An intelligent physiological training parameter optimization and adjustment system and method

Abstract

The application discloses a kind of intelligent physiological training parameter optimization adjustment system and method, belong to physiological training control technical field, including: first, obtain user multi-source baseline physiological signal sequence and training stage parameter, identify skeletal muscle fiber type and construct multi-training target decoupling control model, generate the control configuration data containing muscle fiber targeted execution action parameter interval, dynamic safety threshold curve, body pressure-muscle collaborative constraint rule;Subsequently, start parameter iteration optimization, collect physiological feedback signal, calculate four core indexes, and output muscle basic function diagnosis result after comprehensive difference meets standard;Finally, output rehabilitation action according to preset rhythm, continuously collect signal and identify abnormality such as spasm, muscle weakness, trigger intensity rollback or enhancement mechanism;The application realizes the precision, intelligent adjustment of pelvic floor rehabilitation parameter, improves rehabilitation safety and effectiveness, adapts to the individualized rehabilitation needs of different users.

Description

Technical Field

[0001] This invention belongs to the field of physiological training control technology, specifically an intelligent physiological training parameter optimization and adjustment system and method. Background Technology

[0002] Currently, pelvic floor rehabilitation largely relies on electrical stimulation training with fixed parameters, which has several drawbacks: First, it fails to differentiate the proportion of slow and fast contraction muscle fibers in the user's pelvic floor, resulting in inconsistent rehabilitation outcomes due to a lack of personalized parameter adaptation. Second, it lacks a dynamic safety management mechanism, making it prone to electromyography exceeding limits and body pressure imbalance during training, which can lead to muscle spasms or injuries. Third, it cannot monitor key indicators such as muscle strength, displacement, and fatigue in real time, making it difficult to accurately assess the rehabilitation status and dynamically adjust parameters. Fourth, insufficient coordination between body pressure and pelvic floor muscle control can easily lead to uncoordinated training movements, reducing rehabilitation efficiency.

[0003] Existing technologies struggle to balance personalization, safety, and precise adjustment, failing to meet the diverse needs of users with varying degrees of injury and training stages. Therefore, there is an urgent need for a method to adjust pelvic floor rehabilitation parameters that can identify muscle fiber types, dynamically optimize parameters, provide real-time monitoring and feedback, and intelligently identify abnormalities, in order to improve the safety, effectiveness, and intelligence of rehabilitation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an intelligent physiological training parameter optimization and adjustment system and method. First, it acquires the user's multi-source baseline physiological signal sequence and training phase parameters, identifies the user's skeletal muscle fiber type, and constructs a multi-training target decoupled regulation model. This generates regulation configuration data containing muscle fiber-targeted execution action parameter ranges, dynamic safety threshold curves, and body pressure-muscle synergistic constraint rules. Subsequently, iterative parameter optimization is initiated, physiological feedback signals are collected, four core indicators are calculated, and a basic muscle function diagnosis result is output after the comprehensive difference reaches the target. Finally, rehabilitation movements are output according to a preset rhythm, continuously collecting signals and identifying abnormalities such as spasms and insufficient muscle strength, triggering intensity regression or enhancement mechanisms. This invention achieves precise and intelligent adjustment of pelvic floor rehabilitation parameters, improving rehabilitation safety and effectiveness, and adapting to the personalized rehabilitation needs of different users.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for intelligently optimizing and adjusting physiological training parameters, comprising:

[0007] S1: Obtain the multi-source baseline physiological signal sequence of the target user and the corresponding training phase parameters;

[0008] S2: Based on the multi-source baseline physiological signal sequence and the training phase parameters, identify the user's skeletal muscle fiber type and construct a multi-training target decoupled regulation model to generate muscle regulation configuration data; the muscle regulation configuration data includes muscle fiber targeted execution action parameter range, dynamic safety threshold curve, and body pressure-muscle synergistic constraint rules.

[0009] S3: Based on the muscle regulation configuration data, initiate parameter iterative optimization, simultaneously collect physiological feedback signals, and calculate four indicators: electromyographic amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index. When the comprehensive difference of the four indicators is less than a preset first threshold, output the basic muscle function diagnosis result. The physiological feedback signals include electromyographic signals, muscle strength signals, muscle displacement signals, tissue impedance values, body pressure fluctuation values, and respiratory phase information.

[0010] S4: After receiving the basic muscle function diagnosis results, output physiological training execution actions according to the preset action rhythm, and continuously collect physiological feedback signals during the training execution process to identify abnormal muscle symptoms.

[0011] The specific steps of S2 include:

[0012] Based on the electromyographic baseline signal in the multi-source baseline physiological signal sequence, the first dominant frequency corresponding to slow contraction muscle fibers and the second dominant frequency corresponding to fast contraction muscle fibers are extracted using the sliding window superposition averaging method. At the same time, based on the muscle strength baseline signal and muscle displacement baseline signal in the multi-source baseline physiological signal sequence, the first fatigue slope of slow contraction muscle fibers in continuous contraction state and the second fatigue slope of fast contraction muscle fibers in staged contraction state are calculated using the waveform area integration method.

[0013] Based on the first dominant frequency, the second dominant frequency, the first fatigue slope, the second fatigue slope, and the training phase parameters, the first proportion of slow-twitch muscle fibers and the second proportion of fast-twitch muscle fibers of the user are determined by a preset muscle fiber type discrimination decision tree.

[0014] Based on the first proportion value, the second proportion value, and the training stage parameters, a multi-training-target decoupled regulation model is constructed, which includes a sub-target of enhancing muscle endurance, a sub-target of enhancing explosive power, and a sub-target of enhancing coordination and control. The multi-training-target decoupled regulation model outputs the range of muscle fiber targeted execution action parameters, dynamic safety threshold curves, and body pressure-muscle synergistic constraint rules, which are summarized as muscle regulation configuration data.

[0015] The process of generating the muscle fiber targeted execution action parameter range includes:

[0016] Construct a muscle endurance sub-model corresponding to the muscle endurance enhancement sub-target, set the first proportion value obtained as the first input of the muscle endurance sub-model, set the second proportion value as the second constraint input of the muscle endurance sub-model, and output the first set of execution action parameters adapted to the rehabilitation of slow contraction muscle fibers through the muscle endurance sub-model;

[0017] Construct a sub-model of explosive force corresponding to the sub-target of explosive force enhancement, set the second proportion value as the first input of the sub-model of explosive force, set the first proportion value as the second constraint input of the sub-model of explosive force, and output the second set of execution action parameters adapted to fast contraction muscle fiber rehabilitation through the sub-model of explosive force.

[0018] Construct a coordination control sub-model corresponding to the coordination control enhancement sub-target, input the first proportion value and the second proportion value into the coordination control sub-model, and output a set of temporal coupling parameters for alternately activating slow contraction muscle fibers and fast contraction muscle fibers based on a preset muscle group co-activation temporal template.

[0019] The first set of execution action parameters, the second set of execution action parameters, and the temporal coupling parameter set are spliced ​​together in a non-overlapping temporal domain to form a muscle fiber targeted execution action parameter range.

[0020] The process of generating the dynamic safety threshold curve includes:

[0021] Based on the baseline values ​​of tissue impedance and body pressure fluctuation in the multi-source baseline physiological signal sequence, the lower and upper percentiles of impedance under resting conditions are calculated using percentile statistics. At the same time, the mean and standard deviation of resting body pressure are calculated, and the lower and upper percentiles of impedance, the mean and standard deviation of resting body pressure are used as basic safety boundary parameters.

[0022] The injury severity level parameter and training days parameter are obtained from the training phase parameters. The basic safety boundary parameter, the injury severity level parameter, and the training days parameter are fused using an exponentially weighted moving average method to generate a dynamic safety threshold curve that changes over time. The dynamic safety threshold curve includes a dynamic lower limit threshold curve for impedance, a dynamic upper limit threshold curve for impedance, and a dynamic peak threshold curve for body pressure.

[0023] The generation process of the body pressure-muscle synergistic constraint rule includes:

[0024] Based on the baseline signals of body pressure fluctuations and electromyography (EMG) in the multi-source baseline physiological signal sequence, the Granger causality analysis method was used to calculate the causal influence coefficient of body pressure changes on EMG changes, and the feedback inhibition coefficient of EMG changes on body pressure changes was also calculated.

[0025] Based on the causal influence coefficient and the feedback inhibition coefficient, a collaborative constraint rule base containing multiple fuzzy logic rules is constructed. The antecedent of the fuzzy logic rule includes the membership degree of body pressure fluctuation value and the membership degree of electromyography amplitude value. The consequent of the fuzzy logic rule includes the allowed direction of motion intensity adjustment and the adjustment step size.

[0026] The collaborative constraint rule base is converted into a two-dimensional lookup table, and the two-dimensional lookup table is used as the body pressure-muscle collaborative constraint rule; one dimension of the two-dimensional lookup table is the body pressure fuzziness level, and the other dimension is the electromyography fuzziness level, and each table entry corresponds to an activation action option.

[0027] The calculation process for the electromyography amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index includes:

[0028] The range of muscle fiber targeted execution action parameters is used as the initial iteration parameter space, the dynamic safety threshold curve is used as the single iteration amplitude constraint boundary, and the body pressure-muscle synergistic constraint rule is used as the multi-source signal coupling constraint condition.

[0029] Based on the initial iterative parameter space, the amplitude constraint boundary, and the coupling constraint conditions, the iterative optimization process of the electrical actuation action parameters is initiated. During the iterative optimization process, physiological feedback signals are synchronously acquired at a preset first sampling frequency.

[0030] Based on the electromyographic signal, the electromyographic envelope amplitude sequence is extracted using the sliding window method, and the electromyographic amplitude stability rate is obtained by calculating the variance of the amplitude difference between adjacent cycles.

[0031] Based on the muscle strength signal, the muscle strength achievement rate is obtained by comparing the peak value with the preset training phase muscle strength target value.

[0032] Based on the muscle displacement signal, the time delay and amplitude difference of the displacement waveforms of the two sides of the levator ani muscle are extracted using the cross-correlation function calculation method, and the muscle displacement coordination rate is calculated based on the time delay and amplitude difference.

[0033] Based on the collected tissue impedance values, body pressure fluctuation values, and respiratory phase information, the muscle fatigue index was calculated using the normalized energy integral to recovery time ratio method.

[0034] When the combined difference of the four indicators is less than a preset first threshold, the basic muscle function diagnosis result is output, including:

[0035] The differences between the electromyographic amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index and their respective preset ideal target values ​​are calculated to obtain the amplitude stability difference, muscle strength achievement difference, displacement coordination difference, and fatigue index difference. The first weight coefficient, the second weight coefficient, the third weight coefficient, and the fourth weight coefficient are dynamically allocated according to the training stage parameters.

[0036] The first weighted difference is obtained by multiplying the amplitude stability difference by the first weighting coefficient using the weighted Euclidean distance calculation method; the second weighted difference is obtained by multiplying the muscle strength achievement difference by the second weighting coefficient; the third weighted difference is obtained by multiplying the displacement coordination difference by the third weighting coefficient; and the fourth weighted difference is obtained by multiplying the fatigue index difference by the fourth weighting coefficient. The sum of the squares of the first, second, third, and fourth weighted differences is then taken as the square root to calculate the comprehensive difference of the four indicators.

[0037] Determine whether the combined difference of the four indicators is less than a preset first threshold. If it is less than the preset first threshold, determine that the current user's combined measured state of electromyography, muscle strength, muscle displacement, and anti-fatigue performance is an ideal standard state. Determine that the parameter iterative calibration for muscle function analysis is complete. Based on the four indicators, the differences in each dimension, and the combined difference of the four indicators obtained from this iteration, output the basic muscle function diagnosis result.

[0038] The specific steps of S4 include:

[0039] In response to the diagnostic results of basic muscle function, a pre-stored training execution action sequence template is loaded; the training execution action sequence template includes a sequentially cyclical muscle endurance training phase, a power training phase, and a relaxation interval phase;

[0040] The final pulse intensity and pulse frequency parameters determined at the time of completion of parameter iteration calibration are used as the basic parameters for output action. According to the training execution action time sequence template, the physiological training execution action sequence arranged on the time axis is generated in combination with the basic parameters, and the physiological training execution action is output.

[0041] During the execution of physiological training actions, physiological feedback signals are collected at a preset second sampling frequency to establish a real-time physiological feedback buffer queue.

[0042] Extract electromyography (EMG) signals and muscle displacement signals within the current execution cycle from the real-time physiological feedback buffer queue; calculate the real-time peak amplitude of EMG based on the EMG signals; and calculate the real-time muscle contraction displacement amplitude based on the muscle displacement signals.

[0043] If the peak amplitude of the real-time electromyography exceeds the preset upper limit of the electromyography spasm threshold and the duration exceeds the preset spasm confirmation time window, it is determined that the user's muscles have abnormal spasm and a muscle spasm identifier is generated.

[0044] If the real-time muscle contraction displacement amplitude is lower than the preset displacement insufficiency threshold and the number of consecutive occurrences exceeds the preset number threshold, then it is determined that the user has abnormal muscle strength deficiency and a muscle strength deficiency flag is generated.

[0045] The identification of abnormal muscle conditions during training is accomplished by identifying muscle spasms and muscle weakness.

[0046] After generating the muscle spasm identifier, an automatic motion intensity regression mechanism is triggered, including:

[0047] In response to the muscle spasm indicator, the latest consecutive multiple waveform cycles of the electromyographic signal within the current execution action cycle are obtained, and the root mean square value of the electromyographic peak value of the latest consecutive multiple waveform cycles is calculated;

[0048] Find the impedance dynamic upper limit threshold curve value corresponding to the current training time point from the dynamic safety threshold curve, and determine the smaller of the electromyography peak root mean square value and the impedance dynamic upper limit threshold curve value as the regression target reference value.

[0049] A linear decreasing strategy is adopted to gradually reduce the current pulse intensity parameter according to a preset backtracking step size, and to re-acquire the electromyographic signal after each step of reduction. When the peak amplitude of the real-time electromyographic signal is lower than the preset spasm marker removal threshold, the backtracking is stopped and the muscle spasm marker is cleared.

[0050] After generating the muscle weakness indicator, a progressive enhancement mechanism for movement intensity is triggered, including:

[0051] In response to the muscle weakness indicator, the peak sequence of muscle displacement signal within multiple execution action cycles in the real-time physiological feedback buffer queue is extracted, the increasing trend slope of the peak sequence of muscle displacement signal is calculated, and it is determined whether the increasing trend slope is positive and exceeds a preset trend effective threshold.

[0052] If the slope of the increasing trend is positive and exceeds the effective threshold of the trend, the pulse intensity parameter is gradually increased according to the preset incremental step size. After each increase, wait for a complete execution cycle and re-evaluate the peak value of the muscle displacement signal.

[0053] If the peak value of the muscle displacement signal fails to reach the lower limit of the insufficient displacement threshold and the insufficient muscle strength indicator persists for multiple consecutive execution cycles, the current execution frequency will be adjusted up to the range of the optimal induction frequency for fast-twitch muscle fibers, and the muscle strength attainment status will be reassessed.

[0054] An intelligent physiological training parameter optimization and adjustment system, comprising:

[0055] The parameter acquisition module is used to acquire multi-source baseline physiological signal sequences and corresponding training phase parameters.

[0056] The regulation configuration generation module identifies the user's skeletal muscle fiber type and constructs a multi-training target decoupled regulation model based on multi-source baseline physiological signal sequences and training phase parameters, generating muscle regulation configuration data.

[0057] The functional diagnosis module initiates parameter iterative optimization based on muscle regulation configuration data, synchronously collects physiological feedback signals, calculates four indicators, and outputs basic muscle function diagnosis results when the comprehensive difference of the four indicators is less than the preset first threshold.

[0058] The anomaly recognition module is used to receive the diagnostic results of basic muscle function, output physiological training execution actions according to the preset movement rhythm, continuously collect physiological feedback signals during the training execution process, and identify abnormal muscle symptoms.

[0059] The abnormal adaptive adjustment module is used to dynamically adjust training parameters by triggering an automatic regression mechanism and a progressive enhancement mechanism for movement intensity after identifying abnormal muscle conditions.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] 1. This invention proposes an intelligent physiological training parameter optimization and adjustment method. By accurately identifying the type of pelvic floor muscle fibers through multi-source physiological signals, a multi-training target decoupled regulation model is constructed to realize personalized customization and dynamic iterative optimization of rehabilitation parameters. This effectively solves the problems of fixed parameters and poor adaptability in traditional programs, improves pelvic floor muscle endurance, explosive power and coordination control ability, and enhances the accuracy and effectiveness of rehabilitation.

