Adaptive calibration method and device for surface myoelectric signals representing locally effective conduction

Abstract

The application provides a local effective conduction characteristic surface electromyogram adaptive calibration method and device, belongs to the wearable physiological signal monitoring technical field, collects and pre-processes the resting surface electromyogram and standard contraction electromyogram of a target muscle, extracts a depth sensitive electromyogram feature reflecting a local conduction state, combines target muscle identification and a preset offline muscle conduction model library, estimates the current local effective conduction state, determines the equivalent transfer response corresponding to the current conduction state and a preset reference conduction state from the model library, when a complete calibration waveform needs to be output, solves the theoretical calibration coefficient of each frequency component by a regularization mapping operator, and performs maximum compensation gain limitation; when only the amplitude or activation feature needs to be output, a feature level compensation coefficient is constructed according to the ratio of the model amplitude transfer gain and gain limitation is performed, and the calibrated electromyogram signal and calibration reliability are output. The application can reduce the influence of local tissue conduction difference on the surface electromyogram comparison result.

Description

Technical Field

[0001] This invention belongs to the field of wearable physiological signal monitoring technology, and specifically refers to an adaptive calibration method and device for surface electromyography signals that provide local effective conduction characterization. Background Technology

[0002] Surface electromyography (EMG) signals are the potential differences formed on the skin surface after the action potentials of skeletal muscle motor units are transmitted through tissues such as muscles, fascia, subcutaneous tissue, and skin. The actual measured surface EMG signals are related to the recruitment and firing frequency of motor units and the action potentials of muscle fibers, and are also affected by the tissue conduction path from the electrical activity source to the skin surface, electrode position, contact state, and the activity of surrounding muscles.

[0003] The skin, subcutaneous tissue, fascia, and muscle structure may vary between different muscles of the same subject, between different measurement locations of the same muscle, and between the same muscle of different subjects. As the distance from the target muscle electrical activity source to the skin surface increases, the amplitude, frequency composition, and spatial distribution of the surface electromyography signal may change; subcutaneous tissue thickness, muscle fiber orientation, muscle feather structure, joint posture, and probe pressure may also alter local volume conduction conditions.

[0004] Existing surface electromyography (EMG) devices typically employ fixed analog gain, fixed filtering parameters, and fixed signal interpretation methods. Some devices normalize EMG amplitudes based on maximal voluntary contraction, submaximal contraction, or task peak, while others adjust the gain according to body mass index and body fat percentage.

[0005] However, fixed gain is only used to bring the signal into the range of the analog front-end and analog-to-digital converter. It cannot determine whether the small signal is caused by low muscle activation, strong local tissue attenuation, electrode deviation from the muscle belly, poor contact, or crosstalk from adjacent muscles. Compensation based solely on body mass index or body fat percentage is also insufficient to reflect local differences between different anatomical parts of the same human body.

[0006] Furthermore, directly amplifying or inversely filtering the attenuated surface electromyography (EMG) signal without constraints may simultaneously amplify noise and motion artifacts. When the target muscle is located deep within the skin or is covered by other muscles, the signal from the skin surface may lack sufficient observability of the target muscle, and simply increasing the compensation coefficient cannot recover the lost information. Summary of the Invention

[0007] To address the technical problems existing in the prior art, the present invention provides an adaptive calibration method and device for surface electromyography signals with local effective conduction characterization, the technical solution of which is as follows: On the one hand, an adaptive calibration method for surface electromyography signals, characterized by local effective conduction, is provided, the method comprising: S1. Acquire and preprocess resting surface electromyography (EMG) and at least one standard systolic EMG of the target muscle, wherein the resting surface EMG is used for baseline and noise estimation, and the standard systolic EMG is used as the target task EMG to be calibrated. S2. Extract depth-sensitive electromyographic features that reflect the local conduction state from the electromyographic data of the target task. S3. Based on the depth-sensitive electromyographic features, combined with the target muscle identifier and a pre-set offline muscle conduction model library, estimate the current local effective conduction state; S4. Determine the equivalent transmission response corresponding to the current local effective conduction state and the preset reference conduction state from the offline muscle conduction model library; S5. When a complete calibration waveform needs to be output, the theoretical calibration coefficients of each frequency component are calculated using a regularized mapping operator based on the equivalent transfer response corresponding to the current local effective conduction state and the preset reference conduction state, and the amplitude of the mapping operator is subject to maximum compensation gain limitation; when only the amplitude or activation feature needs to be output, feature-level mapping is used to form feature-level compensation coefficients based on the ratio of the model amplitude transfer gain corresponding to the current local effective conduction state and the preset reference conduction state, and maximum compensation gain limitation is applied. S6. Output the calibrated surface electromyography signal and the calibration reliability.

