Method and system for automatically adjusting stimulation parameters of electric acupuncture apparatus

Abstract

The invention relates to the technical field of electric acupuncture apparatuses, and discloses a method and a system for automatically adjusting stimulation parameters of an electric acupuncture apparatus. According to the method, body surface electromyographic signals, skin impedance and capillary hemodynamic parameters of a user are collected in real time through a sensor set, feature extraction and analysis are conducted through a time sequence prediction model of a Transform architecture, and physiological feature data are generated. And then, constructing a user physiological state characterization model by adopting a multi-modal neural perception fusion algorithm, and carrying out collaborative optimization on the electrical stimulation parameters through an adaptive quantum particle swarm optimization algorithm to generate a stimulation parameter combination matched with the real-time neuromuscular response characteristics of the user. The system drives electrical stimulation output according to the parameter combination, monitors physiological data changes in real time, realizes closed-loop dynamic balance of stimulation parameters and biological feedback signals through the dynamic parameter compensation module, and improves the treatment effect and safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electroacupuncture apparatus, in particular to a method and system for automatically adjusting stimulation parameters of an electroacupuncture apparatus. BACKGROUND

[0002] In the treatment process of a traditional electroacupuncture apparatus, stimulation parameters (such as frequency, pulse width, and intensity) are often preset by doctors according to experience, and it is difficult to dynamically adjust them according to the real-time physiological state of a patient. In addition, long-term use of improper stimulation parameters may cause unnecessary tissue damage or discomfort to the patient. Therefore, developing a method and system capable of automatically adjusting the stimulation parameters of an electroacupuncture apparatus to adapt to the real-time physiological state of different patients and improve the treatment effect and safety has become an important research direction in the technical field of electroacupuncture apparatus. SUMMARY

[0003] In view of the above technical deficiencies, the purpose of the present application is to provide a method and system for automatically adjusting stimulation parameters of an electroacupuncture apparatus to solve the problem that the stimulation parameters cannot be automatically adjusted according to the physiological state of a user in the prior art.

[0004] To solve the above technical problems, the present application adopts the following technical solutions: In a first aspect, the present application provides a method for automatically adjusting stimulation parameters of an electroacupuncture apparatus, which comprises: physiological data of a user is collected in real time by a sensor group, including surface electromyogram data, skin impedance data, and microvascular hemodynamic parameter data, and the physiological data is input into a time series prediction model based on a Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data; Based on the physiological feature data, a user physiological state representation model is constructed using a multi-modal neural perception fusion algorithm, and an adaptive quantum particle swarm optimization algorithm is used to collaboratively optimize the electrical stimulation waveform parameters to generate a combination of stimulation frequency, pulse width, and intensity parameters that matches the real-time neuromuscular response characteristics of the user; According to the parameter combination, electrical stimulation output is driven, and physiological data changes are monitored in real time by a sensor, and surface electromyogram data, skin impedance data, and microvascular hemodynamic parameter data are fed back to a dynamic parameter compensation module in real time, and the dynamic parameter compensation module uses an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width, and intensity, and when the tissue tolerance threshold is detected, a nonlinear attenuation mechanism is triggered to achieve closed-loop dynamic balance of the stimulation parameters and the biological feedback signal.

[0005] Preferably, in a possible implementation form of the first aspect, the sensor group comprises a bipolar surface electromyography electrode array, a four-electrode bioimpedance measurement module, and a laser Doppler flowmeter, and the physiological data acquisition adopts a sliding window mechanism for time alignment processing.

[0006] Preferably, in a possible implementation form of the first aspect, the time prediction model comprises a cascade structure of a multi-head self-attention mechanism layer and a time convolution module. The time convolution module extracts local physiological rhythm features using an inflation causal convolution kernel, and the multi-head self-attention mechanism layer establishes a nonlinear coupling relationship between the phase change of the electromyography signal and the blood flow pulsation through a cross-channel correlation matrix. The cross-channel correlation matrix is generated using a feature tensor outer product operation.

