Pelvic and abdominal muscle group cooperative training adaptive control system based on biological feedback

Abstract

The invention relates to the technical field of automatic control, and discloses a pelvic and abdominal muscle group cooperative training self-adaptive control system based on biological feedback, which comprises a state drift observation module, a dynamic threshold value resolving module and a self-adaptive control module, and is characterized in that the system monitors the amplitude attenuation rate of a main constraint variable in real time to update an attenuation coefficient; constructing a dynamic safety envelope containing slew rate limitation and amplitude saturation; when the first-order time derivative of the disturbance variable exceeds the limit, the controller locks the output slope of the disturbance variable, and dynamic oscillation of the time-varying coupling system caused by high-frequency pulse disturbance is suppressed by establishing a dynamic mapping mechanism of the disturbance loading rate and the main variable response bandwidth. The adaptive compensation of the parameter drift of the controlled object and the steady-state maintenance of the control precision are realized.

Description

Technical Field

[0001] This invention relates to an adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback, belonging to the field of automatic control technology. Background Technology

[0002] Current mainstream technical solutions typically employ a single-loop independent control strategy to apply excitation or monitor biological loads. The controller treats the pelvic floor muscles and abdominal muscles as independent single-input single-output subsystems, generating control commands solely based on the deviation between the real-time acquired signal amplitude and a preset static threshold. Based on the assumption of a linear time-invariant system, the control logic ignores the inherent strong coupling characteristics and time-varying parameters of the biological muscle groups as the controlled object. In actual working conditions, the contraction of the abdominal muscles constitutes a non-stationary load disturbance to the pelvic floor muscles. The viscoelasticity and metabolic fatigue characteristics of biological soft tissues lead to physical lag and nonlinear attenuation of gain in the system response.

[0003] Although the industry has introduced multi-muscle group collaborative monitoring, there is a lag in the conversion of monitoring data into real-time control strategies. For example, Chinese invention patent CN115177274A discloses a device for assessing the coordination and fatigue susceptibility of key and synergistic muscle groups of the pelvic floor muscles. By calculating the activation time difference of multi-channel electromyography signals and the slope of the median frequency linear fitting, it achieves quantitative characterization of muscle coordination and fatigue. However, the technology is still limited to offline diagnosis or status monitoring. The core logic outputs an assessment report for physicians' reference, but it has not built a real-time feedback control closed loop based on fatigue status. When the device detects that pelvic floor muscle fatigue leads to a deterioration of frequency domain indicators or sudden high-frequency disturbances in abdominal muscle groups, such as coughing or sneezing, it cannot automatically reduce the excitation intensity or lock the output slope within a millisecond control cycle. The open-loop mode of diagnosis without control is prone to biological soft tissue overload damage when the physical properties of the controlled object drift rapidly due to the inability of control parameters to adapt in time, making it difficult to maintain steady-state stability in the dynamic training process.

[0004] Therefore, how to construct an adaptive control strategy in the time domain with disturbance feedforward decoupling, transient slew rate limiting and historical state memory capabilities, and maintain the global dynamic stability of the nonlinear time-varying coupled system, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: an adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback, comprising:

[0006] The multi-channel signal acquisition module is used to simultaneously acquire the first state signal representing the main constraint variable and the second state signal representing the disturbance variable. The first state signal corresponds to the real-time contraction voltage value of the pelvic floor muscle group, and the second state signal corresponds to the real-time contraction voltage value of the abdominal muscle group.

[0007] The state drift observation module is connected to the multi-channel signal acquisition module. It is used to monitor the amplitude attenuation rate of the first state signal when it is in the steady state maintenance period, and update the current real-time attenuation coefficient of the system according to the amplitude attenuation rate. The real-time attenuation coefficient is used to characterize the time-varying parameter drift characteristics of the controlled object.

[0008] The dynamic threshold calculation module, connected to the state drift observation module, is used to construct a dynamic safety envelope for the disturbance variable based on the real-time amplitude and real-time attenuation coefficient of the first state signal. The dynamic safety envelope includes the amplitude saturation threshold and slew rate limit threshold allowed by the system at the current moment.

