Electromyographic noise detection method, apparatus, device, and storage medium

Abstract

The application provides an electromyographic noise detection method, device, equipment and storage medium; the method in the application comprises: collecting a surface electromyographic signal to be identified; performing feature extraction on the surface electromyographic signal to obtain standard electromyographic features and surface electromyographic features of the surface electromyographic signal; performing difference feature extraction on the standard electromyographic features and the surface electromyographic features to obtain electromyographic noise features; classifying the electromyographic noise features to determine the noise type of the electromyographic noise features and output a noise optimization reminder. The application can improve the extraction accuracy of electromyographic noise features, effectively classify the electromyographic noise, and accurately identify the electromyographic noise type in the surface electromyographic signal.

Description

Technical Field

[0001] This application relates to the field of surface electromyography (EMG) detection, specifically to an EMG noise detection method, apparatus, device, and storage medium. Background Technology

[0002] Currently, the acquisition of surface electromyography (SEMG) signals is one of the main research directions in wearable devices. SEMG is a direct manifestation of skeletal muscle excitation and contraction during human limb movements, and it has wide applications in sports health, rehabilitation medicine, and human-computer interaction. However, due to its weak and easily interfered-with characteristics, the acquisition environment for SEMG signals requires stringent conditions. Especially for wearable devices, the signal acquisition quality is often poor due to the user's physical condition and behavioral movements during wear. Furthermore, it is difficult to effectively detect wide-bandwidth, high-intensity noise during the acquisition process, such as motion artifacts and external electromagnetic interference, making effective analysis of the acquired signals impossible. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for detecting electromyography (EMG) noise, aiming to solve the technical problem of difficulty in detecting noise types during surface EMG detection in the prior art.

[0004] On one hand, embodiments of this application provide a method for detecting electromyographic noise, the method comprising the following steps:

[0005] Acquire surface electromyography signals to be identified;

[0006] Feature extraction is performed on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal;

[0007] Difference features are extracted from the standard electromyography (EMG) features and the surface EMG features to obtain EMG noise features;

[0008] The electromyographic noise features are classified to determine the noise type of the electromyographic noise features, and a noise optimization reminder is output.

[0009] In one possible implementation of this application, the electromyographic noise detection method is applied to an electromyographic noise detection model, which includes a first feature extraction unit and a second feature extraction unit.

[0010] The step of extracting features from the surface electromyography (EMG) signal to obtain standard EMG features and surface EMG features includes:

[0011] The surface electromyography signal is extracted by the first feature extraction unit to obtain the standard electromyography features of the surface electromyography signal.

[0012] The surface electromyography (EMG) signal is subjected to feature extraction by the second feature extraction unit to obtain the surface EMG features of the surface EMG signal.

[0013] In one possible implementation of this application, the step of extracting differential features from the standard electromyography (EMG) features and surface EMG features to obtain EMG noise features includes:

[0014] The electromyographic noise feature is obtained by subtracting the standard electromyographic feature and the surface electromyographic feature element by element using the electromyographic noise detection model.

[0015] The process of classifying the electromyographic noise features and determining the noise type of the electromyographic noise features includes:

[0016] The electromyographic noise detection model is used to perform convolutional classification on the electromyographic noise features to determine the noise type of the electromyographic noise features.

[0017] In one possible implementation of this application, before performing feature extraction on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal, the method further includes:

[0018] Obtain standard electromyography (EMG) samples, and extract features from the standard EMG samples to obtain standard sample features;

[0019] The preset feature extraction unit is iteratively trained based on the features of the standard sample to obtain the first feature extraction unit.

[0020] In one possible implementation of this application, the step of iteratively training a preset feature extraction unit based on the standard electromyographic features to obtain a first feature extraction unit includes:

[0021] The standard sample features are input into the encoder of the feature extraction unit, and the encoder encodes the standard sample features to obtain the standard electromyography training vector.

[0022] The standard electromyography training vector is input into the decoder of the feature extraction unit, and the training mean feature and training variance feature of the standard electromyography training vector are extracted by the decoder to generate a training feature vector.

[0023] The training feature vector is decoded, and the loss value is calculated on the decoded training feature vector using a preset loss function;

[0024] The encoder and decoder are optimized iteratively based on the loss value to obtain the first feature extraction unit.

[0025] In one possible implementation of this application, the electromyography noise detection model further includes a noise classification unit;

[0026] The step of performing convolutional classification on the electromyographic noise vector using the electromyographic noise detection model to determine the noise type of the electromyographic noise features includes:

[0027] The electromyographic noise features are input into the noise classification unit for convolution operation to extract the high-dimensional noise features of the electromyographic noise features;

[0028] The high-dimensional noise features are nonlinearly transformed according to a preset activation function to obtain a feature fitting vector;

[0029] The feature fitting vector is pooled and compressed to obtain a noise vector of a preset dimension;

[0030] The noise vector is matched with a preset label to determine the noise type of the noise vector.

[0031] In one possible implementation of this application, classifying the electromyographic noise features, determining the noise type of the electromyographic noise features, and outputting a noise optimization reminder includes:

[0032] The electromyographic noise features are classified and identified to obtain the noise type of the electromyographic noise features;

[0033] If the noise type is a background noise type, determine that the noise optimization reminder corresponding to the background noise type is a background noise reduction reminder, and output the background noise reduction reminder;

[0034] If the noise type is a motion artifact type, determine that the noise optimization reminder corresponding to the motion artifact type is an artifact removal reminder, and output the artifact removal reminder.