[0062] 2. This invention proposes an intelligent physiological training parameter optimization and adjustment method, which introduces a dynamic safety threshold curve and body pressure-muscle synergistic constraint rules. Combined with real-time physiological feedback signal monitoring, it can promptly identify abnormalities such as pelvic floor spasm and insufficient muscle strength and trigger an adaptive adjustment mechanism, ensuring the safety of rehabilitation training throughout the process, reducing the risk of injury, and meeting the intelligent and safe rehabilitation needs of users with different degrees of injury and training stages. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of an intelligent physiological training parameter optimization and adjustment method according to the present invention;

[0064] Figure 2 This is a flowchart illustrating the principle of an intelligent physiological training parameter optimization and adjustment method according to the present invention.

[0065] Figure 3 A comparative experimental diagram showing the calculation of fatigue slope;

[0066] Figure 4 Experimental diagram for adaptive adjustment of dynamic safety threshold curve;

[0067] Figure 5 This is a diagram of the architecture of an intelligent physiological training parameter optimization and adjustment system according to the present invention. Detailed Implementation

[0068] Example 1:

[0069] Please see Figure 1 and Figure 2 The present invention provides an embodiment of an intelligent physiological training parameter optimization and adjustment method, applicable to pelvic floor rehabilitation training, comprising the following steps:

[0070] S1: Obtain the multi-source baseline physiological signal sequence of the target user and the corresponding training phase parameters;

[0071] In this embodiment, before the user formally begins pelvic floor rehabilitation training, the user is arranged in a quiet, relaxed, and interference-free standard testing position. This usually involves lying supine with knees bent, legs naturally apart, and all muscles fully relaxed. This ensures that the pelvic floor muscles are in a natural resting state, avoiding active exertion or tension that could cause baseline signal distortion. Under this standard position, an integrated pelvic floor physiological signal acquisition device is used, which includes pelvic floor electromyography electrodes, pelvic floor muscle strength and pressure sensors, pelvic floor muscle displacement ultrasound probes, vaginal impedance detection electrodes, body pressure sensing airbags, and respiratory motion sensors. This ensures that the devices work synchronously on the same time reference, avoiding signal time misalignment or asynchrony.

[0072] Furthermore, within the baseline acquisition period, various raw physiological signals of the user at rest are collected to form a continuous and complete raw physiological signal stream. The raw physiological signal stream contains multiple key signals, including electromyography baseline signals, muscle strength baseline signals, muscle displacement baseline signals, tissue impedance baseline signals, body pressure fluctuation baseline signals, and respiratory phase baseline signals. Each type of signal has a precise timestamp and corresponding sensor identifier. After acquisition, the raw physiological signal stream is automatically preprocessed, including removing common interference components such as high-frequency noise, baseline drift, motion artifacts, and electromagnetic interference. At the same time, the raw physiological signals are segmented for verification, outlier removal, and signal quality assessment to ensure that the final multi-source baseline physiological signal sequence has a high signal-to-noise ratio, high stability, and high consistency, and can truly and accurately reflect the basic physiological characteristics of the pelvic floor muscles and related physiological systems of the user at rest.

[0073] At the same time, the system synchronously inputs and associates the user's corresponding training stage parameters. The training stage parameters are determined by rehabilitation physicians based on a comprehensive assessment of multiple dimensions of clinical information, such as the user's age, gender, type of pelvic floor injury, degree of injury, previous rehabilitation history, and basic physical condition. These parameters include key information such as the initial stage of rehabilitation, the current rehabilitation cycle, the degree of injury, the expected rehabilitation duration, the priority of rehabilitation goals, the upper limit of training tolerance, and the risk warning level. All training stage parameters are stored in the form of structured data and are bound one by one to the multi-source baseline physiological signal sequences.

[0074] For example, this embodiment uses a female user who is 6 months postpartum, has mild pelvic floor dysfunction, mainly manifested as weak pelvic floor muscles, mild urinary incontinence, insufficient coordination of pelvic floor muscles, and no serious organic lesions. Before the rehabilitation begins, the user is arranged in a standard rehabilitation testing room, maintaining a standard baseline data acquisition position of supine with knees bent and the whole body relaxed, avoiding any active exertion or emotional tension. Through an integrated pelvic floor physiological signal acquisition device, pelvic floor electromyography electrodes, muscle strength sensors, displacement ultrasound probes, vaginal impedance detection electrodes, body pressure sensing balloons, and respiratory sensors are simultaneously connected. All devices are calibrated and time-synchronized in advance to ensure that the acquisition process is without delay or deviation. Various physiological signals are continuously collected for 10 minutes in a resting state. During this period, the user is required to maintain calm breathing, relax both physically and mentally, and not perform any pelvic floor contraction or relaxation movements. After the acquisition is completed, the system automatically performs noise removal and artifact elimination. After preprocessing operations such as signal verification, a user-specific multi-source baseline physiological signal sequence is generated. This sequence includes continuous and complete baseline signals for electromyography, muscle strength, muscle displacement, tissue impedance, body pressure fluctuation, and respiratory phase. Each signal segment is labeled with a precise timestamp and signal quality level to ensure data authenticity and reliability. Simultaneously, training phase parameters determined by rehabilitation physicians based on the user's postpartum recovery, pelvic floor muscle strength test results, severity of urinary incontinence symptoms, and physical tolerance are collected. These parameters are: the initial rehabilitation phase is the basic activation period; the current rehabilitation cycle is the first cycle; the injury level is mild; the expected rehabilitation duration is 12 weeks; the priority of rehabilitation goals is, in order: increased muscle endurance, improved coordination control, gradual improvement of explosive power; the upper limit of training tolerance is moderate intensity; and the risk warning level is low risk. All training phase parameters are entered into the system and linked to the multi-source baseline physiological signal sequence.

[0075] S2: Based on the multi-source baseline physiological signal sequence and the training phase parameters, identify the user's skeletal muscle fiber type and construct a multi-training target decoupled regulation model to generate muscle regulation configuration data; the muscle regulation configuration data includes muscle fiber targeted execution action parameter range, dynamic safety threshold curve, and body pressure-muscle synergistic constraint rules.

[0076] The specific steps of S2 include:

[0077] S2.1: Based on the electromyographic baseline signal in the multi-source baseline physiological signal sequence, the first dominant frequency corresponding to slow contraction muscle fibers and the second dominant frequency corresponding to fast contraction muscle fibers are extracted using the sliding window superposition averaging method. At the same time, based on the muscle strength baseline signal and muscle displacement baseline signal in the multi-source baseline physiological signal sequence, the first fatigue slope of slow contraction muscle fibers in continuous contraction state and the second fatigue slope of fast contraction muscle fibers in staged contraction state are calculated respectively using the waveform area integration method.

[0078] Further, the specific process of extracting the first dominant frequency corresponding to slow-twitch muscle fibers and the second dominant frequency corresponding to fast-twitch muscle fibers includes: First, extracting preprocessed, noise-free, artifact-free, and highly stable electromyographic baseline signals from the multi-source baseline physiological signal sequence. The electromyographic baseline signal is a continuous electrical signal generated by the natural discharge of the pelvic floor muscles in the user's resting state, containing the low-frequency, continuous, and stable discharge characteristics of slow-twitch muscle fibers and the high-frequency, intermittent, and fluctuating discharge characteristics of fast-twitch muscle fibers, which is the most core electrophysiological basis for distinguishing the two types of muscle fibers. After extraction, the sliding window superposition averaging method is used to extract features from the electromyographic baseline signal. The core principle of the sliding window superposition averaging method is to set a sliding window of fixed length and fixed step size, slide it segment by segment on the continuous electromyographic baseline signal, and at each step, the electromyographic baseline within the current window is... The baseline electromyography (EMG) signal is processed by superposition and averaging. Through multiple superpositions and averagings, random noise and interference fluctuations in the signal are effectively suppressed, while the unique frequency characteristics of slow-twitch and fast-twitch muscle fibers are amplified to obtain the processed EMG baseline signal. In this embodiment, the window length is set to 200ms, the sliding step size to 50ms, and the number of superpositions to 20. Frequency feature analysis is performed on the processed EMG baseline signal to identify and extract the first dominant frequency corresponding to the slow-twitch muscle fibers. This frequency is in the low-frequency range, with stable characteristics, long duration, and small fluctuation amplitude, which perfectly matches the endurance and sustained contraction physiological characteristics of slow-twitch muscle fibers. At the same time, the second dominant frequency corresponding to the fast-twitch muscle fibers is identified and extracted. This frequency is in the high-frequency range, with obvious fluctuations, short duration, and regular discharge intervals, which perfectly matches the explosive and rapid contraction physiological characteristics of fast-twitch muscle fibers.

[0079] Furthermore, the specific process for calculating the first and second fatigue slopes includes: extracting preprocessed, quality-compliant, and continuous muscle strength baseline signals and muscle displacement baseline signals from the multi-source baseline physiological signal sequence, respectively. The muscle strength baseline signal reflects the pressure changes generated by the user's pelvic floor muscles during rest and simulated contraction, while the muscle displacement baseline signal reflects the changes in the displacement amplitude and velocity of the muscle tissue during pelvic floor muscle contraction and relaxation. For slow-twitch muscle fibers, simulating continuous contraction, a 3-minute contraction phase signal segment is extracted from the muscle strength baseline signal and muscle displacement baseline signal to obtain the first contraction phase signal. The muscle force signal waveform and the first displacement signal waveform are extracted. The trend of these signals reflects the process of force attenuation, displacement reduction, and gradual fatigue accumulation of slow-twitch muscle fibers during prolonged continuous contraction. After extraction, the waveform area integration method is used to integrate the first muscle force signal waveform and the first displacement signal waveform during the 3-minute contraction phase. The integration results quantify the total force output and total displacement change of the slow-twitch muscle fibers during continuous contraction. Combined with the signal attenuation trend over time, the first fatigue slope corresponding to the slow-twitch muscle fibers is calculated. The value of the first fatigue slope directly reflects the endurance level and fatigue tolerance of the slow-twitch muscle fibers. The smaller the slope, the better the endurance of slow-twitch muscle fibers and the slower the fatigue accumulation; the larger the slope, the weaker the endurance of slow-twitch muscle fibers and the faster the fatigue accumulation. For fast-twitch muscle fibers, simulating the phased contraction state of fast-twitch muscle fibers, the system extracts signal segments from 10 fast-twitch muscle fiber contraction and relaxation cycles from the muscle force baseline signal and muscle displacement baseline signal to obtain the second muscle force signal waveform and the second displacement signal waveform. The signal change trend reflects the changes in peak force, displacement amplitude, and gradual fatigue accumulation of fast-twitch muscle fibers during multiple rapid burst contractions. After extraction, the system also uses the waveform area integration method to analyze the second muscle force signal waveform of the phased contraction stage. The waveform of the shape and the second displacement signal are integrated to quantify the total peak force and total displacement amplitude of the fast contraction muscle fiber during multiple rapid contractions. Combined with the signal attenuation trend with the increase of the number of contractions, the second fatigue slope corresponding to the fast contraction muscle fiber is calculated. The value of the second fatigue slope directly reflects the explosive force level and fatigue tolerance of the fast contraction muscle fiber. The smaller the slope, the better the explosive force of the fast contraction muscle fiber and the slower the fatigue accumulation; the larger the slope, the weaker the explosive force of the fast contraction muscle fiber and the faster the fatigue accumulation. The waveform area integration method is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0080] For example, using a female user with mild pelvic floor dysfunction 6 months postpartum as the experimental subject, firstly, preprocessed, high-quality, continuous and complete muscle strength baseline signals and muscle displacement baseline signals were extracted from the user's multi-source baseline physiological signal sequence. For slow-twitch muscle fibers, simulating the continuous contraction state of slow-twitch muscle fibers, a 3-minute contraction phase signal segment was extracted from the muscle strength baseline signal and muscle displacement baseline signal to obtain the first muscle strength signal waveform and the first displacement signal waveform during the contraction phase. The trend of this signal change reflects the process of force attenuation, displacement reduction, and gradual fatigue accumulation of slow-twitch muscle fibers during long-term continuous contraction. After extraction, the waveform area integration method was used to integrate the first muscle strength signal waveform and the first displacement signal waveform of the 3-minute contraction phase. The integration results quantified the total force output and total displacement change of slow-twitch muscle fibers during continuous contraction. Combined with the signal attenuation trend over time, the first fatigue slope corresponding to the slow-twitch muscle fibers was calculated. The experimental results showed that slow-twitch muscle fibers had strong endurance; after 3 minutes of continuous contraction, the force decreased from the initial value to 85%, with an attenuation rate of only 15%, and the first fatigue slope was -0.025. Figure 3 As shown in (a); for fast-twitch muscle fibers, the phased contraction state of fast-twitch muscle fibers is simulated. Signal segments from 10 fast-twitch muscle fiber contraction-relaxation cycles are extracted from the muscle force baseline signal and muscle displacement baseline signal to obtain the second muscle force signal waveform and the second displacement signal waveform. The signal change trend reflects the changes in peak force, displacement amplitude, and gradual fatigue accumulation of fast-twitch muscle fibers during multiple rapid burst contractions. After extraction, the waveform area integration method is used to integrate the second muscle force signal waveform and the second displacement signal waveform of this phased contraction stage. The integration results quantify the total peak force and total displacement amplitude of fast-twitch muscle fibers during multiple rapid contractions. Combined with the signal attenuation trend with increasing contraction count, the corresponding second fatigue slope of the fast-twitch muscle fiber is calculated, as shown in (a). Figure 3 As shown in (b), the experimental results show that fast contraction muscle fibers have strong explosive power but are easily fatigued. After 10 rapid explosive contractions, the peak force decreased from the initial value to 58%, with a decay rate of 42%. The second fatigue slope was -0.085, and the ratio of the first fatigue slope to the second fatigue slope was 0.29, which is in line with physiological laws. Slow contraction muscle fibers have better endurance than fast contraction muscle fibers.

[0081] S2.2: Based on the first dominant frequency, the second dominant frequency, the first fatigue slope, the second fatigue slope and the training phase parameters, determine the first proportion value of the user's slow contraction muscle fibers and the second proportion value of the fast contraction muscle fibers through a preset muscle fiber type discrimination decision tree.

[0082] Further, the specific steps of S2.2 include: acquiring the first dominant frequency of slow-twitch muscle fibers, the second dominant frequency of fast-twitch muscle fibers, the first fatigue slope of slow-twitch muscle fibers, and the second fatigue slope of fast-twitch muscle fibers, and inputting them along with the user's corresponding training stage parameters into a preset muscle fiber type discrimination decision tree. This discrimination decision tree is an intelligent discrimination model built based on a large amount of clinical pelvic floor muscle fiber detection data, rehabilitation case data, and muscle fiber physiological characteristic data. The hierarchical structure of the muscle fiber type discrimination decision tree is sorted according to feature importance. The first layer prioritizes judging the degree of difference in the dominant frequencies of the two types of muscle fibers. By matching the first dominant frequency and the second dominant frequency with preset frequency threshold intervals, the basic distribution tendency of slow-twitch muscle fibers and fast-twitch muscle fibers is initially distinguished. The frequency threshold is set to 10Hz. The second layer judges the degree of difference in the fatigue slope of the two types of muscle fibers. The fatigue slope and the second fatigue slope are matched with the fatigue slope ratio thresholds pre-set based on the physiological fatigue characteristics of slow-twitch and fast-twitch muscles, respectively, to further refine the functional state differences between the two types of muscle fibers. The fatigue slope ratio threshold is set to 1.2. The third layer combines key information such as the degree of injury, priority of rehabilitation goals, and physical tolerance in the training phase parameters to correct and calibrate the preliminary discrimination results of the first and second layers, eliminating the interference of external factors such as the degree of injury, training phase, and physical condition on the muscle fiber characteristics, and ensuring that the final discrimination results can truly reflect the actual distribution ratio and functional state of the user's pelvic floor muscle fibers. The fourth layer integrates the discrimination results of the first three layers to perform the final proportion calculation and output, generating the first proportion value of slow-twitch muscle fibers and the second proportion value of fast-twitch muscle fibers. The sum of the two proportion values ​​is 100%, accurately quantifying the distribution ratio of the two types of muscle fibers in the user's pelvic floor muscle group.

[0083] S2.3: Based on the first proportion value, the second proportion value, and the training stage parameters, construct a multi-training target decoupled regulation model that includes a muscle endurance enhancement sub-target, a power enhancement sub-target, and a coordination control enhancement sub-target; the multi-training target decoupled regulation model outputs the muscle fiber targeted execution action parameter range, dynamic safety threshold curve, and body pressure-muscle synergistic constraint rules, which are summarized into muscle regulation configuration data.

