromyography signal decomposition and optimization processing to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above objectives, the present invention provides a method for accurately identifying lower limb movements based on surface electromyography signal decomposition and optimization processing, comprising:

[0007] Acquire a surface electromyography signal, and decompose the surface electromyography signal to obtain sub-signals;

[0008] The sub-signals are optimized by a two-stage modal selection method, and features are extracted and fused on the optimized sub-signals to obtain feature vectors, which are then dimensionally reduced; wherein the two-stage modal selection method includes an energy contribution selection method and a modal correlation selection method;

[0009] The feature vector after dimension reduction is identified by a classification algorithm to obtain a lower limb movement recognition result.

[0010] Optionally, the process of obtaining the surface electromyogram signal includes:

[0011] Obtain the original surface electromyogram signal, preprocess the original surface electromyogram signal to obtain the surface electromyogram signal; wherein the preprocessing process includes filtering, denoising and normalization processing; wherein the filtering uses a band-pass filter of 20 - 450 Hz and a notch filter of 50 Hz, and the denoising uses the multi-scale principal component analysis method.

[0012] Optionally, decompose the surface electromyogram signal by the empirical mode decomposition method.

[0013] Optionally, the process of optimizing the sub-signals by the two-stage mode selection method includes:

[0014] Calculate the energy of the sub-signals by the energy contribution method to obtain the proportion of the energy contribution of the sub-signals, remove the sub-signals with the proportion of the energy contribution lower than 2% to obtain the energy sub-signals;

[0015] Calculate the correlation between the energy sub-signals and the surface electromyogram signal by the mode correlation selection method to obtain the correlation between the energy sub-signals and the correlation between the energy sub-signals and the surface electromyogram signal, and retain the energy sub-signals with the correlation between the energy sub-signals less than 85% and the correlation between the energy sub-signals and the surface electromyogram signal greater than 10% to obtain the optimized sub-signals.

[0016] Optionally, the process of obtaining the proportion of the energy contribution of the sub-signals includes:

[0017] Calculate the energy of the sub-signals:

[0018]

[0019] E k is the energy of the k-th sub-signal sub, t i is the time series of the sub-signal, and N is the number of data points;

[0020] Calculate the proportion of the energy contribution of the sub-signal:

[0021]

[0022] wherein, E total is the total energy of the signal, which is the summation result of the energies of the sub-signals, and Z k represents the proportion of the energy contribution of the k-th sub-signal.

[0023] Optionally, the process of correlation calculation includes:

[0024]

[0025] Among them, x i , y i are sample points of two signals. During the calculation of the correlation between the energy sub-band signal and the surface electromyogram signal, x i represents the energy sub-band signal, and y i represents the pure surface electromyogram signal. During the calculation of the correlation between energy sub-band signals, x i represents the energy sub-band signal, and y i represents another energy sub-band signal; are the means of the two signals respectively.

[0026] Optionally, the process of feature extraction and fusion for the optimized sub-signals includes:

[0027] Feature extraction is performed on the optimized sub-signals through the k-nearest neighbor estimated entropy feature calculation method to obtain the extracted features, and a feature vector is obtained by combining the extracted features.

[0028] Optionally, the process of obtaining the extracted features includes:

[0029]

[0030] Among them, represents the k-nearest neighbor estimated entropy feature, that is, the extracted feature, ψ is the digamma function, N is the number of samples, ε(i) is the distance between the i-th sample and its k-th nearest neighbor sample, i represents the sample serial number, k represents the serial number of the neighboring sample, c d is the volume of a unit sphere with dimension d, d represents the corresponding dimension, and its c d The Euclidean regularization calculation formula is:

[0031]

[0032] Among them, Γ(.) is the Gamma function.

[0033] Optionally, the feature vector is dimensionally reduced through a feature dimensionality reduction algorithm, and the diffusion mapping feature dimensionality reduction algorithm is used for the feature dimensionality reduction algorithm.