[0035] On the other hand, this application provides an electromyographic noise detection device, the electromyographic noise detection device comprising:

[0036] The electromyography (EMG) acquisition module is configured to acquire surface EMG signals to be identified.

[0037] The feature extraction module is configured to extract features from the surface electromyography (EMG) signal to obtain standard EMG features and surface EMG features of the surface EMG signal.

[0038] The noise identification module is configured to extract differential features from the standard electromyography features and the surface electromyography features to obtain electromyography noise features;

[0039] The noise classification module is configured to classify the electromyographic noise features, determine the noise type of the electromyographic noise features, and output a noise optimization reminder.

[0040] On the other hand, this application also provides an electromyography (EMG) noise detection device, the EMG noise detection device comprising:

[0041] One or more processors;

[0042] Memory; and

[0043] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the electromyographic noise detection method.

[0044] On the other hand, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to perform the steps in the electromyography noise detection method.

[0045] This application acquires the surface electromyography (EMG) signal to be identified and inputs it into different feature extraction branches for feature extraction. This yields the spatial mapping of the EMG signal in a feature space constructed from high signal-to-noise ratio signals, i.e., the standard EMG features and the surface EMG features of the EMG signal. After acquiring the standard EMG features and the surface EMG features, differential feature extraction is performed on them to obtain the feature differences between the standard EMG features and the surface EMG features, identifying these differences as EMG noise features. These EMG noise features are then classified to determine their noise type, and a noise optimization reminder corresponding to that noise type is output. This improves the accuracy of EMG noise feature extraction, effectively classifies the EMG noise, and accurately identifies the EMG noise type in the surface EMG signal. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a schematic diagram of a scenario for the electromyography noise detection method according to an embodiment of this application;

[0048] Figure 2 This is a flowchart illustrating one embodiment of the electromyography noise detection method in this application.

[0049] Figure 3 This is a flowchart illustrating an embodiment of the electromyography noise detection method provided in this application, which iteratively trains the first feature extraction unit.

[0050] Figure 4 This is a flowchart illustrating an embodiment of the electromyographic noise detection method provided in this application, which classifies electromyographic noise features and outputs noise optimization reminders based on the noise type of the electromyographic noise features.

[0051] Figure 5 This is a schematic diagram of one embodiment of the electromyography noise detection device provided in this application.

[0052] Figure 6 This is a schematic diagram of an embodiment of the electromyography noise detection device provided in this application. Detailed Implementation

[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0055] In this application, the term "exemplary" is used to mean "serving as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0056] Currently, the acquisition of surface electromyography (SEMG) signals is one of the main research directions in wearable devices. SEMG is a direct manifestation of skeletal muscle excitation and contraction during human limb movements, and it has wide applications in sports health, rehabilitation medicine, and human-computer interaction. However, due to its weak and easily interfered-with characteristics, the acquisition environment for SEMG signals requires stringent conditions. Especially for wearable devices, the signal acquisition quality is often poor due to the user's physical condition and behavioral movements during wear. Furthermore, it is difficult to effectively detect wide-bandwidth, high-intensity noise during the acquisition process, such as motion artifacts and external electromagnetic interference, making effective analysis of the acquired signals impossible.

[0057] Based on this, this application proposes a method, apparatus, device, and computer-readable storage medium for detecting electromyography noise, in order to solve the technical problem that it is difficult to detect the type of noise in the surface electromyography detection process in the prior art.

[0058] The electromyography (EMG) noise detection method in this embodiment of the invention is applied to an EMG noise detection device. The EMG noise detection device is disposed in an EMG noise detection equipment, which includes one or more processors, a memory, and one or more application programs. The one or more application programs are stored in the memory and configured to be executed by the processor to implement the EMG noise detection method. The EMG noise detection equipment can be a smart terminal, such as a mobile phone, tablet computer, smart wearable device, and smart computer. Optionally, the EMG noise detection equipment can also be a server or a service cluster composed of multiple servers.

[0059] like Figure 1 As shown, Figure 1This is a schematic diagram of a scenario for the electromyography (EMG) noise detection method according to an embodiment of this application. The EMG noise detection scenario in this embodiment includes an EMG noise detection device 100 (which integrates an EMG noise detection unit). The EMG noise detection device 100 is equipped with a computer-readable storage medium corresponding to the EMG noise detection method to execute the steps of the EMG noise detection method.

[0060] Understandable, Figure 1 The electromyography noise detection device in the scenario of the electromyography noise detection method shown, or the device included in the electromyography noise detection device, does not constitute a limitation on the embodiments of the present invention. That is, the number or type of electromyography noise detection device included in the scenario of the electromyography noise detection method, or the number or type of device included in each device, does not affect the overall implementation of the technical solution in the embodiments of the present invention, and can all be considered as equivalent substitutions or derivatives of the technical solutions claimed in the embodiments of the present invention.

[0061] In this embodiment of the invention, the electromyography noise detection device 100 is mainly used for: acquiring surface electromyography signals to be identified;

[0062] Feature extraction is performed on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal;

[0063] Differential feature extraction is performed on the standard electromyography (EMG) features and surface EMG features to obtain EMG noise features;

[0064] The electromyographic noise features are classified to determine the noise type of the electromyographic noise features, and a noise optimization reminder is output.

[0065] The electromyography noise detection device 100 in this embodiment of the invention can be an independent electromyography noise detection device, such as a mobile phone, tablet computer, smart wearable device, server and smart computer, or an electromyography noise detection network or electromyography noise detection cluster composed of multiple electromyography noise detection devices.

