A myoelectric feedback electrostimulation control system and method

Abstract

This invention discloses an electromyography (EMG) feedback electrical stimulation control system and method, relating to the field of electrical stimulation control technology in rehabilitation physiotherapy. The system includes acquiring raw surface EMG signals of the target muscle, extracting the signal envelope after bandpass filtering and rectification to obtain a time-varying amplitude signal reflecting muscle contraction intensity, and obtaining a muscle activation value through a preset muscle activation calculation model. The muscle activation value is compared with a preset target activation range; if the value is lower than the lower limit of the range, electrical stimulation output is initiated and a state inversion process is entered. The electrical stimulation intensity is deduced using the muscle activation value as the current state and the median of the target activation range as the desired state. Based on this intensity, the duty cycle and frequency parameters of the electrical stimulation waveform are configured, and a control signal is generated. This invention achieves precise matching between electrical stimulation intensity and the real-time muscle state, allowing waveform parameters to dynamically adapt to the stimulation intensity, thus optimizing the control accuracy of EMG feedback electrical stimulation.

Description

Technical Field

[0001] This invention belongs to the field of rehabilitation physiotherapy electrical stimulation control technology, specifically an electromyographic feedback electrical stimulation control system and control method. Background Technology

[0002] Existing electromyographic feedback (EMG) stimulation control technologies mostly involve acquiring raw surface EMG signals from the target muscle, performing bandpass filtering and rectification on the signals, extracting the envelope to obtain a signal that characterizes the muscle contraction intensity, and then using a corresponding computational model to obtain a muscle activation value. This value is then compared with a preset threshold to determine whether to initiate electrical stimulation output. In these technologies, the electrical stimulation intensity is mostly determined using a fixed value or a simple ratio, and the duty cycle and frequency parameters of the electrical stimulation waveform are mostly preset fixed parameters that are not adjusted in real time according to actual stimulation needs.

[0003] Existing technologies, in the process of electrical stimulation control, can only determine whether to apply electrical stimulation by comparing the muscle activation level with a preset interval. They cannot perform inversion calculations based on the desired state between the current muscle activation state and the median of the target activation interval, making it difficult to determine the appropriate electrical stimulation intensity to match the muscle state. Even after the electrical stimulation intensity is determined, it is impossible to specifically configure the duty cycle and frequency parameters of the electrical stimulation waveform based on that intensity; the waveform parameters cannot form a matching relationship with the actual required stimulation intensity. This invention aims to solve the problems that the electrical stimulation intensity cannot be accurately determined through state inversion, and that the electrical stimulation waveform parameters cannot be dynamically configured based on the inverted intensity. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a method for controlling electromyographic feedback electrical stimulation, comprising: The raw surface electromyographic (EMG) signal of the target muscle is acquired, and the raw surface EMG signal is subjected to bandpass filtering and rectification to obtain the processed EMG signal. Envelope extraction is performed on the processed electromyographic signal to obtain a time-varying amplitude signal that reflects the intensity of muscle contraction; Based on the time-varying amplitude signal, the current muscle activation value is calculated by reasoning through a preset muscle activation calculation model. The muscle activation values ​​are compared with a pre-set target activation range, and the necessity of electrical stimulation intervention is determined based on the comparison results. When the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and the state inversion process is initiated. In the state inversion process, the muscle activation value is taken as the current state, and the median value of the target activation range is taken as the desired state. The required electrical stimulation intensity is then calculated through inversion. Based on the inverted calculation of the electrical stimulation intensity, the duty cycle and frequency parameters of the electrical stimulation waveform are configured to generate an electrical stimulation control signal.

[0005] Further, the original surface electromyography (EMG) signal is subjected to bandpass filtering and rectification to obtain a processed EMG signal, including: Based on the action potential characteristics of the motor unit of the target muscle, a lower cutoff frequency and an upper cutoff frequency are set to construct a bandpass filter. The original surface electromyography signal is filtered using the bandpass filter to remove power frequency interference and motion artifacts, resulting in a filtered electromyography signal. The filtered electromyographic signal is subjected to full-wave rectification to flip the signal on the negative half-axis to the positive half-axis, thus obtaining a full-wave rectified electromyographic signal. The full-wave rectified electromyographic signal is subjected to low-pass smoothing to eliminate high-frequency jitter in the signal, resulting in the processed electromyographic signal.

[0006] Furthermore, based on the time-varying amplitude signal, the current muscle activation value is calculated using a preset muscle activation calculation model, including: The time-varying amplitude signal is averaged using a sliding window to obtain the average amplitude within the corresponding time window. Acquire the baseline amplitude of the target muscle in a resting state, which is obtained through pre-acquisition; Calculate the gain factor of the average amplitude relative to the reference amplitude; Substituting the gain factor into a preset S-shaped function model, the muscle activation value between zero and one is calculated by using a lookup table method or piecewise linear interpolation method.

[0007] Furthermore, when the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and the state inversion process is entered, including: Define an error variable, which is equal to the median of the target activation interval minus the current muscle activation value; Determine whether the absolute value of the error variable is greater than a preset dead zone threshold; If the absolute value of the error variable is greater than the preset dead zone threshold, the current control mode will be switched to active closed-loop adjustment mode, and the state inversion process will be executed. In the state inversion process, a nonlinear mapping relationship is established that includes muscle activation, electrical stimulation intensity, and nerve excitation conduction rate; Based on the aforementioned nonlinear mapping relationship, the increment of electrical stimulation intensity required to change muscle activation from the current state to the desired state is calculated.

[0008] Furthermore, the step of configuring the duty cycle and frequency parameters of the electrical stimulation waveform based on the inverted calculated electrical stimulation intensity, and generating an electrical stimulation control signal, includes: Read the inverted and calculated electrical stimulation intensity value and map it to the preset stimulation intensity level table; Based on the correspondence in the stimulation intensity level table, a basic pulse frequency and pulse width are selected; A muscle fatigue factor is introduced to correct the selected pulse width, the muscle fatigue factor being calculated based on continuous working time; The modified pulse width is combined with the selected pulse frequency, and clamping protection is performed according to the preset upper voltage limit to generate the electrical stimulation control signal.