[0009] An adaptive controller, connected to both the dynamic threshold calculation module and the multi-channel signal acquisition module, executes dual-constraint control logic: when the first-order time derivative of the second state signal exceeds the slew rate limit threshold, a slope locking command is output to forcibly lock the rise rate of the control output against the disturbance variable at the slew rate limit threshold; when the real-time amplitude of the second state signal exceeds the amplitude saturation threshold, a gain clamping command is output to limit the amplitude of the control output against the disturbance variable; wherein, the slew rate limit threshold and the real-time amplitude of the first state signal are positively correlated, so that the loading rate of the disturbance variable is limited by the current response bandwidth of the main constraint variable.

[0010] Preferably, the logic for updating the real-time attenuation coefficient by the state drift observation module is as follows: setting a drift judgment threshold that characterizes the steady-state error tolerance of the system; continuously calculating the negative change of the first state signal per unit time; when the absolute value of the negative change continuously exceeds the drift judgment threshold, performing a decrement operation on the real-time attenuation coefficient according to a preset step amount until the negative change of the first state signal falls back to within the drift judgment threshold.

[0011] Preferably, the slew rate limiting logic within the adaptive controller includes a nonlinear filter. The nonlinear filter is used to reshape the input step control signal into a ramp signal when the first-order time derivative of the second state signal exceeds the slew rate limiting threshold. The slope value of the ramp signal is dynamically adjusted in real time following the slew rate limiting threshold.

[0012] Preferably, the dynamic threshold calculation module uses the following formula to calculate the amplitude saturation threshold: ,in, The amplitude saturation threshold of the second-state signal allowed at the current moment. This is the preset system safety coupling gain constant. This represents the real-time amplitude of the first-state signal. This is the real-time attenuation coefficient output by the state drift observation module.

[0013] Preferably, the system further includes a cumulative load assessment module, which is used to perform a weighted integral operation on the signal amplitude and residence time to generate a cumulative load value when the amplitude of the second state signal is within a preset critical range of the amplitude saturation threshold. The cumulative load value is used to characterize the degree of fatigue accumulation of the system under critical high load operating conditions.

[0014] Preferably, the adaptive controller further includes a hysteresis recovery logic unit, which performs the following adjustments: when the accumulated load value exceeds a preset fatigue upper limit, a forced derating action is triggered, and the amplitude saturation threshold is lowered to a preset recovery level; when the accumulated load value naturally decays back to a preset reset lower limit, the forced derating action is released; the reset lower limit is lower than the fatigue upper limit, thereby forming a state recovery hysteresis loop in the control logic.

[0015] Preferably, the multi-channel signal acquisition module includes a differential signal preprocessing unit, which performs absolute value detection and moving average filtering on the acquired raw signal to extract the extremely low frequency DC component as the first state signal and the second state signal.

[0016] Preferably, the adaptive controller is used to perform a deviation-based feedback regulation: when the output slope lock command is received, the deviation between the actual output slope and the slew rate limit threshold is calculated; using the deviation as a feedback quantity, the drive signal for the disturbance variable is corrected through a proportional-integral regulation algorithm to suppress the dynamic overshoot of the system response.

[0017] Preferably, the dynamic threshold calculation module is also used to set the slew rate limit threshold to zero when the real-time amplitude of the first state signal is lower than the preset minimum operating point of the system, thereby cutting off the control output for the disturbance variable.

[0018] Preferably, the system includes a first acquisition probe, a second acquisition probe, and a central processing unit; the first acquisition probe and the second acquisition probe are respectively connected to the analog input port of the central processing unit; the central processing unit integrates an execution state drift observation module, a dynamic threshold calculation module, and an arithmetic circuit for adaptive controller logic.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. In the pelvic floor muscle group coordinated training of biofeedback, a dynamic mapping mechanism between the state of the first controlled object and the second disturbance variable is constructed to solve the problem of open-loop instability caused by the load disturbance amplitude exceeding the maintenance gain in a multivariable coupled system. The state drift observation unit is used to monitor the amplitude decay rate of the first time domain signal during the maintenance period in real time, which is used as the feedback quantity to dynamically reconstruct the saturation threshold of the second time domain signal. Based on the real-time state observation dynamic gain clamping logic, the amplitude of the disturbance variable input is locked within the physical carrying capacity range of the system at the current moment. There is no need to introduce high-computing frequency domain analysis, which ensures that the control loop operates within the dynamic safety manifold defined by the master variable when the parameters of the controlled object are time-varying or decaying, and eliminates the risk of system divergence caused by the mismatch between the control command and the actual capacity of the object.