[0066] This application provides a method, apparatus, device, and computer-readable storage medium for detecting electromyographic noise, which will be described in detail below.

[0067] It will be understood by those skilled in the art that Figure 1 The application environment shown is only one application scenario related to the solution of this application and does not constitute a limitation on the application scenario of this application. Other application environments may include more than one application scenario. Figure 1 The number of more or fewer EMG noise detection devices shown, or the network connections for EMG noise detection, for example Figure 1Only one electromyography (EMG) noise detection device is shown in the figure. It is understood that the scenario of this EMG noise detection method may also include one or more EMG noise detection devices, which are not specifically limited here. The EMG noise detection device may also include a memory for storing video sequences and other data.

[0068] It should be noted that, Figure 1 The schematic diagram of the electromyography noise detection method shown is merely an example. The scenario of the electromyography noise detection method described in the embodiments of the present invention is for the purpose of more clearly illustrating the technical solution of the embodiments of the present invention, and does not constitute a limitation on the technical solution provided by the embodiments of the present invention.

[0069] Based on the scenarios described above for electromyographic noise detection methods, various embodiments of the electromyographic noise detection method disclosed in this invention are proposed.

[0070] like Figure 2 As shown, Figure 2 This is a flowchart illustrating one embodiment of the electromyography noise detection method in this application. The image processing method includes the following steps 201-204:

[0071] 201. Acquire the surface electromyography signals to be identified;

[0072] The electromyography (EMG) noise detection method in this embodiment is applied to an EMG noise detection device. The type and number of EMG noise detection devices are not specifically limited. That is, the EMG noise detection device can be one or more smart terminals or servers. In a specific embodiment, the EMG noise detection device is a wearable smart device.

[0073] The electromyography (EMG) noise detection device is configured to acquire surface electromyography (EMG) signals to be identified, perform noise detection on the EMG signals, determine the noise type of the EMG signals, and output a corresponding noise optimization reminder. Optionally, in one specific embodiment, the EMG noise detection device is connected to the person being tested, and acquires the person's EMG signals through an EMG signal acquisition port, thereby determining the person's movement information through the EMG signals. Optionally, the EMG noise detection device can also be a server communicatively connected to the user's EMG signal acquisition terminal, and perform movement analysis on the user by receiving the EMG signals acquired by the terminal.

[0074] Specifically, during operation, the electromyography (EMG) noise detection device receives EMG noise detection requests. These requests can be user-initiated; for example, a user collecting surface EMG signals could actively trigger an EMG noise detection request by clicking the device's noise detection button. Alternatively, the device may have a pre-set automatic noise detection process that automatically triggers the request upon acquiring the surface EMG signal. The EMG noise detection request is an instruction to the device to perform EMG noise detection on the surface EMG signal, obtaining the noise characteristics and type carried within the signal.

[0075] Upon receiving an electromyography (EMG) noise detection request, the EMG noise detection device acquires the surface EMG signal to be identified, as indicated by the request. It then filters the surface EMG signal to remove any potential power frequency interference.

[0076] Specifically, the electromyography (EMG) noise detection device is pre-set with a power frequency filter. After acquiring the surface EMG signal, the signal is input to the power frequency filter for noise reduction and filtering, thereby filtering out power frequency interference in the EMG signal. Optionally, in one specific embodiment, the power frequency filter is a set of third-order notch filters with a center frequency of 50 Hz.

[0077] After filtering out power frequency interference from the surface electromyography (EMG) signal, the EMG noise detection device further buffers the signal to ensure it has a certain data length. Specifically, the device pre-sets a buffer length and, after noise reduction, buffers the filtered signal based on this length. Optionally, in one embodiment, the preset buffer length is greater than or equal to 5 seconds, resulting in a buffered EMG signal with a length greater than or equal to 5 seconds.

[0078] The electromyography (EMG) noise detection device performs noise detection on the buffered EMG signal after filtering and buffering the signal.

[0079] 202. Perform feature extraction on the surface electromyography signal to obtain the standard electromyography features and surface electromyography features of the surface electromyography signal;

[0080] After preprocessing the surface electromyography (EMG) signal through filtering and signal buffering, the EMG noise detection device further extracts features from the EMG signal to obtain the standard EMG features and surface EMG features. The standard EMG features are the spatial mapping features of the feature space constructed by the surface EMG signal and a clean, high signal-to-noise ratio signal; the surface EMG features are the EMG features of the surface EMG signal carrying background noise and / or motion artifact noise.

[0081] Specifically, the electromyography (EMG) noise detection device pre-sets an EMG noise detection model and uses this model to extract features from the cached surface EMG signal. The EMG noise detection model includes at least a first feature extraction unit and a second feature extraction unit. After acquiring the cached surface EMG signal, the EMG noise detection device inputs the signal into the first and second feature extraction units of the EMG noise detection model. The first feature extraction unit extracts standard EMG features from the signal, and the second feature extraction unit extracts surface EMG features. The first feature extraction unit is a neural network model trained using standard EMG samples; the second feature extraction unit is a neural network model obtained by freezing the parameters of the first feature extraction unit and jointly training it with other modules of the EMG noise detection model.