[0009] Furthermore, it also includes: The generated electrical stimulation control signal is applied to the target muscle, and the time-varying amplitude signal is continuously monitored during the electrical stimulation process; During electrical stimulation, the inference calculation and state inversion steps are executed cyclically. The muscle activation value is recalculated based on the real-time time-varying amplitude signal, and the electrical stimulation intensity is adjusted again until the muscle activation value enters and stabilizes within the target activation range. The inference calculation and state inversion steps are performed cyclically during electrical stimulation, including: A fixed sampling period is set, and at the beginning of each sampling period, the time-varying amplitude signal is synchronously acquired once. Using the latest time-varying amplitude signal, the inference calculation step is re-executed to update the current muscle activation value; The difference between the updated muscle activation value and the muscle activation value at the previous sampling time is calculated to obtain the activation change rate. The activation rate change is used as a constraint in the state inversion process to limit the adjustment range of electrical stimulation intensity and prevent overshoot. The process of until the muscle activation value enters and stabilizes within the target activation range includes: During the electrical stimulation output, the fluctuation range of the muscle activation value is continuously monitored; When the muscle activation value first enters the target activation range, a stable timing record begins. If the muscle activation value remains within the target activation range for a continuous preset duration, it is determined that a stable state has been reached. Once a stable state is determined, the current electrical stimulation intensity parameters are locked, the state inversion process is stopped, and only basic electromyographic signal monitoring is retained.

[0010] Furthermore, it also includes a safe exit mechanism under abnormal operating conditions: Before each inference calculation step, the signal-to-noise ratio of the time-varying amplitude signal is first verified. If the signal-to-noise ratio is lower than a preset safety threshold, the currently acquired electromyographic signal is determined to be invalid. When the electromyographic signal is determined to be invalid, the electrical stimulation output should be immediately paused and the control mode should be switched to manual mode. At the same time, an alarm is issued, and the system waits for the operator to intervene and check. Automatic control is only restored after a valid electromyographic signal is obtained again.

[0011] Furthermore, it also includes adaptive parameter adjustment steps for different training stages: Record the maximum and average muscle activation values ​​during each training session; Based on the changing trend of the maximum muscle activation value recorded in multiple training sessions, the upper and lower limits of the target activation range are dynamically adjusted. Based on the changing trend of the average muscle activation value recorded from multiple training sessions, the slope parameter of the nonlinear mapping relationship used in the state inversion process is adaptively modified. Through this adaptive adjustment of parameters, the electromyographic feedback electrical stimulation control system can adapt to functional changes during the patient's rehabilitation process.

[0012] Furthermore, it also includes data backtracking and analysis steps after training is completed: After a single training session, retrieve all muscle activation value sequences and corresponding electrical stimulation intensity sequences from the entire training cycle. Align the muscle activation numerical sequence with the electrical stimulation intensity sequence on the time axis to create a two-dimensional trajectory diagram; Mark the time points where state inversion occurs and the corresponding activation abrupt change points on the two-dimensional trajectory graph; By analyzing the convergence speed and oscillation of the two-dimensional trajectory graph, an evaluation report on the effectiveness of this training is generated. This evaluation report is used to guide the formulation of the next rehabilitation training plan.

[0013] Furthermore, the present invention also includes an electromyographic feedback electrical stimulation control system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the electromyographic feedback electrical stimulation control method described above.

[0014] Compared with the prior art, the beneficial effects of the present invention are: Using muscle activation values ​​as the current state and the median of the target activation range as the desired state, the required electrical stimulation intensity is calculated through inversion. This allows the determination of electrical stimulation intensity to be directly based on the current muscle activation state and the target desired state. The value of electrical stimulation intensity is directly related to the real-time muscle activation level and the target activation level. The inversion calculation method can accurately match the difference between the current muscle state and the target state, making the determination of electrical stimulation intensity real-time adaptable. The output of electrical stimulation intensity can fit the real-time changes in muscle activation state, avoiding deviations between electrical stimulation intensity and actual muscle activation needs. This ensures that the setting of electrical stimulation intensity can accurately correspond to the target requirements of muscle activation, achieving direct linkage between muscle activation state, target desired state, and electrical stimulation intensity.

[0015] Based on the inverted calculated electrical stimulation intensity, the duty cycle and frequency parameters of the electrical stimulation waveform are configured to generate an electrical stimulation control signal. This allows the key parameters of the electrical stimulation waveform to be adjusted synchronously with the inverted intensity. The setting of the duty cycle and frequency parameters is directly based on the electrical stimulation intensity value, and the waveform parameters and stimulation intensity form a fixed adaptation relationship. This eliminates the problem of mismatch between fixed waveform parameter settings and stimulation intensity requirements. The generation of the electrical stimulation control signal can accurately correspond to the inverted calculated stimulation intensity, ensuring that the waveform parameters of the electrical stimulation output are consistent with the actual required stimulation intensity. The output form of the electrical stimulation can fit the parameter requirements corresponding to muscle activation, achieving coordinated adaptation between electrical stimulation intensity and waveform parameters. This ensures that the output parameter combination of the electrical stimulation control signal matches the actual needs of muscle activation. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of an electromyographic feedback electrical stimulation control method according to the present invention. Figure 2 A flowchart for filtering and rectifying raw surface electromyography signals; Figure 3 A flowchart for reasoning and calculating muscle activation values; Figure 4 This is a graph showing the changes in electromyographic signal amplitude and signal-to-noise ratio. Figure 5 A time-series comparison diagram of the entire electromyography signal processing process. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0018] See Figure 1 After system startup, the system acquires raw surface electromyographic (EMG) signals from the target muscle via surface electrodes. Preprocessing, including bandpass filtering and rectification, is performed on the acquired raw EMG signals, resulting in a clearer reflection of muscle electrical activity. Envelope extraction is then performed on the processed EMG signals to obtain a time-varying amplitude signal characterizing muscle contraction intensity. Based on this extracted amplitude signal, a pre-defined muscle activation calculation model is used to derive a quantified current muscle activation value between zero and one. This activation value is compared to a pre-defined target activation range specific to the rehabilitation goal, and the comparison result determines whether electrical stimulation intervention is necessary. When the muscle activation value falls below the lower limit of the target activation range, the system determines that electrical stimulation is required, initiates electrical stimulation output, and simultaneously initiates the state inversion phase of the control process. In the state inversion process, the system uses the currently calculated muscle activation value as the current state of the controlled object and the median of the target activation range as the desired state. The inversion control algorithm then calculates the electrical stimulation intensity required to bring the state from the current value to the desired value. Based on the inverted electrical stimulation intensity value, the system configures the specific parameters of the electrical stimulation waveform, including duty cycle and frequency, and then generates the final electrical stimulation control signal used to drive the stimulation device.