[0021] 2. By introducing state-coupled slew rate constraint logic, the static amplitude control strategy overcomes the bandwidth lag defect in response to transient pulse disturbances. When the first-order rate of change of the second time domain signal exceeds the allowable value calculated based on the first time domain signal, the slope limit of the control command is forcibly applied, reshaping the step-change signal into a ramp signal. The differential control principle is used to match the loading rate of the disturbance variable with the physical response speed of the controlled object, suppressing the dynamic oscillation of the system caused by the excessively steep rise of the signal. This enables the control system to maintain a smooth and stable transient process when facing high-frequency sudden disturbances, and makes up for the time dimension control blind spot of the traditional threshold comparison method.

[0022] 3. By utilizing cumulative load integral and asymmetric hysteresis recovery strategies, the implicit state drift problem of the system under critical high load operation is solved. The signal in the critical interval is time-weighted integral to construct a virtual thermal capacity variable that reflects the historical load history. Based on this, derating control with recovery dead zone is executed. When the cumulative integral value falls below the trigger line reset threshold, the restriction is lifted, forming a forced state recovery hysteresis loop in the control logic. Based on the historical data variable structure control strategy, the system's cumulative error memory and self-adjustment capability are improved, avoiding the cumulative decay of the controlled object's performance over time, which could lead to sudden functional collapse, and ensuring the steady-state stability of the system under long-term continuous operation conditions. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the closed-loop feedback control architecture and module interaction principle of the present invention.

[0024] Figure 2 This is a comparison chart of fatigue recognition accuracy and control overshoot under different setting conditions of the present invention;

[0025] Figure 3 This is a flowchart illustrating the signal monitoring and parameter update logic of the state drift observation module of the present invention. Detailed Implementation

[0026] This detailed description aims to provide a clear and complete description of the technical solution of the present invention. 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.

[0027] An adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback includes a multi-channel signal acquisition module, a state drift observation module, a dynamic threshold calculation module, and an adaptive controller. The system dynamically reconstructs the amplitude saturation and slew rate limit boundaries for the second state signal by real-time monitoring of the amplitude attenuation characteristics of the first state signal, and locks the physical characteristics of the control output when the disturbance variable exceeds the limit. The multi-channel signal acquisition module is connected to a first acquisition probe and a second acquisition probe via hardware interfaces. The first acquisition probe is used to acquire real-time electromyographic signals of the pelvic floor muscles, and the second acquisition probe is used to acquire real-time electromyographic signals of the abdominal muscles. The differential signal preprocessing unit integrated within the module executes a signal conditioning procedure. It uses an instrumentation amplifier to amplify the weak input differential signal, performs absolute value detection processing to convert the bipolar AC signal into a unipolar signal, and extracts the extremely low-frequency DC component of the signal using a moving average filtering algorithm. The moving average filtering algorithm sets the sampling window length. For 20 to 50 sampling points, corresponding to a time window of 20ms to 50ms at a sampling rate of 1kHz, the arithmetic mean of the sampled data within this window is calculated. The DC component output after the above processing constitutes the first state signal representing the main constraint variable and the second state signal representing the disturbance variable, respectively.

[0028] The state drift observation module is connected to the multi-channel signal acquisition module to quantify the time-varying parameter drift characteristics of the controlled object. This module executes real-time update logic. The system presets a drift judgment threshold, which defines the upper limit of the allowable natural fluctuation range of the first state signal during the steady-state maintenance period. When the first state signal is in a non-zero steady-state maintenance phase, this module continuously calculates the rate of change of the first state signal per unit time. If the absolute value of the calculated negative change is continuous... If a sampling period exceeds the drift detection threshold, the system determines that the controlled object has experienced fatigue drift and executes the steady-state maintenance period effectiveness filtering procedure: [Construct depth] The circular FIFO buffer stores the first state signal sequence. Real-time variance of FPGA pipelined parallel computing buffer with the mean Set the effective contraction level. and maximum permissible variance The state drift observation module has built-in digital comparison logic, which compares the mean... With effective contraction level Compare and simultaneously incorporate variance. With maximum permissible variance Based on the judgment result, perform the following operation: only if the condition is met. and The system only sets the flag when the condition is met. The clock signal is used to calculate the conduction attenuation rate; during other periods, the change rate calculation register is forcibly reset by hardware, physically blocking the entry of non-steady-state data from the rising and releasing edges of muscle force into the parameter update loop, thus avoiding dynamic noise affecting the parameter update loop. In case of a false trigger, the state drift observation module adjusts the real-time attenuation coefficient according to a preset linear step size. Perform a decrementing operation until the negative change in the first state signal falls back below the drift detection threshold, and the real-time attenuation coefficient is adjusted accordingly. The initial value is set to 1.0, and its value range is limited to 0 to 1.