[0034] Optionally, the classification algorithm uses a machine learning classification algorithm.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] A precise lower limb movement recognition method based on surface electromyogram signal decomposition and optimization processing provided by the present invention captures rich muscle activity information by collecting surface electromyogram signals from different lower limb muscle parts; through preprocessing techniques such as band-pass filtering, notch filtering, multi-scale principal component analysis, and normalization, it reduces the interference of random noise and motion artifacts on the surface electromyogram signals, ensuring the purity of the surface electromyogram signal data; by using the empirical mode decomposition method to decompose the pure surface electromyogram signals into multiple sub-band signals and adopting a two-stage mode selection method to select multiple sub-component signals, it can more effectively extract the multi-level information of the surface electromyogram signals, improving the accuracy of signal processing and the expression ability of signal features; by extracting the k-nearest neighbor estimation entropy features from the decomposed sub-components, it can accurately quantify the non-linearity and complexity of the surface electromyogram signals, providing more sensitive and accurate features for lower limb movement recognition and improving the accuracy of lower limb movement recognition; by using the diffusion mapping feature dimensionality reduction algorithm to reduce the dimensionality of the extracted k-nearest neighbor estimation entropy features, removing the noise and redundant parts, it improves the efficiency and accuracy of the subsequent classification algorithm; by using machine learning algorithms to recognize lower limb movements, the purpose that the lower limb movement recognition system can maintain high accuracy and stability is achieved.

[0037] The technical solution proposed by the present invention is applicable to various lower limb movement recognition scenarios, and has broad application potential especially in the fields of rehabilitation medicine, intelligent prosthetic control, sports monitoring, etc. By accurately recognizing the lower limb movement state, it can provide personalized rehabilitation feedback for patients, provide real-time control signals for intelligent prosthetics, improve the use effect and comfort of prosthetics, and has good market prospects and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0039] Figure 1 It is a schematic diagram of the overall method flow of the embodiment of the present invention;

[0040] Figure 2 It is a flow chart of surface electromyogram signal processing of the embodiment of the present invention;

[0041] Figure 3 It is a flow chart of surface electromyogram signal decomposition and optimization processing of the embodiment of the present invention;

[0042] Figure 4 It is a flow chart of feature extraction of the embodiment of the present invention;

[0043] Figure 5 It is a flow chart of feature dimensionality reduction of the embodiment of the present invention;

[0044] Figure 6 It is a flowchart for lower limb movement recognition according to an embodiment of the present invention. Specific implementation manners

[0045] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the drawings and in combination with the embodiments.

[0046] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from here.

[0047] A precise lower limb movement recognition method based on surface electromyogram signal decomposition and optimization processing provided by the present invention includes the following steps:

[0048] Install surface electromyogram sensors on different muscle parts of the human lower limbs to collect the original surface electromyogram signals of multiple muscles during the execution of different lower limb movements (such as walking on flat ground, going up stairs, going down stairs, and jumping).

[0049] Preprocess the collected original surface electromyogram signals. Use band-pass filtering to remove low-frequency noise and high-frequency noise to ensure that the effective frequency band of the surface electromyogram signal is retained; use notch filtering to remove power frequency noise in the surface electromyogram signal; use multi-scale principal component analysis to remove random noise in the surface electromyogram signal; perform amplitude normalization processing on the surface electromyogram signal to eliminate differences between different muscle parts or individuals, and obtain pure surface electromyogram signals.

[0050] Apply a data-driven decomposition method to decompose the preprocessed pure surface electromyogram signal, decompose the pure surface electromyogram signal into multiple sub-band signals; perform a two-stage modal selection method on the sub-band signals to obtain multiple sub-component signals. Through multi-level signal decomposition, more dynamic features hidden in the signal can be extracted, and the data-driven decomposition method uses the empirical mode decomposition method.

[0051] Extract entropy features from multiple sub-components and combine them into a feature vector. This feature can effectively reflect the complexity and non-linear characteristics of the signal, thus providing more useful information for lower limb movement recognition.

[0052] Based on a feature dimensionality reduction algorithm, reduce the dimensionality of the extracted entropy features, remove redundant information, and improve the calculation efficiency and recognition accuracy.

[0053] Input the feature set after dimensionality reduction into a classification algorithm to accurately recognize four lower limb movements (such as walking on flat ground, going up stairs, going down stairs, and jumping).

[0054] The present invention can be integrated into applications such as wearable devices, rehabilitation training systems, and prosthetic control systems. Based on the above technical solutions, it is possible to monitor and accurately identify human lower limb movements, which is applicable to various scenarios such as sports training, rehabilitation therapy, and assisted prosthetic control.