[0084] Furthermore, in this embodiment, after obtaining the user's first proportion of slow-twitch muscle fibers and second proportion of fast-twitch muscle fibers, the system combines the user's individualized training stage parameters to construct a multi-training-target decoupled regulation model. This involves decoupling the three core rehabilitation targets in pelvic floor rehabilitation—muscle endurance enhancement, explosive power enhancement, and coordination control enhancement—which are interconnected and mutually influential, and constructing independent sub-models for each. Each sub-model focuses on one rehabilitation target and outputs action parameters tailored to the corresponding muscle fiber rehabilitation needs. Simultaneously, through unified constraint rules and collaborative mechanisms, the system ensures that the parameters output by the three sub-models are compatible, mutually supportive, and do not conflict. Finally, the system integrates and outputs a complete, accurate, safe, and suitable set of muscle regulation configuration data.

[0085] The process of generating the muscle fiber targeted execution action parameter range includes:

[0086] A1: Construct a muscle endurance sub-model corresponding to the muscle endurance enhancement sub-target, set the obtained first proportion value as the first input of the muscle endurance sub-model, set the second proportion value as the second constraint input of the muscle endurance sub-model, and output the first set of execution action parameters adapted to the rehabilitation of slow contraction muscle fibers through the muscle endurance sub-model;

[0087] Furthermore, the specific steps of A1 include:

[0088] (1) Obtain the first proportion value of slow contraction muscle fibers and the second proportion value of fast contraction muscle fibers from the output of the muscle fiber type discrimination decision tree. Taking the female user in this embodiment as an example, assuming that after the decision tree discrimination, the first proportion value of the user's slow contraction muscle fibers is 65% and the second proportion value of fast contraction muscle fibers is 35%, the system directly uses the first proportion value of 65% as the first input parameter of the muscle endurance sub-model, which determines the dominant weight of muscle endurance training. At the same time, the system uses the second proportion value of 35% as the second constraint input parameter of the muscle endurance sub-model to limit the participation of fast contraction muscle fibers in the muscle endurance training process and avoid the training parameters from causing excessive stimulation or interference to fast contraction muscle fibers.

[0089] (2) The first input parameter, namely the first proportion of slow contraction muscle fibers 65%, is compared with the three preset fuzzy levels. The classification criteria for the three fuzzy levels are as follows: low level corresponds to a first proportion of slow contraction muscle fibers less than or equal to 40%, medium level corresponds to a first proportion of slow contraction muscle fibers between 40% and 70%, and high level corresponds to a first proportion of slow contraction muscle fibers greater than or equal to 70%. Since 65% is within the range of 40% to 70%, the system determines that the user's first proportion of slow contraction muscle fibers belongs to the medium level. At the same time, the system compares the second constraint input parameter, namely the second proportion of fast contraction muscle fibers 35%, with the three preset fuzzy levels. The classification criteria for the three fuzzy levels of the second proportion of fast contraction muscle fibers are as follows: low level corresponds to a second proportion of fast contraction muscle fibers less than or equal to 30%, medium level corresponds to a second proportion of fast contraction muscle fibers between 30% and 60%, and high level corresponds to a second proportion of fast contraction muscle fibers greater than or equal to 60%. Since 35% is within the range of 30% to 60%, the system determines that the user's second proportion of fast contraction muscle fibers belongs to the medium level.

[0090] (3) Match the rules corresponding to the fuzzy level and output the fuzzy inference result. The system has nine rules pre-stored. Each rule corresponds to a combination of the fuzzy level of the proportion of slow contraction muscle fibers and the fuzzy level of the proportion of fast contraction muscle fibers. The system uses the medium level corresponding to the first proportion value and the medium level corresponding to the second proportion value. Among these nine rules, the system searches for the rule that completely matches the two levels. Specifically, when the proportion of slow contraction muscle fibers is medium and the proportion of fast contraction muscle fibers is medium, the rule matched by the system specifies that the output action is a standard long-term low-frequency parameter set. The pulse frequency range corresponding to this standard long-term low-frequency parameter set is 10 times per second to 16 times per second, the duration of a single contraction is 2 seconds to 4 seconds, and the interval time is 5 seconds to 7 seconds. The system outputs this set of ranges as the fuzzy inference result.

[0091] (4) The centroid method is used to convert the obtained pulse frequency range, single contraction duration range, and interval time range into specific numerical ranges. For the pulse frequency range of 10 to 16 times per second, the system divides this range into 11 discrete points at equal intervals, namely 10.0 times per second, 10.6 times per second, 11.2 times per second, 11.8 times per second, 12.4 times per second, 13.0 times per second, 13.6 times per second, 14.2 times per second, 14.8 times per second, 15.4 times per second, and 16 times per second. Since the input fuzzy level is At the intermediate level, the membership degree corresponding to each discrete point is set to 1.0. The system calculates the product of each discrete point and its membership degree, that is, multiplying each discrete point value by 1.0 to obtain the original value, and then summing these 11 products to obtain a total of 143 times per second. Then, this total is divided by the total membership degree sum, which is 11.0, to obtain the centroid value of the pulse frequency as 13 times per second. Based on this, the system needs to consider the limiting effect of the second constraint input parameter, namely the second proportion value of fast-twitch muscle fibers. The system calculates a constraint coefficient based on the second proportion value of fast-twitch muscle fibers, and the calculation formula is as follows: ,in, This represents the constraint coefficient of fast contractile muscle fibers. This represents the second percentage of fast-twitch muscle fibers; in this embodiment, it is taken as 35%. This represents the empirical weighting coefficient of the muscle endurance sub-model, which is set to 0.3 in this embodiment. It is used to adjust the compression degree of the second proportion of fast-twitch muscle fibers on the muscle endurance training parameters when calculating the constraint coefficient. Substituting into the formula, we get 0.195, therefore ; Calculate the symmetric expansion factor of the constraint coefficient. The symmetric expansion factor is calculated by multiplying the original range half-width by the constraint coefficient. The original pulse frequency range is from 10 times per second to 16 times per second, and its half-width is 3 times per second. 3 × 0.195 = 0.585. Therefore, the final pulse frequency range is [12.415, 13.585]. After the system retains one decimal place, the final pulse frequency range is [12.4, 13.6].

[0092] For a single contraction duration ranging from 2 to 4 seconds, the same centroid method is used for calculation. This range is divided into 11 discrete points at equal intervals, namely 2.0 seconds, 2.2 seconds, 2.4 seconds, 2.6 seconds, 2.8 seconds, 3.0 seconds, 3.2 seconds, 3.4 seconds, 3.6 seconds, 3.8 seconds, and 4.0 seconds. The membership degree of each discrete point is 1.0. The sum of all discrete points is 33 seconds, which is divided by 11 to obtain a centroid value of 3.3 seconds. The original range center value is 3.3 seconds, and the half-width is 1.0 seconds. Multiplying the half-width by the constraint coefficient 0.195 gives 0.195 seconds. Therefore, the final single contraction duration range is [3.105, 3.495]. The system retains one decimal place and determines it as [3.1, 3.5].

[0093] For the interval time range of 5 to 7 seconds, it is also divided into 11 discrete points at equal intervals, namely 5.0 seconds, 5.2 seconds, 5.4 seconds, 5.6 seconds, 5.8 seconds, 6.0 seconds, 6.2 seconds, 6.4 seconds, 6.6 seconds, 6.8 seconds, and 7.0 seconds. The membership degree of each discrete point is 1.0. The sum of all discrete points is 66.6 seconds. Dividing by 11, we get the centroid value of 6.6 seconds. The original range center value is 6.6 seconds, and the half-width is 1.0 seconds. Multiplying the half-width by the constraint coefficient 0.195, we get 0.195 seconds. Therefore, the final interval time range is [6.405, 6.795]. The system retains one decimal place and determines it as [6.4, 6.8].

[0094] (5) Organize the three calculated numerical ranges. The first dimension is the pulse frequency parameter range, specifically set to [12.4, 13.6]. The second dimension is the single contraction duration parameter range, specifically set to [3.1, 3.5]. The third dimension is the interval time parameter range, specifically set to [6.4, 6.8]. The system determines the execution duration of each muscle endurance training session based on the expected rehabilitation duration and the rehabilitation start stage in the training phase parameters. In this embodiment, the user is in the basic activation period and the rehabilitation cycle is the first cycle. The system sets the execution duration of each muscle endurance training session to 10 minutes. The system packages the above pulse frequency range, single contraction duration range, interval time range and execution duration of 10 minutes together to form the first set of execution action parameters adapted to slow contraction muscle fiber rehabilitation.

[0095] A2: Construct a sub-model of explosive force corresponding to the sub-target of explosive force enhancement, set the second proportion value as the first input of the explosive force sub-model, set the first proportion value as the second constraint input of the explosive force sub-model, and output the second set of execution action parameters adapted to fast contraction muscle fiber rehabilitation through the explosive force sub-model.

[0096] Furthermore, the specific steps of A2 include:

[0097] (1) Obtain the first proportion value of slow contraction muscle fibers and the second proportion value of fast contraction muscle fibers from the output of the muscle fiber type discrimination decision tree. Taking the female user in the example, it is assumed that after the decision tree discrimination, the first proportion value of slow contraction muscle fibers of the user is 65% and the second proportion value of fast contraction muscle fibers is 35%. The system directly uses the second proportion value of 35% as the first input parameter of the explosive power sub-model, which determines the dominant weight of explosive power training. At the same time, the system uses the first proportion value of 65% as the second constraint input parameter of the explosive power sub-model to limit the participation of slow contraction muscle fibers in the explosive power training process.

[0098] (2) The first input parameter, namely the second proportion value of fast-twitch muscle fibers 35%, is compared with the three preset fuzzy levels. The classification criteria for the three fuzzy levels are as follows: low level corresponds to a second proportion value of fast-twitch muscle fibers less than or equal to 30%, medium level corresponds to a second proportion value of fast-twitch muscle fibers between 30% and 60%, and high level corresponds to a second proportion value of fast-twitch muscle fibers greater than or equal to 60%. Since 35% is within the range of 30% to 60%, the system determines that the user's second proportion value of fast-twitch muscle fibers belongs to the medium level. At the same time, the system compares the second constraint input parameter, namely the first proportion value of slow-twitch muscle fibers 65%, with the three preset fuzzy levels. The classification criteria for the three fuzzy levels of slow-twitch muscle fiber proportion values ​​are as follows: low level corresponds to a first proportion value of slow-twitch muscle fibers less than or equal to 40%, medium level corresponds to a first proportion value of slow-twitch muscle fibers between 40% and 70%, and high level corresponds to a first proportion value of slow-twitch muscle fibers greater than or equal to 70%. Since 65% is within the range of 40% to 70%, the system determines that the user's slow-twitch muscle fiber proportion value belongs to the medium level.

[0099] (3) Match the rules corresponding to the fuzzy level and output the fuzzy inference result. The system has nine rules pre-stored. Each rule corresponds to a combination of the fuzzy level of the proportion of fast contraction muscle fibers and the fuzzy level of the proportion of slow contraction muscle fibers. The system uses the obtained medium-level second proportion value of fast contraction muscle fibers and medium-level first proportion value of slow contraction muscle fibers to search for the rule that completely matches the two levels among these nine rules. Specifically, when the second proportion value of fast contraction muscle fibers is medium and the first proportion value of slow contraction muscle fibers is medium, the rule matched by the system specifies that the output action is a standard short-term high-frequency parameter set. The pulse frequency range corresponding to the standard short-term high-frequency parameter set is 20 to 30 times per second, the duration of a single contraction is 0.5 to 1.0 seconds, the interval time is 2.0 to 3.0 seconds, the number of repetitions per set is 10 to 15, and the rest time between sets is 15 to 20 seconds. The system outputs this set of parameter ranges as the fuzzy inference result.

[0100] (4) The system uses the centroid method to convert the pulse frequency range, single contraction duration range, interval time range, number of repetitions per group range and rest time range between groups obtained in the previous step into specific numerical ranges;

[0101] For a pulse frequency range of 20 to 30 times per second, the system divides this range into 11 discrete points at equal intervals of once per second, corresponding to 20.0 times per second, 21.0 times per second, and so on up to 30.0 times per second. Since the input fuzziness level is medium, the membership degree corresponding to each discrete point is set to 1.0. The system calculates the product of each discrete point and its membership degree, that is, multiplying each discrete point value by 1.0 to obtain the original value. Then, these 11 products are summed to obtain a total of 275.0 times per second. This total is then divided by the total membership degree, which is 11.0, to obtain the centroid value of the pulse frequency as 25 times per second. Based on this, the system needs to consider the limiting effect of the second constraint input parameter, namely the first proportion value of slow-twitch muscle fibers. A constraint coefficient is calculated based on the first proportion value of slow-twitch muscle fibers, using the following formula: ,in, This represents the constraint coefficient of slow-twitch muscle fibers. This represents the percentage of slow-twitch muscle fibers; in this embodiment, it is taken as 65%. The empirical weighting coefficient of the explosive force sub-model is 0.4 in this embodiment, which is used to adjust the compression degree of the first proportion of slow-twitch muscle fibers on the explosive force training parameters when calculating the constraint coefficient. In this embodiment, the first proportion of slow-twitch muscle fibers of 65% is substituted into the formula to obtain 0.14, so the constraint coefficient of slow-twitch muscle fibers is 0.14. The symmetry expansion factor of the constraint coefficient of slow-twitch muscle fibers is calculated, where the half-width is 5. Similarly, 0.14×5=0.7 is obtained, so the final pulse frequency range is [24.3, 25.7].

[0102] Similarly, for a single contraction duration ranging from 0.5 seconds to 1.0 seconds, the same centroid method is used for calculation. This range is divided into 11 discrete points at equal intervals, namely 0.50 seconds, 0.55 seconds, 0.60 seconds, and up to 1.0 seconds. The membership degree of each discrete point is 1.0. The sum of all discrete points is 8.25 seconds. Dividing by 11, we get a centroid value of 0.75 seconds. The original range center value is 0.75 seconds, and the half-width is 0.25 seconds. Therefore, the symmetry expansion factor is 0.25 × 0.14 = 0.035 seconds. Thus, the final single contraction duration range is [0.715, 0.785]. The system retains one decimal place and determines it as [0.7, 0.8].

[0103] Similarly, for the interval time range of 2.0 seconds to 3.0 seconds, it is divided into 11 discrete points at equal intervals, namely 2.0 seconds, 2.1 seconds, and up to 3.0 seconds. The membership degree of each discrete point is 1.0. The sum of all discrete points is 28.5 seconds. Dividing by 11, we get the centroid value of 2.59 seconds. The original range center value is 2.59 seconds, and the half-width is 0.5 seconds. Therefore, the symmetry expansion factor is 0.5 × 0.14 = 0.07 seconds. Thus, the final interval time range is [2.52, 2.66]. After rounding to one decimal place, it is determined to be [2.5, 2.7].

[0104] Similarly, for each group of repetitions ranging from 10 to 15 times, it is divided into 11 discrete points at equal intervals, namely 10.0 times, 10.5 times, 11.0 times, up to 15 times. The membership degree of each discrete point is 1.0. The sum of all discrete points is 137.5 times. Dividing by 11, we get the centroid value of 12.5 times. The original range center value is 12.5 times, and the half-width is 2.5 times. Therefore, the symmetry expansion factor is 2.5 × 0.14 = 0.35 times. Thus, the final repetition range for each group is [12.15, 12.85]. In actual execution, the integer number of times is taken, so it is determined to be 12 to 13 times.

[0105] Similarly, for the inter-group rest time range of 15 to 20 seconds, it is divided into 11 discrete points at equal intervals, namely 15.0 seconds, 15.5 seconds, and up to 20 seconds. The membership degree of each discrete point is 1.0. The sum of all discrete points is 192.5 seconds. Dividing by 11, we get the centroid value of 17.5 seconds. The original range center value is 17.5 seconds, and the half-width is 2.5 seconds. Therefore, the symmetry expansion factor is 2.5 × 0.14 = 0.35 seconds. Thus, the final inter-group rest time interval is [17.15, 17.85]. After rounding to one decimal place, it is determined as [17.2, 17.9].

[0106] (5) Organize the five calculated numerical ranges. The first dimension is the pulse frequency parameter range, specifically set to [24.3, 25.7]; the second dimension is the single contraction duration parameter range, specifically set to [0.7, 0.8]; the third dimension is the interval time parameter range, specifically set to [2.5, 2.7]; the fourth dimension is the number of repetitions per set parameter range, specifically set to 12 to 13 times; and the fifth dimension is the rest time between sets parameter range, specifically set to [17.2, 17.9]. The system determines the burst power training for each session based on the rehabilitation goal priority and training tolerance upper limit in the training phase parameters. In this embodiment, the user's priority in rehabilitation goals is to gradually improve explosive power, with the training tolerance limit set to moderate intensity. The system sets the number of execution sets for each explosive power training session to be three. Each set completes a contraction-interval cycle according to the repetition parameter range. After each set is completed, the user rests according to the rest time parameter range between sets. The explosive power training session ends after all three sets are completed. The system packages the above pulse frequency range, single contraction duration range, interval range, repetition range per set, rest time range between sets, and execution set together to form a second set of execution action parameters adapted to fast-twitch muscle fiber rehabilitation.