[0019] In one embodiment of the present invention, the original surface electromyography (EMG) signal is subjected to bandpass filtering and rectification to obtain a processed EMG signal. The process includes: (See reference) Figure 2 Based on the action potential characteristics of the target muscle's motor units, a lower cutoff frequency and an upper cutoff frequency are set to construct a bandpass filter. This bandpass filter is used to filter the original surface electromyography (EMG) signal to effectively remove power frequency interference and motion artifacts, resulting in a filtered EMG signal. A full-wave rectification operation is then performed on the filtered EMG signal, flipping the negative half-axis to the positive half-axis, thus obtaining a fully rectified EMG signal. The fully rectified EMG signal is then low-pass smoothed to eliminate high-frequency jitter components, ultimately outputting a smoothed EMG signal.

[0020] In practice, the raw surface electromyography (EMG) signal is bandpass filtered and rectified to obtain the processed EMG signal. This process includes: based on the action potential characteristics of the target muscle's motor units, a lower cutoff frequency and an upper cutoff frequency are set to construct a bandpass filter. The transfer function of the bandpass filter can be expressed as: in: Indicates the filter at frequency Complex frequency response at , It is the imaginary unit. It is the quality factor of the filter, used to control the width of the passband. This is the center frequency of the bandpass filter, and its value is determined by both the lower and upper cutoff frequencies. The constructed bandpass filter is used to filter the original surface electromyography (EMG) signal, removing power frequency interference and motion artifacts to obtain the filtered EMG signal.

[0021] In some embodiments, a full-wave rectification operation is performed on the filtered electromyography (EMG) signal, flipping the negative half-axis of the signal to the positive half-axis, thereby obtaining a fully rectified EMG signal. The full-wave rectification operation is implemented using an absolute value function, causing all sample points with values ​​less than zero in the original signal to become their opposites, while all sample points with values ​​greater than or equal to zero remain unchanged. The fully rectified EMG signal is then low-pass smoothed to eliminate high-frequency jitter components, ultimately outputting a smoothed EMG signal. The low-pass smoothing is implemented using a finite impulse response (FIR) filter, whose difference equation does not involve a feedback path and only performs a weighted average of the input signal. Optionally, the lower cutoff frequency of the bandpass filter is set to 20 Hz, and the upper cutoff frequency is set to 500 Hz; this frequency band effectively covers the main energy distribution of the surface EMG signal. Quality factor The value of is between 0.7 and 1.0 to strike a balance between passband selectivity and transition band steepness. The cutoff frequency for low-pass smoothing is set to 5 Hz to extract the envelope contour of the electromyographic signal while suppressing high-frequency noise components remaining after full-wave rectification.

[0022] It is understandable that full-wave rectification is a necessary preprocessing step for obtaining the signal envelope, converting the electromyographic signal containing positive and negative oscillations into a single-polarity signal. The finite impulse response filter for low-pass smoothing is chosen to be of order 50, and its unit impulse response coefficient is designed using a Hamming window function to obtain a smooth roll-off characteristic in the frequency domain, avoiding the Gibbs phenomenon. In some embodiments, the bandpass filtering and rectification processes are completed in real time in the embedded signal processor. The processed electromyographic signal is stored in a circular buffer at a fixed sampling rate for subsequent envelope extraction module reading. The signal processing flow adopts a modular design, with the bandpass filtering module, full-wave rectification module, and low-pass smoothing module cascaded sequentially. The output of each module serves as the input of the next module, and the data flow is unidirectional.

[0023] In one embodiment of the present invention, based on a time-varying amplitude signal, the current muscle activation value is calculated by inference using a preset muscle activation calculation model. The process includes: (See below) Figure 3 The time-varying amplitude signal is averaged using a sliding window to obtain the average amplitude within the corresponding time window. A baseline amplitude of the target muscle at rest is obtained, acquired during the system's pre-acquisition phase. The gain factor of the average amplitude relative to the baseline amplitude is calculated. This calculated gain factor is then substituted into a preset sigmoid function model, and the muscle activation value between zero and one is inferred using a lookup table method or piecewise linear interpolation. In specific implementation, based on the time-varying amplitude signal, the current muscle activation value is inferred and calculated using a preset muscle activation calculation model. This process includes: averaging the time-varying amplitude signal using a sliding window to obtain the average amplitude within the corresponding time window. The sliding window averaging calculation uses a fixed-length... A window of sampling points slides along the time-varying amplitude signal data stream, moving at intervals of one sampling point. The sum of the amplitudes of all sampling points within the window is divided by the window length. The output is the average amplitude at the center of the window. Obtain the baseline amplitude of the target muscle at rest. The signals are obtained through pre-acquisition, which requires continuously acquiring electromyographic signals for a period of time under conditions of complete relaxation and no voluntary contraction of the target muscle, and then calculating the average amplitude. The average amplitude is then calculated. Relative to the reference amplitude Gain factor The calculation relationship is Increase the gain factor. Substituting the values ​​into the pre-defined S-shaped function model, the muscle activation values ​​between zero and one are calculated using either a lookup table method or piecewise linear interpolation. The sigmoid function model is used to describe the nonlinear saturation relationship between the gain factor and muscle activation, and its expression is: in: This represents the calculated muscle activation value, with a value range of [value range missing]. ; It is the base of the natural logarithm; It is the kurtosis coefficient of the curve, which controls the transition rate of the function from 0 to 1; It is the input gain factor; It is the offset of the function, corresponding to the midpoint of the gain factor when the activation is 0.5.