[0008] On the other hand, an adaptive calibration device for surface electromyography (EMG) signals, characterized by local effective conduction, is provided. The device includes: surface EMG electrodes, a memory, and a processor. The surface electromyography electrode, using a wet electrode, dry electrode, fabric electrode, or flexible electrode, collects resting surface electromyography and at least one standard systolic electromyography of the target muscle through one or more electromyography channels. The memory stores at least one instruction and saves the target muscle identifier, recommended electrode configuration, offline muscle conduction model library, reference conduction state, model applicability range, compensation limit, historical calibration records, and raw surface electromyography data. The processor loads the instructions and executes them to implement the adaptive calibration method for surface electromyography signals as described in any one of claims 1-7.

[0009] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described adaptive calibration method for surface electromyography signals with local effective conduction characterization.

[0010] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement the above-described adaptive calibration method for surface electromyography signals with local effective conduction characterization.

[0011] The beneficial effects of the technical solution provided by this invention include at least the following: This invention can reduce the impact of local tissue conduction differences on surface electromyography comparison results, and limit or reject unreliable calibrations when signal quality, model applicability, or target muscle observability are insufficient. Specifically: (1) This invention clarifies that resting electromyography is used for baseline and noise estimation, and standard contraction electromyography is used to form controlled and repeatable target task inputs, reducing the impact of baseline drift, contact changes and random contraction differences on calibration parameters.

[0012] (2) The present invention changes from fixed correction to adaptive calibration driven by local conduction state, which reduces the influence of the mutual confusion between muscle activation changes and tissue conduction differences.

[0013] (3) This invention establishes a unified and traceable reference conduction state mapping benchmark, which enhances the comparability of the same target muscle across subjects, time periods and after rewearing.

[0014] (4) This invention improves calibration robustness by regularization mapping and maximum compensation gain constraint, and suppresses noise amplification, inverse operation instability and overcompensation caused by model extrapolation.

[0015] (5) This invention achieves active degradation processing by calibrating the confidence output and failure gating, avoiding misjudging unobservable, uncertain or lost information as recovered. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0017] Figure 1 This is a flowchart of an adaptive calibration method for surface electromyography signals, which provides a local effective conduction characterization method according to an embodiment of the present invention. Figure 2 This is a general block diagram of an adaptive calibration method for surface electromyography signals, which provides a local effective conduction characterization method according to an embodiment of the present invention. Figure 3 This is a block diagram of the overall structure of the electromyography adaptive calibration device provided in the embodiments of the present invention; Figure 4This is a schematic diagram of local effective conduction provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the surface electromyography probe structure provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of a calibration implementation method for surface electromyography data provided in this invention. Figure 7 This is a schematic diagram of offline model establishment and online electromyography calibration provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] The embodiments of this invention divide computational quantities into three categories: the first category consists of observational quantities that are directly measured online or calculated from measured signals, including raw surface electromyography, resting noise, standard contractile electromyography, power spectrum, band energy ratio, and repeatability; the second category consists of model quantities that are established and stored offline, including tissue conductivity parameter ranges, electrode spatial weighting functions, target muscle conduction model libraries, reference conduction states, parameter priors, model uncertainties, and compensation upper limits; the third category consists of online state quantities to be estimated, including local effective conduction states such as local effective conduction depth, amplitude attenuation parameters, frequency selectivity parameters, waveform or spatial diffusion parameters, crosstalk risk, and target muscle observability.

[0020] Single-segment arbitrary surface electromyography (EMG) simultaneously encompasses muscle-source activation and tissue conduction, and these two processes cannot be uniquely separated from a single observation. Therefore, this embodiment of the invention does not directly invert the actual tissue thickness or actual source EMG from the target task EMG. Instead, it forms relatively stable EMG characteristics through a prescribed resting and standard contraction process, and then performs constrained matching with a pre-set offline muscle conduction model library to obtain an equivalent local conduction state for calibration.

[0021] The calibration output is not an unconstrained recovery of electromyography from unknown real sources, but rather maps the current surface observation to a predefined reference conduction state. The reference conduction state can be a system-preset standard state, the first valid individual baseline state for the same subject, a baseline state for a specific site, or a population statistical reference state.

[0022] When the standard contraction repeatability is insufficient, the current features are beyond the model's applicability, the target muscle is not observable, the required compensation exceeds the allowable upper limit, or the surface electromyography signal quality is below the threshold, the system performs at least one of the following operations: limiting compensation, maintaining the original output, using the most recent valid calibration parameters, reducing confidence, marking the calibration as invalid, or prompting for re-acquisition, in order to avoid overcompensating for unobservable information.