[0055] The present invention proposes a method for accurately identifying lower limb movements based on surface electromyogram (sEMG) signal decomposition and optimization processing. By using the sEMG signal acquisition technology of multiple lower limb muscles, raw sEMG signals with richer and more comprehensive motion information are obtained; through signal preprocessing technology, noise is removed from the raw sEMG signals to obtain purer and more reliable sEMG signals; through signal decomposition technology, the pure sEMG signals are decomposed into multiple sub-band signals, and then a two-stage mode selection method is used to obtain multiple sub-band signals to extract multi-level sEMG signal information, improving the expression ability of sEMG signals and the richness of information; entropy features are used to quantify the sEMG signals; a feature dimensionality reduction method is adopted to reduce redundant features and improve the classification efficiency and accuracy; a classification algorithm is used to achieve efficient and accurate lower limb movement recognition.

[0056] The above technical solutions of the present invention will be described in detail in conjunction with the relevant drawings. As Figure 1 shown, a method for accurately identifying lower limb movements based on surface electromyogram (sEMG) signal decomposition and optimization processing provided by the present invention specifically includes the following contents:

[0057] Step 1: sEMG signal acquisition:

[0058] Surface electromyogram sensors are installed at multiple key muscle sites of the human lower limb, and sEMG signals are synchronously acquired at a sampling frequency of 1000 Hz to obtain raw sEMG signals with richer and more comprehensive motion information.

[0059] The muscle sites include: rectus femoris, vastus medialis, vastus lateralis, semitendinosus, tibialis anterior, lateral gastrocnemius, and medial gastrocnemius. These muscle groups play an important role in motion control during lower limb movements.

[0060] As Figure 2 shown, Step 2: sEMG signal preprocessing:

[0061] The process of sEMG signal processing includes filtering, denoising, and normalization of sEMG signals. This step ensures the purity of sEMG signals and lays a foundation for subsequent analysis. Specifically, it includes:

[0062] For the acquired raw sEMG signals, a band-pass filter with a frequency range of 20 - 450 Hz is used to remove low-frequency and high-frequency noises, and the frequency band related to the sEMG signals of lower limb movements is retained.

[0063] For the frequency band related to the surface electromyogram (sEMG) signals of lower limb movements, a notch filter with a frequency of 50 Hz is used to remove power frequency noise, and the filtered sEMG signals are obtained.

[0064] For the filtered sEMG signals, multi-scale principal component analysis is used to remove the random noise and artifacts existing in the sEMG signals, and the denoised sEMG signals are obtained.

[0065] The denoised sEMG signals are normalized so that the signal amplitudes between different electrodes are consistent. The normalization process selects to normalize the signals to the range of [0, 1].

[0066] The calculation formula for the normalization process is:

[0067]

[0068] where x min is the minimum value of the sEMG signal, and x max is the maximum value of the sEMG signal.

[0069] As Figure 3 shown, Step 3: Decomposition and optimization of sEMG signals:

[0070] The pure sEMG signals are decomposed into multiple sub-signals through a two-stage mode selection method. Specifically, it includes:

[0071] The empirical mode decomposition method is performed on the preprocessed pure sEMG signals to decompose the pure sEMG signals into multiple sub-band signals.

[0072] The two-stage mode selection method is applied to multiple sub-band signals to select the sub-components that can effectively characterize the lower limb movements.

[0073] The first stage of the two-stage mode selection method is the energy contribution selection method: calculate the energy of each sub-band signal, and count the proportion of the energy contribution of each sub-band signal. Discard the sub-band signals with an energy contribution proportion lower than 2%, and obtain the energy sub-band signals;

[0074] The calculation formula for the energy of each sub-band signal is:

[0075]

[0076] where E k is the energy of the k-th sub-band signal sub, t i is the time series of the signal, N is the number of data points, and sub k () represents the k-th sub-band signal sub under the time series t i of the signal.

[0077] The calculation formula for the proportion of the energy contribution of each sub-band signal is:

[0078]

[0079] Among them, E total is the total energy of the signal, and Z k represents the proportion of the energy contribution of the k-th sub-band signal.