[0082] The electromyography (EMG) noise detection device inputs the cached surface EMG signal into a first feature extraction unit for feature extraction. Specifically, the first feature extraction unit includes a first coarse-grained extraction module and a second fine-grained feature extraction module. The first coarse-grained extraction module includes at least a one-dimensional convolutional layer and an activation layer, wherein the activation layer uses the ReLU activation function. The EMG noise detection device inputs the surface EMG signal into the first coarse-grained extraction module to perform coarse-grained feature extraction on the surface EMG signal, extracting coarse-grained features of the surface EMG signal. These coarse-grained features are feature vectors with a feature dimension of 4. After obtaining the coarse-grained features of the surface EMG signal, the coarse-grained features are input into the first fine-grained extraction module of the first feature extraction unit. The first fine-grained extraction module further extracts features from the coarse-grained features to obtain the standard EMG features corresponding to the surface EMG signal. These standard EMG features are feature vectors with a feature dimension of 32 associated with the surface EMG signal. The first fine-grained feature extraction module includes a one-dimensional depth convolutional layer, a first activation layer, a one-dimensional channel convolutional layer, a normalization layer, and a second activation layer.

[0083] Specifically, the electromyography (EMG) noise detection device also inputs the surface EMG signal into a second feature extraction unit for feature extraction to obtain the surface EMG features corresponding to the surface EMG signal. Specifically, the second feature extraction unit includes a second coarse-grained extraction module and a second fine-grained feature extraction module. The second coarse-grained extraction module includes at least a one-dimensional convolutional layer and an activation layer. The EMG noise detection device inputs the surface EMG signal into the second coarse-grained extraction module to perform coarse-grained feature extraction on the surface EMG signal, extracting the coarse-grained features of the surface EMG signal. These coarse-grained features are feature vectors with a feature dimension of 4. After obtaining the coarse-grained features of the surface EMG signal, these features are input into the second fine-grained extraction module of the second feature extraction unit. The second fine-grained extraction module further extracts features from the coarse-grained features to obtain the surface EMG features corresponding to the surface EMG signal. These surface EMG features are feature vectors with a feature dimension of 32 associated with the surface EMG signal. The second fine-grained feature extraction module includes a one-dimensional depth convolutional layer, a first activation layer, a one-dimensional channel convolutional layer, a normalization layer, and a second activation layer.

[0084] After acquiring standard electromyographic features and surface electromyographic features, the electromyographic noise detection device obtains the electromyographic noise features of the surface electromyographic signal based on the standard electromyographic features and surface electromyographic features.

[0085] 203. Extract the differential features from the standard electromyography features and the surface electromyography features to obtain electromyography noise features;

[0086] After acquiring the surface electromyography (EMG) features and standard EMG features of the surface EMG signal, the EMG noise detection device extracts the difference features between the standard EMG features and the surface EMG features to obtain the feature difference vector between the surface EMG features and the standard EMG features. This feature difference vector is then determined as the EMG noise feature of the surface EMG signal.

[0087] Specifically, the electromyography (EMG) noise detection model in the EMG noise detection device also includes a noise extraction unit. After acquiring the surface EMG features and standard EMG features of the surface EMG signal, the device inputs these features into the noise extraction unit. The noise extraction unit then extracts the difference features between the surface EMG features and the standard EMG features. Specifically, after acquiring the surface EMG features and the standard EMG features, the noise extraction unit subtracts them element-wise to obtain a difference feature vector. This difference vector represents the EMG noise feature of the surface EMG signal, and the EMG noise feature is a 32-dimensional feature vector.

[0088] 204. Classify the electromyographic noise features, determine the noise type of the electromyographic noise features, and output a noise optimization reminder.

[0089] After acquiring the characteristics of electromyographic noise, the electromyographic noise detection device classifies the noise to determine the noise type of the noise characteristic, and outputs a corresponding noise optimization reminder based on the noise type.

[0090] Specifically, the electromyography noise detection model in the electromyography noise detection device also includes a noise classification unit, wherein the noise classification unit is a neural network model configured to classify electromyography noise features. In a specific embodiment, the noise classification unit includes a first convolutional layer, a first activation layer, a second convolutional layer, a pooling layer, and a second activation layer.

[0091] Specifically, the noise classification unit in the EMG noise detection model uses two convolutional layers as classification layers to classify EMG noise features. After acquiring the EMG noise features, the noise classification unit inputs these features into the first convolutional layer, where convolution operations are performed to extract the high-dimensional noise features. These high-dimensional noise features are then input into the first activation layer for nonlinear transformation, and then undergo a second convolution in the second convolutional layer to obtain the corresponding feature fitting vector. Finally, the feature fitting vector is input into a pooling layer for pooling compression, resulting in a noise vector with a feature length of 1 and a feature width of 3.

[0092] After acquiring a noise vector, the noise classification unit matches the noise vector with preset labels to determine the noise type corresponding to the noise vector. The preset labels are noise labels obtained through training on different noise samples, with different noise labels representing different noise types. Optionally, in a specific embodiment, the noise labels include motion artifact labels, background noise labels, and other noise labels.

[0093] After determining the noise type of the electromyographic noise characteristics corresponding to the surface electromyographic signal, the electromyographic noise detection device obtains the corresponding noise optimization reminder based on the noise type and outputs the noise optimization reminder to instruct the user to perform specified actions to improve the acquisition environment, thereby reducing noise interference during the surface electromyographic signal acquisition process.