[0023] The following is a detailed explanation with reference to the accompanying drawings: This invention provides an adaptive calibration method for surface electromyography signals with local effective conduction characterization. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart of this method is shown below. Figure 2 The diagram shown is an overall block diagram of the method. The processing flow may include the following steps: S1. Acquire and preprocess resting surface electromyography (EMG) and at least one standard systolic EMG of the target muscle, wherein the resting surface EMG is used for baseline and noise estimation, and the standard systolic EMG is used as the target task EMG to be calibrated. like Figures 3 to 6 As shown, in this embodiment of the invention, a surface electromyography (EMG) signal adaptive calibration device with a locally effective conduction characterization acquires resting surface EMG and at least one standard systolic EMG of the target muscle using surface EMG electrodes. The surface EMG electrodes can be wet electrodes, dry electrodes, fabric electrodes, or flexible electrodes. The device can employ one or more EMG channels, but the number of electrodes is not a core limitation of this invention.

[0024] An adaptive calibration device for surface electromyography (EMG) signals, which provides local effective conduction characterization according to an embodiment of the present invention, further includes a memory that stores at least one instruction and saves the target muscle identifier, recommended electrode configuration, offline muscle conduction model library, reference conduction state, model applicability range, compensation upper limit, historical calibration records, and original surface EMG data (the original data will not be overwritten by calibration). An adaptive calibration device for surface electromyography (EMG) signals with local effective conduction characterization according to an embodiment of the present invention further includes a processor, which loads the instructions and executes them to implement the adaptive calibration method for surface EMG signals with local effective conduction characterization.

[0025] An adaptive calibration device for surface electromyography signals, which provides local effective conduction characterization according to an embodiment of the present invention, further includes a low-noise analog front-end, an anti-aliasing filter and analog-to-digital conversion circuit, a communication module, and a power supply.

[0026] Optionally, the user first determines the target muscle, measurement side, and electrode configuration. The device records the target muscle identifier, electrode location or configuration number, sampling rate, analog gain, and filtering parameters. This information is part of an online known configuration library, used to select a subset of parameters from a pre-built offline muscle conduction model library that matches the current target muscle and electrode configuration.

[0027] The device first acquires resting surface electromyography (SEMG signal of the target muscle in a resting state) to estimate baseline drift, power frequency interference, effective noise level, saturation ratio, and signal-to-noise ratio baseline; if the resting EMG does not meet the quality requirements, it will not proceed to standard contraction calibration.

[0028] The subject is then guided to complete at least one prescribed standard contraction, and standard contraction electromyography (EMG) (surface EMG signal of the target muscle in at least one standard contraction state) is collected as the target task EMG to be calibrated. Preferably, this is repeated at least twice, or at least two nominal contraction levels (low and moderate) are completed in a preset sequence. The device extracts amplitude, normalized spectrum, high-low frequency energy ratio, spectral slope, waveform broadening, and repeatability from the repetitive segments. If repeatability is insufficient, a re-acquisition prompt is given. If the analog gain changes, it should first be converted back to a unified physical dimension.

[0029] Specifically, taking the biceps brachii and other target muscles suitable for surface electromyography observation as examples, the user fixes the probe in the recommended position, the device first collects resting electromyography and completes the quality check, and then guides the user to complete three short standard contractions with consistent duration and interval.

[0030] Preprocessing is performed on resting surface electromyography and standard contractile electromyography, including DC removal, bandpass filtering, power frequency suppression, windowing, abnormal segment removal, and amplitude physical quantity conversion.

[0031] S2. Extract depth-sensitive electromyographic features that reflect the local conduction state from the electromyographic data of the target task. Optionally, the depth-sensitive electromyographic features are represented as follows: In the formula, is the depth-sensitive electromyographic feature vector obtained from the current standard contraction; A is the amplitude feature of the standard contraction window; This represents the normalized energy ratio between the high-frequency band and the low-frequency band. The median frequency; This represents the spectral slope or the waveform broadening as calibrated by the model. As an indicator of stability or repeatability between repeated standard contractions; The signal quality component is obtained from resting noise and standard contraction signal.

[0032] In practice, a subset of the above features can be used, or time correlation, channel correlation, or spatial diffusion features can be added.

[0033] S3. Based on the depth-sensitive electromyographic features, combined with the target muscle identifier and a pre-set offline muscle conduction model library, estimate the current local effective conduction state; First, let me explain the volume conduction mechanism of this invention: Surface electromyography (EMG) is the observation result of muscle electrical activity propagating through volumetric conductors such as muscle, fascia, subcutaneous tissue, and skin, and then spatially filtered by surface electrodes. To illustrate the core input-output relationship, the current measurement site can be simplified as follows: In the formula, The surface electromyography signal (frequency domain) actually measured by the device is an online known quantity; The unknown equivalent source signal for the target muscle; The equivalent transmission relationship is formed by the current local tissue structure, equivalent source depth, muscle fiber direction and electrode configuration; Crosstalk caused by adjacent muscles or non-target sources; As noise, its statistical level can be estimated by resting electromyography. For time frequency; This represents the set of parameters that affect local conduction.

