A method of long-term monitoring of myoelectric activity and related devices

Abstract

This application discloses a method and related equipment for long-term monitoring of electromyographic activity. The method includes: segmenting a first surface electromyographic signal, a first muscle electromyographic signal, and a first muscle sound signal acquired within a first preset time period by a preset time length to obtain multiple time segments of the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal; determining the mapping relationship between the first surface electromyographic signal and the first muscle electromyographic signal based on the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal within each time segment; acquiring the second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period; and predicting the second muscle electromyographic signal of the target muscle within the second preset time period based on the mapping relationship. This application can achieve long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness, and can be widely applied in the field of electromyographic signal processing technology.

Description

Technical Field

[0001] This application relates to the field of electromyography signal processing technology, and in particular to a method and related equipment for long-term monitoring of electromyography activity. Background Technology

[0002] Currently, electromyography (EMG) monitoring technologies mainly include two methods: surface electromyography (sEMG) and implantable electromyography (EMG). sEMG records muscle activity by placing electrodes on the skin surface, offering advantages such as being non-invasive and allowing for long-term monitoring. However, the signal is attenuated and aliased by skin, fat, and other tissues during transmission, resulting in low spatial resolution and difficulty in accurately reflecting the electrical activity characteristics of specific muscles. While implantable EMG, by directly placing electrodes inside the target muscle, can obtain high-precision, high-signal-to-noise ratio signals and accurately capture the electrical activity of specific muscles, this method is invasive, carrying risks such as infection, tissue damage, and electrode displacement. Furthermore, the implantation time is limited, making it unsuitable for long-term continuous monitoring.

[0003] However, existing technologies that simultaneously acquire surface electromyography (EMG) signals and muscle EMG signals, and analyze the correlation between them and the relationship between surface EMG and implanted EMG, primarily aim to describe whether the signals are "correlated" or "what kind of coupling exists." They cannot characterize the true and complex mapping relationship from surface mixed signals to deep specific source signals. Therefore, existing technologies can essentially only prove that surface EMG signals and muscle EMG signals are correlated under certain conditions. They cannot, after the EMG electrodes are removed, rely solely on continuously acquired surface EMG signals and myocardial signals to reconstruct or infer the transient electrical activity of a specific muscle with high fidelity and stability. Thus, existing technologies cannot achieve long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness. Summary of the Invention

[0004] The main objective of this application is to propose a method and related equipment for long-term monitoring of electromyographic activity, which can achieve long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness.

[0005] To achieve the above objectives, one aspect of this application provides a method for long-term monitoring of electromyographic activity, comprising: Acquire the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of the target muscle within a first preset time period; The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period are divided into multiple time segments to obtain the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. Based on the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within each time segment, determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal; Acquire the second surface electromyographic signal and the second myocardial sound signal of the target muscle within a second preset time period; Based on the mapping relationship, the second muscle electromyographic signal of the target muscle within the second preset time period is predicted.

[0006] In some embodiments, before the step of segmenting the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period by a preset time length, the method further includes: The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal are preprocessed. The preprocessing of the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal includes: The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal are respectively filtered; Artifact detection is performed on the filtered first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. If an artifact is detected, the signal segment containing the artifact is removed.

[0007] In some embodiments, determining the mapping relationship between the first surface electromyography (SEMG) signal and the first muscle EEMG signal based on the first surface electromyography (SEMG) signal, the first muscle EEMG signal, and the first muscle sound signal within each time segment includes: The first surface electromyography signal and the first muscle sound signal of each time segment are processed to obtain the input sequence of the preset mapping model, and the first muscle electromyography signal in the corresponding time segment is used as the target output sequence of the preset mapping model. Determine the dimension of the target output sequence, wherein the dimension is used to characterize the number of measurement channels contained in the first muscle electromyography signal; The preset mapping model performs global temporal modeling on the input sequence, and based on the result of the global temporal modeling, generates a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence. The preset mapping model determines the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the muscle electromyography signal prediction sequence.

[0008] In some embodiments, processing the first surface electromyography signal and the first muscle sound signal for each time segment to obtain the input sequence of a preset mapping model includes: Determine the time step for each of the time segments; Linear projection is performed on the first surface electromyography signal and the first muscle sound signal corresponding to each time step within each time segment to embed the first surface electromyography signal and the first muscle sound signal into a dimensional space adapted to the target output sequence, thereby obtaining an embedded signal sequence; The positions of each time step of the embedded signal sequence are superimposed and encoded to obtain an input sequence containing time information. The input sequence containing time information is then used as the input sequence of the preset mapping model.

[0009] In some embodiments, the preset mapping model performs global temporal modeling on the input sequence, and based on the result of the global temporal modeling, generates a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, including: The preset mapping model performs global temporal modeling on the input sequence to obtain a feature sequence containing temporal information and contextual information; Using the dimension of the target output sequence as the target dimension, linear mapping is performed on each feature corresponding to each time step in the feature sequence to obtain the predicted value of the muscle electromyography signal for each time step. The predicted muscle electromyography (EMG) signals for each time step are combined in chronological order to obtain a predicted muscle EMG sequence that matches the dimension of the target output sequence.

[0010] In some embodiments, after obtaining a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, the method further includes: Calculate the loss function value between the predicted muscle electromyography signal sequence and the target output sequence; Based on the loss function value, the parameters of the preset mapping model are updated to obtain an optimized preset mapping model.