[0029] The dynamic threshold calculation module calculates the threshold based on the real-time amplitude of the first state signal. and real-time attenuation coefficient Calculate the dynamic safety envelope for the second state signal, which includes an amplitude saturation threshold. and slew rate limit threshold Among them, amplitude saturation threshold Follow the formula Calculations are performed, in which, The system safety coupling gain constant is determined through offline calibration experiments. Specifically, when the controlled object is in a non-fatigue state, the ratio of the maximum allowable amplitude of the second state signal to the amplitude of the first state signal is measured to maintain the stability of the first state signal. This represents the current real-time voltage amplitude of the first state signal. The real-time decay coefficient at the current moment, and the slew rate limit threshold. Through linear mapping relationship Confirmed, among which The rate coupling coefficient is used to define the maximum rate of change of the disturbance variable that can be tolerated per unit amplitude of the main variable, when the real-time amplitude of the first state signal is detected. When the threshold is lower than the preset minimum operating point of the system, the dynamic threshold calculation module will... Set to zero.

[0030] The adaptive controller connects to the dynamic threshold calculation module and the multi-channel signal acquisition module to perform dual-constraint control on the second state signal. This controller calculates the first-order time derivative of the second state signal. ,when Greater than the slew rate limit threshold At that time, the nonlinear filter in the controller triggers a slope locking action, which discards the actual sampled value of the second state signal at the current moment and outputs a value superimposed on the output value at the previous moment. The reconstructed value will lock the rate of increase of the control output for the disturbance variable at [value]. Low-level microinstruction-level slope circuit breaker logic: In In the servo interrupt service routine (ISR), the CPU reads the current disturbance input register value. Output buffer from the previous cycle Call the ALU to calculate the step difference. The comparator determines if... Jump to the safe branch and force write the clamp value to the DAC data bus. Synchronous set integral separation flag bit In the current instruction cycle, a no-operation (NOP) is used to replace the regular PID calculation instruction, physically preventing system divergence caused by integral saturation. The adaptive controller calculates the actual output slope and... The deviation between the two states is input to the proportional-integral control unit, which corrects the drive signal through integral calculation to eliminate steady-state error. When the real-time amplitude of the second state signal exceeds the amplitude saturation threshold... At this time, the controller outputs a gain clamping command to limit the control output amplitude to no more than .

[0031] The system also includes a cumulative load assessment module and a hysteresis recovery logic unit to handle critical high load conditions. The cumulative load assessment module monitors the amplitude of the second state signal, and when the amplitude is at the amplitude saturation threshold... When the signal amplitude is within 90% to 100% of its range, a weighted integral calculation of the signal amplitude and residence time is initiated to generate the cumulative load value. The hysteresis recovery logic unit will Compared to the preset fatigue limit When comparing, Exceed When this is triggered, a forced derating action is taken, reducing the amplitude saturation threshold. Lowered to the preset recovery level, when After natural decay, it falls back to the preset reset lower limit. At that time, the delayed recovery logic unit releases the forced derating action, in which, The value is lower than This forms a state recovery hysteresis loop in the control logic.

[0032] Example 1: In a scenario involving pelvic floor muscle strengthening training for patients with stress urinary incontinence, the system operates in the latter half of a long-duration contraction task. Due to the decreased neuromuscular transmission efficiency caused by lactic acid accumulation, the first state signal characterizing the contraction intensity of the pelvic floor muscles exhibits a continuous amplitude decay trend. The state drift observation module monitors that the absolute value of the negative rate of change of this signal stabilizes within 10 consecutive sampling cycles (10ms). The level is such that this value exceeds the system's preset drift detection threshold. This module uses a linear stepping procedure to calculate the real-time attenuation coefficient. The value was lowered from the initial value of 1.0 to 0.82 to represent the decrease in the actual bearing capacity of the controlled object. The dynamic threshold calculation module then calculated the value according to the formula. and Lower the amplitude saturation threshold for abdominal muscle groups With slew rate limit threshold Thus, a safe envelope for perturbations is allowed based on the fatigue state contraction of the main variables.