[0034] because and At the same time unknown, based solely on Since source activation and tissue conduction cannot be uniquely separated, this embodiment of the invention does not employ [this method]. , Instead of directly calculating the transfer function, we first establish an offline muscle conduction model library and then perform online electromyography calibration, such as... Figure 7 As shown.

[0035] Optionally, the offline muscle conduction model library is established using a hierarchical analytical model, a hierarchical cylindrical model, a finite element model, an experimental calibration model, a physical constraint regression model, or a combination thereof, and scans at least one of the following: equivalent source depth, overlying layer conditions, tissue conductivity, muscle fiber orientation, electrode configuration, and noise level. The surface electromyography obtained from simulation or experiment is processed using the same preprocessing and feature extraction procedures as the device-side model. , , Reference state and model applicable scope, the Let be the model feature function, representing the conduction state when the target muscle is m. At that time, the electromyographic characteristics predicted by the model; the The equivalent transmission response corresponding to the conduction state is used to describe the relative amplitude attenuation and frequency selectivity of surface electromyography within a specified analysis frequency band. It represents the equivalent transmission relationship formed by local tissue structure, equivalent source depth, muscle fiber direction, and electrode configuration, and is a state-related quantity indirectly determined through a model. The gain is transferred to the model amplitude corresponding to the conduction state.

[0036] The tissue conductivity, geometric range, muscle fiber orientation, and electrode spatial parameters in the model are pre-given or calibrated model quantities and are not used as variables in the online free inversion. Their value range can be determined based on literature, material testing, imaging measurements, finite element sensitivity analysis, and repeated human experiments.

[0037] The determination of model parameters follows the principle of separating training, validation, and testing. Training data is used to establish the mapping between features and states, while validation data is used to select feature combinations, , , , , Quality weights and rejection thresholds are used, and test data is used to evaluate stability under cross-person, cross-day, and re-wearing conditions. The device does not need to run a complete finite element model in real time; a validated model can be compressed into a lookup table, interpolator, low-order surrogate model, or physical constraint regression model.

[0038] Optionally, the current local effective conduction state Includes at least one of the following: local effective conduction depth ( The equivalent conduction parameter (not limited to the actual subcutaneous tissue thickness or the geometric distance of any actual tissue interface) represents the equivalent conduction depth formed by the combined effects of factors such as the current local tissue, the depth of the electroactive source, the direction of muscle fibers, and electrode spatial filtering; amplitude attenuation parameter. This indicates the degree to which the amplitude of electromyographic activity is attenuated during its transmission from the inside of the muscle to the skin surface; frequency selectivity parameter. This indicates that local tissues have different abilities to transmit electromyographic components of different frequencies; waveform or spatial diffusion parameters. This indicates the degree of broadening or spatial diffusion of the electromyographic signal after transmission through tissue volume; crosstalk risk parameter. This indicates the risk that the current electrode acquisition results may be affected by nearby non-target muscles; target muscle observability parameters. This indicates the effective sensing capability of the current surface electromyography configuration for target muscle activity; In one implementation, θ includes at least an amplitude transfer parameter. and frequency selectivity parameters In multi-channel implementations, waveform or spatial diffusion parameters may also be included. Local effective conduction depth This can serve as an equivalent label for the state or as additional output obtained from the state through a preset mapping, and is not limited to the actual subcutaneous tissue thickness. Target muscle observability parameters. and crosstalk risk parameters Preferred in The estimated value is then used as the quality evaluation metric.

[0039] The current local effective conduction state The value is determined based on the degree of matching between the current depth-sensitive electromyographic features and the model, combined with the consistency of the prior range and repetitive standard contractions, and is expressed by the formula: In the formula, The estimated value of the current local effective conduction state represents the state that minimizes the sum of electromyographic feature matching loss, state prior loss, and repetitive standard contraction consistency loss. The electromyography feature matching loss represents the current depth-sensitive electromyography features. The difference between the predicted electromyographic features and those predicted by the model; For state prior loss, it represents the penalty term for state parameters deviating from the allowable prior range of the target muscle; The consistency loss for repeated standard shrinkage represents the penalty for inconsistent states obtained from different repeated standard shrinkages, and is a preset value. Consistency weights are used; characteristic uncertainties are characterized by the covariance matrix R; the prior propagation state of muscle m and its uncertainty are respectively represented by... And P characterization.