[0007] wherein is an element of the cross-channel correlation matrix, represents the dynamic coupling strength of the electromyography signal of the i th channel and the blood flow signal of the j th channel, represents a tensor product operation, is a sliding window time length, is the surface electromyography signal data of the i th channel collected at time t, is the microvascular hemodynamic parameter data of the j th channel collected at time t. Preferably, in a possible implementation form of the first aspect, the multi-modal neural perception fusion algorithm adopts feature tensor high-order singular value decomposition to project the electromyography time-frequency features, impedance phase angles, and blood flow harmonic components into a low-dimensional latent space, and the adaptive quantum particle swarm optimization algorithm maps the stimulus parameter combination to a Hilbert space through quantum bit coding, introduces an entangled state update mechanism in the parameter optimization process, and enables the particle swarm to explore multiple optimal solution regions simultaneously. Preferably, in a possible implementation form of the first aspect, the quantum bit coding satisfies:

[0008] wherein represents the phase angle of the stimulation frequency,

[0009] corresponds to the rotation phase of the pulse width parameter, and the particle position update formula is:

[0010] wherein represents the phase angle of the stimulation frequency, corresponds to the rotation phase of the pulse width parameter, and the particle position update formula is:

[0011] wherein is the surface electromyography signal data of the i th channel collected at time t,​​​ The particle in the first Position at the next iteration For the first The particle in the first Position at the next iteration For the first Adaptive inertia weights for each particle. , The quantum tunneling coefficient, For the first The optimal position of each individual particle. The optimal position globally. , for Random numbers generated between intervals.

[0012] Preferably, in one possible implementation of the first aspect, the adaptive inertia weights Dynamically adjust based on particle fitness:

[0013] in For the first The fitness value of each particle. This represents the minimum value of the inertia weight. This represents the maximum value of the inertia weight. and The particle fitness function is defined as follows: (The maximum and minimum fitness values ​​in the population are given by the given values.)

[0014] in , , These are weighting coefficients. The signal-to-noise ratio of the electromyographic signal. The gradient of skin impedance change, This represents the coefficient of variation of blood flow.

[0015] Preferably, in one possible implementation of the first aspect, the dynamic parameter compensation module includes a parallel structure of a fuzzy rule knowledge base and an incremental PID controller. The fuzzy rule knowledge base generates compensation factors based on the slope of the electromyographic signal, the differential value of the impedance, and the blood flow acceleration. When the tissue tolerance threshold detection module detects that the rate of change in skin temperature exceeds a preset safety range, it triggers a nonlinear decay based on the hyperbolic tangent function.

[0016] in For time Adjusted stimulus intensity The initial stimulus intensity, is the decay coefficient, is the difference between real-time temperature and baseline, is the initial time point, is the integral variable.

[0017] Preferably, in a possible implementation form of the first aspect, the fuzzy rules employ Takagi-Sugeno type inference system, the input variable membership functions are defined as Gaussian distribution:

[0018] wherein is the input variable is the membership value under the m-th fuzzy rule, is the cluster center of the myoelectric signal feature value under the m-th fuzzy rule, is the variance parameter under the m-th fuzzy rule, the output compensation factor is calculated by weighted average:

[0019] wherein is the weight of the m-th fuzzy rule, is the rule triggering strength, is the total number of fuzzy rules.

[0020] Preferably, in a possible implementation form of the first aspect, the activation condition of the nonlinear decay mechanism satisfies:

[0021] wherein is the second derivative of the skin temperature is the tissue thermal damage threshold coefficient, is the real-time skin impedance value, is the initial skin impedance value, the decay coefficient is adaptively adjusted according to the stimulation duration: wherein

[0022] is the decay coefficient adjusted over time, is the initial decay coefficient, is the decay gain factor, is the tolerance time threshold, is the normalization function.