[0094] In this embodiment, the electromyography (EMG) noise detection device acquires the surface EMG signal to be identified and inputs it into different feature extraction branches for feature extraction. This yields a spatial mapping of the surface EMG signal in a feature space constructed by a high signal-to-noise ratio signal, namely, the standard EMG feature and the surface EMG feature of the surface EMG signal. After acquiring the standard EMG feature and the surface EMG feature, differential feature extraction is performed on the standard EMG feature and the surface EMG feature to obtain the feature difference between the standard EMG feature and the surface EMG feature, which is then identified as an EMG noise feature. The EMG noise feature is then classified to determine the noise type, and a noise optimization reminder corresponding to the noise type is output. This improves the noise feature extraction capability of EMG noise features and effectively classifies the EMG noise, reducing the acquisition of invalid signals such as noise signals in subsequent acquisition processes and improving the accuracy of surface EMG signal acquisition.

[0095] like Figure 3 As shown, Figure 3 This is a flowchart illustrating an embodiment of the electromyography noise detection method provided in this application, which iteratively trains the first feature extraction unit.

[0096] Based on the above embodiments, in this application embodiment, a first feature extraction unit is obtained by iteratively training a preset feature extraction unit using standard electromyography samples. The first feature extraction unit is then used to extract features from the surface electromyography signal, specifically including steps 301-304:

[0097] 301. Input the standard sample features into the encoder of the feature extraction unit, and encode the standard sample features through the encoder to obtain the standard electromyography training vector;

[0098] 302. Input the standard electromyography training vector into the decoder of the feature extraction unit, and extract the training mean feature and training variance feature of the standard electromyography training vector through the decoder to generate a training feature vector;

[0099] 303. Decode the training feature vector and calculate the loss value of the decoded training feature vector using a preset loss function;

[0100] 304. The encoder and the decoder are optimized and iterated based on the loss value to obtain the first feature extraction unit.

[0101] In this embodiment, before receiving surface electromyography (EMG) signals, the EMG noise detection device performs iterative training on a preset model to obtain an EMG noise detection model. Specifically, the EMG noise detection device performs iterative training on the first feature extraction unit and other units in the EMG noise detection model.

[0102] Specifically, the electromyography (EMG) noise detection device acquires standard EMG samples and extracts features from them to obtain standard sample features. The standard EMG samples are surface EMG signal samples labeled "good signal quality." After acquiring the standard sample features, the preset feature extraction unit is iteratively trained based on these features to obtain the first feature extraction unit.

[0103] Specifically, the preset feature extraction unit corresponding to the first feature extraction unit includes an encoder and a decoder. After acquiring standard sample features, the electromyography (EMG) noise detection device inputs the standard sample features into the encoder of the feature extraction unit. The encoder encodes the standard sample features to obtain the standard EMG training vector corresponding to the standard feature sample. The EMG noise detection device then inputs this standard EMG training vector into the decoder for decoding.

[0104] Specifically, after receiving the standard EMG training vector, the decoder inputs this standard EMG training feature vector into the mean extraction module and the variance extraction module. These modules then extract the training mean and training variance features of the standard EMG training vector, and generate a training feature vector based on these features. Both the mean and variance extraction modules consist of one-dimensional convolutional layers and activation layers.

[0105] After generating the training feature vector, the decoder further parameterizes the training feature vector. Specifically, the decoder uses parameterization formulas. The training feature vector is further parameterized, where feat2 is the training feature vector, mean is the training mean feature, and var is the training variance feature. (Operator) This involves adding the features element by element.

[0106] After parameterizing the training feature vector, the decoder decodes the training features and calculates the loss value of the decoded training feature vector using a preset loss function. This determines the error between the training feature vector and the output standard electromyography (EMG) features. Specifically, the preset loss function is:

[0107] Where, mean i var i This represents the value of the i-th dimension of the two sets of feature vectors.

[0108] After calculating the loss value corresponding to the training feature vector based on the loss function, the decoder optimizes and iterates the encoder and decoder based on the loss value to obtain the first feature extraction unit.

[0109] After training the first feature extraction unit, the electromyography (EMG) noise detection device further trains the other units of the EMG noise detection model. Specifically, the EMG noise detection device freezes the parameters of the first feature extraction unit and trains the second feature extraction unit, the noise extraction unit, and the noise classification unit to obtain the trained second feature extraction unit, noise extraction unit, and noise classification unit.

[0110] After all units of the electromyography (EMG) noise detection model have been trained, the EMG noise detection device uses the EMG noise detection model to detect noise in the surface EMG signal, determines the type of noise carried by the surface EMG signal, and outputs a noise optimization reminder.

[0111] In this embodiment, the electromyography (EMG) noise detection device inputs the standard sample features into the encoder of the feature extraction unit. The encoder encodes the standard sample features to obtain a standard EMG training vector. The standard EMG training vector is then input into the decoder of the feature extraction unit. The decoder extracts the training mean and training variance features of the standard EMG training vector to generate a training feature vector. The training feature vector is then decoded, and a loss value is calculated on the decoded training feature vector using a preset loss function. Based on the loss value, the encoder and decoder are iteratively optimized to obtain the first feature extraction unit. This iterative training of the first feature extraction unit enables it to extract standard EMG signals from surface EMG signals, improving the noise recognition accuracy.

[0112] like Figure 4 As shown, Figure 4 This is a flowchart illustrating an embodiment of the electromyographic noise detection method provided in this application, which classifies electromyographic noise features and outputs noise optimization reminders based on the noise type of the electromyographic noise features.