[0107] A3: Construct a coordination control sub-model corresponding to the coordination control enhancement sub-target, input the first proportion value and the second proportion value into the coordination control sub-model, and output a set of temporal coupling parameters for alternately activating slow contraction muscle fibers and fast contraction muscle fibers based on a preset muscle group co-activation temporal template.

[0108] Furthermore, the specific steps in A3 include:

[0109] (1) Obtain the first proportion value of slow contraction muscle fibers and the second proportion value of fast contraction muscle fibers from the output of the muscle fiber type discrimination decision tree. Taking the female user in the embodiment as an example, it is assumed that after the decision tree discrimination, the first proportion value of slow contraction muscle fibers of the user is 65% and the second proportion value of fast contraction muscle fibers is 35%. The system uses the first proportion value of 65% and the second proportion value of 35% as input parameters of the coordinated control sub-model. These two parameters together determine the activation weight of slow contraction muscle fibers and fast contraction muscle fibers and the switching rhythm between the two activation modes during the alternating activation process.

[0110] (2) Load the preset muscle group synergistic activation timing template. The system internally stores a basic timing template, which defines the basic time frame for the alternating activation of slow-twitch and fast-twitch muscle fibers without personalized adjustments. The structure of the basic timing template is as follows: A complete coordinated control cycle has a total duration of 60 seconds, with the first 30 seconds occupied by the slow-twitch muscle fiber priority activation phase and the last 30 seconds occupied by the fast-twitch muscle fiber priority activation phase; the first 10 seconds of the slow-twitch muscle fiber priority activation phase is the slow-twitch muscle fiber individual activation period, and the system only outputs the action parameters used to activate the slow-twitch muscle fibers; the following 10 seconds is the slow-twitch muscle fiber main activation and fast-twitch muscle fiber auxiliary activation period, and the system outputs the slow-twitch muscle fiber activation parameters while superimposing the low-intensity fast-twitch muscle fiber activation parameters; the last 10 seconds is the slow-twitch and fast-twitch muscle fiber synergistic activation period, and the system follows a 1: The system alternately outputs activation parameters for two types of muscle fibers at a 1:1 time ratio. In each alternation cycle, slow-twitch muscle fiber activation lasts for 3 seconds, and fast-twitch muscle fiber activation lasts for 3 seconds. During the first 10 seconds of the fast-twitch muscle fiber priority activation phase, the system is in the fast-twitch muscle fiber individual activation phase, and only outputs the motion parameters used to activate the fast-twitch muscle fibers. The following 10 seconds is the fast-twitch muscle fiber primary activation phase with slow-twitch muscle fiber secondary activation, and the last 10 seconds is the fast-twitch and slow-twitch muscle fiber synergistic activation phase. The system alternately outputs activation parameters for the two types of muscle fibers at a 1:1 time ratio, with fast-twitch muscle fiber activation lasting for 3 seconds and slow-twitch muscle fiber activation lasting for 3 seconds in each alternation cycle.

[0111] (3) The system uses the first proportion of slow-twitch muscle fibers (65%) and the second proportion of fast-twitch muscle fibers (35%) to calculate three key timing adjustment coefficients. The first adjustment coefficient is the slow-twitch muscle fiber phase duration adjustment coefficient, calculated as follows: the slow-twitch muscle fiber phase duration adjustment coefficient equals the first proportion of slow-twitch muscle fibers divided by 50%. Dividing 65% by 50% yields 1.3, meaning that the base duration of the slow-twitch muscle fiber preferential activation phase (30 seconds) needs to be multiplied by 1.3. The second adjustment coefficient is the fast-twitch muscle fiber phase duration adjustment coefficient, calculated as follows: The fast-twitch muscle fiber phase duration adjustment factor is equal to the second proportion of fast-twitch muscle fibers divided by 50%. Dividing 35% by 50% gives 0.7, which means that the base duration of the fast-twitch muscle fiber preferential activation phase of 30 seconds needs to be multiplied by 0.7. The third adjustment factor is the alternation cycle ratio adjustment factor, which is calculated as the ratio of the first proportion of slow-twitch muscle fibers to the second proportion of fast-twitch muscle fibers, i.e., 65% ÷ 35% = 1.857. It is used to adjust the ratio of the activation duration of the two types of muscle fibers during the synergistic activation period.

[0112] (4) The system modifies each time parameter in the basic time sequence template according to the calculated adjustment coefficient. First, the total duration of the slow contraction muscle fiber priority activation phase is modified. The 30 seconds in the basic template is multiplied by the slow contraction muscle fiber phase duration adjustment coefficient of 1.3 to get 39 seconds. In this 39-second phase, the division ratio of the three 10-second sub-phases remains unchanged. Therefore, the duration of each sub-phase also needs to be adjusted accordingly. The specific adjustment method is: multiply the basic duration of each sub-phase of 10 seconds by the slow contraction muscle fiber phase duration adjustment coefficient of 1.3 to get the actual duration of each sub-phase of 13 seconds. Therefore, the detailed time structure of the slow contraction muscle fiber priority activation phase becomes: the first 13 seconds is the slow contraction muscle fiber single activation period, the second 13 seconds is the slow contraction muscle fiber main activation and fast contraction muscle fiber auxiliary activation period, and the third 13 seconds is the slow contraction muscle fiber and fast contraction muscle fiber synergistic activation period.

[0113] Then, the total duration of the fast-twitch muscle fiber preferential activation phase is modified by multiplying the 30 seconds in the basic template by the fast-twitch muscle fiber phase duration adjustment coefficient of 0.7, resulting in 21 seconds. Within this 21-second phase, the division ratio of the original three 10-second sub-phases remains unchanged, that is, the basic duration of each sub-phase is 10 seconds multiplied by 0.7 to get 7 seconds. Therefore, the detailed time structure of the fast-twitch muscle fiber preferential activation phase becomes: the first 7 seconds is the fast-twitch muscle fiber single activation period, the second 7 seconds is the fast-twitch muscle fiber main activation and slow-twitch muscle fiber auxiliary activation period, and the third 7 seconds is the fast-twitch muscle fiber and slow-twitch muscle fiber synergistic activation period.

[0114] Finally, the ratio of alternating activation time within the synergistic activation phase was modified. In the third 13-second synergistic activation phase of the slow-twitch muscle fiber preferential activation stage, the original ratio of slow-twitch muscle fiber activation to fast-twitch muscle fiber activation was 3 seconds and 3 seconds respectively in each alternating cycle (1:1 ratio). This has now been modified based on an adjustment coefficient of 1.857, keeping the total alternating cycle duration unchanged at 6 seconds, but adjusting the slow-twitch muscle fiber activation duration to: 6 × 1.857 ÷ (1.857 + 1) = 3.899, which is approximately 3.9 seconds. The fast-twitch muscle fiber activation duration is adjusted to 6 - 3.9 = 2.1 seconds. Similarly, in the third 7-second synergistic activation phase of the fast-twitch muscle fiber preferential activation stage, the original ratio of fast-twitch muscle fiber activation to slow-twitch muscle fiber activation was 3 seconds and 3 seconds respectively in each alternating cycle (also 1:1 ratio). Since fast-twitch muscle fibers are dominant at this time, it is necessary to adjust the duration of fast-twitch muscle fiber activation according to the reciprocal of the alternation cycle ratio adjustment coefficient. That is, use 1÷1.857=0.538 to adjust the duration of fast-twitch muscle fiber activation to 6×0.538÷(0.538+1)=2.1 seconds, and the duration of slow-twitch muscle fiber activation to 6-2.1=3.9 seconds.

[0115] (5) When the coordinated control sub-model outputs the temporal coupling parameter set, it needs to obtain specific pulse frequency values ​​from the action parameters already output by the muscular endurance sub-model and the explosive force sub-model. The first set of execution action parameters output by the muscular endurance sub-model contains a pulse frequency range of [12.4, 13.6], and the second set of execution action parameters output by the explosive force sub-model contains a pulse frequency range of [24.3, 25.7]. The coordinated control sub-model takes the median value of the pulse frequency range of the muscular endurance sub-model as the base frequency for slow-twitch muscle fiber activation, and the calculation method is: (12.4 + 13.6) / 2. )÷2=13 times; Similarly, the median of the pulse frequency range of the explosive force model is taken as the base frequency for fast contraction muscle fiber activation, and the calculation method is: (24.3+25.7)÷2=25; For the superposition parameters of the auxiliary activation period, the system is set that during the period of primary activation of slow contraction muscle fibers and auxiliary activation of fast contraction muscle fibers, the activation frequency of fast contraction muscle fibers is 30% of its base frequency, that is, 25×0.3=7.5 times; during the period of primary activation of fast contraction muscle fibers and auxiliary activation of slow contraction muscle fibers, the activation frequency of slow contraction muscle fibers is 30% of its base frequency, that is, 13×0.3=3.9 times;

[0116] (6) Integrate the modified time structure in (4) and the activation parameters determined in (5) to form a structured time axis parameter set. This parameter set contains the following in chronological order: The first time period is the slow-twitch muscle fiber activation period, lasting 13 seconds. During this time period, the system continuously outputs slow-twitch muscle fiber activation parameters with a pulse frequency of 13 times per second, and does not output fast-twitch muscle fiber activation parameters; The second time period is the period of primary activation of slow-twitch muscle fibers and secondary activation of fast-twitch muscle fibers, lasting 13 seconds. During this time period... The system continuously outputs slow-twitch muscle fiber activation parameters at a pulse frequency of 13 times per second, while simultaneously outputting fast-twitch muscle fiber activation parameters at a pulse frequency of 7.5 times per second. The third time period is the synergistic activation period of slow-twitch and fast-twitch muscle fibers, lasting 13 seconds. During this time period, the system outputs parameters in an alternating cycle, with each alternating cycle lasting 6 seconds. The first 3.9 seconds output slow-twitch muscle fiber activation parameters at a pulse frequency of 13 times per second, and the last 2.1 seconds output fast-twitch muscle fiber activation parameters at a pulse frequency of 25 times per second, and so on. The process continues until 13 seconds have elapsed; the fourth time period is the fast-twitch muscle fiber activation phase, lasting 7 seconds. During this period, the system continuously outputs fast-twitch muscle fiber activation parameters at a pulse frequency of 25 times per second, without outputting slow-twitch muscle fiber activation parameters; the fifth time period is the primary activation phase for fast-twitch muscle fibers and the secondary activation phase for slow-twitch muscle fibers, lasting 7 seconds. During this period, the system continuously outputs fast-twitch muscle fiber activation parameters at a pulse frequency of 25 times per second, while simultaneously outputting slow-twitch muscle fiber activation parameters at a pulse frequency of 3.9 times per second; the sixth time period is... The synergistic activation phase of fast-twitch and slow-twitch muscle fibers lasts for 7 seconds. During this period, the system outputs parameters in an alternating cycle, with each cycle lasting 6 seconds. The first 2.1 seconds output fast-twitch muscle fiber activation parameters at a pulse frequency of 25 times per second, and the next 3.9 seconds output slow-twitch muscle fiber activation parameters at a pulse frequency of 13 times per second. This cycle continues until the 7-second period ends. The system combines the parameters from the above six time periods in sequence to form a complete temporally coupled parameter set, which has a total duration of 39 + 21 = 60 seconds.

[0117] A4: The first set of execution action parameters, the second set of execution action parameters, and the temporal coupling parameter set are spliced ​​together in a non-overlapping temporal domain to form a muscle fiber targeted execution action parameter interval.

[0118] The process of generating the dynamic safety threshold curve includes:

[0119] B1: Based on the baseline values ​​of tissue impedance and body pressure fluctuation in the multi-source baseline physiological signal sequence, the lower and upper percentiles of impedance under resting conditions are calculated using percentile statistics. At the same time, the mean and standard deviation of resting body pressure are calculated, and the lower and upper percentiles of impedance, the mean and standard deviation of resting body pressure are used as basic safety boundary parameters.

[0120] Furthermore, the specific steps for B1 include:

[0121] (1) After the user completes the baseline signal acquisition under the standard position, the tissue impedance baseline signal acquired by the vaginal impedance detection electrode and the body pressure fluctuation baseline signal acquired by the body pressure sensing balloon are read from the stored multi-source baseline physiological signal sequence. Taking the female user in the example, during the continuous 10-minute resting acquisition process, the vaginal impedance detection electrode continuously records the tissue impedance value in the vagina at a sampling frequency of 10 times per second, and the body pressure sensing balloon continuously records the pressure fluctuation value in the abdominal cavity at the same sampling frequency. After the acquisition is completed, the system extracts the complete 10-minute data segment from the two signal streams to ensure that each signal segment contains 6,000 sampling points, each sampling point has a precise timestamp and signal quality label, and has completed preprocessing operations such as noise removal, artifact removal and signal verification.

[0122] (2) The 6000 sampling points extracted from the 10-minute tissue impedance baseline signal are sorted in ascending order of value. The system sets the lower percentile of impedance to the 5th percentile and the upper percentile of impedance to the 95th percentile. The 5th percentile is calculated by multiplying the total number of the 6000 sampling points after sorting by 5%, which gives 300. Therefore, the 5th percentile corresponds to the impedance value of the 300th sampling point in the sorted sequence. The 95th percentile is calculated by multiplying the total number of the 6000 sampling points after sorting by 95%, which gives 5700. Therefore, the 95th percentile corresponds to the impedance value of the 300th sampling point in the sorted sequence. The digit corresponds to the impedance value of the 5700th sampling point in the sorted sequence. Assuming that the impedance value of the 300th sampling point in the sorted tissue impedance baseline signal of this user is 80Ω and the impedance value of the 5700th sampling point is 120Ω, then the system records the lower percentile of impedance as 80Ω and the upper percentile of impedance as 120Ω. These two values ​​represent the normal fluctuation range of tissue impedance of this user in a completely resting and relaxed state. 90% of the sampling points fall within this range, while those below 80Ω and above 120Ω each account for 5%, which are considered to be the boundaries of normal physiological fluctuations.

[0123] (3) Extract all 6000 sampling points from the 10-minute baseline signal of body pressure fluctuation. First, calculate the resting mean body pressure. The method for calculating the resting mean body pressure is as follows: add up the body pressure values ​​of all 6000 sampling points to get the sum, and then divide the sum by 6000. Assuming that the sum of the body pressure values ​​of the user within 10 minutes is 36000 mmH2O, then the resting mean body pressure is equal to 36000 divided by 6000, which equals 6 mmH2O. Then calculate the resting standard deviation of body pressure. The method for calculating the resting standard deviation of body pressure is as follows: first calculate the difference between the body pressure value of each sampling point and the resting mean body pressure of 6 mmH2O, square each difference, sum the squared differences of all sampling points to get the sum of squared differences, then divide the sum of squared differences by the total number of sampling points 6000 to get the variance, and finally take the square root of the variance to get the standard deviation. Assuming that the calculated sum of squared differences is 3600 mmH2O, 2 If H2O is present, then the variance is 3600 ÷ 6000 = 0.6 mm. 2 H2O, the standard deviation is equal to the square root of 0.6, approximately 0.775 mmH2O. After rounding to one decimal place, the resting standard deviation of body pressure is determined to be 0.8 mmH2O.

[0124] (4) The calculated impedance lower percentile of 80Ω, impedance upper percentile of 120Ω, body pressure resting mean of 6mmH2O and body pressure resting standard deviation of 0.8mmH2O are packaged and stored as a structured set of basic safety boundary parameters. Each value in the set of basic safety boundary parameters has a clear physical unit. The unit of impedance-related parameters is Ω and the unit of body pressure-related parameters is mmH2O.

[0125] B2: Obtain the injury severity level parameter and training days parameter from the training phase parameters. First, divide the injury severity into three fixed levels: mild injury, moderate injury, and severe injury, and assign a fixed initial correction coefficient to each level. The initial correction coefficient is 0.9 for mild injury, 0.7 for moderate injury, and 0.5 for severe injury. Following the logic that the corrected static parameter equals the product of the basic safety boundary parameter and the initial correction coefficient, perform linear scaling calculations on the four basic safety boundary parameters: lower percentile of impedance, upper percentile of impedance, resting mean of body pressure, and resting standard deviation of body pressure. The corrected static boundary parameters are obtained; then, using the number of training days as a time-series variable, the corrected static boundary parameters are iteratively updated hourly using an exponentially weighted moving average method. Relying on the time-series smoothing characteristics of the exponentially weighted moving average, the parameter values ​​are dynamically and gradually amplified and the safety threshold range is widened as the number of training days gradually increases, ultimately generating a dynamic safety threshold curve that changes continuously with the training time. The exponentially weighted moving average method is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here. The dynamic safety threshold curve includes the impedance dynamic lower limit threshold curve, the impedance dynamic upper limit threshold curve, and the body pressure dynamic peak threshold curve.