[0023] ​​​​​In a second aspect, the present invention provides a system for automatically adjusting stimulation parameters of an electroacupuncture device, the system comprising: The data acquisition module collects the user's physiological data in real time through the sensor group, including surface electromyography signal data, skin impedance data and microvascular hemodynamic parameter data. The physiological signals are input into the time-series prediction model based on the Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data. The parameter optimization module, based on physiological feature data, uses a multimodal neural perception fusion algorithm to construct a user physiological state representation model. It then uses an adaptive quantum particle swarm optimization algorithm to collaboratively optimize the electrical stimulation waveform parameters, generating a combination of stimulation frequency, pulse width, and intensity parameters that match the user's real-time neuromuscular response characteristics. The dynamic adjustment module drives the electrical stimulation output according to the parameter combination, and monitors the changes in physiological data in real time through sensors. It feeds back the electromyographic signal data, skin impedance data and microvascular hemodynamic parameter data to the dynamic parameter compensation module. It uses an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width and intensity. When the tissue tolerance threshold is detected, a nonlinear decay mechanism is triggered to achieve a closed-loop dynamic balance between stimulation parameters and biofeedback signals.

[0024] The beneficial effects of this invention are as follows: This method can dynamically adjust the stimulation parameters of the electroacupuncture device by collecting and analyzing the user's physiological data in real time, such as surface electromyography signals, skin impedance and microvascular hemodynamic parameters, so as to accurately match the user's real-time neuromuscular response characteristics.

[0025] This not only improves the effectiveness of electroacupuncture treatment but also enhances the safety of the treatment process, avoiding tissue damage or patient discomfort caused by improper stimulation parameters.

[0026] Furthermore, the adaptive quantum particle swarm optimization algorithm and multimodal neural perception fusion technology of this invention have promoted the development of intelligent control technology in the biomedical engineering industry, provided strong support for the formulation of personalized medical solutions, and promoted innovation and progress in the biomedical engineering industry. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This application provides a flowchart of a method for automatically adjusting stimulation parameters of an electroacupuncture device. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1: As Figure 1 As shown, the present invention provides a method for automatically adjusting stimulation parameters of an electroacupuncture device, comprising: The sensor array collects users' physiological data in real time, including surface electromyography data, skin impedance data, and microvascular hemodynamic parameters. The physiological data is then input into a time-series prediction model based on the Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data.

[0031] In this embodiment, a bipolar surface electromyography (EMG) electrode array (using Ag / AgCl electrodes, electrode spacing 20 mm, sampling frequency 2 kHz) configured at the patient's treatment site is used to collect surface EMG signal data in real time. Simultaneously, a four-electrode bioimpedance measurement module (operating frequency 50 kHz - 1 MHz, excitation current 100 Nm) is used. Skin impedance data was obtained and used in conjunction with a laser Doppler flowmeter (wavelength 785nm, sampling rate 32Hz) to monitor microvascular hemodynamic parameters.

[0032] The three sensors start acquisition in a synchronous trigger mode. In the data preprocessing stage, a sliding window mechanism is used for time alignment processing. The window length is set to 5 seconds and the step size is 1 second. The time deviation caused by the difference in sampling rate of each sensor is eliminated by cubic spline interpolation, ensuring the spatiotemporal consistency between the phase information of electromyography signal and the blood flow pulsation cycle.

[0033] Before the raw physiological data enters the time-series prediction model based on the Transformer architecture, it needs to undergo multiple levels of preprocessing: electromyographic signals are filtered by 20-500Hz bandpass to eliminate power frequency interference and baseline drift, skin impedance data are extracted for frequency domain features using the Cole-Cole model, and hemodynamic parameters are removed by using an adaptive thresholding method to remove motion artifacts.

[0034] Preprocessed multi-source data are input into a time-series prediction model, which consists of a cascaded structure of a temporal convolutional module and a multi-head self-attention mechanism layer. The temporal convolutional module comprises four layers of dilated causal convolutional networks (dilation coefficients of 1, 2, 4, and 8, with a kernel size of 3), employing residual connections and gated activation mechanisms to extract local physiological rhythm features and expand the temporal receptive field. The multi-head self-attention mechanism layer uses eight independent attention heads to calculate the dynamic coupling relationship between electromyographic signals and blood flow parameters through a cross-channel correlation matrix.

[0035] Specifically, the first Electromyographic signal fragments of the channel Blood flow signal fragments from channel n In the sliding window Perform tensor product operations within the channel to generate cross-channel correlation matrix elements. ,in This represents the tensor product operation. The value is set to 50 sampling points (corresponding to a 2.5-second time window). This matrix, after being normalized by Softmax, serves as the attention weight, guiding the model to capture the nonlinear interaction between the phase changes of electromyographic signals and the harmonic components of blood flow.