[0011] In some embodiments, predicting the second muscle electromyographic signal of the target muscle within the second preset time period based on the mapping relationship includes: Based on the optimized preset mapping model, the mapping relationship is optimized; The second surface electromyographic signal and the second myophone signal of the target muscle within the second preset time period are input into the optimized preset mapping model; The optimized preset mapping model predicts the second muscle electromyographic signal of the target muscle within the second preset time period based on the optimized mapping relationship.

[0012] To achieve the above objectives, another aspect of this application provides a long-term electromyography (EMG) activity monitoring device, the device comprising: The first signal acquisition module is used to acquire the first surface electromyographic signal, the first muscle electromyographic signal and the first muscle sound signal of the target muscle within a first preset time period. The signal segmentation module is used to segment the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period by a preset time length to obtain multiple time segments of the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. The relationship determination module is used to determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the first surface electromyography signal, the first muscle electromyography signal and the first muscle sound signal in each time segment; The second signal acquisition module is used to acquire the second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period. The second muscle electromyography (EMG) signal prediction module is used to predict the second muscle EMG signal of the target muscle within the second preset time period based on the mapping relationship.

[0013] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned long-term electromyography activity monitoring method.

[0014] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned long-term electromyographic activity monitoring method.

[0015] This application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned method.

[0016] The embodiments of this application include at least the following beneficial effects: This application provides a method for long-term monitoring of electromyographic activity. The method first acquires the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of a target muscle within a first preset time period; it then segments the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal within the first preset time period by a preset time length to obtain multiple time segments of the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal; based on the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal within each time segment, it determines the mapping relationship between the first surface electromyographic signal, the first muscle sound signal, and the first muscle electromyographic signal; it acquires the second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period; and based on the mapping relationship, it predicts the second muscle electromyographic signal of the target muscle within the second preset time period. This application constructs a nonlinear mapping relationship from non-invasive / minimally invasive EMG to specific muscle EMG by using surface EMG signals, muscle EMG signals, and myophone signals within a limited time. Based on this mapping relationship, after the EMG electrodes are removed or no longer used, the muscle EMG signal of the target muscle can be obtained by relying only on surface EMG signals and myophone signals. This enables long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application; Figure 2 This is a flowchart of a long-term electromyography (EMG) activity monitoring method provided in an embodiment of this application; Figure 3 This is a waveform diagram of multi-channel surface electromyography signals acquired by a dry electrode array according to an embodiment of this application; Figure 4 This is a waveform diagram of muscle electromyography signals acquired by an implanted flexible electrode wire according to an embodiment of this application; Figure 5 This is a waveform diagram of the muscle sound signal acquired by a tensile strain sensor according to an embodiment of this application; Figure 6 This is a flowchart illustrating the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal, provided in an embodiment of this application. Figure 7 This is a flowchart illustrating the generation of a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, as provided in an embodiment of this application. Figure 8 This is a schematic diagram of the process of deriving electromyographic signals from surface electromyographic signals and myophone signals, provided in an embodiment of this application. Figure 9This is a schematic diagram of the structure of the long-term electromyography activity monitoring device provided in the embodiments of this application; Figure 10 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] It is understood that the terms "first," "second," "third," "fourth," etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] Existing technologies involve simultaneously acquiring surface electromyography (EMG) signals and muscle EMG signals, and analyzing the correlation between them. The relationship between surface EMG and implanted EMG is primarily aimed at describing whether signals are correlated or what kind of coupling exists. However, these methods cannot characterize the true, complex mapping relationship from surface mixed signals to specific source signals in deeper layers. Therefore, existing technologies can essentially only demonstrate the correlation between surface EMG and muscle EMG signals under certain conditions. They cannot, after removing the EMG electrodes, rely solely on continuously acquired surface EMG and myocardial signals to reconstruct or infer the transient electrical activity of a specific muscle with high fidelity and stability. Thus, existing technologies cannot achieve long-term, accurate monitoring of specific muscle activity while maintaining low invasiveness.

[0023] In view of this, this application provides a method and related equipment for long-term monitoring of electromyography (EMG) activity. The method first acquires the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal of the target muscle within a first preset time period; then, it segments the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal within the first preset time period by a preset time length to obtain multiple time segments of the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal; based on the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal within each time segment, it determines the mapping relationship between the first surface EMG signal, the first muscle sound signal, and the first muscle EMG signal; it acquires the second surface EMG signal and the second muscle sound signal of the target muscle within a second preset time period; and based on the mapping relationship, it predicts the second muscle EMG signal of the target muscle within the second preset time period. This application constructs a nonlinear mapping relationship from non-invasive / minimally invasive EMG to specific muscle EMG by using surface EMG signals, muscle EMG signals, and myophone signals within a limited time. Based on this mapping relationship, after the EMG electrodes are removed or no longer used, the muscle EMG signal of the target muscle can be obtained by relying only on surface EMG signals and myophone signals. This enables long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness.

[0024] The long-term electromyography (EMG) activity monitoring method provided in this application relates to the field of EMG signal processing technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the long-term EMG activity monitoring method, but is not limited to the above forms.

[0025] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0026] like Figure 1 The diagram shown is a schematic representation of an implementation environment provided in an embodiment of this application. (Refer to...) Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected via a network, either wirelessly or via a wired connection, to complete data transmission and exchange.

[0027] Server 101 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0028] Additionally, server 101 can also be a node server in a blockchain network. Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms.

[0029] Terminal 102 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc. It can also be a vehicle-mounted terminal of the various device types described above, but is not limited to these. Terminal 102 and server 101 can be directly or indirectly connected via wired or wireless communication, and this embodiment does not impose any limitations.