[0024] In some embodiments, the length of the sliding window The corresponding time span is 250 milliseconds, which matches the physiological response time scale of human voluntary movement control. The pre-acquisition process lasts for 3 seconds, during which the system prompts the user to keep their muscles relaxed, and directly calculates the arithmetic mean of the acquired time-varying amplitude signal, storing the calculation result as the reference amplitude. Parameters of the S-shaped function model and Preset parameters based on the type of target muscle and individual user differences. The typical value range is 3.0 to 8.0, parameter The typical value range is from 1.5 to 4.0. Optionally, the lookup table method can be implemented by pre-calculating a mapping table based on the sigmoid function model, storing a series of discrete gain factors. Input values ​​and corresponding muscle activation levels Output value. During real-time calculations, the system will output the currently calculated gain factor. By comparing the discrete input values ​​in the mapping table, the muscle activation value is obtained either directly or through linear interpolation of neighboring points by finding the closest discrete point. Piecewise linear interpolation divides the complete S-curve into several continuous linear segments. Each segment is defined by a straight line equation with a start point and an end point. Real-time calculations are performed based on the gain factor. The section that falls into is calculated using the corresponding straight line equation.

[0025] It is understandable that sliding window averaging can smooth out instantaneous fluctuations in time-varying amplitude signals, yielding an amplitude that better represents the average level of muscle contraction over a period of time. (Baseline amplitude) The introduction of this method normalizes individual differences and variations in electrode placement, thus improving the accuracy of subsequent gain calculations. It becomes a dimensionless relative quantity, focusing more on reflecting the relative intensity changes of muscle activity. The nonlinear characteristics of the type function model make the gain factor... Lower and higher regions, muscle activation The changes are relatively gradual, but significant in the intermediate gain region, which aligns with the physiological response characteristics of muscles from rest to maximal voluntary contraction. In some embodiments, the muscle activation calculation is performed once per control cycle, the duration of which corresponds to the sliding window's step size. The average amplitude required for each calculation... All data are from the latest complete sliding window, with a baseline amplitude. The mapping table used by the lookup method or the segment parameters used by the piecewise linear interpolation method are determined at the beginning of a single training session and remain unchanged throughout the session. The mapping table used by the lookup method or the segment parameters used by the piecewise linear interpolation method are loaded from non-volatile memory into memory during system initialization to ensure the speed of real-time computation.

[0026] Considering that factors such as sweating and electrode micro-movements can easily cause baseline drift in the reference amplitude during actual long-term rehabilitation training, thus affecting the accuracy of subsequent activation calculations, a dynamic update mechanism for the reference amplitude and a slow baseline drift filtering logic are added throughout the training cycle. During training, at preset intervals (the preset interval can be set according to the duration of rehabilitation training and actual needs, such as 5-10 minutes), the system automatically triggers a resting state detection. If the current resting state meets the reference amplitude acquisition conditions of the pre-acquisition stage, the reference amplitude is updated. Simultaneously, the electromyography (EMG) signal processing module performs a slow baseline drift filtering operation in real time. Through a moving average filtering algorithm, the acquired EMG signals are smoothed to suppress interference caused by slow baseline drift, ensuring the stability of the reference amplitude and thus guaranteeing the accuracy of subsequent activation calculations. This dynamic update mechanism and drift filtering logic can be integrated into the EMG signal processing module and work in conjunction with the original signal acquisition and activation calculation process.

[0027] In one embodiment of the present invention, when the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and a state inversion process is entered. This process includes: defining an error variable equal to the median of the target activation range minus the current muscle activation value; determining whether the absolute value of the error variable is greater than a preset dead zone threshold; if the absolute value of the error variable is greater than the preset dead zone threshold, switching the current control mode to an active closed-loop adjustment mode, and executing the state inversion process. In the state inversion process, a nonlinear mapping relationship is established, including muscle activation, electrical stimulation intensity, and nerve excitation conduction rate. Based on this nonlinear mapping relationship, the electrical stimulation intensity increment required for the muscle activation to change from the current state to the desired state is calculated. The inverted electrical stimulation intensity value is read and mapped to a preset stimulation intensity level table. A basic pulse frequency and pulse width are selected according to the correspondence in the stimulation intensity level table. A muscle fatigue factor is introduced to correct the selected pulse width; the muscle fatigue factor is calculated based on the continuous working time of the electrical stimulation. The modified pulse width is combined with the selected pulse frequency, and clamping protection is performed according to a preset upper voltage limit to generate an electrical stimulation control signal. The generated electrical stimulation control signal is applied to the target muscle, and the time-varying amplitude signal during the electrical stimulation process is continuously monitored.

[0028] During electrical stimulation, the inference calculation and state inversion steps are executed cyclically. Based on the real-time time-varying amplitude signal, the muscle activation value is recalculated, and the electrical stimulation intensity is adjusted again until the muscle activation value enters and stabilizes within the target activation range. A fixed sampling period is set, and a time-varying amplitude signal is synchronously acquired at the beginning of each sampling period. Using the latest time-varying amplitude signal, the inference calculation step is executed again to update the current muscle activation value. The difference between the updated muscle activation value and the muscle activation value at the previous sampling time is calculated to obtain the activation change rate. The activation change rate is used as a constraint in the state inversion process to limit the adjustment range of the electrical stimulation intensity and prevent overshoot. During the electrical stimulation output, the fluctuation range of the muscle activation value is continuously monitored. When the muscle activation value first enters the target activation range, stabilization timing begins. If the muscle activation value remains within the target activation range for a continuous preset duration, it is determined that a stable state has been reached. Once a stable state is determined, the current electrical stimulation intensity parameters are locked, the state inversion process is stopped, and only basic electromyographic signal monitoring is retained.

[0029] In practical implementation, when the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and the state inversion process begins. This process includes: defining an error variable equal to the median of the target activation range minus the current muscle activation value; determining whether the absolute value of the error variable is greater than a preset dead zone threshold; if the absolute value of the error variable is greater than the preset dead zone threshold, switching the current control mode to active closed-loop regulation mode and executing the state inversion process; and establishing a nonlinear mapping relationship encompassing muscle activation, electrical stimulation intensity, and neural conduction rate; and calculating the electrical stimulation intensity increment required to change muscle activation from the current state to the desired state based on this nonlinear mapping relationship. The nonlinear mapping relationship describes the electrical stimulation intensity at a specific electrical stimulation intensity. Under input, the nerve excitation conduction rate With final muscle activation The dynamic coupling between them can be simplified as follows: in: Indicates muscle activation level Over time rate of change, It concerns the current level of muscle activation. and electrical stimulation intensity A function that characterizes the conduction rate of nerve excitation. It is the excitation gain coefficient. It is the attenuation coefficient. Solve for the increment of electrical stimulation intensity. The process involves combining the difference between the desired muscle activation and the current muscle activation, the rate of change of the current muscle activation, and the aforementioned mapping relationships to solve for the equation that will cause the system state to evolve in the desired direction. Value, and then calculate .