[0040] The model's predictive relationships, feature uncertainties, and state priors are all determined offline using training data, simulation data, and repeated human measurement data. R can be determined by the feature residuals generated by repeated measurements or simulation perturbations under the same state, and P can be determined by the allowable parameter range of the target muscle or the population statistical distribution. The parameters can be selected using independent validation data. All of the above parameters are known offline and do not require retraining for each online measurement.

[0041] S4. Determine the equivalent transmission response corresponding to the current local effective conduction state and the preset reference conduction state from the offline muscle conduction model library; After obtaining the current local conduction state estimate, the equivalent transmission responses corresponding to the current effective local conduction state and the preset reference conduction state are determined from the offline muscle conduction model library, respectively. and .

[0042] in, This is the equivalent transfer response at the current measurement site under specified target muscle, specified electrode configuration, and estimated local conduction state. The equivalent transfer response describes the relative amplitude attenuation and frequency selectivity of surface electromyography (EMG) within a specified analysis frequency band. It is not limited to a unique transfer function of real human tissue, nor is it directly obtained through the ratio of measured surface EMG to unknown source EMG. The system uses the target muscle identifier m, electrode configuration c, frequency f, and... (Estimated value) Input the offline target muscle conduction model and obtain it by calculation or table lookup. .

[0043] The reference conduction state θ ref Input the same model and obtain .

[0044] More specifically: In the offline phase, for each target muscle and electrode configuration, the allowable range of local conduction states is pre-determined, and a correspondence is established between the local conduction states and the model-predicted electromyographic characteristics and equivalent transfer responses. This correspondence can be stored as parameterized function coefficients, discrete lookup tables, interpolation nodes, regression models, or combinations thereof. In one embodiment, the equivalent transfer response uses a non-negative amplitude-frequency function determined by amplitude transfer parameters and frequency selectivity parameters; the system traverses preset state nodes, generates the equivalent transfer response for each state node, and forms model-predicted features according to the feature extraction process consistent with online acquisition. Model coefficients, feature uncertainties, state priors, and model applicability range are determined using simulation data, tissue phantom data, human repeated measurement data, or combinations thereof. In the online phase, the model is not retrained, but rather... Call the parameterized function, or interpolate between adjacent state nodes to obtain .

[0045] S5. When a complete calibration waveform needs to be output, the theoretical calibration coefficients of each frequency component are calculated using a regularized mapping operator based on the equivalent transfer response corresponding to the current local effective conduction state and the preset reference conduction state. The amplitude of the mapping operator is then subject to maximum compensation gain limitation (to generate a restricted mapping operator while maintaining its phase relationship, thereby obtaining the equivalent surface electromyography waveform under the reference conduction state). When only the amplitude or activation feature needs to be output, feature-level mapping is used to form feature-level compensation coefficients based on the ratio of the model amplitude transfer gain corresponding to the current local effective conduction state and the preset reference conduction state, and maximum compensation gain limitation is applied. The complete waveform path and feature-level path in the embodiments of the present invention are two selectable calibration methods, and simultaneous execution is not required.

[0046] Optionally, the regularization mapping operator is expressed by the formula: In the formula, For regularized mapping operators without maximum compensation gain constraints; This is the equivalent transfer response corresponding to the preset reference conduction state; For the present The equivalent transitive response in the model library; This is a regularization parameter used to suppress excessive inverse compensation when the current propagation response is weak, the noise is high, or the model uncertainty is large. It is determined based on the resting noise power, the current signal-to-noise ratio, and the model prediction uncertainty.

[0047] Optionally, the maximum compensation gain is expressed as: In the formula, This represents the maximum allowable compensation gain at frequency f; The upper limit of the hardware is determined by the dynamic range of data acquisition and numerical stability. The upper limit is determined by the current signal-to-noise ratio and noise floor; It is an upper limit determined by the distance of the current state from the boundary of the model's applicable range or the model's uncertainty; This is the upper limit determined by the standard shrinkage repeatability.

[0048] In actual output, the gain of the mapping operator is limited to Within.

[0049] Optionally, the feature-level mapping is expressed by the formula: In the formula, The equivalent amplitude characteristics after mapping to the reference conduction state; This represents the original amplitude characteristics; The equivalent transfer gain of the model with respect to a specified amplitude characteristic under the reference conduction state; ε is the model magnitude transfer gain corresponding to the current local conduction state; ε is a small regularization quantity used to avoid numerical instability when the current model magnitude transfer gain is too small, and to make the compensation ratio more conservative when both the current local conduction state and the reference conduction state are in the low observable range.