[0126] For example, using a female user 6 months postpartum with mild pelvic floor dysfunction as the experimental subject, firstly, the user's training phase parameters were obtained, where the degree of impairment was mild, corresponding to an initial correction coefficient of 0.9; secondly, baseline safety boundary parameters were extracted from the user's multi-source baseline physiological signal sequence, including a lower percentile impedance of 80Ω, an upper percentile impedance of 120Ω, a mean resting body pressure of 6mmH2O, and a resting body pressure standard deviation of 0.8mmH2O; then, according to the principle that the corrected static boundary parameters are equal to the product of the baseline safety boundary parameters and the initial correction coefficient, the four baseline safety parameters were respectively... The boundary parameters were calculated using linear scaling, resulting in the following corrected static boundary parameters: lower bound of corrected impedance 72Ω, upper bound of corrected impedance 108Ω, mean body pressure 5.4mmH2O, and standard deviation of corrected body pressure 0.72mmH2O. Finally, using training time as the time-series variable, the corrected static boundary parameters were iteratively updated hourly using an exponentially weighted moving average method. Leveraging the time-series smoothing properties of the exponentially weighted moving average, the parameter values ​​were dynamically and gradually increased and the safety threshold range was widened as the training time increased, ultimately generating a dynamic safety threshold curve that continuously changes with training time. Figure 4 As shown in the figure. Experimental results show that the dynamic safety threshold curve generated in this application can adjust the safety boundary in real time according to the training process, ensuring safety in the early stages of training while providing sufficient training space for later rehabilitation. Compared with existing fixed threshold methods, training safety is improved by 47% and training efficiency by 32%. Figure 4The dynamic upper limit threshold curve and the dynamic lower limit threshold curve of impedance show a smooth upward trend as the training time increases. The dynamic peak threshold curve of body pressure also gradually increases with the training time. In contrast, the existing fixed threshold remains constant and cannot adapt to the physiological changes during the training process. This verifies the significant advantages of the dynamic safety threshold curve of this invention in terms of personalized adaptation, dynamic self-adaptation and risk warning.

[0127] The generation process of the body pressure-muscle synergistic constraint rule includes:

[0128] C1: Based on the baseline signal of body pressure fluctuation and the baseline signal of electromyography in the multi-source baseline physiological signal sequence, the Granger causality analysis method is used to calculate the causal influence coefficient of body pressure change on electromyography change, and the feedback inhibition coefficient of electromyography change on body pressure change is calculated. The Granger causality analysis method is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0129] Furthermore, the specific steps of C1 include:

[0130] (1) Read the baseline signal of body pressure fluctuation collected by the body pressure sensing balloon and the baseline signal of electromyography collected by the electromyography acquisition electrode from the multi-source baseline physiological signal sequence, respectively;

[0131] (2) The ADF test method is used to test the stationarity of the baseline signal of body pressure fluctuation and the baseline signal of electromyography. In the ADF test, the lag order is set to 10 and the significance level is 5%. For the baseline signal of body pressure fluctuation, the system calculates the test statistic to be -3.8, while the critical value under this significance level is -2.68. Since -3.8 < -2.68, the system determines that the baseline signal of body pressure fluctuation is a stationary sequence. For the baseline signal of electromyography, the system calculates the test statistic to be -3.5. Similarly, -3.5 < -2.68, the system determines that the baseline signal of electromyography is also a stationary sequence. However, if a signal is determined to be non-stationary, the system will perform first-order difference processing on the signal, that is, calculate the difference between each sampling point and the previous sampling point to form a new difference sequence, and then perform stationarity test again until the sequence reaches stationarity. The ADF test method and first-order difference are existing technologies in this field and are not the inventive solution of this application. They will not be described in detail here.

[0132] (3) The optimal lag order is determined by using the AIC criterion. Specifically, the values ​​of the Akaike Information Criterion from 1 to 20 are calculated respectively, and the lag order that minimizes the value of the Akaike Information Criterion is selected as the final lag order. Specifically, the system first sets the maximum candidate lag order to 20. Then, for each candidate lag order, the system constructs a vector autoregression model of the body pressure fluctuation baseline signal and the electromyography baseline signal respectively, and calculates the Akaike Information Criterion value of the model. After calculation, the system finds that the Akaike Information Criterion value reaches the minimum when the lag order is 5. Therefore, the system determines that the optimal lag order is 5. The AIC criterion and autoregression model are existing technologies in this field and are not the inventive solution of this application. They will not be described in detail here.

[0133] (4) Calculate the causal influence coefficient of body pressure change on electromyography (EMG) change. Body pressure change is used as the cause variable and EMG change is used as the result variable. The system first constructs a restricted model. The restricted model only uses the historical value of the EMG variable itself to predict the current value of the EMG variable. The mathematical form of the restricted model is: the current EMG value is equal to the constant term plus the linear combination of the first five EMG lag values ​​plus the error term. Specifically, the system represents the value of each sampling point in the EMG baseline signal as the weighted sum of the first to fifth EMG values ​​before that sampling point, plus a constant term, plus a random error term. The system uses the least squares method to estimate each coefficient in the restricted model and calculates the residual sum of squares of the model. After calculation, the residual sum of squares of the restricted model is 1250. The least squares method is the prior art in this field and is not the inventive solution of this application. It will not be described in detail here.

[0134] (5) Then an unrestricted model is constructed. The unrestricted model uses the historical values ​​of the electromyography (EMG) variable itself and the historical values ​​of the body pressure variable to predict the current value of the EMG variable. The mathematical form of the unrestricted model is: the current EMG value is equal to the constant term plus the linear combination of the first five EMG lag values, plus the linear combination of the first five body pressure lag values, plus the error term. Specifically, the system represents the value of each sampling point in the EMG baseline signal as the weighted sum of the first to fifth EMG values ​​before that sampling point, plus the weighted sum of the first to fifth body pressure values ​​before that sampling point, plus a constant term, plus a random error term. The system also uses the least squares method to estimate each coefficient in the unrestricted model and calculates the residual sum of squares of the unrestricted model. After calculation, the residual sum of squares of the unrestricted model is 840.

[0135] (6) Substitute the sum of squared residuals of the restricted model and the sum of squared residuals of the unrestricted model into the formula for calculating the F-statistic. The formula for calculating the F-statistic is: ,in, Represents the squared residuals of the constrained model. Let p represent the sum of squared residuals of the unrestricted model, p represent the lag order, and n represent the sample size. In this embodiment, p=5 and n=300000. Substituting the specific values, the numerator is 1250-840=410, which is then divided by the lag order of 5 to get 82. The denominator is 840÷(300000-2×5-1)=0.0028. Therefore, the F-statistic is equal to 82÷0.0028=29286. The system compares this F-statistic with the preset significance level of 5%, which is approximately 3.84. Since 29286 is much greater than 3.84, the system determines that changes in body pressure have a significant impact on changes in electromyography. The Granger causality relationship was established. The causal influence coefficient of body pressure change on electromyography (EMG) change was defined as the square root of the sum of squares of the coefficients of the body pressure lag terms in the unrestricted model, followed by taking the absolute value. After calculation, the coefficients of the five body pressure lag terms in the unrestricted model were 0.32, 0.28, 0.25, 0.22 and 0.18, respectively. After squaring them, the coefficients were 0.1024, 0.0784, 0.0625, 0.0484 and 0.0324, respectively. The sum was 0.3241, and the square root was 0.569. After retaining two decimal places, the causal influence coefficient of body pressure change on EMG change was determined to be 0.57.

[0136] (7) Construct a restricted model and an unrestricted model for calculating the feedback inhibition coefficient of electromyography (EMG) changes on body pressure changes; using EMG changes as the cause variable and body pressure changes as the result variable, repeat the calculation process of (4)-(5). First, construct a restricted model, using only the historical value of the body pressure variable itself to predict the current value of the body pressure variable. The mathematical form of the restricted model is that the body pressure value at the current moment is equal to the constant term plus the linear combination of the first five lag values ​​of body pressure plus the error term. The system uses the least squares method to estimate the model, and the residual sum of squares of the restricted model is calculated to be 960. Then, construct an unrestricted model, using both the historical value of the body pressure variable itself and the historical value of the EMG variable to predict the current value of the body pressure variable. The mathematical form of the unrestricted model is that the body pressure value at the current moment is equal to the constant term plus the linear combination of the first five lag values ​​of body pressure plus the linear combination of the first five lag values ​​of EMG plus the error term. The system uses the least squares method to estimate the unrestricted model, and the residual sum of squares of the unrestricted model is calculated to be 720.

[0137] (8) Substituting the sum of squared residuals of the restricted model (960) and the sum of squared residuals of the unrestricted model (720) into the formula for calculating the F-statistic, the F-statistic is obtained as 20000. This F-statistic is also much greater than 3.84, indicating that the system determines that there is a significant Granger causal relationship between electromyography (EMG) changes and body pressure changes. The system defines the feedback inhibition coefficient of EMG changes on body pressure changes as the square root of the sum of squares of the EMG lag term coefficients in the unrestricted model, followed by the absolute value, but it needs to be multiplied by a negative sign to reflect the inhibition relationship. After calculation, the five EMG lag terms in the unrestricted model are... The coefficients are -0.15, -0.12, -0.10, -0.08, and -0.06, respectively. After squaring them, we get 0.0225, 0.0144, 0.0100, 0.0064, and 0.0036. The sum is 0.0569, the square root is 0.239, and the negative sign is -0.239. After keeping two decimal places, the feedback inhibition coefficient of electromyography changes on body pressure changes is determined to be -0.24. The negative sign indicates that the increase in electromyography activity will lead to a decrease in body pressure, that is, pelvic floor muscle contraction has an inhibitory effect on the increase in body pressure.

[0138] C2: Based on the causal influence coefficient and the feedback inhibition coefficient, a collaborative constraint rule base containing multiple fuzzy logic rules is constructed. The antecedent of the fuzzy logic rule includes the membership degree of body pressure fluctuation value and the membership degree of electromyography amplitude value. The consequent of the fuzzy logic rule includes the allowed direction of motion intensity adjustment and the adjustment step size.

[0139] Furthermore, the specific steps of C2 include:

[0140] (1) Determine the fuzzy level classification and membership function of body pressure fluctuation values. Based on the causal influence coefficient of body pressure change on electromyography change of 0.57 obtained from Granger causality analysis, the system determines the fuzzy level boundary of body pressure fluctuation values ​​by combining the resting mean body pressure and the resting standard deviation of body pressure. The system sets three fuzzy levels of body pressure fluctuation values, namely low level, medium level and high level. The body pressure fluctuation range corresponding to the low level is from the resting mean body pressure minus one time the resting standard deviation of body pressure to the resting mean body pressure minus 0.5 times the resting standard deviation of body pressure. Substituting the resting mean body pressure of 6 mmH2O and the resting standard deviation of body pressure of 0.8 mmH2O into the system, The standard deviation is 0.8 mmH2O, and 0.5 times the standard deviation is 0.4 mmH2O. Therefore, the body pressure fluctuation range corresponding to the low level is 5.2 mmH2O to 5.6 mmH2O; the body pressure fluctuation range corresponding to the medium level is from the resting body pressure mean minus 0.5 times the resting body pressure standard deviation to the resting body pressure mean plus 0.5 times the resting body pressure standard deviation, i.e., 5.6 mmH2O to 6.4 mmH2O; the body pressure fluctuation range corresponding to the high level is from the resting body pressure mean plus 0.5 times the resting body pressure standard deviation to the resting body pressure mean plus one times the resting body pressure standard deviation, i.e., 6.4 mmH2O to 6.8 mmH2O.

[0141] (2) Define a triangular membership function. For the low level, the membership function is 0 when the body pressure fluctuation value is 5.2 mmH2O, reaches a maximum value of 1 at 5.4 mmH2O, and drops to 0 at 5.6 mmH2O. For the medium level, the membership function is 0 when the body pressure fluctuation value is 5.6 mmH2O, reaches a maximum value of 1 at 6.0 mmH2O, and drops to 0 at 6.4 mmH2O. For the high level, the membership function is 0 when the body pressure fluctuation value is 6.4 mmH2O, reaches a maximum value of 1 at 6.6 mmH2O, and drops to 0 at 6.8 mmH2O. When the real-time collected body pressure fluctuation value falls within any interval, the system calculates the membership degree of the body pressure fluctuation value to each fuzzy level according to the membership function.

[0142] (3) Based on the feedback inhibition coefficient of electromyography changes on body pressure changes of -0.24 obtained from Granger causality analysis, and combined with the characteristics of the electromyography baseline signal, the fuzzy level boundary of the electromyography amplitude is determined. The system first calculates the mean amplitude and standard deviation of all sampling points in the electromyography baseline signal. Assuming that the mean amplitude of the electromyography baseline signal is 50 μV and the standard deviation is 10 μV, the system sets three fuzzy levels of electromyography amplitude, namely low level, medium level and high level. The electromyography amplitude range corresponding to the low level is from the mean amplitude minus one standard deviation to the mean amplitude minus 0.5 standard deviation, that is, 40 μV to 45 μV; the electromyography amplitude range corresponding to the medium level is from the mean amplitude minus 0.5 standard deviation to the mean amplitude plus 0.5 standard deviation, that is, 45 μV to 55 μV; the electromyography amplitude range corresponding to the high level is from the mean amplitude plus 0.5 standard deviation to the mean amplitude plus one standard deviation, that is, 55 μV to 60 μV.

[0143] (4) Similarly, a triangular membership function is defined for each fuzzy level. For the low level, the membership function is 0 at 40 microvolts, reaches a maximum value of 1 at 42.5 microvolts, and drops to 0 at 45 microvolts. For the medium level, the membership function is zero at 45 microvolts, reaches a maximum value of 1 at 50 microvolts, and drops to 0 at 55 microvolts. For the high level, the membership function is 0 at 55 microvolts, reaches a maximum value of 1 at 57.5 microvolts, and drops to 0 at 60 microvolts. When the real-time collected electromyography amplitude falls within any interval, the system calculates the membership degree of the electromyography amplitude to each fuzzy level according to the triangular membership function.

[0144] (5) Nine fuzzy logic rules are set according to the inherent physiological logic of causality. Each rule corresponds to a combination of the fuzzy level of body pressure fluctuation value and the fuzzy level of electromyography amplitude value. The consequent of each rule contains two elements, namely the adjustment direction and the adjustment step size. The adjustment direction is divided into three types: positive direction enhancement, negative direction weakening, and zero direction maintenance. The adjustment step size is divided into three types: small step size, medium step size, and large step size. The small step size is defined as 5% of the current pulse intensity parameter, the medium step size is defined as 10%, and the large step size is defined as 15%.

[0145] The specific contents of the nine rules are as follows: Rule 1 stipulates that when the body pressure fluctuation value is low and the electromyography amplitude value is low, it indicates that the body pressure is low and the electromyography activity is also low. At this time, the driving effect of body pressure on electromyography is weak, and the feedback inhibition of electromyography on body pressure is also weak. The system determines that the current state is that both body pressure and electromyography are in an inhibited state. Therefore, the consequent is set to adjust the positive direction to enhance and the adjustment step size is medium step size.

[0146] Rule 2 stipulates that when the body pressure fluctuation value is low and the electromyography amplitude value is medium, it indicates that the body pressure is low but the electromyography activity is within the normal range. At this time, the driving effect of body pressure on electromyography is insufficient, but the pelvic floor muscles themselves maintain a normal activity level. The system determines that the current state is insufficient body pressure and good pelvic floor compensation. Therefore, the posterior setting is to increase the positive direction and the adjustment step size is small step size.

[0147] Rule 3 states that when the body pressure fluctuation value is low and the electromyography amplitude value is high, it indicates that the body pressure is low but the electromyography activity is high. At this time, the pelvic floor muscles may be in an over-tense state to compensate for insufficient body pressure. The system judges the current state as a risk of over-compensation of the pelvic floor muscles. Therefore, the consequent is set to adjust the direction of negative weakening and the adjustment step size is small.

[0148] Rule 4 stipulates that when the body pressure fluctuation value is at a medium level and the electromyography amplitude value is at a low level, it indicates that the body pressure is normal but the electromyography activity is low. At this time, the driving effect of body pressure on electromyography has not been effectively converted into electromyography response. The system determines that the current state is insufficient pelvic floor muscle response. Therefore, the posterior setting is to adjust the positive direction to enhance and the adjustment step size is medium step size.

[0149] Rule 5 stipulates that when the body pressure fluctuation value is at a medium level and the electromyography amplitude value is at a medium level, it means that both body pressure and electromyography are within the normal range and the two have reached a coordinated balance. The system determines that the current state is an ideal coordinated state. Therefore, the consequent is set to zero direction and zero adjustment step.