[0030] In some embodiments, the inverted electrical stimulation intensity value is read and mapped to a preset stimulation intensity level table. The stimulation intensity level table defines a discrete mapping relationship from the electrical stimulation intensity value to the basic pulse frequency and basic pulse width. Based on the correspondence in the stimulation intensity level table, a basic pulse frequency and pulse width are selected. A muscle fatigue factor is introduced to correct the selected pulse width; this muscle fatigue factor is calculated based on the continuous working time of the electrical stimulation. The corrected pulse width is combined with the selected pulse frequency, and clamping protection is performed according to a preset voltage upper limit to generate an electrical stimulation control signal. The generated electrical stimulation control signal is applied to the target muscle, and the time-varying amplitude signal during the electrical stimulation process is continuously monitored. During the electrical stimulation, the inference calculation and state inversion steps are executed cyclically, the muscle activation value is recalculated based on the real-time time-varying amplitude signal, and the electrical stimulation intensity is adjusted again until the muscle activation value enters and stabilizes within the target activation range. Optionally, the dead zone threshold is set to 0.05 to avoid frequent switching of the control mode when the muscle activation value approaches the boundary of the target range. The stimulus intensity rating scale contains 10 levels, with level 1 corresponding to the lowest baseline stimulus intensity and level 10 corresponding to the highest. (Muscle fatigue factor) Continuous working time The function is represented as ,in It is the fatigue factor; the corrected pulse width is equal to the base pulse width multiplied by the muscle fatigue factor. The voltage upper limit clamping protection ensures that the peak voltage of the generated electrical stimulation control signal does not exceed a safe limit, such as 30 volts.

[0031] It is understood that a fixed sampling period is set, and a time-varying amplitude signal is synchronously acquired at the beginning of each sampling period. Using the latest time-varying amplitude signal, the inference calculation step is re-executed to update the current muscle activation value. The difference between the updated muscle activation value and the muscle activation value at the previous sampling time is calculated to obtain the activation change rate. The activation change rate is used as a constraint in the state inversion process to limit the adjustment range of electrical stimulation intensity and prevent overshoot. During electrical stimulation output, the fluctuation range of the muscle activation value is continuously monitored. When the muscle activation value first enters the target activation range, stabilization timing begins. If the muscle activation value remains within the target activation range for a continuous preset duration, it is determined that a stable state has been reached. Once a stable state is determined, the current electrical stimulation intensity parameter is locked, the state inversion process is stopped, and only basic electromyography signal monitoring is retained. In some embodiments, the fixed sampling period is set to 100 milliseconds. The activation change rate is calculated by subtracting the muscle activation value of the previous sampling period from the current muscle activation value and then dividing by the sampling period duration. The preset stabilization time is set to 2 seconds, which requires the muscle activation value to remain within the target activation range for 2 consecutive seconds. After reaching a stable state, the electrical stimulation intensity parameters locked, including pulse frequency, corrected pulse width, and voltage amplitude, will remain unchanged in the subsequent maintenance stimulation phase, unless the muscle activation value falls below the lower limit of the target range again and triggers a new adjustment cycle.

[0032] In one embodiment of the present invention, the method further includes a safety exit mechanism for abnormal operating conditions. Before each inference calculation step, the signal-to-noise ratio (SNR) of the time-varying amplitude signal is first checked. If the SNR is lower than a preset safety threshold, the currently acquired electromyographic (EMG) signal is determined to be invalid. When the EMG signal is determined to be invalid, the electrical stimulation output is immediately paused, and the control mode is switched to manual mode. Simultaneously, an alarm is issued, awaiting operator intervention and inspection, until a valid EMG signal is reacquired before resuming automatic control. In a specific implementation, the EMG feedback electrical stimulation control method further includes a safety exit mechanism for abnormal operating conditions. Before each inference calculation step, the SNR of the time-varying amplitude signal is first checked. The SNR check is performed by segmenting the time-varying amplitude signal, extracting the latest segment containing... For each sampled data frame, calculate the signal-to-noise ratio (SNR) of that data frame. The calculation formula is defined as the ratio of signal power to noise power in decibels, and the expression is: in: Signal-to-noise ratio, expressed in decibels (dB); Represents logarithmic operations to base 10; This represents the power of the signal components that are considered valid. This represents the power identified as noise. Effective signal power. The mean square value is calculated by applying a bandpass filter that matches the characteristics of the electromyographic signal to the data frame. Noise power. The mean square value of the difference between the original data frame and the signal after filtering by the bandpass filter is obtained by calculating the mean square value. This difference mainly includes power frequency interference, motion artifacts and high-frequency thermal noise.

[0033] In some embodiments, if the calculated signal-to-noise ratio If the signal falls below a preset safety threshold, the currently acquired electromyographic signal is considered invalid. The safety threshold is preset based on the clinical application scenario and the typical electrical characteristics of the electrode-skin interface. The logic for determining signal validity is closely related to the preset safety threshold; different threshold settings correspond to different system sensitivity and anti-interference requirements. A possible threshold and response action correspondence is shown in Table 1.

[0034] Table 1: Correspondence between Signal-to-Noise Ratio Threshold and System Response Action In Table 1, This represents a high-confidence effective threshold. This represents the minimum acceptable threshold. When the electromyographic signal is determined to be invalid, the electrical stimulation output is immediately paused, and the control mode is switched from automatic closed-loop adjustment mode to manual mode. In manual mode, the system stops automatic adjustment based on electromyographic feedback, and all parameters of the electrical stimulation output are manually set and controlled by the operator through the human-machine interface. The system only maintains basic safety monitoring functions. Optional, high-confidence effective threshold. The typical value is 15dB, and the lowest acceptable threshold is... The typical value is 8dB. The length of the data frame... The corresponding time window is 500 milliseconds to ensure that the signal-to-noise ratio assessment is based on a signal of sufficient length to be statistically significant. The action of pausing the electrical stimulation output is instantaneous; the system stops generating pulse waves within the same control cycle when the signal is deemed invalid. Switching between control modes is achieved through state transitions in the internal state machine, from "automatic operation" to "manual standby."