[0050] S6. Output the calibrated surface electromyography signal and the calibration reliability.

[0051] Optionally, the calibration confidence (avoiding overly certain calibration results when signal quality is poor, model extrapolation is incomplete, or target muscle observability is insufficient) is formed based on multiple mass components, expressed by the formula: In the formula, To calibrate credibility; The i-th normalized mass component has a value range of 0 to 1 and represents resting noise, power frequency interference, low frequency drift, saturation ratio, standard contraction repeatability, model applicability, target muscle observability and compensation safety, respectively. For the corresponding weights ( It can be determined based on the calibration error and failure risk on the independent validation set), and the sum of each weight is 1; M is the number of quality components participating in the credibility evaluation.

[0052] Target muscle observability is used to evaluate the effectiveness of the current electrode configuration in sensing target muscle activity. In one implementation, the expected contributions of the target muscle and major neighboring muscles to the current electrode configuration are determined based on an offline conduction model, and the noise contribution is determined based on resting electromyography (EMG). Target muscle observability is determined by the proportion of the expected contribution of the target muscle to the sum of the expected contributions of the target muscle, neighboring muscles, and noise. The expected contribution is a model estimate and does not represent the actual energy of the target muscle and neighboring muscles that has been uniquely isolated from single-channel measured surface EMG.

[0053] Compensation safety is used to evaluate the relationship between the theoretically required calibration amount and the currently allowed calibration amount. In one implementation, for full waveform calibration, the frequency compensation margin is determined based on the ratio of the maximum allowable compensation gain at each frequency to the gain of the unlimited mapping operator, and the compensation margins within the analysis band are weighted and summarized; for feature-level calibration, the feasibility of compensation is determined based on the ratio of the maximum allowable feature-level compensation amount to the theoretically required feature compensation amount. When the required compensation exceeds the hard upper limit, the maximum over-limit multiple exceeds the threshold, or the proportion of the effective frequency band requiring limiting exceeds the threshold, the system limits compensation or marks the calibration as invalid.

[0054] The confidence score in this embodiment of the invention does not replace critical failure gating. When the observability of the target muscle is below the threshold, the standard contraction repeatability is significantly insufficient, the current features are far beyond the applicable range of the model, the signal is saturated, or the required compensation exceeds the hard upper limit, the system directly performs at least one of the following operations: limiting compensation, maintaining the original output, using the most recent valid parameters, reducing confidence, marking the calibration as invalid, or prompting for re-acquisition.

[0055] The embodiments of the present invention perform the following online updates based on the electromyography window: During the target task, the device can recalculate signal quality and depth sensitivity characteristics according to preset time windows. Updates are only permitted if the current window is sufficiently consistent with the activation range, spectral shape, and quality conditions of the calibration phase. ; Update weights as needed It reduces noise and decreases the impact. When window noise is high, repeatability is insufficient, or the model range is exceeded, the most recent effective state is maintained, and no new core calculation formula is required.

[0056] The embodiments of the present invention can also perform the following multi-level standard shrinkage calibration: Users follow the prompts to complete two types of standard contraction: low-level and medium-level, repeating each type at least twice. The system generates the amplitude, normalized spectrum, frequency band ratio, and repeatability characteristics for each level, and uses repeatability within the same level and relative response between different levels to form consistency constraints.

[0057] When both contraction levels are stable, the model uses the amplitude scale, which mainly changes with the nominal contraction level, as evidence of changes in source activation, and uses the spectrum shape, broadening, and model matching results that remain relatively stable at different levels as evidence of local conduction status, thereby reducing the risk of misjudging "stronger contraction" as "better tissue conduction"; when either level is unstable, only the reliable level is used and the confidence level is reduced.

[0058] The embodiments of the present invention can also perform the following cross-day individual reference calibration: For the first effective measurement, save the target muscle, electrode configuration, standard contraction characteristics, and reference state. and model version. Subsequent measurements involved re-acquiring resting and standard contraction electromyography to estimate the current... It is also mapped back to the initial reference state to improve comparability across days and after rewearing.

[0059] When subsequent features deviate from the initial reference and cannot be explained by the conduction state changes allowed by the model, or are significantly beyond the scope of the model's applicability, the system prompts for checking the electrode position and skin contact state or re-establishing the reference state, instead of forcibly performing compensation.

[0060] The embodiments of the present invention also include the following alternative solutions: Standard contractions can be guided by device voice prompts, interface animations, or preset timing programs. Different target muscles can use different standard contraction movements, but the calibration protocol for the same target muscle should remain consistent. The device can output only the characteristic-level calibration results or output the complete calibration waveform. Regardless of the output, the raw surface electromyography data that has not undergone local conduction calibration must be retained.