[0036] During model training, a bidirectional teacher-mandated strategy was employed, using a multimodal physiological dataset containing 200 clinically collected cases for supervised learning. The loss function combined with mean squared error and dynamic time warping algorithm enhanced temporal alignment capabilities. The trained temporal convolutional module identified recruitment patterns of electromyographic signals (such as the burst emission characteristics of motor unit action potentials), while the multi-head self-attention layer revealed cross-modal correlations between impedance phase angle changes and blood flow harmonic components (especially the 0.1Hz metabolic oscillations and the 0.3Hz neurogenic oscillations). The final output physiological feature data contained a 128-dimensional temporal embedding vector, where the first 64 dimensions represented the time-frequency characteristics of electromyographic signals (such as wavelet packet entropy and Teager-Kaiser energy operator), the middle 32 dimensions encoded the relaxation time constant of the impedance spectrum, and the last 32 dimensions described the harmonic energy distribution of the blood flow signal.

[0037] Based on physiological feature data, a multimodal neural perception fusion algorithm is used to construct a user physiological state representation model. An adaptive quantum particle swarm optimization algorithm is used to collaboratively optimize the electrical stimulation waveform parameters to generate a combination of stimulation frequency, pulse width and intensity parameters that match the user's real-time neuromuscular response characteristics.

[0038] In this embodiment, the multimodal neural perception fusion algorithm first performs high-order tensor reconstruction on the 128-dimensional physiological feature data to construct a three-dimensional feature tensor (channel number × time step × feature dimension) containing electromyographic time-frequency features, impedance phase angle, and blood flow harmonic components. This tensor is then projected into a low-dimensional latent space using high-order singular value decomposition. Specifically, the time-frequency feature matrix (64-dimensional), impedance phase angle feature vector (32-dimensional), and blood flow harmonic component matrix (32-dimensional) of the electromyographic signal are modally expanded into three subspaces. The core tensor and each modal factor matrix are iteratively calculated using an alternating least squares algorithm, retaining principal components with a cumulative energy percentage of over 95%. Finally, the original feature dimension is compressed to an 18-dimensional fusion feature vector, achieving effective representation of cross-modal physiological information.

[0039] In the parameter optimization phase, the adaptive quantum particle swarm optimization algorithm combines electrical stimulation parameters (frequency) Pulse width ,strength This is mapped to the Hilbert space encoded by qubits. Each particle position consists of three qubits, and its quantum state is represented as... ,in Phase angle of encoded stimulus frequency (via linear mapping) (Frequency parameterization) Rotation phase corresponding to pulse width parameter ( ), while the square of the magnitude of the quantum state It is positively correlated with stimulus intensity A. During the initialization phase, the particle swarm size is set to 50, and the quantum state parameters of each particle are... Particles are randomly generated within the defined domain, and the correlation between particles is established through a quantum entanglement gate, so that 20% of the particles in the population form entangled pairs and share optimal position information to enhance global search capabilities.

[0040] During particle position updates, a strategy combining an improved quantum tunneling mechanism with classical particle swarm optimization rules is employed. The particle in the first The position update formula for the next iteration is: , For the first The particle in the first Position at the next iteration For the first The particle in the first Position at the next iteration, adaptive inertia weight Dynamically adjust based on the real-time fitness of the particles: In this embodiment , , This represents the current particle fitness value. and These represent the maximum and minimum fitness values ​​in the population, respectively. Quantum tunneling coefficient. and Solving the Schrödinger equation yields values ​​of 1.2 and 1.6, respectively. For the first The optimal position of each individual particle. The optimal position globally. , for Uniformly distributed random numbers. Fitness function. Defined as The weighting coefficients , , , The signal-to-noise ratio of the electromyographic signal (calculated by wavelet packet decomposition, which is the ratio of signal energy to noise energy). The gradient of skin impedance change (take the mean of the absolute values ​​of the impedance difference within the sliding window). This represents the blood flow coefficient of variation (the ratio of standard deviation to mean). After each iteration, the particle state is collapsed according to the quantum measurement principle, mapping the quantum state to the classical solution space, and an elite-preserving strategy is used to update the individual optimality. and global optimal untie.