[0035] It is understandable that issuing alarm prompts is an important part of the safe exit mechanism. Alarm prompts are implemented through various means, including visual, auditory, and tactile. Visual alarms are displayed on the user interface with a flashing red icon and the text message "Poor signal quality"; auditory alarms are emitted through a built-in speaker; and tactile alarms are triggered by a short vibration generated by a vibration motor on the control device's casing. Operators are required to check for proper contact between the electrodes and skin, whether the electrode pads are detached or dry, and whether the connecting wires are loose, and take appropriate measures. Automatic control is only restored after a valid electromyographic signal is re-acquired. The valid criterion is the signal-to-noise ratio calculated over three consecutive evaluation data frames. All are above the high-confidence effective threshold. Once this condition is met, the system automatically switches from "manual standby" to "automatic operation" and restarts the closed-loop control process based on the latest valid electromyographic signal. In some embodiments, the signal-to-noise ratio (SNR) verification module runs independently of the main control loop, evaluating the latest time-varying amplitude signal data at a fixed cycle. The evaluation result is input as a flag bit into the main control logic. Before executing the inference calculation step, the main control logic first reads this flag bit. If the flag bit indicates "invalid signal," it skips the current muscle activation calculation, state inversion, and electrical stimulation control signal generation steps and executes the safety exit procedure instead. The safety exit procedure includes immediately resetting the electrical stimulation intensity parameter to zero, generating an alarm event log, and updating the system status display. The alarm event log records the timestamp, duration, and final SNR value of the invalid signal event for subsequent analysis.

[0036] See Figure 4 This is a graph showing the changes in electromyographic (EMG) signal amplitude and signal-to-noise ratio (SNR), primarily illustrating the dynamic changes in time-varying amplitude signals and SNR. During the first 0–15 seconds, the signal amplitude stabilizes at around 2mV with relatively small fluctuations. From 15–35 seconds, the amplitude fluctuates dramatically, peaking at nearly 9mV, representing the main phase of active muscle contraction or external interference. After 35 seconds, the amplitude gradually decreases and stabilizes at around 2mV, returning to baseline levels. Regarding SNR changes, from 0–15 seconds, the SNR is at a low level with poor signal quality. From 15–38 seconds, the SNR significantly improves, achieving optimal signal quality, which highly coincides with the period of dramatic amplitude fluctuations. From 38–60 seconds, the SNR drops again to -18 to -12 dB, with signal quality declining, synchronized with the amplitude returning to baseline. The high SNR window of 15–38 seconds is the only interval approaching an effective signal and can be used as a reference period for control logic.

[0037] In one embodiment of the present invention, the method further includes a parameter adaptive adjustment step for different training stages. The maximum and average muscle activation values ​​are recorded during each training session. Based on the changing trend of the maximum muscle activation values ​​recorded across multiple training sessions, the upper and lower limits of the target activation range are dynamically adjusted. Based on the changing trend of the average muscle activation values ​​recorded across multiple training sessions, the slope parameter of the nonlinear mapping relationship used in the state inversion process is adaptively modified. The method also includes a data backtracking and analysis step after training. After a single training session, all muscle activation value sequences and corresponding electrical stimulation intensity sequences from the entire training cycle are retrieved. The muscle activation value sequences and electrical stimulation intensity sequences are aligned on the time axis and plotted as a two-dimensional trajectory graph. The time points where state inversion occurs and the corresponding activation mutation points are marked on the two-dimensional trajectory graph. By analyzing the convergence speed and oscillation of the two-dimensional trajectory graph, an evaluation report on the training effect is generated.

[0038] To prevent parameter abrupt changes in the differential equations upon which state inversion relies during a single training session, which could trigger control oscillations, a parameter disturbance rejection control logic is added during each training session. Specifically, during the parameter operation of the differential equations, amplitude limiting and smoothing adjustment logic are added. Amplitude limiting is used to set the value range of the excitation gain coefficient and attenuation coefficient. When a parameter abrupt change is detected and exceeds the preset value range, the parameter is immediately limited to the preset range. Smoothing adjustment uses an exponential smoothing algorithm to smooth the parameter change process, suppressing abrupt parameter fluctuations and ensuring the continuity and stability of parameter changes. This amplitude limiting and smoothing adjustment logic works in conjunction with the original parameter adaptive adjustment mechanism between multiple training sessions, ensuring parameter adaptability between multiple training sessions and avoiding control oscillations within a single training session, thus ensuring the stability of state inversion.

[0039] In practical implementation, the electromyographic feedback electrical stimulation control method also includes an adaptive parameter adjustment step for different training stages, recording the maximum and average muscle activation values ​​during each training session. The recording process occurs throughout the entire duration of a single training session. The system samples and records the current muscle activation values ​​at fixed time intervals. At the end of the training session, the maximum value is extracted from the recorded muscle activation value sequence as the maximum muscle activation value for this training session, and the arithmetic mean of the entire sequence is calculated as the average muscle activation value for this training session. These two values ​​are then associated with the date and duration of the training session and stored in the training history database. Based on the changing trend of the maximum muscle activation values ​​recorded across multiple training sessions, the upper and lower limits of the target activation range are dynamically adjusted. The adjustment logic relies on the analysis of historical maximum muscle activation value sequences. Assuming the most recent... The maximum muscle activation value during a training session is denoted as... The adjusted upper limit of the target activation range and lower limit value It can be calculated according to the following rules: in: It's recently The moving average of the maximum muscle activation value for each training session. It is the upper limit coefficient ( ), It is the lower limit coefficient ( Through this calculation, the target activation range will increase as the user's maximum shrinkage ability increases ( (Increase) and move upwards synchronously. Based on the changing trend of the average muscle activation value from multiple training records, the slope parameter of the nonlinear mapping relationship used in the state inversion process is adaptively modified. The specific form of the nonlinear mapping relationship may contain one or more slope parameters, such as parameters controlling response sensitivity. The adjustment method is as follows: calculate the most recent... Moving average of average muscle activation values ​​over training sessions and a preset reference value. Compare the results and proceed according to the preset step size. Increase or decrease the slope parameter The value of , if It shows an upward trend and is higher than Then it will decrease slightly. To make control smoother, if Showing a downward trend or remaining below for a long period of time Then it will increase slightly. To enhance control response.