[0150] Rule 6 states that when the body pressure fluctuation value is at a medium level and the electromyography amplitude value is at a high level, it indicates that the body pressure is normal but the electromyography activity is high. At this time, there may be unnecessary overactivation of the pelvic floor muscles. The system determines that the current state is at risk of overactivation of the pelvic floor muscles. Therefore, the consequent is set to adjust the direction of negative weakening and the adjustment step size is medium step size.

[0151] Rule 7 states that when the body pressure fluctuation value is high and the electromyography amplitude value is low, it indicates that the body pressure is high but the electromyography activity is low. At this time, the body pressure has a greater impact on the pelvic floor muscles and the pelvic floor muscles cannot effectively resist it. The system determines that the current state is a risk of insufficient protection of the pelvic floor muscles. Therefore, the posterior setting is to adjust the direction of positive enhancement and the adjustment step size is a large step size.

[0152] Rule 8 states that when the body pressure fluctuation value is high and the electromyography amplitude value is medium, it indicates that the body pressure is high but the electromyography activity is within the normal range. At this time, the pelvic floor muscles are trying to resist the increase in body pressure, but may be close to the upper limit of tolerance. The system determines that the current state is the critical load state of the pelvic floor muscles. Therefore, the posterior setting is to adjust the direction of negative weakening and the adjustment step size is small step size.

[0153] Rule 9 stipulates that when the body pressure fluctuation value is high and the electromyography amplitude value is high, it indicates that both body pressure and electromyography are at a high level. At this time, the driving force of body pressure on the pelvic floor muscles is too strong, and the pelvic floor muscles are also in a state of high tension. The two form a positive feedback loop. The system determines that the current state is a risk of pelvic floor muscle overload. Therefore, the consequent is set to adjust in the negative direction to weaken, and the adjustment step size is a large step size.

[0154] (6) The causal influence coefficient of 0.57 and the feedback inhibition coefficient of -0.24 obtained from Granger causality analysis were used to numerically correct the set adjustment step size. The correction method was as follows: for rules that strengthen in the positive direction, the original step size was multiplied by 1 and the sum of the causal influence coefficients was added, i.e., multiplied by 1.57. For rules that weaken in the negative direction, the original step size was multiplied by 1 and the absolute value of the difference between the feedback inhibition coefficient and the original step size was subtracted, i.e., multiplied by 0.76. The step size of rule five remained unchanged at zero. Specifically, for each rule, the original adjustment step size of rule one was a medium step size of 10%, which was multiplied by 1.57 to get 15.7%. After rounding, the adjustment step size was determined to be 16%. The original adjustment step size of rule two was a small step size of 5%, which was multiplied by 1.57 to get 7.85%. After rounding, the adjustment step size was determined to be 8%. The original adjustment step size of rule three was a small step size of 5%. The original step size for Rule 4 was 5%, multiplied by 0.76 to get 3.8%, and rounded to determine the adjustment step size as 4%. The original adjustment step size for Rule 4 was a medium step size of 10%, multiplied by 1.57 to get 15.7%, and rounded to determine the adjustment step size as 16%. The original adjustment step size for Rule 6 was a medium step size of 10%, multiplied by 0.76 to get 7.6%, and rounded to determine the adjustment step size as 8%. The original adjustment step size for Rule 7 was a large step size of 15%, multiplied by 1.57 to get 23.55%, and rounded to determine the adjustment step size as 24%. The original adjustment step size for Rule 8 was a small step size of 5%, multiplied by 0.76 to get 3.8%, and rounded to determine the adjustment step size as 4%. The original adjustment step size for Rule 9 was a large step size of 15%, multiplied by 0.76 to get 11.4%, and rounded to determine the adjustment step size as 11%.

[0155] (7) The nine revised fuzzy logic rules are encoded and stored in a unified format. Each rule contains four fields: fuzzy level of antecedent body pressure fluctuation value, fuzzy level of antecedent electromyography amplitude value, consequent adjustment direction, and consequent adjustment step size. When Rule 1 is encoded as an antecedent body pressure low level and antecedent electromyography low level, the consequent adjustment direction is positively enhanced and the consequent adjustment step size is 16%. When Rule 2 is encoded as an antecedent body pressure low level and antecedent electromyography medium level, the consequent adjustment direction is positively enhanced and the consequent adjustment step size is 8%. When Rule 3 is encoded as an antecedent body pressure low level and antecedent electromyography high level, the consequent adjustment direction is negative. When the direction of adjustment is weakened and the step size of the consequent is 4%, the direction of adjustment for the consequent is strengthened and the step size of the consequent is 16%, and the direction of adjustment for the consequent is maintained at zero and the step size of the consequent is 0%, when the direction of adjustment for the consequent is maintained at zero and the step size of the consequent is 0%, when the direction of adjustment for the consequent is weakened and the step size of the consequent is 8%, and when the direction of adjustment for the consequent is strengthened and the step size of the consequent is 24%, the direction of adjustment for the consequent is strengthened and the step size of the consequent is 4%. When Rule 8 is coded as a high level of antecedent body pressure and a medium level of antecedent electromyography, the adjustment direction of the consequent is negative and the adjustment step size of the consequent is 4%; when Rule 9 is coded as a high level of antecedent body pressure and a high level of antecedent electromyography, the adjustment direction of the consequent is negative and the adjustment step size of the consequent is 11%. The system stores these nine coded rules together as a rule base, which is a collaborative constraint rule base constructed based on the causal influence coefficient and the feedback inhibition coefficient.

[0156] C3: Convert the collaborative constraint rule base into an executable two-dimensional lookup table, and use the two-dimensional lookup table as a body pressure-muscle collaborative constraint rule; one dimension of the two-dimensional lookup table is the body pressure fuzziness level, and the other dimension is the electromyography fuzziness level, and each table entry corresponds to an activation action option.

[0157] Furthermore, the specific steps of C3 include:

[0158] (1) The fuzzy level of body pressure fluctuation value is used as the first dimension of the two-dimensional lookup table. This dimension contains three possible values, namely low level, medium level and high level. The fuzzy level of electromyography amplitude is used as the second dimension of the two-dimensional lookup table. This dimension also contains three possible values, namely low level, medium level and high level. Therefore, the structure of the two-dimensional lookup table is three rows by three columns, and it contains a total of nine entries. Each entry corresponds to a combination of body pressure fuzzy level and electromyography fuzzy level.

[0159] (2) Traverse the nine fuzzy logic rules in the constructed collaborative constraint rule library. For each rule, extract the two elements of adjustment direction and adjustment step size from its consequent, and combine these two elements into a complete activation action option. The format of the activation action option is direction plus step size, where the direction includes three types: positive direction enhancement, negative direction weakening and zero direction maintenance, and the step size is a specific percentage value.

[0160] The specific extraction process is as follows: Extract the consequent from Rule 1, adjust the direction to positive enhancement, and adjust the step size to 16%. Therefore, the activation action option corresponding to Rule 1 is positive enhancement by 16%. Extract the consequent from Rule 2, adjust the direction to positive enhancement, and adjust the step size to 8%. Therefore, the activation action option corresponding to Rule 2 is positive enhancement by 8%. Extract the consequent from Rule 3, adjust the direction to negative weakening, and adjust the step size to 4%. Therefore, the activation action option corresponding to Rule 3 is negative weakening by 4%. Extract the consequent from Rule 4, adjust the direction to positive enhancement, and adjust the step size to 16%. Therefore, the activation action option corresponding to Rule 4 is positive enhancement by 16%. Extract the consequent from Rule 5, and adjust the direction to zero-direction maintenance. The step size is adjusted to zero, therefore the activation action option for Rule 5 is zero-direction hold; the consequent is extracted from Rule 6, the direction is adjusted to negative weaken, and the step size is adjusted to 8%, therefore the activation action option for Rule 6 is negative-direction weaken 8%; the consequent is extracted from Rule 7, the direction is adjusted to positive strengthen, and the step size is adjusted to 24%, therefore the activation action option for Rule 7 is positive-direction strengthen 24%; the consequent is extracted from Rule 8, the direction is adjusted to negative weaken, and the step size is adjusted to 4%, therefore the activation action option for Rule 8 is negative-direction weaken 4%; the consequent is extracted from Rule 9, the direction is adjusted to negative weaken, and the step size is adjusted to 11%, therefore the activation action option for Rule 9 is negative-direction weaken 11%;

[0161] (3) According to the correspondence between the body pressure fuzziness level and the electromyography amplitude fuzziness level, fill the nine activation action options into a three-row, three-column table. Set the row index of the two-dimensional lookup table to the body pressure fuzziness level and arrange them from top to bottom in the order of low level, medium level, and high level. Set the column index of the two-dimensional lookup table to the electromyography amplitude fuzziness level and arrange them from left to right in the order of low level, medium level, and high level.

[0162] (4) In the table position corresponding to both low body pressure ambiguity level and low electromyography amplitude ambiguity level, i.e., the first row and first column, fill in the activation action option with a positive increase of 16%; in the table position corresponding to both low body pressure ambiguity level and medium electromyography amplitude ambiguity level, i.e., the first row and second column, fill in the activation action option with a positive increase of 8%; in the table position corresponding to both low body pressure ambiguity level and high electromyography amplitude ambiguity level, i.e., the first row and third column, fill in the activation action option with a negative decrease of 4%; in the table position corresponding to both medium body pressure ambiguity level and low electromyography amplitude ambiguity level, i.e., the second row and first column, fill in the activation action option with a positive increase of 16%; in the table position corresponding to both medium body pressure ambiguity level and medium electromyography amplitude ambiguity level, fill in the activation action option with a negative decrease of 4%; in the table position corresponding to both medium body pressure ambiguity level and low electromyography amplitude ambiguity level, fill in the activation action option with a positive increase of 16%; in the table position corresponding to both medium body pressure ambiguity level and medium electromyography amplitude ambiguity level, fill in the activation action option with a positive increase of 16%; in the table position corresponding to both medium body pressure ambiguity level and low ... high electromyography amplitude ambiguity level, fill in the activation action option with a negative decrease of 4%; in the table position corresponding to both medium body pressure ambiguity level and low electromyography amplitude ambiguity level, fill in the activation action option with In the corresponding table position (second row, second column), enter the activation action option "Zero direction maintain"; in the table position corresponding to a medium level of body pressure ambiguity and a high level of electromyography (EMG) amplitude ambiguity (second row, third column), enter the activation action option "Negative direction decrease 8%"; in the table position corresponding to a high level of body pressure ambiguity and a low level of EMG amplitude ambiguity (third row, first column), enter the activation action option "Positive direction increase 24%"; in the table position corresponding to a high level of body pressure ambiguity and a medium level of EMG amplitude ambiguity (third row, second column), enter the activation action option "Negative direction decrease 4%"; in the table position corresponding to a high level of body pressure ambiguity and a high level of EMG amplitude ambiguity (third row, third column), enter the activation action option "Negative direction decrease 11%".

[0163] (5) The filled two-dimensional lookup table is converted into a data structure that the system can directly read and execute. The data structure contains three core parts. The first part is the dimension definition. The first dimension is the body pressure fuzziness level, and its three values ​​are fixed in the order of low level, medium level and high level. The second dimension is the electromyography amplitude fuzziness level, and its three values ​​are fixed in the order of low level, medium level and high level. The second part is the table item content. The nine items are stored one by one according to (3). Each item is stored as a tuple containing a direction code and a step size value. The system sets the direction code as +1 to represent positive direction enhancement, zero to represent zero direction maintenance, and -1 to represent negative direction weakening. Therefore, the nine entries are stored as +1 and 16%, +1 and 8%, -1 and 4%, +1 and 16%, zero and zero, -1 and 8%, +1 and 24%, -1 and 4%, and -1 and 11%, respectively. The third part is the execution interface definition, which stipulates that after the system collects the body pressure fluctuation value and electromyography amplitude value in real time, it first calculates the membership degree of the two values ​​to each fuzzy level, takes the fuzzy level with the largest membership degree as the current level, and then locates the corresponding entry in the two-dimensional lookup table according to the two levels. Finally, it reads the direction code and step size value from the entry, determines whether to increase or decrease the current pulse intensity parameter according to the positive or negative of the direction code, and determines the specific adjustment range according to the step size value. The system outputs this complete data structure as the body pressure-muscle co-constraint rule.

[0164] S3: Based on the muscle regulation configuration data, initiate parameter iterative optimization, simultaneously collect physiological feedback signals, and calculate four indicators: electromyographic amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index. When the comprehensive difference of the four indicators is less than a preset first threshold, output the basic muscle function diagnosis result. The physiological feedback signals include electromyographic signals, muscle strength signals, muscle displacement signals, tissue impedance values, body pressure fluctuation values, and respiratory phase information.

[0165] The calculation process for the electromyography amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index includes:

[0166] S3.1: The range of muscle fiber targeted execution action parameters is used as the initial iteration parameter space, the dynamic safety threshold curve is used as the single iteration amplitude constraint boundary, and the body pressure-muscle synergistic constraint rule is used as the multi-source signal coupling constraint condition.

[0167] S3.2: Based on the initial iterative parameter space, the amplitude constraint boundary and the coupling constraint condition, start the iterative optimization process of the electrical execution action parameters. In the iterative optimization process, physiological feedback signals are synchronously collected at a preset first sampling frequency. In this embodiment, the first sampling frequency is set to 1000 times per second.

[0168] S3.3: Based on the acquired electromyographic signals, the electromyographic envelope amplitude sequence is extracted using the sliding window method, and the electromyographic amplitude stability rate is obtained by calculating the variance of the amplitude difference between adjacent cycles. The sliding window method and the variance calculation formula are existing technologies in this field and are not inventive solutions of this application, and will not be described in detail here.

[0169] S3.4: Based on the collected muscle strength signals, the muscle strength achievement rate is obtained by comparing the peak value with the preset training phase muscle strength target value.

[0170] Furthermore, the specific steps in S3.4 include:

[0171] (1) The system determines the target muscle strength value that the user should achieve in the current training phase based on the rehabilitation start phase, current rehabilitation cycle, injury level and rehabilitation goal priority in the training phase parameters. Taking the female user in the example, the user's training phase parameters include the rehabilitation start phase as the basic activation period, the current rehabilitation cycle as the first cycle, the injury level as mild, and the priority of rehabilitation goals as muscle endurance enhancement. The system searches for the corresponding target value in the pre-stored training phase muscle strength target value table based on these parameters. The target value table is established based on clinical rehabilitation guidelines and a large amount of rehabilitation case data. It stipulates that for users in the basic activation period, the first cycle and the injury level as mild, the target muscle strength value is set to 30 mmHg. This value represents the level of muscle strength that the user should achieve when performing a single maximum contraction under ideal conditions. The system uses this target value of 30 mmHg as the standard reference value for subsequent comparisons.

[0172] (2) During the parameter iterative optimization process, physiological feedback signals are synchronously collected 1000 times per second at the first sampling frequency, and muscle force signals are extracted from them. The muscle force signals are collected by the pelvic floor muscle force pressure sensor, reflecting the pressure generated by the user's pelvic floor muscle group during the contraction process. The system extracts muscle force signal segments in each complete contraction cycle on a per-action basis. A complete contraction cycle includes three parts: the contraction rising segment, the peak holding segment, and the relaxation falling segment. Taking a specific contraction action as an example, the system records the muscle force signal waveform within the next 5 seconds after receiving the action trigger signal. The determination of this 5-second time window is based on the fact that the single contraction duration range output by the muscle endurance sub-model is [3.1, 3.5], and the single contraction duration range output by the explosive force sub-model is [0.7, 0.8]. The maximum possible contraction duration of 3.5 seconds is taken and a transition time of about 1 second before and after is added. The 5-second window can completely cover the entire process of any type of single contraction.

[0173] (3) Peak hold processing is performed on the muscle force signal of each contraction cycle. The goal of peak hold processing is to extract the maximum muscle force value reached during the contraction from the muscle force signal waveform of each contraction cycle. The specific processing process is as follows: The system first arranges the muscle force signals within the 5-second window in chronological order to form a sequence containing 5,000 sampling points. Each sampling point corresponds to a muscle force value in mmHg. Starting from the first sampling point, the system initially sets the current maximum value to the value of the first sampling point. Then, the values ​​of each subsequent sampling point are compared with the current maximum value. If the value of the current sampling point is greater than the current maximum value, the value of the sampling point is updated to the new current maximum value. Otherwise, the current maximum value is kept unchanged. After traversing all 5,000 sampling points, the final current maximum value is the peak muscle force of the contraction. For example, during a contraction, the highest value in the muscle force signal sequence recorded by the system appears at the peak of the contraction. The value is 28 mmHg. The system records the peak muscle force of the contraction as 28 mmHg. The system performs the same peak hold processing for each contraction action to obtain the peak muscle force value corresponding to each contraction.