[0041] During the collaborative optimization process, the algorithm introduces a multi-objective Pareto front screening mechanism. When the particle swarm converges to the Pareto optimal solution set, a fuzzy decision system selects the parameter combination with the highest overall fitness from the non-dominated solutions. In specific implementation, the maximum number of iterations is set to 100, and the convergence condition is that the change in the global optimal solution is less than 1% for 10 consecutive iterations. The final output electrical stimulation parameter combination must meet clinical safety constraints: frequency... (Avoid high-frequency thermal effects), pulse width (Ensure neuronal depolarization), strength (To prevent tissue damage).

[0042] The electrical stimulation output is driven by the parameter combination. The physiological data changes are monitored in real time by the sensor. The electromyographic signal data, skin impedance data and microvascular hemodynamic parameter data of the body surface are fed back to the dynamic parameter compensation module in real time. The dynamic parameter compensation module adopts an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width and intensity. When the tissue tolerance threshold is detected, a nonlinear decay mechanism is triggered to achieve a closed-loop dynamic balance between stimulation parameters and biofeedback signals.

[0043] In this embodiment, the optimized parameter combination drives the electrical stimulation output module to generate a biphasic rectangular pulse waveform, with the pulse amplitude accuracy controlled within... Frequency resolution up to 0.1Hz, pulse width adjustment step size The dynamic parameter compensation module employs a parallel architecture of a fuzzy rule knowledge base and an incremental PID controller to achieve closed-loop regulation. The fuzzy rule knowledge base is configured as a Takagi-Sugeno inference system, containing 36 fuzzy rules. Input variables include the electromyographic signal slope (calculated using the five-point difference method), the impedance differential value (the differential signal after eliminating high-frequency noise using a Butterworth low-pass filter), and blood flow acceleration (extracted based on the first derivative of the Doppler frequency shift signal). Each input variable defines three fuzzy subsets (low, medium, and high), and the membership function adopts a Gaussian distribution. ,in Based on historical data cluster analysis, the cluster centers for electromyography slope were determined as follows: variance parameter The maximum likelihood estimation method was used to optimize the result to 0.2; the cluster centers of the impedance differential values ​​were... , Blood flow acceleration cluster centers are , Output compensation factor Calculated using the weighted average method: Weight After optimization by genetic algorithm, the distribution is in Range, rule trigger strength Take the algebraic product of the membership degrees of each antecedent.

[0044] The incremental PID controller employs a three-loop structure: a current loop update frequency of 10kHz, a voltage loop of 5kHz, and a parameter loop of 1kHz. The proportional gain... The initial value is set to 0.8, and the integration time is... Differential time Output increment ,in To account for the deviation between the set parameters and the measured parameters, the final compensation amount of the fuzzy-PID parallel output is determined by a weighted fusion algorithm, with the weights of the fuzzy system varying with the stimulus time. according to Attenuation, PID weights This enables a smooth transition in control strategies from fuzzy dominance to precise adjustment.

[0045] The tissue tolerance threshold detection module monitors skin temperature changes at a sampling rate of 200 Hz. Second derivative Time triggers a nonlinear decay mechanism, in which The calibration is 0.3, and the basic impedance is... Take the initial 10-second average. Decay function. The integral term in the calculation calculates the cumulative heat load in real time, where For time Adjusted stimulus intensity The initial stimulus intensity, The attenuation coefficient is... This is the difference between the real-time temperature and the baseline. At the initial time point, For the integral variable, the attenuation coefficient initial value Gain factor Time threshold Once activated, the stimulus intensity decreases exponentially with a hyperbolic tangent constraint, reducing it to a safe threshold within 120 seconds. ), and the frequency is according to Gradually decrease the pulse width, reducing it every 10 seconds. Stepwise adjustments are made to ensure that the rate of tissue temperature recovery does not exceed [a certain threshold]. .