[0030] Exemplary based on Figure 1 The implementation environment shown in this application embodiment provides a method for long-term monitoring of electromyographic activity. The following description uses the application of this method in server 101 as an example. It can be understood that this method can also be applied in terminal 102.

[0031] Reference Figure 2 , Figure 2 The flowchart illustrates a method for long-term monitoring of electromyographic activity applied to a server, as provided in this application embodiment. The execution entity of this method can be any of the aforementioned computer devices (including a server or terminal). (Refer to...) Figure 2 The method may include the following steps: S100: Acquire the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of the target muscle within a first preset time period.

[0032] Among them, the first muscle sound signal refers to the low-frequency mechanical vibration signal generated by the vibration, thickness change and overall shape change of the internal muscle fibers during the contraction and relaxation of the muscle.

[0033] For example, embodiments of this application can acquire the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of the target muscle through a multi-channel electromyographic synchronous acquisition system; in embodiments of this application, the multi-channel electromyographic synchronous acquisition system acquires the first surface electromyographic signal of the target muscle through a dry electrode wristband, such as... Figure 3 As shown, Figure 3 A schematic diagram of the waveforms of multi-channel surface electromyography (EMG) signals acquired via a dry electrode array, provided for implementation of this application. The EMG signal of the first muscle in the target muscle is acquired via an implanted flexible electrode wire, such as... Figure 4 As shown, Figure 4 This is a waveform diagram of muscle electromyography (EMG) signals acquired via an implanted flexible electrode wire, as provided in an embodiment of this application. The first muscle sound signal of the target muscle is acquired using a stretchable liquid metal sensor, such as... Figure 5As shown, Figure 5 This is a waveform diagram of the myophone signal acquired by a stretchable strain sensor, as provided in an embodiment of this application. The multi-channel synchronous electromyography (EMG) acquisition system ensures the alignment of EMG signals at different levels across time scales through a unified sampling clock and triggering mechanism. Specifically, during the calibration phase of the multi-channel synchronous EMG acquisition system, subjects covering both normal and abnormal EMG states are selected, and EMG signals are acquired under various experimental conditions. These conditions include at least a resting state, a specified motor task execution state, a drug intervention state, and an external stimulus intervention state. Based on the above multi-channel synchronous acquisition method, time-aligned surface EMG signal-muscle EMG signal paired data, as well as synchronously acquired myophone signal data, can be obtained, providing sufficient and effective sample data support for the subsequent model training process.

[0034] After acquiring the first surface electromyography (EMG) signal, the first muscle EMG signal, and the first muscle sound signal of the target muscle, this embodiment of the application further includes preprocessing the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal. For example, the preprocessing step of the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal includes: firstly, filtering the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal respectively; further, bandpass filtering (e.g., 10-600 Hz) is performed on the first surface EMG signal, the muscle sound signal, and the muscle EMG signal channels, and a 50 Hz and its harmonic notch filter is added to remove power frequency noise. After filtering, the mean of each channel is removed, and normalization is performed according to its standard deviation to weaken the influence of individual differences and electrode contact differences on the amplitude, so that the subsequent preset mapping model pays more attention to the temporal morphology and relative changes. Secondly, artifact detection is performed on the filtered first surface electromyography signal, first muscle electromyography signal, and first muscle sound signal. If artifacts are detected, signal segments containing artifacts are removed. Furthermore, robust artifact detection algorithms (e.g., based on amplitude, instantaneous energy, or derivative thresholds) can be used to mark and remove time segments containing large-amplitude motion artifacts, saturation, or dropouts, and to exclude or interpolate bad channels with long-term distortion.

[0035] S200: The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period are divided into multiple time segments to obtain the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal.

[0036] In this embodiment of the application, the continuously recorded first surface electromyography (EMG), first muscle EMG, and first muscle sound signal can be segmented by a preset length sliding window (preset time length) to obtain multiple time segments of the first surface EMG, first muscle EMG, and first muscle sound signal; for example, the preset length sliding window can be 5-10 seconds, and the sliding step size can be 1-3 seconds.

[0037] S300. Based on the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within each time segment, determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal.

[0038] For example, Figure 6 This is a flowchart illustrating the mapping relationship between a first surface electromyography (EMG) signal and a first muscle EMG signal, provided in an embodiment of this application. For example... Figure 6 As shown, the steps for determining the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal specifically include S310-S340: S310. Process the first surface electromyography signal and the first muscle sound signal of each time segment to obtain the input sequence of the preset mapping model, and take the first muscle electromyography signal in the corresponding time segment as the target output sequence of the preset mapping model.

[0039] In this embodiment, the preset mapping model uses the Transformer model as the core temporal modeling framework. For example, for each time segment, its corresponding first surface electromyography signal is used as the input sequence for the preset mapping, and this input sequence is expressed as: The first muscle sound signal, as the input sequence of the preset mapping, is represented as follows: ;in, This represents the number of non-invasive channels for the first surface electromyography signal. denoted as the number of non-invasive channels for the first muscle sound signal, and T as the time step.

[0040] The electromyographic signal of the first muscle is used as the target output sequence of the preset mapping model. This target output sequence can be represented as: ,in, This represents the number of non-invasive channels for the electromyographic signal of the first muscle.