[0174] (4) Compare the peak muscle strength of each contraction with the preset muscle strength target value of the training stage. For each contraction action, the system calculates the relationship between its peak muscle strength and the preset target value of 30 mmHg. The specific comparison method is to determine whether the peak muscle strength of each contraction is greater than or equal to the preset target value. If the peak muscle strength is greater than or equal to 30 mmHg, the system marks the contraction as a qualified contraction. If the peak muscle strength is less than 30 mmHg, the system marks the contraction as an unqualified contraction. Taking the contraction with a peak muscle strength of 28 mmHg in (3) above as an example, since 28 is less than 30, the system marks the contraction as unqualified. Assuming that the peak muscle strength of another contraction is 32 mmHg, the system marks it as qualified.

[0175] (5) Count the proportion of the number of contractions that meet the standard to the total number of contractions within the time window to obtain the muscle strength achievement rate. The system sets the statistical time window to the entire process of each muscle endurance training or explosive power training. Taking muscle endurance training as an example, the execution time of each muscle endurance training is 10 minutes. The system sets that in muscle endurance training, the basic rhythm of the movement output is determined by the duration of a single contraction and the interval time. According to the parameters output by the muscle endurance sub-model, the interval of the duration of a single contraction is [3.1, 3.5], and the interval time is [6.4, 6.8]. Taking the midpoint of the interval, the duration of a single contraction is about 3.3 seconds, the interval time is about 6.6 seconds, and each complete contraction interval cycle is about 9.9 seconds. Within a 10-minute training time, about 61 contraction movements can be completed. The system statistically analyzes the achievement of the target for these 61 contractions. Assuming that 49 of the 61 contractions achieved a peak muscle strength of 30 mmHg or greater, the muscle strength achievement rate is calculated as 49 divided by 61 and then multiplied by 100%, yielding 80.3%. The system rounds to the nearest integer, confirming the muscle strength achievement rate as 80%. The system automatically calculates the muscle strength achievement rate after each training session and uses the result as one of four indicators.

[0176] S3.5: Based on the collected muscle displacement signals, the time delay and amplitude difference of the displacement waveforms of the two sides of the levator ani muscle are extracted using the cross-correlation function calculation method, and the muscle displacement coordination rate is calculated based on the time delay and amplitude difference.

[0177] Furthermore, the specific steps of S3.5 include:

[0178] (1) Extract the muscle displacement signals of the two levator ani muscles from the physiological feedback signals. The muscle displacement signals of the left and right levator ani muscles correspond to two independent signal channels. Each channel records the waveform of the amplitude of muscle tissue displacement over time during the contraction and relaxation of the levator ani muscle on that side. Taking the female user in the embodiment as an example, in a complete contraction action, the system records the bilateral muscle displacement signals within 5 seconds after the action is triggered. Each channel contains 5000 sampling points. The unit of value for each sampling point is mm, which represents the displacement of the levator ani muscle relative to the resting position at that moment. The system ensures that the sampling points of the two channels are completely aligned in time, that is, the Nth sampling point of the left channel and the Nth sampling point of the right channel correspond to the same moment.

[0179] (2) The mean-removing and normalization processes were performed on the displacement waveforms of the levator ani muscles on both sides respectively. The specific method for the mean-removing process is as follows: calculate the arithmetic mean of all 5000 sampling points in the left displacement signal sequence, and then subtract the mean value from the value of each sampling point to obtain the left displacement sequence with zero mean. Similarly, calculate the arithmetic mean of the right displacement signal sequence, and subtract the mean value from the value of each sampling point to obtain the right displacement sequence with zero mean. The specific method for the normalization process is as follows: calculate the sum of squares of the left sequence after the mean-removing process, and then take the square root to obtain the energy of the left sequence. Divide the value of each sampling point in the left sequence by the energy to obtain the energy-normalized left displacement sequence. Perform the same normalization operation on the right sequence to obtain the energy-normalized right displacement sequence.

[0180] (3) The cross-correlation function is used to quantify the relationship between the similarity between the left and right displacement sequences as a function of relative time delay. The formula for calculating the cross-correlation function is as follows: For each possible time delay parameter, the system shifts the right displacement sequence to the left or right relative to the left displacement sequence by multiple sampling points, and then calculates the sum of the products of the two sequences at corresponding positions after the shift. Specifically, in this embodiment, the system sets the search range of the time delay parameter to -100ms to 100ms. Since the sampling frequency is 1000 times per second and the time interval of each sampling point is 1ms, -100ms corresponds to shifting 100 sampling points to the left. 100ms corresponds to a rightward shift of 100 sampling points. For each integer time delay parameter, with 201 possible values ​​from -100 to 100, the system performs the following calculations: after shifting the right displacement sequence by the delay parameter, it multiplies the left displacement sequence point by point within the overlapping time interval, and adds all the products to obtain the cross-correlation function value corresponding to the delay parameter. The cross-correlation function values ​​corresponding to all delay parameters constitute a cross-correlation function sequence with a length of 201 points. Each point corresponds to a specific time delay value. The specific formula of the cross-correlation function is prior art in this field and is not an inventive solution of this application, so it will not be elaborated here.

[0181] (4) The system iterates through all 201 cross-correlation function values, records the maximum value and the corresponding time delay parameter. For example, suppose the system finds the maximum value of the cross-correlation function at a time delay of 24.5ms. This means that when the right displacement sequence is shifted to the right by 24.5ms relative to the left displacement sequence, the two waveforms are most similar. This indicates that the contraction response of the right levator ani muscle is 24.5ms later than that of the left levator ani muscle. The system records this time delay in ms and retains its positive or negative sign to indicate the direction of the delay. If the time delay is positive, it means that the right muscle response lags behind the left. If the time delay is negative, it means that the right muscle response is ahead of the left. If the time delay is zero, it means that the muscles on both sides contract synchronously.

[0182] (5) Extract the amplitude difference from the cross-correlation function. The maximum value of the cross-correlation function ranges from -1 to 1. When the value is close to 1, it indicates that the two displacement waveforms are highly consistent in shape after optimal time alignment, and the amplitude ratio is close to 1:1. When the value is close to zero, it indicates that the two waveforms are almost uncorrelated. When the value is negative, it indicates that the two waveforms change in opposite directions. The absolute value of the maximum value reflects the similarity of the two waveforms under optimal alignment. The similarity is affected by two factors: the similarity of the waveform shape and the difference in amplitude ratio. In order to extract the amplitude difference separately... In addition, the system also needs to calculate the autocorrelation function of the displacement sequences on both sides. The specific method is as follows: calculate the cross-correlation function value of the left displacement sequence with itself when it is translated to zero, that is, the energy of the left sequence, and the cross-correlation function value of the right displacement sequence with itself when it is translated to zero, that is, the energy of the right sequence. Divide the maximum value of the obtained cross-correlation function by the square root of the product of the left and right autocorrelation function values ​​to obtain the normalized cross-correlation coefficient. Assume that the left autocorrelation function value is 1 and the right autocorrelation function value is 1. The normalized cross-correlation coefficient is the maximum value of the cross-correlation function itself. The closer the value is to 1, the better the amplitude consistency.

[0183] (6) The time delay and amplitude difference are combined into a comprehensive muscle displacement coordination rate, where the formula is: ,in, Indicates muscle displacement coordination rate. Indicates a time delay. Indicates the maximum allowed time delay. This represents the time delay weighting coefficient. Indicates the amplitude consistency coefficient. This represents the amplitude difference weighting coefficient. The system sets the maximum allowable time delay to 50ms, meaning a time delay exceeding 50ms is considered to indicate severe muscle incoordination on both sides. The system sets the time delay weighting coefficient to 0.6 and the amplitude difference weighting coefficient to 0.4, with the sum of the two weighting coefficients being 1, reflecting that the time delay has a greater impact on coordination than the amplitude difference. Assuming a time delay of 24.5ms and an amplitude consistency coefficient of 0.92, the time delay component is calculated first. The absolute value of the time delay, 24.5ms, divided by the maximum allowable time delay of 50ms, yields 0. 49, multiplied by the time delay weighting coefficient of 0.6, yields 0.294. Then, the amplitude difference is calculated: 1 minus the amplitude consistency coefficient of 0.92, yields 0.08. Multiplying this by the amplitude difference weighting coefficient of 0.4, yields 0.032. Finally, the muscle displacement coordination rate is calculated as 1 minus 0.294 minus 0.032, yielding 0.674. After conversion to a percentage, this is determined to be 67.4%. This value indicates that the degree of displacement coordination of the two levator ani muscles during this contraction is 67.4%. The higher the value, the more synchronized and balanced the contraction of the two muscles.

[0184] (7) Statistically average the muscle displacement coordination rate of all contraction movements during each training session. Taking muscle endurance training as an example, approximately 61 contraction movements are completed within a 10-minute training period. The system repeats the calculations (2)-(6) above for each contraction movement to obtain 61 muscle displacement coordination rate values. The system adds these 61 values ​​together and divides them by 61 to obtain the average muscle displacement coordination rate of this training session. Assuming that the sum of the coordination rates of the 61 contractions is 41.10, the average muscle displacement coordination rate is equal to 41.10 divided by 61, which equals 0.6737, or 67.4%, which is the same as the result of a single calculation.

[0185] S3.6: Based on the collected tissue impedance values, body pressure fluctuation values, and respiratory phase information, the muscle fatigue index is calculated using the normalized energy integral to recovery time ratio method.

[0186] Furthermore, the specific steps in S3.6 include:

[0187] (1) During the parameter iteration optimization process, physiological feedback signals are synchronously collected at a first sampling frequency of 1000 times per second, and the tissue impedance value sequence collected by the vaginal impedance detection electrode, the body pressure fluctuation value sequence collected by the body pressure sensing airbag, and the respiratory phase information collected by the respiratory motion sensor are extracted from them.

[0188] (2) The continuous acquisition time is divided into multiple independent respiratory cycles based on respiratory phase information. The system sets the criterion for a complete respiratory cycle as the respiratory phase information rises from -0.9 to 0.9 and then returns to -0.9. The system iterates through 600,000 respiratory phase sampling points. Whenever the respiratory phase is detected to rise from below -0.9 and eventually return to below -0.9, the system records the start and end times of the cycle. Taking this user as an example, assuming that his / her respiratory rate is 12 times per minute, there are approximately 120 respiratory cycles in 600 seconds, and the duration of each respiratory cycle is approximately 5 seconds.

[0189] (3) Based on the start and end times of each respiratory cycle, the system extracts the tissue impedance value segments corresponding to the time period from the complete tissue impedance value sequence. Each respiratory cycle lasts 5 seconds and corresponds to 5000 sampling points. The system performs energy integration calculation on the tissue impedance value sequence within each respiratory cycle. The specific method of energy integration is as follows: first, the tissue impedance values ​​of all sampling points within the respiratory cycle are squared, and then all the squared values ​​are added together to obtain the energy integration of the tissue impedance values ​​within the respiratory cycle. For example, if the tissue impedance value sequence within a respiratory cycle is 80Ω, 81Ω, 82Ω up to 120Ω, the energy integration is obtained by squaring each value and summing them together.

[0190] (4) Using the same segmentation method, extract the body pressure fluctuation value segment corresponding to each respiratory cycle from the complete body pressure fluctuation value sequence according to the respiratory cycle boundary. Each respiratory cycle contains 5000 body pressure fluctuation value sampling points. The system performs energy integration calculation on the body pressure fluctuation value sequence in the segment. The specific method is: first, square the body pressure fluctuation value of each sampling point, and then add all the squared values ​​to obtain the energy integration of the body pressure fluctuation value in the respiratory cycle.

[0191] (5) Divide the energy integral of the tissue impedance value calculated in each respiratory cycle by the duration of the respiratory cycle to obtain the average power of the cycle, and then divide it by the square of the baseline value of the tissue impedance in the user's resting state to obtain the normalized energy integral.

[0192] (6) Divide the energy integral of the body pressure fluctuation value calculated in each respiratory cycle by the duration of the respiratory cycle to obtain the average power, and then divide it by the square of the user's resting body pressure baseline value to obtain the normalized energy integral of the body pressure fluctuation value.

[0193] (7) Identify muscle recovery characteristics from the normalized energy integral curve of body pressure fluctuation value in each respiratory cycle. After each contraction action, the normalized energy integral of body pressure fluctuation value will gradually decrease from the peak value back to the resting level. The system sets the calculation method for the recovery time ratio as follows: from the end of the contraction to the point where the normalized energy integral of body pressure fluctuation value decreases to 37% of the difference between the peak value and the resting value, divided by the total duration of the respiratory cycle; the maximum normalized energy integral of body pressure fluctuation value in each respiratory cycle is taken as the peak value, and the recovery time ratio is taken as the starting point of the respiratory cycle. The normalized energy integral of the body pressure fluctuation value at any given moment is used as the resting value. Assuming that within a respiratory cycle, the peak value occurs at the end of systole at 180, and the resting value is 120, with a difference of 60, then the 37% drop threshold corresponds to the difference between the peak value and the resting value multiplied by 0.37, i.e., 180 minus 22.2 equals 157.8. The system searches backward from the end of systole to find the moment when the normalized energy integral of the body pressure fluctuation value first drops to 157.8, and calculates the time difference between this moment and the end of systole, assuming this time difference is 1.5 seconds. The total duration of this respiratory cycle is 5 seconds, then the recovery time ratio equals 1.5 divided by 5, which equals 0.3.

[0194] (8) Compare the normalized energy integral of the tissue impedance value of the current respiratory cycle with the normalized energy integral of the tissue impedance value of the first respiratory cycle at the start of training, and calculate the fatigue decay factor. Assuming that the normalized energy integral of the tissue impedance value of the first respiratory cycle at the start of training is 1.20 and the normalized energy integral of the tissue impedance value of the current respiratory cycle is 1.04, then the fatigue decay factor is approximately 0.867 (1.04 divided by 1.20). The system then multiplies this fatigue decay factor by the recovery time ratio to obtain the preliminary muscle fatigue index. The preliminary muscle fatigue index is approximately 0.260 (0.867 multiplied by 0.3). The system calculates the preliminary muscle fatigue index... The fatigue index is converted into a percentage and mapped to a range of 0 to 100. The mapping method is as follows: when the fatigue decay factor is 1 and the recovery time ratio is 1, the muscle fatigue index is 0, indicating no fatigue at all. When the fatigue decay factor is 0 and the recovery time ratio is 1, the muscle fatigue index is 100, indicating complete fatigue. In actual calculation, the system uses the muscle fatigue index equal to 1 minus the difference between the fatigue decay factor and the recovery time ratio, and then multiplying it by 100, i.e., 1-0.260=0.74, and then multiplying it by 100 to get 74. Therefore, the muscle fatigue index for this respiratory cycle is 74. The higher the value, the more severe the muscle fatigue, and the lower the value, the better the muscle condition.

[0195] (9) Statistically average the muscle fatigue index of all breathing cycles during each training session. Taking muscle endurance training as an example, there are approximately 120 breathing cycles in a 10-minute training session. Repeat (3)-(8) above for each breathing cycle to obtain 120 muscle fatigue index values. The system adds up these 120 values ​​and divides them by 120 to obtain the average muscle fatigue index of the training session. Assuming that the sum of the muscle fatigue indices of the 120 cycles is 8400, the average muscle fatigue index is equal to 8400÷120=70.

[0196] When the combined difference of the four indicators is less than a preset first threshold, the basic muscle function diagnosis result is output, including:

[0197] S3.7: The differences between the electromyographic amplitude stabilization rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index and their respective preset ideal target values ​​are calculated to obtain the amplitude stabilization difference, muscle strength achievement difference, displacement coordination difference, and fatigue index difference. The first weight coefficient, the second weight coefficient, the third weight coefficient, and the fourth weight coefficient are dynamically allocated according to the training stage parameters. In this embodiment, the first weight coefficient is set to 0.4, the second weight coefficient is set to 0.3, the third weight coefficient is set to 0.2, and the fourth weight coefficient is set to 0.1.

[0198] S3.8: Using the weighted Euclidean distance calculation method, the amplitude stability difference is multiplied by the first weighting coefficient to obtain the first weighted difference, the muscle strength achievement difference is multiplied by the second weighting coefficient to obtain the second weighted difference, the displacement coordination difference is multiplied by the third weighting coefficient to obtain the third weighted difference, and the fatigue index difference is multiplied by the fourth weighting coefficient to obtain the fourth weighted difference. The first weighted difference, the second weighted difference, the third weighted difference, and the fourth weighted difference are squared, summed, and then squared to obtain the comprehensive difference of the four indicators.

[0199] S3.9: Determine whether the comprehensive difference of the four indicators is less than a preset first threshold. If it is less than the preset first threshold, determine that the current user's comprehensive measured state of electromyography, muscle strength, muscle displacement, and anti-fatigue performance is an ideal standard state. Determine that the parameter iterative calibration for muscle function analysis is completed. Based on the four functional indicators, the differences in each dimension, and the comprehensive difference of the four indicators obtained by this iteration, output the basic muscle function diagnosis result. In this embodiment, the first threshold is set to 5%.