[0061] The specific feature subsets, model forms, optimization methods, lookup table methods, surrogate models, regularization methods, and confidence combination methods in this invention can all be replaced with equivalent ones. As long as the core relationship of "using only surface electromyography calibration data to form depth-sensitive features—combining offline target muscle conduction models to estimate the local equivalent conduction state—mapping to the reference conduction state—implementing restricted calibration" is maintained, all of these fall under the technical concept of this invention.

[0062] This invention also provides an adaptive calibration device for surface electromyography signals, which is used to characterize local effective conduction. The device includes: surface electromyography electrodes, a memory, and a processor. The surface electromyography electrode, using a wet electrode, dry electrode, fabric electrode, or flexible electrode, collects resting surface electromyography and at least one standard systolic electromyography of the target muscle through one or more electromyography channels. The memory stores at least one instruction and saves the target muscle identifier, recommended electrode configuration, offline muscle conduction model library, reference conduction state, model applicability range, compensation limit, historical calibration records, and raw surface electromyography data. The processor loads the instructions and executes them to implement the adaptive calibration method for surface electromyography signals as described in any one of claims 1-7.

[0063] The adaptive calibration device for surface electromyography (EMG) signals with local effective conduction characterization provided in this embodiment of the invention has a functional structure that corresponds to the adaptive calibration method for surface EMG signals with local effective conduction characterization provided in this embodiment of the invention, and will not be described again here.

[0064] Figure 8 This is a schematic diagram of the structure of an electronic device 800 provided in an embodiment of the present invention. The electronic device 800 may vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 801 and one or more memories 802. The memory 802 stores at least one instruction, which is loaded and executed by the processor 801 to implement the steps of the above-mentioned adaptive calibration method for surface electromyography signals with local effective conduction characterization.

[0065] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to perform the aforementioned adaptive calibration method for surface electromyography signals with local effective conduction characterization. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.

[0066] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive calibration method for surface electromyography signals, characterized by local effective conduction, wherein... The method includes: S1. Acquire and preprocess resting surface electromyography (EMG) and at least one standard systolic EMG of the target muscle, wherein the resting surface EMG is used for baseline and noise estimation, and the standard systolic EMG is used as the target task EMG to be calibrated. S2. Extract depth-sensitive electromyographic features that reflect the local conduction state from the electromyographic data of the target task. S3. Based on the depth-sensitive electromyographic features, combined with the target muscle identifier and a pre-set offline muscle conduction model library, estimate the current local effective conduction state; S4. Determine the equivalent transmission response corresponding to the current local effective conduction state and the preset reference conduction state from the offline muscle conduction model library; S5. When a complete calibration waveform needs to be output, the theoretical calibration coefficients of each frequency component are calculated using a regularized mapping operator based on the equivalent transfer response corresponding to the current local effective conduction state and the preset reference conduction state, and the amplitude of the mapping operator is subject to maximum compensation gain limitation; when only the amplitude or activation feature needs to be output, feature-level mapping is used to form feature-level compensation coefficients based on the ratio of the model amplitude transfer gain corresponding to the current local effective conduction state and the preset reference conduction state, and maximum compensation gain limitation is applied. S6. Output the calibrated surface electromyography signal and the calibration reliability.

2. The method according to claim 1, characterized in that, The depth-sensitive electromyographic features are represented as follows: In the formula, is the depth-sensitive electromyographic feature vector obtained from the current standard contraction; A is the amplitude feature of the standard contraction window; This represents the normalized energy ratio between the high-frequency band and the low-frequency band. The median frequency; This represents the spectral slope or the waveform broadening as calibrated by the model. As an indicator of stability or repeatability between repeated standard contractions; The signal quality component is obtained from resting noise and standard contraction signal.

3. The method according to claim 1, characterized in that, The offline muscle conduction model library is established using hierarchical analytical models, hierarchical cylindrical models, finite element models, experimental calibration models, physical constraint regression models, or combinations thereof. It scans at least one of the following: equivalent source depth, overlying layer conditions, tissue conductivity, muscle fiber orientation, electrode configuration, and noise level. The surface electromyography (EMG) data obtained from simulations or experiments undergoes the same preprocessing and feature extraction procedures as those used on the device. , , Reference state and model applicable scope, the Let be the model feature function, representing the conduction state when the target muscle is m. At that time, the electromyographic characteristics predicted by the model; The The equivalent transmission response corresponding to the conduction state is used to describe the relative amplitude attenuation and frequency selectivity of surface electromyography within a specified analysis frequency band. It represents the equivalent transmission relationship formed by local tissue structure, equivalent source depth, muscle fiber direction, and electrode configuration, and is a state-related quantity indirectly determined through a model. The gain is transferred to the model amplitude corresponding to the conduction state.