[0040] In some embodiments, the number of historical training iterations used to calculate the moving average is... Set to 5 iterations to ensure adaptive adjustments are based on sufficient recent training performance. Upper limit coefficient. The typical value is 0.8, and the lower limit coefficient is... A typical value is 0.6, meaning the system encourages users to maintain activation levels between 60% and 80% of their recent maximum capabilities during training. (Reference value) It can be set to a fixed percentage of the midpoint of the target activation range from the previous stage, for example, 70%. Parameter adjustment step size. Set to a small value, such as 0.05, to ensure that parameter changes are gradual after each training session, avoiding drastic disturbances to the control system. The adaptive adjustment process is executed automatically before the start of each training session, with the system loading the most recent data from the training history database. Calculate the new target activation interval from each record. The updated slope parameters are then applied to the closed-loop control logic of this training session.

[0041] It is understandable that electromyographic feedback electrical stimulation control methods also include data retrospection and analysis steps after training. After a single training session, all muscle activation value sequences and corresponding electrical stimulation intensity sequences from the entire training cycle are retrieved. The muscle activation value sequence is a set of time-series data recorded at a fixed sampling period, while the electrical stimulation intensity sequence is a sequence of electrical stimulation intensity levels or mapped actual physical intensity values ​​recorded at the same timestamp. The muscle activation value sequence and the electrical stimulation intensity sequence are aligned on the time axis and plotted as a two-dimensional trajectory graph. The two-dimensional trajectory graph uses time as the horizontal axis and typically employs dual vertical axes. The left vertical axis represents the muscle activation value, and the right vertical axis represents the electrical stimulation intensity. The two trajectory lines are superimposed on the same time coordinate system and distinguished by different colors and line types.

[0042] Optionally, the moment of state inversion and the corresponding activation mutation point are marked on the two-dimensional trajectory graph. The moment of state inversion is obtained by querying the control log, which records the timestamp of each entry into the state inversion process. The corresponding activation mutation point refers to a local extreme point or inflection point near the moment of state inversion where the muscle activation value changes significantly, which can be identified by analyzing the first difference of the muscle activation value sequence. The marking method can be to use a vertical dashed line to mark the moment of state inversion and mark the activation mutation point with a special symbol at the corresponding position on the muscle activation trajectory line. An evaluation report on the training effect is generated by analyzing the convergence speed and oscillation of the two-dimensional trajectory graph. The convergence speed is evaluated by calculating the time from the first initiation of electrical stimulation to the first entry and stabilization of the muscle activation value within the target activation range. The oscillation is evaluated by analyzing the fluctuation amplitude and frequency of the muscle activation value after entering the target range. The fluctuation amplitude can be measured by calculating the standard deviation of the muscle activation value during the stable phase. In some embodiments, the two-dimensional trajectory graph, along with the annotation information and key indicators, automatically generates a structured evaluation report. The evaluation report is saved in document format and includes basic training information, key parameter settings, performance indicators, and thumbnail diagrams of the trajectory. The evaluation report guides the development of the next rehabilitation training plan. The therapist can review the report to understand the stability of muscle activation during the training, the response speed of system control, and whether there is over- or under-tuning. This information helps determine whether to maintain, increase, or decrease training goals for the next training session, or whether manual fine-tuning of control parameters is necessary.

[0043] See Figure 5 This is a time-series comparison diagram of the entire electromyography (EMG) signal processing flow, clearly showing the evolution from the original signal to the processed signal and then to the time-varying amplitude signal. Its core purpose is to verify the effectiveness of the signal preprocessing algorithm. The original signal contains numerous dense high-frequency spikes and alternating positive and negative fluctuations, including power frequency interference and motion artifacts. The processed signal, through bandpass filtering and full-wave rectification, completely eliminates the negative half-axis signal and smooths high-frequency noise, retaining only the effective components of the positive half-axis, laying the foundation for subsequent envelope extraction. The time-varying amplitude signal is calculated based on the processed red signal through low-pass smoothing and sliding window averaging. Its waveform completely follows the overall trend of the processed signal, but removes all high-frequency details, outputting a smooth curve that reflects the slow change in muscle contraction intensity over time. The signal is generally stable from 0-2 seconds, with low amplitude, corresponding to the muscle resting state. From 2-10 seconds, the signal fluctuation amplitude increases significantly, corresponding to the active muscle contraction phase, and the time-varying amplitude signal synchronously shows a periodic upward trend.

[0044] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for controlling electromyographic feedback electrical stimulation, characterized in that, include: The raw surface electromyographic (EMG) signal of the target muscle is acquired, and the raw surface EMG signal is subjected to bandpass filtering and rectification to obtain the processed EMG signal. Envelope extraction is performed on the processed electromyographic signal to obtain a time-varying amplitude signal that reflects the intensity of muscle contraction; Based on the time-varying amplitude signal, the current muscle activation value is calculated by reasoning through a preset muscle activation calculation model. The muscle activation values ​​are compared with a pre-set target activation range, and the necessity of electrical stimulation intervention is determined based on the comparison results. When the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and the state inversion process is initiated. In the state inversion process, the muscle activation value is taken as the current state, and the median value of the target activation range is taken as the desired state. The required electrical stimulation intensity is then calculated through inversion. Based on the inverted calculation of the electrical stimulation intensity, the duty cycle and frequency parameters of the electrical stimulation waveform are configured to generate an electrical stimulation control signal.