[0041] The steps of processing the first surface electromyography signal and the first muscle sound signal to obtain the input sequence of the preset mapping model include S311-S313: S311. Determine the time step for each time segment; S312. Linearly project the first surface electromyography signal and the first muscle sound signal corresponding to each time step within each time segment to embed the first surface electromyography signal and the first muscle sound signal into a dimensional space adapted to the target output sequence to obtain an embedded signal sequence. For each time step, the non-invasive channel vectors of the corresponding first surface electromyography (EMG) signal and first myocardial sound signal are linearly projected from their original dimensions to a dimension dmodel that fits the target output sequence. The formula is as follows:

[0042] in, This represents the embedded signal sequence at time step t after linear projection. This represents the signal vector corresponding to the time step, namely the non-invasive channel vector of the first surface electromyography signal and the first myocardial sound signal at time step t. The weight matrix, representing the linear projection, is a learnable parameter used to project the input signal from the original dimension to the adapted dimension dmodel of the target. The bias vector representing the linear projection is a learnable parameter used to adjust the offset of the projected signal.

[0043] This linear projection operation maps the multi-channel sequences of surface electromyography (EMG) and myophone signals to the dimensional space required by the Transformer model, enabling it to match the target output sequence (muscle EMG signals) in dimension, which facilitates subsequent Transformer modeling.

[0044] S313. The positions of each time step of the embedded signal sequence are superimposed and encoded to obtain an input sequence containing time information, and the input sequence containing time information is used as the input sequence of the preset mapping model.

[0045] Since the Transformer model itself does not have a built-in sense of temporal order, this embodiment of the application explicitly introduces temporal order information by superimposing a learnable positional code on the embedded signal sequence obtained after projection at each time step, thereby enabling the Transformer model to understand the temporal relationship between the preceding and following signals.

[0046] S320. Determine the dimension of the target output sequence, wherein the dimension is used to characterize the number of measurement channels contained in the first muscle electromyography signal.

[0047] S330. The preset mapping model performs global temporal modeling on the input sequence, and based on the result of the global temporal modeling, generates a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence.

[0048] For example, Figure 7 This is a flowchart illustrating the generation of a muscle electromyography (EMG) signal prediction sequence that matches the dimension of the target output sequence, as provided in an embodiment of this application. Figure 7 As shown, the steps for generating a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence specifically include S331-S333: S331. The preset mapping model performs global temporal modeling on the input sequence to obtain a feature sequence containing temporal information and contextual information; For example, the preset mapping model (Transformer model) adopts a multi-layer stacked Transformer encoder. Each Transformer encoder performs global modeling of surface electromyography and myophone sequences in the time dimension through multi-head self-attention, so that the representation at any time point can comprehensively consider the surface electromyography and myophone activities at other times within the entire time window. At the same time, it enhances the nonlinear representation ability and training stability by using feedforward networks and residual structures. Finally, after encoding by several layers of encoders, a high-level surface electromyography feature representation sequence and a high-level myophone feature representation sequence containing temporal information and contextual information are obtained with a length of T and a dimension of dmodel.

[0049] S332. Using the dimension of the target output sequence as the target dimension, perform linear mapping processing on each feature corresponding to each time step in the feature sequence to obtain the predicted value of the muscle electromyography signal corresponding to each time step.

[0050] For example, in this embodiment of the application, a shared linear output layer decodes the high-level surface electromyography feature representation sequence and the high-level myophone feature representation sequence into specific electromyography signal prediction values ​​through a decoder. Its formula can be expressed as:

[0051] in, This represents the predicted value of the muscle electromyography (EMG) signal at each time step t (i.e., the predicted value of the muscle EMG signal at time step t). This represents the feature at time step t in the feature sequence, specifically the representation of the high-level surface electromyography (EMG) feature and the high-level myophone feature at time step t, output by the Transformer encoder. This represents the weight matrix of the linear output layer. It is a learnable parameter used to map the feature representation from the target-fit dimension dmodel to the number of channels of the muscle electromyography signal. The bias vector representing the linear output layer is a learnable parameter used to adjust the offset of the output features.

[0052] This linear mapping operation converts the high-level feature representation generated by the Transformer model into specific muscle electromyography (EMG) signal predictions, enabling the model to output predictions that match the dimensions of the target output sequence (real muscle EMG signals).

[0053] S333. Combine the predicted values ​​of the muscle electromyography (EMG) signals at each time step in chronological order to obtain a predicted sequence of muscle EMG signals that matches the dimension of the target output sequence.

[0054] The goal of the preset mapping model (Transformer model) in this application embodiment is to learn a parameterized mapping f. θ， Make = f θ (X) Approximates the actual minimally invasive muscle electromyography (EMG) signals as closely as possible (target output sequence Y). Furthermore, in terms of structure, the Transformer model in this embodiment can be equipped with one or more convolutional networks at the front end to pre-extract local waveform features and short-time spectral features. In scenarios with special requirements for real-time performance or resources, lightweight Transformer, TCN, or GRU sequence models can be used to replace or simplify the original Transformer module to achieve similar specific muscle EMG derivation effects.

[0055] In this embodiment, a multi-head self-attention mechanism is employed to capture the correlations between different frequency bands, time segments, and channels in the electromyography (EMG) time series. Each attention head can focus on different time scales or different muscle coupling patterns, thereby simultaneously encoding rapid fluctuations and slow rhythmic changes, local anomalous activities, and global synchronization processes within the same pre-defined mapping model. Finally, by stacking multi-layer Transformer encoder-decoder structures, the pre-defined mapping model abstracts higher-order features from sEMG (surface electromyography) signals and MMG (muscle phonation) signals layer by layer, ultimately generating a predicted muscle EMG signal sequence at the output that is highly similar to the recorded EMG signal. Figure 8 As shown, Figure 8 The schematic diagram provided in the application embodiment shows the process of deriving electromyographic signals from surface electromyographic signals and muscle sound signals. By inputting the collected muscle sound signals and surface electromyographic signals into a preset Transformer model, the muscle electromyographic signals and movements of the target muscle can be predicted.