[0033] When the controlled object coughs, causing abdominal muscle contraction and generating a high-amplitude step-type second state signal, the adaptive controller calculates its first-order time derivative. Greater than the currently reduced slew rate limit threshold The controller triggers the slew rate locking logic, discarding the original sampled value at the current moment and outputting the slope locked at... The linear climbing signal reshapes the high-frequency perturbation into a ramp signal that conforms to the current hysteresis response characteristics of the pelvic floor muscles. As the amplitude of the reconstructed signal increases and reaches a certain level, the signal becomes more responsive. Corrected amplitude saturation threshold Gain clamping logic intervenes and limits the amplitude of the control output. This process establishes a dynamic mapping between the disturbance loading rate and amplitude and the real-time response capability of the main variable, ensuring that the control output is always maintained within the physical safety boundary of the pelvic floor muscles under the condition that the parameters of the controlled object decay over time and face external high-frequency disturbances.

[0034] Example 2: This example aims to verify the steady-state maintenance capability and anti-disturbance performance of the aforementioned biofeedback-based adaptive control system for pelvic and abdominal muscle synergistic training under long-cycle dynamic loads. To realistically simulate the complex working conditions in clinical rehabilitation training, a hardware-in-the-loop (HIL) test platform was constructed, including a physical simulation model and a real-time signal processing unit. A high-precision servo motor was used to simulate the dynamic contraction behavior of the pelvic floor and abdominal muscles. The pelvic floor muscle model was set as a viscoelastic load with time-varying stiffness characteristics, and its output voltage exhibited nonlinear fatigue decay over time. The abdominal muscle model was programmed to generate random high-frequency pulse sequences to simulate sudden abdominal pressure disturbances such as coughing and sneezing. During the experiment, the system sampling frequency was set to 1kHz to ensure aliasing-free acquisition of rapidly changing electromyographic signals; the drift judgment threshold was set to... The value was selected based on statistical analysis of a large amount of real patient electromyography data, aiming to balance the system’s sensitivity to fatigue state with the misjudgment rate.

[0035] After the experiment started, a baseline test was conducted. The system operated under ideal conditions with no external disturbances and no fatigue of the controlled object. The steady-state output waveform was recorded as a reference. Simulated fatigue decay was introduced, causing the amplitude of the first-state signal to decrease linearly at a preset rate. At the same time, random high-frequency disturbance signals were injected. During the 30-minute test window, the system monitored the rate of change of the first-state signal in real time. When the absolute value of the negative rate of change of the signal continuously exceeded the drift judgment threshold, the state drift observation module activated the parameter update mechanism and dynamically adjusted the real-time decay coefficient. Meanwhile, the adaptive controller updates in real time. In sync with the amplitude of the first-state signal, the amplitude saturation threshold of the second-state signal is adjusted. and slew rate limit threshold To quantify the control effect of the system, Table 1 lists the key performance indicators at different operating conditions.

[0036] Table 1: Comparison of Key Performance Indicators of the System under Different Operating Conditions

[0037]

[0038] Referring to Table 1, the data clearly demonstrates the system's adaptive adjustment process in response to parameter drift of the controlled object and strong external disturbances. In the early stage of fatigue, when the attenuation rate of the first-state signal changes from... Rise to The system responded quickly and... Lowered to 0.92, accordingly, Depend on Reduce to As fatigue progresses into the middle and late stages, the attenuation rate further increases... , The depth was reduced to 0.55. Then tightened to Especially during the strong disturbance phase, although the peak slew rate of the input disturbance variable is as high as It far exceeded the limit at the time. However, thanks to the intervention of the slew rate lock-in logic, the system's control output overshoot is only 4.2%, and the steady-state error is maintained at... At extremely low levels, even when the physical capabilities of the controlled object decline and it faces external shocks far exceeding its current bearing capacity, the system can still suppress dynamic oscillations by dynamically shrinking the safety envelope and shaping control commands.

[0039] Example 3: This example combines Figures 1 to 3 The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback is described, such as... Figure 1 As shown, at the physical level, the system consists of a first acquisition probe as a signal source for the pelvic floor muscles and a second acquisition probe as a signal source for the abdominal muscles. Both are connected to a multi-channel signal acquisition module for synchronous acquisition / preprocessing to output a first state signal and a second state signal. At the logical level, the first state signal is transmitted to a state drift observation module to monitor the amplitude attenuation rate and update the real-time attenuation coefficient. This real-time attenuation coefficient is input to a dynamic threshold calculation module. This module performs calculations based on the amplitude of the first state signal to construct a dynamic safety envelope, and then calculates the amplitude saturation threshold / slew rate limit threshold. The aforementioned safety envelope threshold and the second state signal are simultaneously input to an adaptive controller. The controller generates a drive signal for the disturbance variable as a control output command based on dual-constraint control logic, including slope locking / gain clamping, thereby achieving adaptive adjustment of the system.