2. The electromyographic feedback electrical stimulation control method as described in claim 1, characterized in that, The original surface electromyography (EMG) signal is bandpass filtered and rectified to obtain the processed EMG signal, including: Based on the action potential characteristics of the motor unit of the target muscle, a lower cutoff frequency and an upper cutoff frequency are set to construct a bandpass filter. The original surface electromyography signal is filtered using the bandpass filter to remove power frequency interference and motion artifacts, resulting in a filtered electromyography signal. The filtered electromyographic signal is subjected to full-wave rectification to flip the signal on the negative half-axis to the positive half-axis, thus obtaining a full-wave rectified electromyographic signal. The full-wave rectified electromyographic signal is subjected to low-pass smoothing to eliminate high-frequency jitter in the signal, resulting in the processed electromyographic signal.

3. The electromyographic feedback electrical stimulation control method as described in claim 2, characterized in that, Based on the time-varying amplitude signal, the current muscle activation value is calculated using a preset muscle activation calculation model, including: The time-varying amplitude signal is averaged using a sliding window to obtain the average amplitude within the corresponding time window. Acquire the baseline amplitude of the target muscle in a resting state, which is obtained through pre-acquisition; Calculate the gain factor of the average amplitude relative to the reference amplitude; Substituting the gain factor into a preset S-shaped function model, the muscle activation value between zero and one is calculated by using a lookup table method or piecewise linear interpolation method.

4. The electromyographic feedback electrical stimulation control method as described in claim 3, characterized in that, When the muscle activation value is lower than the lower limit of the target activation range, electrical stimulation output is initiated, and the state inversion process begins, including: Define an error variable, which is equal to the median of the target activation interval minus the current muscle activation value; Determine whether the absolute value of the error variable is greater than a preset dead zone threshold; If the absolute value of the error variable is greater than the preset dead zone threshold, the current control mode will be switched to active closed-loop adjustment mode, and the state inversion process will be executed. In the state inversion process, a nonlinear mapping relationship is established that includes muscle activation, electrical stimulation intensity, and nerve excitation conduction rate; Based on the aforementioned nonlinear mapping relationship, the increment of electrical stimulation intensity required to change muscle activation from the current state to the desired state is calculated.

5. The electromyographic feedback electrical stimulation control method as described in claim 4, characterized in that, The step of configuring the duty cycle and frequency parameters of the electrical stimulation waveform based on the inverted calculated electrical stimulation intensity and generating an electrical stimulation control signal includes: Read the inverted and calculated electrical stimulation intensity value and map it to the preset stimulation intensity level table; Based on the correspondence in the stimulation intensity level table, a basic pulse frequency and pulse width are selected; A muscle fatigue factor is introduced to correct the selected pulse width, the muscle fatigue factor being calculated based on continuous working time; The modified pulse width is combined with the selected pulse frequency, and clamping protection is performed according to the preset upper voltage limit to generate the electrical stimulation control signal.

6. The electromyographic feedback electrical stimulation control method as described in claim 5, characterized in that, Also includes: The generated electrical stimulation control signal is applied to the target muscle, and the time-varying amplitude signal is continuously monitored during the electrical stimulation process; During electrical stimulation, the inference calculation and state inversion steps are executed cyclically. The muscle activation value is recalculated based on the real-time time-varying amplitude signal, and the electrical stimulation intensity is adjusted again until the muscle activation value enters and stabilizes within the target activation range. The inference calculation and state inversion steps are performed cyclically during electrical stimulation, including: A fixed sampling period is set, and at the beginning of each sampling period, the time-varying amplitude signal is synchronously acquired once. Using the latest time-varying amplitude signal, the inference calculation step is re-executed to update the current muscle activation value; The difference between the updated muscle activation value and the muscle activation value at the previous sampling time is calculated to obtain the activation change rate. The activation rate change is used as a constraint in the state inversion process to limit the adjustment range of electrical stimulation intensity and prevent overshoot. The process of until the muscle activation value enters and stabilizes within the target activation range includes: During the electrical stimulation output, the fluctuation range of the muscle activation value is continuously monitored; When the muscle activation value first enters the target activation range, a stable timing record begins. If the muscle activation value remains within the target activation range for a continuous preset duration, it is determined that a stable state has been reached. Once a stable state is determined, the current electrical stimulation intensity parameters are locked, the state inversion process is stopped, and only basic electromyographic signal monitoring is retained.

7. The electromyographic feedback electrical stimulation control method as described in claim 6, characterized in that, It also includes a safe exit mechanism under abnormal operating conditions: Before each inference calculation step, the signal-to-noise ratio of the time-varying amplitude signal is first verified. If the signal-to-noise ratio is lower than a preset safety threshold, the currently acquired electromyographic signal is determined to be invalid. When the electromyographic signal is determined to be invalid, the electrical stimulation output should be immediately paused and the control mode should be switched to manual mode. At the same time, an alarm is issued, and the system waits for the operator to intervene and check. Automatic control is only restored after a valid electromyographic signal is obtained again.

8. The electromyographic feedback electrical stimulation control method as described in claim 7, characterized in that, It also includes steps for adaptive parameter tuning for different training phases: Record the maximum and average muscle activation values ​​during each training session; Based on the changing trend of the maximum muscle activation value recorded in multiple training sessions, the upper and lower limits of the target activation range are dynamically adjusted. Based on the changing trend of the average muscle activation value recorded from multiple training sessions, the slope parameter of the nonlinear mapping relationship used in the state inversion process is adaptively modified. Through this adaptive adjustment of parameters, the electromyographic feedback electrical stimulation control system can adapt to functional changes during the patient's rehabilitation process.

9. The electromyographic feedback electrical stimulation control method as described in claim 8, characterized in that, It also includes data backtracking and analysis steps after training: After a single training session, retrieve all muscle activation value sequences and corresponding electrical stimulation intensity sequences from the entire training cycle. Align the muscle activation numerical sequence with the electrical stimulation intensity sequence on the time axis to create a two-dimensional trajectory diagram; Mark the time points where state inversion occurs and the corresponding activation abrupt change points on the two-dimensional trajectory graph; By analyzing the convergence speed and oscillation of the two-dimensional trajectory graph, an evaluation report on the effectiveness of this training is generated. This evaluation report is used to guide the formulation of the next rehabilitation training plan.

10. An electromyographic feedback electrical stimulation control system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the electromyographic feedback electrical stimulation control method as described in any one of claims 1 to 9.

Images