[0113] Based on the above embodiments, in this application embodiment, after determining the noise type of electromyographic noise characteristics, different noise optimization reminders are output according to the noise type, thereby improving the accuracy of surface electromyographic signal acquisition. Specifically, this includes steps 401-403:

[0114] 401. Classify and identify the electromyographic noise features to obtain the noise type of the electromyographic noise features;

[0115] 402. If the noise type is a background noise type, determine that the noise optimization reminder corresponding to the background noise type is a background noise reduction reminder, and output the background noise reduction reminder;

[0116] 403. If the noise type is a motion artifact type, determine that the noise optimization reminder corresponding to the motion artifact type is an artifact removal reminder, and output the artifact removal reminder.

[0117] In this embodiment, the electromyography noise detection device classifies and identifies the electromyography noise features of the surface electromyography signal through an electromyography noise detection model to obtain the noise type corresponding to the electromyography noise feature. Optionally, the electromyography noise feature includes noise types such as background noise and motion artifact noise. The electromyography noise detection device obtains the acquisition strategy of the surface electromyography signal corresponding to the noise type.

[0118] Specifically, the electromyography (EMG) noise detection device is pre-set with acquisition strategies for surface EMG signals corresponding to different noise types. These acquisition strategies instruct the user to perform specified acquisition environment improvement operations. Different acquisition strategies correspond to different noise optimization prompts. Specifically, in one embodiment, the acquisition strategies include a background noise optimization strategy and an artifact elimination strategy, with corresponding noise optimization prompts of reducing background noise and artifact elimination, respectively.

[0119] Specifically, if the electromyography noise detection device detects that the noise type of the surface electromyography signal is background noise, then the acquisition strategy is determined to be a background noise optimization strategy. The electromyography noise detection device outputs a corresponding background noise reduction reminder according to the background noise optimization strategy. The noise optimization reminder corresponding to the background noise optimization strategy is to instruct the user to apply an appropriate amount of water to the electrode where the surface electromyography signal is acquired.

[0120] Specifically, if the electromyography noise detection device detects that the noise type of the surface electromyography signal is motion artifact noise, then the acquisition strategy is determined to be an artifact elimination strategy. The electromyography noise detection device outputs a corresponding artifact elimination reminder based on the background noise optimization strategy, prompting the user to reduce limb movements when acquiring surface electromyography signals.

[0121] In this embodiment, the electromyography (EMG) noise detection device classifies and identifies the EMG noise features to obtain the noise type of the EMG noise features, and acquires the acquisition strategy for the surface EMG signal corresponding to the noise type. If the noise type is background noise, the acquisition strategy is determined to be a background noise optimization strategy, and a background noise reduction reminder is output according to the background noise optimization strategy. If the noise type is motion artifact, the acquisition strategy is determined to be an artifact elimination strategy, and an artifact elimination reminder is output according to the artifact reduction strategy. This effectively improves the accuracy of surface EMG signal acquisition and reduces the acquisition of invalid signals such as noise signals in subsequent acquisition processes.

[0122] To better implement the electromyography (EMG) noise detection method in the embodiments of this application, an EMG noise detection device is also provided in the embodiments of this application, such as... Figure 5 As shown, Figure 5 This is a schematic diagram of an embodiment of the electromyography (EMG) noise detection device provided in this application. The EMG noise detection device 500 includes the following modules 501-504:

[0123] The electromyography acquisition module 501 is configured to acquire surface electromyography signals to be identified.

[0124] The feature extraction module 502 is configured to extract features from the surface electromyography signal to obtain the standard electromyography features and surface electromyography features of the surface electromyography signal.

[0125] The noise recognition module 503 is configured to extract differential features from the standard electromyography features and the surface electromyography features to obtain electromyography noise features;

[0126] The noise classification module 504 is configured to classify the electromyographic noise features, determine the noise type of the electromyographic noise features, and output a noise optimization reminder.

[0127] In some embodiments of this application, the electromyography noise detection device performs feature extraction on the surface electromyography signal to obtain standard electromyography features and surface electromyography features of the surface electromyography signal, including:

[0128] The surface electromyography signal is extracted by the first feature extraction unit to obtain the standard electromyography features of the surface electromyography signal.

[0129] The surface electromyography (EMG) signal is subjected to feature extraction by the second feature extraction unit to obtain the surface EMG features of the surface EMG signal.

[0130] In some embodiments of this application, the electromyography noise detection device extracts differential features from the standard electromyography features and surface electromyography features to obtain electromyography noise features, including:

[0131] The electromyographic noise feature is obtained by subtracting the standard electromyographic feature and the surface electromyographic feature element by element using the electromyographic noise detection model.

[0132] The process of classifying the electromyographic noise features and determining the noise type of the electromyographic noise features includes:

[0133] The electromyographic noise detection model is used to perform convolutional classification on the electromyographic noise features to determine the noise type of the electromyographic noise features.

[0134] In some embodiments of this application, before the electromyography noise detection device extracts features from the surface electromyography signal to obtain the standard electromyography features and surface electromyography features of the surface electromyography signal, it further includes:

[0135] Obtain standard electromyography (EMG) samples, and extract features from the standard EMG samples to obtain standard sample features;

[0136] The preset feature extraction unit is iteratively trained based on the features of the standard sample to obtain the first feature extraction unit.

[0137] In some embodiments of this application, the electromyography noise detection device iteratively trains a preset feature extraction unit based on the standard electromyography features to obtain a first feature extraction unit, including:

[0138] The standard sample features are input into the encoder of the feature extraction unit, and the encoder encodes the standard sample features to obtain the standard electromyography training vector.