[0046] The real-time feedback system utilizes a fiber-optic electromyography (EMG) sensor (bandwidth 10-500Hz) and a miniaturized impedance spectroscopy analysis module (frequency scanning range 1kHz-100kHz) to form a dual-redundant data channel, updating the physiological feature vector every 500ms. The dynamic equilibrium algorithm employs the Lyapunov stability criterion, constructing a cost function that includes stimulus parameter bias, physiological response indices, and cumulative energy input. By setting the rolling optimization window of Model Predictive Control (MPC) to 5 seconds, the solution is made... Minimize the parameter adjustment sequence. When the signal-to-noise ratio of electromyography signal decreases by more than 40% or the coefficient of variation of blood flow increases to 2.5 times the baseline, the system automatically initiates the parameter re-optimization process. While maintaining the current stimulation output, the quantum particle swarm optimization algorithm runs in the background to achieve seamless parameter switching during the treatment process.

[0047] Example 2: This invention provides a system for automatically adjusting stimulation parameters of an electroacupuncture device, comprising: The data acquisition module collects the user's physiological data in real time through the sensor group, including surface electromyography signal data, skin impedance data and microvascular hemodynamic parameter data. The physiological signals are input into the time-series prediction model based on the Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data. The parameter optimization module, based on physiological feature data, uses a multimodal neural perception fusion algorithm to construct a user physiological state representation model. It then uses an adaptive quantum particle swarm optimization algorithm to collaboratively optimize the electrical stimulation waveform parameters, generating a combination of stimulation frequency, pulse width, and intensity parameters that match the user's real-time neuromuscular response characteristics. The dynamic adjustment module drives the electrical stimulation output according to the parameter combination, while monitoring changes in physiological data in real time through sensors. It feeds back the electromyographic signal data, skin impedance data and microvascular hemodynamic parameter data to the dynamic parameter compensation module. It uses an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width and intensity. When the tissue tolerance threshold is detected, a nonlinear decay mechanism is triggered to achieve a closed-loop dynamic balance between stimulation parameters and biofeedback signals.

[0048] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for automatically adjusting stimulation parameters of an electroacupuncture device, characterized in that, The method includes: The sensor array collects users' physiological data in real time, including surface electromyography data, skin impedance data and microvascular hemodynamic parameters. The physiological data is then input into a time-series prediction model based on the Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data. Based on physiological feature data, a multimodal neural perception fusion algorithm is used to construct a user physiological state representation model. An adaptive quantum particle swarm optimization algorithm is used to collaboratively optimize the electrical stimulation waveform parameters to generate a combination of stimulation frequency, pulse width and intensity parameters that match the user's real-time neuromuscular response characteristics. The electrical stimulation output is driven by the parameter combination. The physiological data changes are monitored in real time by the sensor. The electromyographic signal data, skin impedance data and microvascular hemodynamic parameter data of the body surface are fed back to the dynamic parameter compensation module in real time. The dynamic parameter compensation module adopts an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width and intensity. When the tissue tolerance threshold is detected, a nonlinear decay mechanism is triggered to achieve a closed-loop dynamic balance between stimulation parameters and biofeedback signals.

2. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 1, characterized in that, The sensor group includes a bipolar surface electromyography electrode array, a four-electrode bioimpedance measurement module, and a laser Doppler flowmeter. A sliding window mechanism is used for time-series alignment during physiological data acquisition.

3. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 2, characterized in that, The temporal prediction model includes a cascaded structure of a multi-head self-attention mechanism layer and a temporal convolution module; The temporal convolution module uses dilated causal convolution kernels to extract local physiological rhythm features, and the multi-head self-attention mechanism layer establishes a nonlinear coupling relationship between electromyographic signal phase changes and blood flow pulsation through a cross-channel correlation matrix. The cross-channel correlation matrix is ​​generated using the outer product operation of the feature tensor: in For each element of the cross-channel correlation matrix, the element represents the first element. Channel electromyography signals and the first The dynamic coupling strength of the blood flow signal in the channel. This represents the tensor product operation. The duration of the sliding window. For the first Each channel in time Collected electromyography (EMG) signal data from the body surface. No. Each channel in time Collected microvascular hemodynamic parameter data.

4. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 1, characterized in that, The multimodal neural perception fusion algorithm uses high-order singular value decomposition of feature tensors to project electromyographic time-frequency features, impedance phase angle and blood flow harmonic components to a low-dimensional latent space. The adaptive quantum particle swarm optimization algorithm maps the combination of stimulation parameters to Hilbert space through quantum bit encoding and introduces an entangled state update mechanism in the parameter optimization process, so that the particle swarm can explore multiple optimal solution regions at the same time.

5. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 4, characterized in that, The quantum bit encoding satisfies: in The phase angle representing the stimulus frequency, The particle position update formula is based on the rotation phase corresponding to the pulse width parameter: in For the first The particle in the first Position at the next iteration For the first The particle in the first Position at the next iteration For the first Adaptive inertia weights for each particle. , The quantum tunneling coefficient, For the first The optimal position of each individual particle. The optimal position globally. , for Random numbers generated between intervals.

6. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 5, characterized in that, The adaptive inertia weight Dynamically adjust based on particle fitness: in For the first The fitness value of each particle. This represents the minimum value of the inertia weight. This represents the maximum value of the inertia weight. and The particle fitness function is defined as follows: (The maximum and minimum fitness values ​​in the population are given by the given values.) in , , These are weighting coefficients. The signal-to-noise ratio of the electromyographic signal. The gradient of skin impedance change, This represents the coefficient of variation of blood flow.

7. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 1, characterized in that, The dynamic parameter compensation module includes a parallel structure of a fuzzy rule knowledge base and an incremental PID controller. The fuzzy rule knowledge base generates compensation factors based on the slope of the electromyographic signal, the differential value of the impedance, and the blood flow acceleration. When the tissue tolerance threshold detection module detects that the rate of change in skin temperature exceeds the preset safety range, it triggers a nonlinear decay based on the hyperbolic tangent function. in For time Adjusted stimulus intensity The initial stimulus intensity, The attenuation coefficient is... This is the difference between the real-time temperature and the baseline. At the initial time point, It is the integral variable.

8. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 7, characterized in that, The fuzzy rules employ a Takagi-Sugeno inference system, and the membership function of the input variables is defined as a Gaussian distribution. in Input variables In the Membership values ​​under fuzzy rules For the first Cluster centers of electromyographic signal feature values ​​under fuzzy rules For the first The variance parameter under the fuzzy rule is output as the compensation factor. Calculated by weighted average: in For the first Fuzzy rule weights, For rule trigger strength, This represents the total number of fuzzy rules.

9. The method for automatically adjusting stimulation parameters of an electroacupuncture device as described in claim 8, characterized in that, The activation condition of the nonlinear decay mechanism satisfies: in Skin temperature The second derivative, This is the threshold coefficient for tissue thermal damage. This is the real-time skin impedance value. Initial skin impedance value, attenuation coefficient Adaptive adjustment based on stimulus duration: in The attenuation coefficient is adjusted over time. The initial attenuation coefficient, This is the attenuation gain factor. The tolerance time threshold, This is the normalization function.

10. A system for automatically adjusting stimulation parameters of an electroacupuncture device, characterized in that, The system includes: The data acquisition module collects the user's physiological data in real time through the sensor group, including surface electromyography signal data, skin impedance data and microvascular hemodynamic parameter data. The physiological signals are input into the time-series prediction model based on the Transformer architecture for multi-dimensional feature extraction and dynamic correlation analysis to generate physiological feature data. The parameter optimization module, based on physiological feature data, uses a multimodal neural perception fusion algorithm to construct a user physiological state representation model. It then uses an adaptive quantum particle swarm optimization algorithm to collaboratively optimize the electrical stimulation waveform parameters, generating a combination of stimulation frequency, pulse width, and intensity parameters that match the user's real-time neuromuscular response characteristics. The dynamic adjustment module drives the electrical stimulation output according to the parameter combination, and monitors the changes in physiological data in real time through sensors. It feeds back the electromyographic signal data, skin impedance data and microvascular hemodynamic parameter data to the dynamic parameter compensation module. It uses an adaptive control algorithm based on fuzzy rules to incrementally adjust the stimulation frequency, pulse width and intensity. When the tissue tolerance threshold is detected, a nonlinear decay mechanism is triggered to achieve a closed-loop dynamic balance between stimulation parameters and biofeedback signals.

Images