[0040] like Figure 2 As shown, the horizontal axis represents different setting conditions, and the vertical axis represents percentages (%). The data indicates that under the oversensitive region setting, the system's fatigue identification accuracy (%) is low and accompanied by a certain control overshoot (%). Under the sluggish region setting, although the control overshoot increases, the fatigue identification accuracy drops significantly. Only under the calibrated setting conditions does the system achieve an extremely high fatigue identification accuracy exceeding 95%. Figure 3 As shown in the diagram, this illustrates the detailed interaction flow of the state drift observation module: The process begins with the input of the first state signal. During the steady-state maintenance signal monitoring phase, this signal is processed by the rate of change monitoring unit. By inputting real-time amplitude data and calculating the rate of change per unit time, the absolute value of the negative change is transmitted to the subsequent stage. The drift judgment logic then performs drift judgment and parameter adjustment, comparing the change with the drift judgment threshold. If the change is determined to be greater than the drift judgment threshold and confirmed after M consecutive cycles, the attenuation coefficient is triggered to decrease. At this time, the coefficient update unit performs a preset step decrease operation and outputs the updated coefficient through the attenuation coefficient output port. (t), where (t) Range: 0~1 and initial value: 1.0. If the change is less than or equal to the drift threshold, then maintain the current value. (t) The system remains unchanged and then enters the next sampling period for cyclic monitoring.

[0041] Example 4: This example aims to address the key parameter in the aforementioned state drift observation module—the drift determination threshold. and its update step size The determination process is supplemented with information to ensure its sufficiency and transparency. By constructing standardized offline calibration experimental procedures, the black box nature of parameter setting is eliminated, and its impact on the accuracy of system fatigue identification is quantitatively verified. In practical applications, This determines the system's sensitivity to the fatigue state of the controlled object, and The adaptive adjustment rate of the control parameters is determined, and together they constitute the core parameter set ensuring the dynamic stability of the system. This is obtained through a calibration process based on objective physiological data. The calibration environment is constructed and the initial state is defined. A group of subjects clinically diagnosed with pelvic floor muscle strength grade 3 (Oxford muscle strength classification) are selected. A multi-channel signal acquisition module is connected, and the subjects are kept in a supine resting position. At this time, the baseline noise amplitude of both the first-state signal (pelvic floor muscles) and the second-state signal (abdominal muscles) should be confirmed to be below [value missing]. Perform a maximal fatigue induced experiment: Instruct subjects to maintain pelvic floor muscle strength at 60% of maximum voluntary contraction (MVC) until they can no longer maintain this strength (defined as an amplitude drop exceeding 20% ​​of the initial value). Time-series data of the first-state signal are recorded throughout the experiment. Based on this experimental data, the following procedures are performed: and The parameters were calculated and the boundary verification was performed. The acquired first-state signal was differentiated to extract the average negative rate of change during the fatigue stage, from the beginning of the monotonous decrease in amplitude to the end of the test. To avoid misjudgments caused by random noise, a drift detection threshold is set. Simultaneously calculate the total duration of the fatigue stage. and the corresponding total attenuation ratio For example, 0.2, then the single-step adjustment amount Defined as ,in To verify the effectiveness of this calibration procedure, Table 2 shows system response data of a typical group of subjects under different parameter settings, with the system parameter update frequency set to 10Hz.

[0042] Table 2: Comparison of Fatigue Identification and Control Effects under Different Drift Judgment Parameter Settings

[0043]

[0044] See Table 2, where and Based on the calibration values ​​calculated according to the above procedures, the data shows that when Setting too low (overly sensitive area) and When the overshoot is too large, the system may misinterpret the normal physiological fluctuations of the controlled object as fatigue drift, leading to frequent and large fluctuations in control parameters and causing an 8.5% control overshoot; when Setting too high (sluggish zone) and When the value is too small, the system lags in recognizing actual fatigue, resulting in untimely intervention of the protection mechanism and a control overshoot of up to 12.4%. Only when the parameter is set to the calibration value can the system achieve a high fatigue recognition accuracy of 96.8% and control the control overshoot at an extremely low level of 2.3%.