[0139] The standard electromyography training vector is input into the decoder of the feature extraction unit, and the training mean feature and training variance feature of the standard electromyography training vector are extracted by the decoder to generate a training feature vector.

[0140] The training feature vector is decoded, and the loss value is calculated on the decoded training feature vector using a preset loss function;

[0141] The encoder and decoder are optimized iteratively based on the loss value to obtain the first feature extraction unit.

[0142] In some embodiments of this application, the electromyographic noise detection device performs convolutional classification on the electromyographic noise features using the electromyographic noise detection model to determine the noise type of the electromyographic noise features, including:

[0143] The electromyographic noise features are input into the noise classification unit for convolution operation to extract the high-dimensional noise features of the electromyographic noise features;

[0144] The high-dimensional noise features are nonlinearly transformed according to a preset activation function to obtain a feature fitting vector;

[0145] The feature fitting vector is pooled and compressed to obtain a noise vector of a preset dimension;

[0146] The noise vector is matched with a preset label to determine the noise type of the noise vector.

[0147] In some embodiments of this application, the electromyographic noise detection device classifies the electromyographic noise features, determines the noise type of the electromyographic noise features, and outputs a noise optimization reminder, including:

[0148] The electromyographic noise features are classified and identified to obtain the noise type of the electromyographic noise features;

[0149] If the noise type is a background noise type, determine that the noise optimization reminder corresponding to the background noise type is a background noise reduction reminder, and output the background noise reduction reminder;

[0150] If the noise type is a motion artifact type, determine that the noise optimization reminder corresponding to the motion artifact type is an artifact removal reminder, and output the artifact removal reminder.

[0151] In this embodiment, the electromyography (EMG) noise detection device acquires the surface EMG signal to be identified and inputs it into different feature extraction branches for feature extraction. This yields a spatial mapping of the surface EMG signal in a feature space constructed by a high signal-to-noise ratio signal, namely, the standard EMG feature and the surface EMG feature of the surface EMG signal. After acquiring the standard EMG feature and the surface EMG feature, differential feature extraction is performed on the standard EMG feature and the surface EMG feature to obtain the feature difference between the standard EMG feature and the surface EMG feature, which is then identified as an EMG noise feature. The EMG noise feature is then classified to determine the noise type, and a noise optimization reminder corresponding to the noise type is output. This improves the noise feature extraction capability of EMG noise features and effectively classifies the EMG noise, reducing the acquisition of invalid signals such as noise signals in subsequent acquisition processes and improving the accuracy of surface EMG signal acquisition.

[0152] This invention also provides an electromyography noise detection device, such as... Figure 6 As shown, Figure 6 This is a schematic diagram of an embodiment of the electromyography noise detection device provided in this application.

[0153] The electromyographic noise detection device integrates any of the electromyographic noise detection devices provided in the embodiments of the present invention, and the electromyographic noise detection device includes:

[0154] One or more processors;

[0155] Memory; and

[0156] One or more applications, wherein the one or more applications are stored in the memory and configured by the processor to perform the steps of the electromyographic noise detection method described in any of the embodiments of the above-described electromyographic noise detection method.

[0157] Specifically, an electromyography (EMG) noise detection device may include components such as a processor 601 with one or more processing cores, a memory 602 with one or more computer-readable storage media, a power supply 603, and an input unit 604. Those skilled in the art will understand that... Figure 6 The structure of the electromyography (EMG) noise detection device shown does not constitute a limitation on the EMG noise detection device. It may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0158] The processor 601 is the control center of the electromyography (EMG) noise detection device. It connects various parts of the device via interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 602, and by calling data stored in the memory 602, thereby providing overall monitoring of the EMG noise detection device. Optionally, the processor 601 may include one or more processing cores; preferably, the processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 601.

[0159] The memory 602 can be used to store software programs and modules. The processor 601 executes various functional applications and data processing by running the software programs and modules stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electromyography noise detection device, etc. In addition, the memory 602 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 602 may also include a memory controller to provide the processor 601 with access to the memory 602.

[0160] The electromyography noise detection device also includes a power supply 603 that supplies power to the various components. Preferably, the power supply 603 can be logically connected to the processor 601 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 603 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0161] The electromyography noise detection device may also include an input unit 604, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0162] Although not shown, the electromyography (EMG) noise detection device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 601 in the EMG noise detection device loads the executable files corresponding to the processes of one or more application programs into the memory 602 according to the following instructions, and the processor 601 runs the application programs stored in the memory 602 to realize various functions, as follows:

[0163] Acquire surface electromyography signals to be identified;

[0164] Feature extraction is performed on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal;

[0165] Difference features are extracted from the standard electromyography (EMG) features and the surface EMG features to obtain EMG noise features;

[0166] The electromyographic noise features are classified to determine the noise type of the electromyographic noise features, and a noise optimization reminder is output.

[0167] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0168] Therefore, embodiments of the present invention provide a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc. A computer program is stored thereon, which is loaded by a processor to execute the steps in any of the electromyography noise detection methods provided in the embodiments of the present invention. For example, the computer program loaded by the processor can execute the following steps:

[0169] Acquire surface electromyography signals to be identified;

[0170] Feature extraction is performed on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal;

[0171] Difference features are extracted from the standard electromyography (EMG) features and the surface EMG features to obtain EMG noise features;

[0172] The electromyographic noise features are classified to determine the noise type of the electromyographic noise features, and a noise optimization reminder is output.