[0056] S340. The preset mapping model determines the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the muscle electromyography signal prediction sequence.

[0057] This application uses Transformer as the core temporal modeling framework. Compared with traditional convolutional networks or LSTM, Transformer can model the dependency between any two moments in a time series globally through its self-attention mechanism. It is especially suitable for processing complex temporal data such as electromyography (EMG) signals, which have long-term correlation and multi-scale dynamics. This application encodes the multi-channel sequences of surface EMG signals and myophone signals as the input of Transformer and uses the sequence of muscle EMG signals as the target output sequence. Through end-to-end training, it learns the mapping from surface signals to specific muscle signals.

[0058] After obtaining the muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, this application further includes: The loss function value between the predicted muscle electromyography (EMG) signal sequence and the target output sequence is calculated. Based on the loss function value, the parameters of the preset mapping model are updated to obtain an optimized preset mapping model. Specifically, the regression error step-by-step and channel-by-channel is used as the core constraint, and the mean squared error (MSE) loss is used to measure the difference between the prediction and the actual minimally invasive EMG signal.

[0059] in, This represents the predicted value of the muscle electromyography signal at time step t and channel k. This represents the value of the actual muscle electromyography signal at time step t and channel k. Indicates the number of time steps (i.e., the sequence length). This indicates the number of channels for muscle electromyography (EMG) signals. This represents the sum of the squared differences between the predicted and actual values ​​for all time steps and all channels. This indicates that the summation result is normalized to obtain the mean squared error.

[0060] In this embodiment, the loss function is used as the optimization objective to guide the updating of model parameters, so that the model's predicted muscle electromyography (EMG) signals are as close as possible to the actual muscle EMG signals. By minimizing the MSE, the model can learn the mapping relationship between surface EMG signals, myophone signals, and muscle EMG signals.

[0061] Furthermore, to improve the reconstruction quality of the waveform morphology and spectral structure of minimally invasive muscle electromyography (EMG) signals, this application can add simple frequency domain consistency constraints or correlation regularization to the mean squared error loss. In this embodiment, a gradient descent-based optimization algorithm is used, with the loss function as the optimization objective, to iteratively update all learnable parameters in the mapping model. The optimization algorithm includes, but is not limited to, the Adam algorithm. During the training of the mapping model, all paired sample data are collected and divided into training, validation, and test sets, with the division performed on an individual animal basis. This ensures that the generalization evaluation of the mapping model reflects its actual ability to "infer deep activity in new individuals based solely on surface EMG / myophone recordings," thereby providing an end-to-end Transformer methodology framework for reducing deep invasive electrodes and reconstructing or replacing deep information using surface EMG / myophone signals. This embodiment employs a supervised learning strategy during training, using real muscle EMG signals as the target to optimize the parameters of the Transformer model, making it approximate muscle EMG signals in multiple dimensions. Furthermore, the loss function for training the Transformer model not only includes the traditional waveform mean square error, but also incorporates indicators such as spectral error and consistency of detection of specific muscle activity events. This allows the Transformer model to align muscle electromyography signals simultaneously in the time domain, frequency domain, and event-level features. By jointly training with data from normal and abnormal states, the Transformer model can simultaneously learn the mapping relationship between physiological and pathological electromyography activities in surface electromyography signals, muscle sound signals, and specific muscle signals.

[0062] S400: Acquire the second surface electromyographic signal and the second muscle sound signal of the target muscle within the second preset time period.

[0063] S500. Based on the mapping relationship, the second muscle electromyographic signal of the target muscle within the second preset time period is predicted.

[0064] For example, the step of predicting the second muscle electromyography signal of the target muscle within a second preset time period based on the mapping relationship includes: optimizing the mapping relationship based on an optimized preset mapping model; inputting the second surface electromyography signal and the second muscle sound signal of the target muscle within the second preset time period into the optimized preset mapping model; and the optimized preset mapping model predicting the second muscle electromyography signal of the target muscle within the second preset time period based on the optimized mapping relationship.

[0065] Specifically, when the EMG electrodes are removed or no longer used, this embodiment only requires continuous acquisition of surface electromyography (EMG) signals and myophone signals of the target muscle. After the surface EMG signals and myophone signals are processed in the same way as described above, they are input into the optimized Transformer model (preset mapping model), and the corresponding virtual specific muscle EMG signals can be derived in real time. Therefore, the high-value information provided by EMG is "temporarily created" into the non-invasive / minimally invasive EMG recording and mapping model, realizing long-term, low-invasive monitoring of specific muscle activities. This effectively solves the problems in the prior art where EMG cannot be retained for a long time and sEMG reflects specific muscle activities indirectly. It can continuously output the corresponding specific muscle EMG estimates to help observe the relative change trend of specific muscle activities with time and intervention, such as whether specific muscle activities tend to increase or decrease. This embodiment can significantly reduce the invasiveness and discomfort of long-term EMG monitoring, improve patient compliance, and is particularly suitable for scenarios that require long-term monitoring, such as rehabilitation training, chronic muscle disease monitoring, and EMG prosthetic control.

[0066] To explain in detail the principles of the technical solution of this application, the overall process of this application will be described below with reference to some specific embodiments. It is easy to understand that the following is an explanation of the technical principles of this application and should not be regarded as a limitation of this application.