4. The method according to claim 1, characterized in that, The current local effective conduction state Includes at least one of the following: local effective conduction depth This represents the equivalent conduction depth formed by the combined effects of factors such as the current local tissue, the depth of the electrical activity source, the direction of the muscle fibers, and the spatial filtering of the electrodes; amplitude attenuation parameter. This indicates the degree to which the amplitude of electromyographic activity is attenuated as it travels from the inside of the muscle to the skin surface; frequency selectivity parameter. This indicates that local tissues have different abilities to transmit electromyographic components of different frequencies; waveform or spatial diffusion parameters. This indicates the degree of broadening or spatial diffusion of the electromyographic signal after transmission through tissue volume; crosstalk risk parameter. This indicates the risk that the current electrode acquisition results may be affected by nearby non-target muscles; target muscle observability parameters. This indicates the effective sensing capability of the current surface electromyography configuration for target muscle activity; The current local effective conduction state The value is determined based on the degree of matching between the current depth-sensitive electromyographic features and the model, combined with the consistency of the prior range and repetitive standard contractions, and is expressed by the formula: In the formula, The estimated value of the current local effective conduction state represents the state that minimizes the sum of electromyographic feature matching loss, state prior loss, and repetitive standard contraction consistency loss. The electromyography feature matching loss represents the current depth-sensitive electromyography features. The difference between the predicted electromyographic features and those predicted by the model; For state prior loss, it represents the penalty term for state parameters deviating from the allowable prior range of the target muscle. The consistency loss for repeated standard shrinkage represents the penalty for inconsistent states obtained from different repeated standard shrinkages, and is a preset value. Consistency weights are used; characteristic uncertainties are characterized by the covariance matrix R; the prior propagation state of muscle m and its uncertainty are respectively represented by... And P characterization.

5. The method according to claim 1, characterized in that, In step S4, the equivalent transmission responses corresponding to the current local effective transmission state and the preset reference transmission state are determined from the offline muscle conduction model library, respectively. and .

6. The method according to claim 1, characterized in that, The regularization mapping operator is expressed by the following formula: In the formula, For regularized mapping operators without maximum compensation gain constraints; This is the equivalent transfer response corresponding to the preset reference conduction state; For the present The equivalent transitive response in the model library; This is a regularization parameter used to suppress excessive inverse compensation when the current propagation response is weak, the noise is high, or the model uncertainty is large. It is determined based on the resting noise power, the current signal-to-noise ratio, and the model prediction uncertainty.

7. The method according to claim 1, characterized in that, The maximum compensation gain is expressed as: In the formula, This represents the maximum allowable compensation gain at frequency f; The upper limit of the hardware is determined by the dynamic range and numerical stability of the data acquisition. The upper limit is determined by the current signal-to-noise ratio and noise floor; It is an upper limit determined by the distance of the current state from the boundary of the model's applicable range or the model uncertainty; This is the upper limit determined by the standard shrinkage repeatability.

8. The method according to claim 1, characterized in that, The feature-level mapping is expressed by the formula: In the formula, The equivalent amplitude characteristics after mapping to the reference conduction state; This represents the original amplitude characteristics; The equivalent transfer gain of the model with respect to a specified amplitude characteristic under the reference conduction state; Transfer gain to the model amplitude corresponding to the current local conduction state; ε is a small regularization quantity used to avoid numerical instability when the current model's magnitude propagation gain is too small, and to make the compensation ratio more conservative when both the current local propagation state and the reference propagation state are in the low observable range.

9. The method according to claim 1, characterized in that, The calibration reliability is formed based on multiple quality components and is expressed by the formula: In the formula, To calibrate credibility; The i-th normalized mass component has a value range of 0 to 1 and represents resting noise, power frequency interference, low frequency drift, saturation ratio, standard contraction repeatability, model applicability, target muscle observability and compensation safety, respectively. These are the corresponding weights, and the sum of all weights is 1; M represents the number of quality components involved in the credibility evaluation.

10. An adaptive calibration device for surface electromyography signals, characterized by local effective conduction, is characterized in that, The device includes: surface electromyography electrodes, a memory, and a processor; The surface electromyography electrode, using a wet electrode, dry electrode, fabric electrode, or flexible electrode, collects resting surface electromyography and at least one standard systolic electromyography of the target muscle through one or more electromyography channels. The memory stores at least one instruction and saves the target muscle identifier, recommended electrode configuration, offline muscle conduction model library, reference conduction state, model applicability range, compensation limit, historical calibration records, and raw surface electromyography data. The processor loads the instructions and executes them to implement the adaptive calibration method for surface electromyography signals as described in any one of claims 1-7.

Images