[0173] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the detailed descriptions of other embodiments above, which will not be repeated here.

[0174] In practice, each of the above units or structures can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units or structures, please refer to the previous method embodiments, which will not be repeated here.

[0175] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0176] The above provides a detailed description of an electromyography noise detection method provided by the embodiments of this application. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting electromyographic noise, characterized in that, The electromyographic noise detection method includes: Acquire surface electromyography signals to be identified; Feature extraction is performed on the surface electromyography (EMG) signal to obtain the standard EMG feature and surface EMG feature of the surface EMG signal. The standard EMG feature is the spatial mapping feature of the feature space constructed by the surface EMG signal and the clean, high signal-to-noise ratio signal. The surface EMG feature is the EMG feature of the surface EMG signal carrying background noise and / or motion artifact noise. Difference features are extracted from the standard electromyography (EMG) features and the surface EMG features to obtain EMG noise features; The electromyographic noise features are classified to determine the noise type of the electromyographic noise features, and a noise optimization reminder is output.

2. The electromyographic noise detection method as described in claim 1, characterized in that, The electromyographic noise detection method is applied to an electromyographic noise detection model, which includes a first feature extraction unit and a second feature extraction unit. The step of extracting features from the surface electromyography (EMG) signal to obtain standard EMG features and surface EMG features includes: The surface electromyography signal is extracted by the first feature extraction unit to obtain the standard electromyography features of the surface electromyography signal. The surface electromyography (EMG) signal is subjected to feature extraction by the second feature extraction unit to obtain the surface EMG features of the surface EMG signal.

3. The electromyographic noise detection method as described in claim 2, characterized in that, The differential feature extraction of the standard electromyography (EMG) features and surface EMG features to obtain EMG noise features includes: The electromyographic noise feature is obtained by subtracting the standard electromyographic feature and the surface electromyographic feature element by element using the electromyographic noise detection model. The process of classifying the electromyographic noise features and determining the noise type of the electromyographic noise features includes: The electromyographic noise detection model is used to perform convolutional classification on the electromyographic noise features to determine the noise type of the electromyographic noise features.

4. The electromyographic noise detection method as described in claim 3, characterized in that, Before performing feature extraction on the surface electromyography (EMG) signal to obtain the standard EMG features and surface EMG features of the surface EMG signal, the method further includes: Obtain standard electromyography (EMG) samples, and extract features from the standard EMG samples to obtain standard sample features; The preset feature extraction unit is iteratively trained based on the features of the standard sample to obtain the first feature extraction unit.

5. The electromyographic noise detection method as described in claim 4, characterized in that, The step of iteratively training a preset feature extraction unit based on the standard electromyographic features to obtain a first feature extraction unit includes: The standard sample features are input into the encoder of the feature extraction unit, and the encoder encodes the standard sample features to obtain the standard electromyography training vector. The standard electromyography training vector is input into the decoder of the feature extraction unit, and the training mean feature and training variance feature of the standard electromyography training vector are extracted by the decoder to generate a training feature vector. The training feature vector is decoded, and the loss value is calculated on the decoded training feature vector using a preset loss function; The encoder and decoder are optimized iteratively based on the loss value to obtain the first feature extraction unit.

6. The electromyographic noise detection method as described in claim 2, characterized in that, The electromyographic noise detection model also includes a noise classification unit; The step of performing convolutional classification on the electromyographic noise features using the electromyographic noise detection model to determine the noise type of the electromyographic noise features includes: The electromyographic noise features are input into the noise classification unit for convolution operation to extract the high-dimensional noise features of the electromyographic noise features; The high-dimensional noise features are nonlinearly transformed according to a preset activation function to obtain a feature fitting vector; The feature fitting vector is pooled and compressed to obtain a noise vector of a preset dimension; The noise vector is matched with a preset label to determine the noise type of the noise vector.

7. The electromyographic noise detection method as described in claim 1, characterized in that, The process of classifying the electromyographic noise features, determining the noise type of the electromyographic noise features, and outputting noise optimization reminders includes: The electromyographic noise features are classified and identified to obtain the noise type of the electromyographic noise features; If the noise type is a background noise type, determine that the noise optimization reminder corresponding to the background noise type is a background noise reduction reminder, and output the background noise reduction reminder; If the noise type is a motion artifact type, determine that the noise optimization reminder corresponding to the motion artifact type is an artifact removal reminder, and output the artifact removal reminder.

8. A device for detecting electromyographic noise, characterized in that, The electromyographic noise detection device includes: The electromyography (EMG) acquisition module is configured to acquire surface EMG signals to be identified. The feature extraction module is configured to extract features from the surface electromyography (EMG) signal to obtain standard EMG features and surface EMG features of the surface EMG signal. The standard EMG features are spatial mapping features of the feature space constructed by the surface EMG signal and a clean, high signal-to-noise ratio signal. The surface EMG features are EMG features of the surface EMG signal carrying background noise and / or motion artifact noise. The noise identification module is configured to extract differential features from the standard electromyography features and the surface electromyography features to obtain electromyography noise features; The noise classification module is configured to classify the electromyographic noise features, determine the noise type of the electromyographic noise features, and output a noise optimization reminder.

9. A device for detecting electromyographic noise, characterized in that, The electromyographic noise detection device includes: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the electromyographic noise detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It contains a computer program that is loaded by a processor to perform the steps of the electromyography noise detection method according to any one of claims 1 to 7.

Images