[0080] The second stage is the modal correlation selection method: calculate the correlation between the energy sub-band signals and the correlation between the energy sub-band signals and the pure surface electromyogram signal, and select the energy sub-band signals that simultaneously satisfy that the correlation between the energy sub-band signals is less than 85% and the correlation between the energy sub-band signals and the pure surface electromyogram signal is greater than 10% to obtain the sub-components that can effectively characterize the lower limb movements.

[0081] The correlation calculation formula is:

[0082]

[0083] Among them, x i , y i are the sample points of two signals. In the process of calculating the correlation between the energy sub-band signal and the surface electromyogram signal, x i represents the energy sub-band signal, and y i represents the pure surface electromyogram signal. In the process of calculating the correlation between the energy sub-band signals, x i represents the energy sub-band signal, and y i represents another energy sub-band signal; x and y are the means of the two signals respectively.

[0084] As Figure 4 shown, Step 4: Feature extraction:

[0085] Extract features from the multiple sub-signals obtained by decomposition and combine them into a feature vector. Specifically, it includes:

[0086] Extract the k-nearest neighbor estimation entropy features that are helpful for lower limb movement recognition from the decomposed sub-components. The calculation formula is as follows:

[0087]

[0088] Among them, represents the k-nearest neighbor estimation entropy feature, that is, the extracted feature, ψ is the digamma function, N is the number of samples, ε(i) is the distance between the i-th sample and its k-th nearest neighbor sample, i represents the sample number, k represents the nearest neighbor sample number, and c d is the volume of the unit sphere with dimension d, and its Euclidean regularization calculation formula is as follows:

[0089]

[0090] Among them, Γ(.) is the Gamma function.

[0091] A k-nearest neighbor estimated entropy feature calculated based on each sub-component. The k-nearest neighbor estimated entropy features calculated from all sub-components are combined into a feature vector as the input feature for lower limb movement recognition.

[0092] Such as Figure 5 shown, Step Five: Feature Dimensionality Reduction Algorithm:

[0093] Reduce the dimension of the feature vector through a feature dimensionality reduction algorithm to improve the efficiency and accuracy of lower limb movement recognition. Specifically, it includes:

[0094] Use the diffusion mapping feature dimensionality reduction algorithm to reduce the dimension of the feature vector, remove redundant information, and retain the principal components most discriminative for lower limb movement recognition to obtain the reduced-dimensional feature set.

[0095] Among them, the diffusion mapping feature dimensionality reduction algorithm includes the following steps:

[0096] (1) For the feature vector F = {f 1 ,..., f N}, where i represents the feature vector label, i = 1, 2,... N, establish a corresponding finite graph, and use the positive definite symmetric Gaussian kernel function to define the weights of the edges in the graph to obtain the weight matrix W, and its element w(f i , f j ) is:

[0097]

[0098] Among them, σ is the scale parameter of the Gaussian kernel.

[0099] (2) Construct the Markov weight matrix P, and its element p(f i , f j ) is:

[0100]

[0101] Among them represents the degree of the data point f i .

[0102] Each row in the matrix is the transition probability of a one-step random movement between any two data points, and the transition probability matrix P t (f i , f j ) after t-step random movement is:

[0103] P t (f i , f j ) = (P(fi , f j )) t

[0104] (3) Determine the diffusion distance

[0105] The diffusion distance D between two points t (f i , f j ) is defined as follows:

[0106]

[0107] where represents the characteristic that the greater the weight value of the high-density area in the figure.

[0108] (4) On the premise of maintaining the diffusion distance between data points, extract the low-dimensional manifold Y, which can be obtained from the following formula:

[0109] P t Y = λY

[0110] where λ represents the eigenvalue of matrix P t , discard the obtained largest eigenvalue (trivial eigenvalue) λ 0 = 1 and the corresponding eigenvector v 0 , take the eigenvectors corresponding to the remaining D eigenvalues as the low-dimensional embedding result to obtain the low-dimensional manifold Y, and complete the dimensionality reduction of the eigenvector F. Low-dimensional manifold Y: Y = {λ 1 v 1 , λ 2 v 2 ,..., λ D v D}, where v represents the eigenvector, and the subscripts 0, 1, 2, D represent the corresponding eigenvector numbers, and D represents the total number of eigenvectors extracted.