[0067] In one specific embodiment, the long-term electromyographic activity monitoring process of this application includes the following steps: S1. Implanting flexible electrode wires into the target muscle and fixing sEMG (surface electromyography) electrodes in the target muscle: For example, this application uses bullfrogs as experimental subjects and performs chronic electrode implantation. After single-medullary destruction, the bullfrogs are fixed on an experimental table. After skin cleaning and disinfection, the implantation point is determined based on the anatomical location of the target muscles (such as the semimembranosus and vastus medialis). A flexible electrode wire is connected to a radiofrequency ablation device to ensure minimal muscle damage. A multi-channel flexible electrode wire (approximately 0.076 mm in diameter, Pt / Ir) is implanted vertically and slowly into the target muscle using radiofrequency ablation, ensuring a certain length distribution of the electrode wire within the muscle (e.g., approximately 10-15 mm) and that the tip of the electrode wire is located within the muscle fiber group. During insertion, short-term recording is used to observe the electromyographic signal characteristics to ensure correct electrode placement. After successful implantation, the electrode is fixed with bio-adhesive to prevent dislodgement due to movement.

[0068] Simultaneously, a dry electrode wristband (containing a multi-channel sEMG electrode array) was worn on the bullfrog's thigh, with the electrode positions corresponding to the target muscle areas (semimembranosus, vastus lateralis, rectus abdominis, gracilis, etc.). Good contact between the sEMG electrodes and the skin was ensured to guarantee signal quality.

[0069] Then, the liquid metal stretchable strain sensor was wrapped around the base of the bullfrog's thigh and fixed with ergo glue to avoid introducing unnecessary noise artifacts due to sensor slippage.

[0070] Finally, the implanted electrode wire, dry electrode wristband, and liquid metal stretchable strain sensor were connected to the external connector and secured with insulating tape to ensure stability of the implanted flexible electrode and strain sensor on the bullfrog's thigh and good contact between the sEMG electrode and the skin. The entire implantation process was performed with the bullfrog's limbs immobilized to avoid strenuous movement. Immediately after implantation, the signal recording equipment was activated to ensure signal quality and quantity, providing a high-quality, synchronous input-output data foundation for subsequent "sEMG signal derivation of specific muscle signals".

[0071] S2. Acquire electromyographic activity signals: First, after fixing all electrodes / sensors, Ringer's solution was sprayed onto the bullfrog's skin to ensure its activity. Once signal stability was confirmed, electromyographic (EMG) activity was collected, including surface EMG signals, myosalpingography (MGM) signals, and muscle EMG signals. Second, thermal stimulation was used to stimulate receptors in the bullfrog's feet, triggering a reflex arc that resulted in leg retraction / kickback movements, thus obtaining the correspondence between surface EMG signals and muscle EMG signals under normal conditions. Then, specific abnormal muscle activity states were induced, for example, through brief muscle fatigue (such as repetitive elbow flexion until fatigue) or by using specific drugs (such as neostigmine) to induce neuromuscular junction dysfunction, establishing a specific abnormal muscle activity model. Furthermore, during recording, the positional changes of key movement sites (hip, knee, and ankle joints) and the delay from stimulus to response were simultaneously recorded for later stratified analysis of different states.

[0072] S3. Preprocess the acquired surface electromyography (EMG) signals, muscle sound signals, and muscle EMG signals. The raw data consisted of continuous multi-channel acquisitions, with each experiment simultaneously including surface electromyography (sEMG), myophone signals, and muscle EMG signals. First, the data from each experiment were organized and standardized for storage: surface EMG signal matrices, myophone signal matrices, and muscle EMG signal matrices (in the form of channel number × time point) were saved for each individual bullfrog and each single recording, along with metadata such as sampling rate, channel name, movement state, and abnormal state markers. Then, a uniform preprocessing procedure was applied to all channels: bandpass filtering (e.g., 10-600 Hz) was applied to the sEMG and EMG channels, and a 50 Hz and its harmonic notch filter was added to remove power frequency noise. After filtering, the acquired signals from each channel were mean-triggered and normalized according to their standard deviation to reduce the influence of individual differences and electrode contact differences on amplitude, allowing the mapping model to focus more on temporal morphology and relative changes. Secondly, after filtering, artifact detection algorithms (e.g., based on amplitude, instantaneous energy, or derivative thresholds) can be used to mark and remove time segments containing large-amplitude motion artifacts, saturation, or detachment, thus eliminating or interpolating bad channels with long-term distortion. Continuous records are sliced ​​using a fixed-length sliding window, for example, a window length of 5-10 seconds and a sliding step of 1-3 seconds. Strictly time-aligned MMG, sEMG, and EMG segments are extracted within each window. For each segment, the sEMG signal is organized into a T×C_sEMG sequence (T being the number of time points), and the MMG signal is organized into a T×C_MMG sequence, which serves as the model input; simultaneously, the EMG signal of the corresponding time window is organized into a T×C_EMG sequence, serving as the target output for the same sample.

[0073] For different functional states (normal state, mild anomaly, moderate anomaly, severe anomaly, etc.), state labels can be retained to analyze the model's performance under different states. Ultimately, a training set, validation set, and test set are obtained, consisting of a large number of "sEMG sequence → EMG sequence" paired samples, providing a clean and well-structured data foundation for sequence-to-sequence Transformer modeling.

[0074] S3, Transformer model setup: This application's embodiments formalize the "deriving minimally invasive muscle electromyography (EMG) signals from non-invasive myocardial sound signals and multi-channel surface EMG signals" into a sequence-to-sequence regression problem: the corresponding surface EMG signals are used as the input sequence of a preset mapping, and this input sequence is represented as: The muscle sound signal, as the input sequence of the preset mapping, is represented as follows: ;in, This represents the number of non-invasive channels for surface electromyography (EMG) signals. denoted as the number of non-invasive channels for the muscle sound signal, and T as the number of time steps.