[0045] Example 5: In a system initialization deployment scenario involving the intervention of a new individual, to ensure the universality and safety of the control logic, this example describes a pre-deployment calibration procedure. After the system connects to the subject, it executes a resting baseline calibration procedure. While the subject is in a supine, relaxed state, it continuously collects first and second state signals for 30 seconds. The system calculates the mean and standard deviation of the signal amplitude within this time window, and defines the mean plus three times the standard deviation as the system's background noise threshold. When the real-time signal amplitude of any channel is lower than this When an input is deemed invalid, the system sets the control output to zero to prevent environmental electromagnetic interference from falsely triggering the control logic. The system guides the subject to perform three standardized maximum voluntary contractions (MVC), recording the peak value of the first state signal during each contraction. The arithmetic mean of the three peak values ​​is defined as the subject's maximum muscle strength reference value. This reference value is used as a normalization factor to convert all subsequent control thresholds, including drift determination thresholds and amplitude saturation thresholds, into percentages relative to individual capabilities, ensuring adaptive matching of system parameters among individuals with different muscle strength levels.

[0046] To address the potential for sensor contact impedance changes during long-term system operation, the embodiment includes an offline calibration and data filling procedure for constructing an electrode contact quality compensation model. This procedure is performed in a controlled laboratory environment, utilizing a standard simulated skin and a precision impedance analyzer to simulate changes in sensor contact impedance over extended periods. to The system measures the range of contact impedance variation. At each discrete impedance point, a standard sinusoidal signal with known amplitude and frequency is input, and the actual attenuation amplitude at the system acquisition end is recorded. Based on these measured data, a mapping function relationship between contact impedance and signal gain compensation coefficient is established by fitting using the least squares method. This function is then discretized into a lookup table and stored in the controller's non-volatile memory. In actual operation, the system periodically injects a weak high-frequency test current to calculate the current electrode contact impedance value in real time and retrieves the corresponding gain compensation coefficient according to the lookup table to perform real-time amplitude correction on the first and second state signals. This mechanism eliminates signal drift caused by electrode aging, sweat secretion, or skin laxity, ensuring the objectivity and authenticity of the input data on which state observation and threshold calculation depend, and maintaining the control accuracy of the system throughout its entire life cycle.

[0047] Example 6: This example addresses the sliding window length of the moving average filtering algorithm in the aforementioned multi-channel signal acquisition module. The process of determining the optimal operating point involves constructing a standardized engineering calibration procedure. The setting of this parameter directly determines the trade-off between the ability to suppress high-frequency electromyography artifacts and the real-time tracking of rapidly changing muscle force signals. The optimal operating point is determined through quantitative experiments, and a calibration platform including a precision function generator and a data acquisition card is built. The function generator generates a simulated electromyography signal with superimposed noise of a specific frequency, which is input to the system's acquisition probe. This simulated signal consists of two parts: a low-frequency sine wave with a frequency of 0.5Hz as the reference, simulating the normal muscle contraction envelope; and superimposed high-frequency Gaussian white noise with a bandwidth limited to 50Hz to 500Hz, simulating electromyography artifacts.

[0048] With the system sampling frequency fixed at 1kHz, the sliding window length is set. Starting from 5, scan tests are performed with increments of 5 to 100. For each... The system records the processed output signal and calculates two key metrics: signal-to-noise ratio improvement. With phase delay time , Defined as the difference between the signal-to-noise ratio of the output signal and the signal-to-noise ratio of the input signal, it characterizes the filtering effect; Defined as the time lag of the output sine wave relative to the input sine wave, it characterizes real-time performance. Based on measured data, it is plotted... and Follow The changing bi-coordinate curve, analysis of the curve reveals that, Follow The increase shows a non-linear trend of first rising rapidly and then leveling off. A clear inflection point in returns appears nearby, after which increasing the window length contributes less to noise suppression. Follow It grows linearly, when When the delay exceeds 50ms, it begins to affect the closed-loop stability of human-computer interaction. Taking into account both filtering efficiency and real-time constraints, and based on the principles of maximizing inflection point gains and the upper limit of delay tolerance, the sliding window length is finally determined. The optimal value range is 20 to 50. In subsequent batch product consistency testing, this procedure is directly used to verify and solidify the filter parameters of each device.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions 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 solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback, characterized in that, include: The multi-channel signal acquisition module is used to simultaneously acquire the first state signal representing the main constraint variable and the second state signal representing the disturbance variable. The first state signal corresponds to the real-time contraction voltage value of the pelvic floor muscle group, and the second state signal corresponds to the real-time contraction voltage value of the abdominal muscle group. The state drift observation module is connected to the multi-channel signal acquisition module. It is used to monitor the amplitude attenuation rate of the first state signal when it is in the steady state maintenance period, and update the current real-time attenuation coefficient of the system according to the amplitude attenuation rate. The real-time attenuation coefficient is used to characterize the time-varying parameter drift characteristics of the controlled object. The dynamic threshold calculation module, connected to the state drift observation module, is used to construct a dynamic safety envelope for the disturbance variable based on the real-time amplitude and real-time attenuation coefficient of the first state signal. The dynamic safety envelope includes the amplitude saturation threshold and slew rate limit threshold allowed by the system at the current moment. An adaptive controller, connected to both the dynamic threshold calculation module and the multi-channel signal acquisition module, executes dual-constraint control logic: when the first-order time derivative of the second state signal exceeds the slew rate limit threshold, a slope locking command is output to forcibly lock the rise rate of the control output against the disturbance variable at the slew rate limit threshold; when the real-time amplitude of the second state signal exceeds the amplitude saturation threshold, a gain clamping command is output to limit the amplitude of the control output against the disturbance variable; wherein, the slew rate limit threshold and the real-time amplitude of the first state signal are positively correlated, so that the loading rate of the disturbance variable is limited by the current response bandwidth of the main constraint variable.

2. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The logic for updating the real-time attenuation coefficient by the state drift observation module is as follows: set a drift judgment threshold that characterizes the steady-state error tolerance of the system; continuously calculate the negative change of the first state signal per unit time; when the absolute value of the negative change exceeds the drift judgment threshold, perform a decrement operation on the real-time attenuation coefficient according to a preset step amount until the negative change of the first state signal falls back to within the drift judgment threshold.

3. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The slew rate limiting logic within the adaptive controller includes a nonlinear filter. When the first-order time derivative of the second-state signal is detected to exceed the slew rate limiting threshold, the nonlinear filter reshapes the input step control signal into a ramp signal. The slope value of the ramp signal is dynamically adjusted in real time following the slew rate limiting threshold.

4. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The dynamic threshold calculation module uses the following formula to calculate the amplitude saturation threshold: ,in, The amplitude saturation threshold of the second-state signal allowed at the current moment. This is the preset system safety coupling gain constant. This represents the real-time amplitude of the first-state signal. This is the real-time attenuation coefficient output by the state drift observation module.

5. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 4, characterized in that, The system also includes a cumulative load assessment module. When the amplitude of the second state signal is within a preset critical range of the amplitude saturation threshold, the cumulative load assessment module performs a weighted integral operation on the signal amplitude and residence time to generate a cumulative load value. The cumulative load value is used to characterize the degree of fatigue accumulation of the system under critical high load operating conditions.

6. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 5, characterized in that, The adaptive controller also includes a hysteresis recovery logic unit, which performs the following adjustments: when the accumulated load value exceeds a preset fatigue upper limit, a forced derating action is triggered, lowering the amplitude saturation threshold to a preset recovery level; when the accumulated load value naturally decays back to a preset reset lower limit, the forced derating action is released; the reset lower limit is lower than the fatigue upper limit, thus forming a state recovery hysteresis loop in the control logic.

7. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The multi-channel signal acquisition module includes a differential signal preprocessing unit, which performs absolute value detection and moving average filtering on the acquired raw signal to extract the extremely low frequency DC component as the first state signal and the second state signal.

8. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The adaptive controller is used to perform a bias-based feedback regulation: when the output slope lock command is issued, the deviation between the actual output slope and the slew rate limit threshold is calculated; using the deviation as feedback, the drive signal for the disturbance variable is corrected through a proportional-integral regulation algorithm to suppress dynamic overshoot of the system response.

9. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The dynamic threshold calculation module is also used to set the slew rate limit threshold to zero when the real-time amplitude of the first state signal is lower than the preset minimum operating point of the system, thereby cutting off the control output for the disturbance variable.

10. The adaptive control system for pelvic and abdominal muscle group coordinated training based on biofeedback according to claim 1, characterized in that, The system includes a first acquisition probe, a second acquisition probe, and a central processing unit; the first acquisition probe and the second acquisition probe are respectively connected to the analog input port of the central processing unit; the central processing unit integrates an execution state drift observation module, a dynamic threshold calculation module, and an adaptive controller logic operation circuit.

Images