[0111] As Figure 6 shown, Step 6: Lower limb movement recognition:

[0112] Based on the dimensionality-reduced feature set, combine with a classification algorithm to complete the accurate recognition of lower limb movements. Specifically, it includes: inputting the dimensionality-reduced feature set into the support vector machine classification algorithm for lower limb movement recognition, realizing the accurate recognition of four lower limb movements. At the same time, it should be noted that the classification algorithm can also adopt machine learning classifiers such as linear discriminant analysis classification algorithm, ensemble bagging classification algorithm, K-nearest neighbor classification algorithm, and naive Bayes classification algorithm.

[0113] The four lower limb movements include: walking on flat ground, going up stairs, going down stairs, and jumping.

[0114] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for accurately identifying lower limb movements based on surface electromyography signal decomposition and optimization processing, characterized in that: include: Acquire a surface electromyography signal, and decompose the surface electromyography signal to obtain sub-signals; The sub-signals are optimized by a two-stage modal selection method, and features are extracted and fused on the optimized sub-signals to obtain feature vectors, which are then dimensionally reduced; wherein the two-stage modal selection method includes an energy contribution selection method and a modal correlation selection method; The feature vector after dimension reduction is identified by a classification algorithm to obtain a lower limb movement recognition result.

2. The method according to claim 1, characterized in that The acquisition process of the surface electromyography signal includes: The original surface electromyographic signal is obtained, and the original surface electromyographic signal is preprocessed to obtain a surface electromyographic signal; wherein the preprocessing process includes filtering, denoising and normalization processing; wherein the filtering adopts a 20-450Hz bandpass filter and a 50Hz notch filter, and the denoising adopts a multi-scale principal component analysis method.

3. The method according to claim 1, characterized in that The surface electromyography signal is decomposed by an empirical mode decomposition method.

4. The method according to claim 1, characterized in that: The process of optimizing the sub-signals through the two-stage mode selection method includes: The energy of the sub-signal is calculated by the energy contribution method to obtain the energy contribution ratio of the sub-signal, and the sub-signal with an energy contribution ratio less than 2% is removed to obtain the energy sub-signal; The correlation between energy sub-signals and surface electromyography signals is calculated by modal correlation selection method to obtain the correlation between energy sub-signals and the correlation between energy sub-signals and surface electromyography signals. Energy sub-signals with correlation between energy sub-signals less than 85% and correlation between energy sub-signals and surface electromyography signals greater than 10% are retained to obtain optimized sub-signals.

5. The method according to claim 4, characterized in that The process of obtaining the energy contribution ratio of the sub-signal includes: Calculate the energy of the sub-signal: E k is the energy of the kth sub-signal sub, t i is the time series of the sub-signal, and N is the number of data points; The energy contribution ratio of the sub-signal is calculated: Among them, E total is the total energy of the signal, which is the sum of the energies of the sub-signals, Z k Indicates the energy contribution ratio of the kth sub-signal.

6. The method according to claim 4, characterized in that The process of correlation calculation includes: Among them, x i ,y i are the sample points of the two signals. In the correlation calculation process between the energy sub-band signal and the surface electromyography signal, x i represents the energy subband signal, y i represents the pure surface electromyographic signal. In the correlation calculation process between energy sub-band signals, x i represents the energy subband signal, y i represents another energy sub-band signal; are the means of the two signals respectively.

7. The method according to claim 1, characterized in that The process of feature extraction and fusion of the optimized sub-signals includes: The optimized sub-signal is subjected to feature extraction by the k-nearest neighbor estimation entropy feature calculation method to obtain the extracted features, and the feature vector is obtained by combining the extracted features.

8. The method according to claim 1, characterized in that: The process of obtaining the extracted features includes: in, represents the k-nearest neighbor estimated entropy feature, i.e., the extracted feature, ψ is the digamma function, N is the number of samples, ε(i) is the distance between the i-th sample and its k-th nearest neighbor sample, i represents the sample number, k represents the adjacent sample number, c d is the volume of a unit sphere of dimension d, where d represents the corresponding dimension, and c d The Euclidean regularization calculation formula is: Among them, Γ(.) is the Gamma function.

9. The method according to claim 1, characterized in that: The feature vector is reduced in dimension by a feature dimensionality reduction algorithm, wherein the feature dimensionality reduction algorithm adopts a diffusion mapping feature dimensionality reduction algorithm.

10. The method according to claim 1, characterized in that The classification algorithm adopts a machine learning classification algorithm.

Images