[0075] The muscle electromyography (EMG) signal is used as the target output sequence of a predefined mapping model. This target output sequence can be represented as: ,in, Let f be the number of non-invasive channels for muscle electromyography (EMG) signals. The goal of the Transformer model (i.e., the aforementioned pre-defined mapping model) is to learn a parameterized mapping f. θ , making = f θ (X) Approximates the actual minimally invasive EMG signal (target output sequence Y) as closely as possible. Specifically, this involves first generating the non-invasive channel vector x of the corresponding surface electromyography (EMG) and myocardial sound signals for each time step. t Perform a linear projection and embed it into a dimension d that is consistent with the target output sequence. model In space:

[0076] Learnable positional encodings are superimposed on each time step to explicitly introduce temporal order information, thus obtaining the input sequence for the Transformer. The main network employs a multi-layered stacked Transformer encoder. Each layer uses multi-head self-attention to globally model the surface electromyography (EMG) and myophone signal sequences in the time dimension, enabling the representation at any given time point to comprehensively consider the EMG and myophone activities at other times within the entire time window. Simultaneously, feedforward networks and residual structures enhance nonlinear representation capabilities and training stability. After several layers of encoding, a sequence of length T and dimension d is obtained. model The high-level surface electromyography and myophonation representation sequence, and the representation of z for each time step t. t Apply a shared linear output header, mapping it to a deep multichannel prediction at the corresponding time step:

[0077] in, Thus, the predicted deep sequence can be obtained over the entire time window. .

[0078] In the training phase of this application, the regression error step-by-step and channel-by-channel is used as the core constraint, and the mean squared error (MSE) loss is used to measure the difference between the prediction and the actual minimally invasive EMG signal. , Furthermore, simple frequency domain consistency constraints or correlation regularizations can be added as needed to improve the reconstruction quality of the waveform morphology and spectral structure of minimally invasive muscle electromyography (EMG) signals. Overall training employs optimization algorithms such as Adam, conducted on animal-specific training, validation, and test sets. This ensures that the model's generalization evaluation reflects its actual ability to infer deep activity in novel individuals solely based on surface EMG / myophone recordings. This provides an end-to-end Transformer methodology framework for reducing deep invasive electrodes and reconstructing or replacing deep information using surface EMG / myophone signals.

[0079] S4. The Transformer model predicts the electromyographic signals of the target muscle: The optimized Transformer network parameters and its model, pre-trained, are loaded. Subsequently, the multi-channel surface electromyography (EMG) and phonation signals obtained from multiple days of recording are processed according to the aforementioned preprocessing steps, including filtering and artifact detection, to ensure that the distribution of the EMG and phonation signal data input to the Transformer model remains consistent with that during training. For each time period, the EMG and phonation signals are organized into a T×CMMG / sEMG multi-channel EMG / phonation signal sequence and input into the Transformer model to obtain the predicted multi-channel minimally invasive muscle EMG signal output for the corresponding time period. The same input processing and prediction process was used in the experimental group, control group, and at different time points to obtain the deep activity trajectories derived from surface electromyography (EMG) and myophone signals under different states. Subsequently, the predicted muscle EMG signals output by the Transformer model were quantitatively analyzed, including calculating the amplitude, power spectral density, and energy distribution of characteristic frequency bands of each deep channel at different time periods, as well as the correlation coefficient and mean square error between the predicted muscle EMG signals and the actual deep-recorded muscle EMG signals. These metrics were used to evaluate the prediction accuracy and stability of the Transformer model in different animals and under different pathological states, and further analyzed the feasibility and limitations of relying solely on surface EMG and myophone signal recordings to infer deep activity in functional state differentiation and pathological detection.

[0080] In summary, this application provides a method for long-term monitoring of electromyographic activity. This method first acquires the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of a target muscle within a first preset time period. The first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal within the first preset time period are then segmented by a preset time length to obtain multiple time segments of the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal. Based on the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal within each time segment, a mapping relationship between the first surface electromyographic signal, the first muscle sound signal, and the first muscle electromyographic signal is determined. The second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period are then acquired. Based on the mapping relationship, the second muscle electromyographic signal of the target muscle within the second preset time period is predicted. This application constructs a nonlinear mapping relationship from non-invasive / minimally invasive EMG to specific muscle EMG by using surface EMG signals, muscle EMG signals, and myophone signals within a limited time. Based on this mapping relationship, after the EMG electrodes are removed or no longer used, the muscle EMG signal of the target muscle can be obtained by relying only on surface EMG signals and myophone signals. This enables long-term and accurate monitoring of specific muscle activity while maintaining low invasiveness.

[0081] like Figure 9 As shown in the diagram, this application also provides a structural schematic of a long-term electromyography (EMG) activity monitoring device, which can implement the above-described method. The device may include: The first signal acquisition module 21 is used to acquire the first surface electromyographic signal, the first muscle electromyographic signal and the first muscle sound signal of the target muscle within a first preset time period. The signal segmentation module 22 is used to segment the first surface electromyography signal, the first muscle electromyography signal and the first muscle sound signal within the first preset time period by a preset time length to obtain the first surface electromyography signal, the first muscle electromyography signal and the first muscle sound signal in multiple time segments. The relationship determination module 23 is used to determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the first surface electromyography signal, the first muscle electromyography signal and the first muscle sound signal in each time segment; The second signal acquisition module 24 is used to acquire the second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period. The second muscle electromyography (EMG) signal prediction module 25 is used to predict the second muscle EMG signal of the target muscle within the second preset time period based on the mapping relationship.

[0082] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0083] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described long-term electromyography activity monitoring method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0084] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0085] Please see Figure 10 , Figure 10 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 902 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and called and executed by the processor 901 using the long-term electromyography activity monitoring method of the embodiments of this application. The input / output interface 903 is used to implement information input and output; The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904); The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0086] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described long-term electromyography activity monitoring method.

[0087] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0088] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0089] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0090] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0091] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0094] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0096] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for long-term monitoring of electromyographic activity, characterized in that, include: Acquire the first surface electromyographic signal, the first muscle electromyographic signal, and the first muscle sound signal of the target muscle within a first preset time period; The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period are divided into multiple time segments to obtain the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. Based on the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within each time segment, determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal; Acquire the second surface electromyographic signal and the second myocardial sound signal of the target muscle within a second preset time period; Based on the mapping relationship, the second muscle electromyographic signal of the target muscle within the second preset time period is predicted.

2. The method for long-term monitoring of electromyographic activity according to claim 1, characterized in that, Before the step of segmenting the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period by a preset time length, the method further includes: The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal are preprocessed. The preprocessing of the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal includes: The first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal are respectively filtered; Artifact detection is performed on the filtered first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. If an artifact is detected, the signal segment containing the artifact is removed.

3. The method for long-term monitoring of electromyographic activity according to claim 1, characterized in that, The step of determining the mapping relationship between the first surface electromyography (SEMG) signal and the first muscle electromyography (EMG) signal based on the first surface EMG signal, the first muscle EMG signal, and the first muscle sound signal within each time segment includes: The first surface electromyography signal and the first muscle sound signal of each time segment are processed to obtain the input sequence of the preset mapping model, and the first muscle electromyography signal in the corresponding time segment is used as the target output sequence of the preset mapping model. Determine the dimension of the target output sequence, wherein the dimension is used to characterize the number of measurement channels contained in the first muscle electromyography signal; The preset mapping model performs global temporal modeling on the input sequence, and based on the result of the global temporal modeling, generates a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence. The preset mapping model determines the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the muscle electromyography signal prediction sequence.

4. The method for long-term monitoring of electromyographic activity according to claim 3, characterized in that, The process of processing the first surface electromyography signal and the first muscle sound signal for each time segment to obtain the input sequence of the preset mapping model includes: Determine the time step for each of the time segments; Linear projection is performed on the first surface electromyography signal and the first muscle sound signal corresponding to each time step within each time segment to embed the first surface electromyography signal and the first muscle sound signal into a dimensional space adapted to the target output sequence, thereby obtaining an embedded signal sequence; The positions of each time step of the embedded signal sequence are superimposed and encoded to obtain an input sequence containing time information. The input sequence containing time information is then used as the input sequence of the preset mapping model.

5. The method for long-term monitoring of electromyographic activity according to claim 4, characterized in that, The preset mapping model performs global temporal modeling on the input sequence, and based on the result of the global temporal modeling, generates a muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, including: The preset mapping model performs global temporal modeling on the input sequence to obtain a feature sequence containing temporal information and contextual information; Using the dimension of the target output sequence as the target dimension, linear mapping is performed on each feature corresponding to each time step in the feature sequence to obtain the predicted value of the muscle electromyography signal for each time step. The predicted muscle electromyography (EMG) signals for each time step are combined in chronological order to obtain a predicted muscle EMG sequence that matches the dimension of the target output sequence.

6. The method for long-term monitoring of electromyographic activity according to claim 5, characterized in that, After obtaining the muscle electromyography signal prediction sequence that matches the dimension of the target output sequence, the method further includes: Calculate the loss function value between the predicted muscle electromyography signal sequence and the target output sequence; Based on the loss function value, the parameters of the preset mapping model are updated to obtain an optimized preset mapping model.

7. The method for long-term monitoring of electromyographic activity according to claim 6, characterized in that, The step of predicting the second muscle electromyographic signal of the target muscle within the second preset time period based on the mapping relationship includes: Based on the optimized preset mapping model, the mapping relationship is optimized; The second surface electromyographic signal and the second myophone signal of the target muscle within the second preset time period are input into the optimized preset mapping model; The optimized preset mapping model predicts the second muscle electromyographic signal of the target muscle within the second preset time period based on the optimized mapping relationship.

8. A long-term electromyographic activity monitoring device, characterized in that, The device includes: The first signal acquisition module is used to acquire the first surface electromyographic signal, the first muscle electromyographic signal and the first muscle sound signal of the target muscle within a first preset time period. The signal segmentation module is used to segment the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal within the first preset time period by a preset time length to obtain multiple time segments of the first surface electromyography signal, the first muscle electromyography signal, and the first muscle sound signal. The relationship determination module is used to determine the mapping relationship between the first surface electromyography signal and the first muscle electromyography signal based on the first surface electromyography signal, the first muscle electromyography signal and the first muscle sound signal in each time segment; The second signal acquisition module is used to acquire the second surface electromyographic signal and the second muscle sound signal of the target muscle within a second preset time period. The second muscle electromyography (EMG) signal prediction module is used to predict the second muscle EMG signal of the target muscle within the second preset time period based on the mapping relationship.

9. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement a method for long-term monitoring of electromyographic activity as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a method for long-term monitoring of electromyographic activity as described in any one of claims 1 to 7.

Images