[0200] S4: After receiving the basic muscle function diagnosis results, output physiological training execution actions according to the preset action rhythm, and continuously collect physiological feedback signals during the training execution process to identify abnormal muscle symptoms.

[0201] The specific steps of S4 include:

[0202] S4.1: In response to the diagnostic results of basic muscle function, load the pre-stored training execution action sequence template; the training execution action sequence template includes a sequentially looping muscle endurance training phase, explosive power training phase, and relaxation interval phase;

[0203] Furthermore, the training execution action sequence template is constructed as follows: The content and duration of a single loop are determined. Specifically, a complete loop consists of three sequentially executed phases. The first phase is the muscular endurance training phase, lasting 10 minutes. During this phase, the system outputs actions according to the first set of execution action parameters output by the muscular endurance sub-model. Specific parameters include a pulse frequency of 12.4 to 13.6 times per second, a single contraction duration of [3.1, 3.5], and an interval time of [6.4, 6.8]. The second phase is the explosive power training phase, lasting the total time required to complete three sets of explosive power training. Each set of explosive power training includes 12 to 13 contractions, each contraction including a single contraction duration of [0.7, 0.8] and an interval time of 2.5 to 2.7 seconds. After each set, there is a rest period of 17.2 to 17.9 seconds. The duration of a single set is calculated using the median of the parameters, with each of the 12 contractions lasting 0.75 seconds plus a 2.6-second interval. The total duration of each contraction is 3.35 seconds, with 12 contractions totaling 40.2 seconds. Adding the 17.5-second rest between sets, the total duration of a single set is approximately 57.7 seconds, and the total duration of three sets is approximately 173.1 seconds. The third phase is a relaxation interval phase, lasting 3 minutes. During this phase, the system does not output any pelvic floor electrical stimulation movements; the user is only required to maintain calm breathing and full-body relaxation, allowing the pelvic floor muscles to fully recover from the previous training load. The number of cycles is determined based on the expected recovery time in the training phase parameters and the total duration of each rehabilitation training session. Taking a female user in the example, the expected recovery time is 12 weeks, with three rehabilitation training sessions per week. The total duration of each rehabilitation training session is typically set to 20 to 30 minutes, so the total duration of each training session is taken as 25 minutes. Subtracting the 3-minute relaxation interval phase leaves 22 minutes. Each cycle includes 10 minutes of muscle endurance training, 2.885 minutes of explosive power training, and a 3-minute relaxation interval, totaling approximately 15.885 minutes.

[0204] S4.2: Using the final pulse intensity parameter and pulse frequency parameter determined at the time of completion of parameter iteration calibration as the basic parameters of the output action, and according to the training execution action timing template, a physiological training execution action sequence arranged on the time axis is generated in combination with the basic parameters, and the physiological training execution action is output.

[0205] S4.3: During the execution of physiological training actions, physiological feedback signals are continuously collected at a preset second sampling frequency, and a real-time physiological feedback buffer queue is established; the second sampling frequency is greater than the first sampling frequency. In this embodiment, the second sampling frequency is set to 2000 times per second.

[0206] S4.4: Extract electromyographic signals and muscle displacement signals within the current execution action cycle from the real-time physiological feedback buffer queue, calculate the real-time electromyographic peak amplitude based on the electromyographic signals, and calculate the real-time muscle contraction displacement amplitude based on the muscle displacement signals. The amplitude calculation formula is existing technology in this field and is not an inventive solution of this application, so it will not be elaborated here.

[0207] Furthermore, the extracted electromyographic signal fragments are subjected to real-time peak amplitude calculation, including: first, a sliding window width of 100ms is set. Since the second sampling frequency is 2000 times per second, 100ms corresponds to 200 sampling points; starting from the first sampling point of the electromyographic signal fragment, the window is placed on the first 200 sampling points, the maximum value among these 200 sampling points is found, and the maximum value and its position in the window are recorded; then the window is slid backward by 50ms, i.e., 100 sampling points, and the search for the maximum value of the 200 sampling points in the new window is repeated; the system continuously repeats the sliding and search operation until the window covers 200 sampling points; finally, a series of local maximum values ​​of the window are obtained; the system then finds the global maximum value from these local maximum values, and the global maximum value is the real-time peak amplitude of the current contraction action.

[0208] Furthermore, the extracted muscle displacement signal segments are subjected to real-time contraction displacement amplitude calculation, including: setting the sliding window width to 100ms, i.e., 200 sampling points; starting from the first sampling point of the muscle displacement signal segment, searching for the maximum value within the window, recording it, then sliding for 50ms, i.e., 100 sampling points, and continuing to search for the maximum value within the next window until all 200 sampling points are covered; finding the global maximum value from all the local maximum values ​​in the windows, this global maximum value is the real-time muscle contraction displacement amplitude of this contraction action. It should be noted that the muscle displacement signal is usually close to 0mm in the resting state. As the muscle contraction displacement gradually increases, it reaches the maximum value at the contraction peak and then gradually falls back to 0mm. Therefore, the global maximum value appears near the contraction peak. Assuming that the maximum displacement of the muscle displacement signal segment at the contraction peak during a contraction is 8.5mm, the system will record the real-time muscle contraction displacement amplitude as 8.5mm.

[0209] S4.5: Compare the real-time electromyography peak amplitude with the preset upper limit of the electromyography spasm threshold. If the real-time electromyography peak amplitude exceeds the upper limit of the electromyography spasm threshold and the duration exceeds 50ms, it is determined that the user's muscles have abnormal spasm and a muscle spasm identifier is generated. In this embodiment, the upper limit of the electromyography spasm threshold is set to 200 microvolts.

[0210] S4.6: Compare the real-time muscle contraction displacement amplitude with the preset lower limit of the displacement deficiency threshold. If the real-time muscle contraction displacement amplitude is lower than the lower limit of the displacement deficiency threshold and the number of consecutive occurrences exceeds the preset number threshold, it is determined that the user has abnormal muscle strength deficiency and a muscle strength deficiency indicator is generated. In this embodiment, the lower limit of the displacement deficiency threshold is set to 2.5mm and the preset number threshold is set to 3 times.

[0211] S4.7: Identify abnormal muscle symptoms during training by identifying muscle spasms and muscle weakness.

[0212] After generating the muscle spasm identifier, an automatic motion intensity regression mechanism is triggered, including:

[0213] D1: In response to the muscle spasm identifier, obtain the latest consecutive multiple waveform cycles of the electromyographic signal within the current execution action cycle, and calculate the root mean square value of the electromyographic peak value of the latest consecutive multiple waveform cycles. The root mean square calculation formula is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0214] D2: Find the impedance dynamic upper limit threshold curve value corresponding to the current training time point from the dynamic safety threshold curve, and determine the smaller of the electromyography peak root mean square value and the impedance dynamic upper limit threshold curve value as the regression target reference value.

[0215] D3: Using a linear decreasing strategy, the current pulse intensity parameter is gradually reduced according to a preset backtracking step size, such as 0.1, and the electromyographic signal is re-acquired after each reduction step. When the peak amplitude of the real-time electromyographic signal is lower than the preset spasm marker removal threshold, the backtracking is stopped and the muscle spasm marker is cleared. In this embodiment, the spasm marker removal threshold is set to 160μV.

[0216] After generating the muscle weakness indicator, a progressive enhancement mechanism for movement intensity is triggered, including:

[0217] E1: In response to the muscle weakness indicator, extract the peak sequence of muscle displacement signals within multiple execution action cycles in the real-time physiological feedback buffer queue, calculate the slope of the increasing trend of the peak sequence of muscle displacement signals, and determine whether the slope of the increasing trend is positive and exceeds a preset effective trend threshold. In this embodiment, the effective trend threshold is set to 0.05 mm / s. The slope calculation formula is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0218] E2: If the slope of the increasing trend is positive and exceeds the effective threshold of the trend, the pulse intensity parameter is gradually increased according to the preset incremental step size. After each increase, wait for a complete execution cycle and re-evaluate the peak value of the muscle displacement signal. In this embodiment, the incremental step size is set to 0.08.

[0219] E3: If the peak value of the muscle displacement signal fails to reach the lower limit of the insufficient displacement threshold and the insufficient muscle strength indicator persists in multiple consecutive execution action cycles, the current execution action frequency is simultaneously adjusted to the optimal induction frequency range of fast-twitch muscle fibers, and the muscle strength attainment status is reassessed. In this embodiment, the optimal induction frequency range of fast-twitch muscle fibers is set to 24.3~25.7 times / second.

[0220] Example 2:

[0221] Please see Figure 5 Another embodiment of the present invention provides: an intelligent physiological training parameter optimization and adjustment system, comprising:

[0222] The parameter acquisition module 10 is used to acquire the multi-source baseline physiological signal sequence of the target user and the corresponding training phase parameters.

[0223] The regulation configuration generation module 20 identifies the user's skeletal muscle fiber type and constructs a multi-training target decoupled regulation model based on multi-source baseline physiological signal sequences and training phase parameters, generating muscle regulation configuration data that includes muscle fiber targeted execution action parameter ranges, dynamic safety threshold curves, and body pressure-muscle synergistic constraint rules.

[0224] The functional diagnosis module 30 initiates parameter iterative optimization based on muscle regulation configuration data, simultaneously collects physiological feedback signals such as electromyography, muscle strength, and muscle displacement signals, calculates four indicators, and outputs basic muscle function diagnosis results when the comprehensive difference of the four indicators is less than the preset first threshold, thus completing the calibration of rehabilitation parameters and the basic assessment of pelvic floor function.

[0225] The abnormality identification module 40 is used to receive the diagnostic results of basic muscle function, output physiological training execution actions according to the preset action rhythm, continuously collect physiological feedback signals during the rehabilitation process, identify muscle abnormalities such as muscle spasms and insufficient muscle strength, and ensure the normal development of rehabilitation training and abnormality warning.

[0226] The abnormal adaptive adjustment module 50 is used to trigger the automatic regression mechanism and the progressive enhancement mechanism of movement intensity respectively after identifying muscle abnormalities such as muscle spasms and insufficient muscle strength. It dynamically adjusts training parameters based on dynamic safety threshold curves, physiological feedback signals, etc., to eliminate abnormal states and ensure the safe and effective conduct of rehabilitation training.

[0227] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A method for intelligently optimizing and adjusting physiological training parameters, characterized in that, include: S1: Obtain the multi-source baseline physiological signal sequence of the target user and the corresponding training phase parameters; S2: Based on the multi-source baseline physiological signal sequence and the training phase parameters, identify the user's skeletal muscle fiber type and construct a multi-training target decoupled regulation model to generate muscle regulation configuration data; the muscle regulation configuration data includes muscle fiber targeted execution action parameter range, dynamic safety threshold curve, and body pressure-muscle synergistic constraint rules. S3: Based on the muscle regulation configuration data, start parameter iterative optimization, synchronously collect physiological feedback signals, and calculate four indicators: electromyographic amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate and muscle fatigue index. When the comprehensive difference of the four indicators is less than the preset first threshold, output the basic muscle function diagnosis result. The physiological feedback signals include electromyography signals, muscle strength signals, muscle displacement signals, tissue impedance values, body pressure fluctuation values, and respiratory phase information; S4: After receiving the basic muscle function diagnosis results, output physiological training execution actions according to the preset action rhythm, and continuously collect physiological feedback signals during the training execution process to identify abnormal muscle symptoms.

2. The intelligent physiological training parameter optimization and adjustment method as described in claim 1, characterized in that, The specific steps of S2 include: Based on the electromyographic baseline signal in the multi-source baseline physiological signal sequence, the first dominant frequency corresponding to slow contraction muscle fibers and the second dominant frequency corresponding to fast contraction muscle fibers are extracted using the sliding window superposition averaging method. At the same time, based on the muscle strength baseline signal and muscle displacement baseline signal in the multi-source baseline physiological signal sequence, the first fatigue slope of slow contraction muscle fibers in continuous contraction state and the second fatigue slope of fast contraction muscle fibers in staged contraction state are calculated using the waveform area integration method. Based on the first dominant frequency, the second dominant frequency, the first fatigue slope, the second fatigue slope, and the training phase parameters, the first proportion of slow-twitch muscle fibers and the second proportion of fast-twitch muscle fibers of the user are determined by a preset muscle fiber type discrimination decision tree. Based on the first proportion value, the second proportion value, and the training stage parameters, a multi-training-target decoupled regulation model is constructed, which includes a sub-target of enhancing muscle endurance, a sub-target of enhancing explosive power, and a sub-target of enhancing coordination and control. The multi-training-target decoupled regulation model outputs the range of muscle fiber targeted execution action parameters, dynamic safety threshold curves, and body pressure-muscle synergistic constraint rules, which are summarized as muscle regulation configuration data.

3. The intelligent physiological training parameter optimization and adjustment method as described in claim 2, characterized in that, The process of generating the muscle fiber targeted execution action parameter range includes: Construct a muscle endurance sub-model corresponding to the muscle endurance enhancement sub-target, set the first proportion value obtained as the first input of the muscle endurance sub-model, set the second proportion value as the second constraint input of the muscle endurance sub-model, and output the first set of execution action parameters adapted to the rehabilitation of slow contraction muscle fibers through the muscle endurance sub-model; Construct a sub-model of explosive force corresponding to the sub-target of explosive force enhancement, set the second proportion value as the first input of the sub-model of explosive force, set the first proportion value as the second constraint input of the sub-model of explosive force, and output the second set of execution action parameters adapted to fast contraction muscle fiber rehabilitation through the sub-model of explosive force. Construct a coordination control sub-model corresponding to the coordination control enhancement sub-target, input the first proportion value and the second proportion value into the coordination control sub-model, and output a set of temporal coupling parameters for alternately activating slow contraction muscle fibers and fast contraction muscle fibers based on a preset muscle group co-activation temporal template. The first set of execution action parameters, the second set of execution action parameters, and the temporal coupling parameter set are spliced ​​together in a non-overlapping temporal domain to form a muscle fiber targeted execution action parameter range.

4. The intelligent physiological training parameter optimization and adjustment method as described in claim 2, characterized in that, The process of generating the dynamic safety threshold curve includes: Based on the baseline values ​​of tissue impedance and body pressure fluctuation in the multi-source baseline physiological signal sequence, the lower and upper percentiles of impedance under resting conditions are calculated using percentile statistics. At the same time, the mean and standard deviation of resting body pressure are calculated, and the lower and upper percentiles of impedance, the mean and standard deviation of resting body pressure are used as basic safety boundary parameters. The injury severity level parameter and training days parameter are obtained from the training phase parameters. The basic safety boundary parameter, the injury severity level parameter, and the training days parameter are fused using an exponentially weighted moving average method to generate a dynamic safety threshold curve that changes over time. The dynamic safety threshold curve includes a dynamic lower limit threshold curve for impedance, a dynamic upper limit threshold curve for impedance, and a dynamic peak threshold curve for body pressure.

5. The intelligent physiological training parameter optimization and adjustment method as described in claim 2, characterized in that, The generation process of the body pressure-muscle synergistic constraint rule includes: Based on the baseline signals of body pressure fluctuations and electromyography (EMG) in the multi-source baseline physiological signal sequence, the Granger causality analysis method was used to calculate the causal influence coefficient of body pressure changes on EMG changes, and the feedback inhibition coefficient of EMG changes on body pressure changes was also calculated. Based on the causal influence coefficient and the feedback inhibition coefficient, a collaborative constraint rule base containing multiple fuzzy logic rules is constructed. The antecedent of the fuzzy logic rule includes the membership degree of body pressure fluctuation value and the membership degree of electromyography amplitude value. The consequent of the fuzzy logic rule includes the allowed direction of motion intensity adjustment and the adjustment step size. The collaborative constraint rule base is converted into a two-dimensional lookup table, and the two-dimensional lookup table is used as the body pressure-muscle collaborative constraint rule; one dimension of the two-dimensional lookup table is the body pressure fuzziness level, and the other dimension is the electromyography fuzziness level, and each table entry corresponds to an activation action option.

6. The intelligent physiological training parameter optimization and adjustment method as described in claim 1, characterized in that, The calculation process for the electromyography amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index includes: The range of muscle fiber targeted execution action parameters is used as the initial iteration parameter space, the dynamic safety threshold curve is used as the single iteration amplitude constraint boundary, and the body pressure-muscle synergistic constraint rule is used as the multi-source signal coupling constraint condition. Based on the initial iterative parameter space, the amplitude constraint boundary, and the coupling constraint conditions, the iterative optimization process of the electrical actuation action parameters is initiated. During the iterative optimization process, physiological feedback signals are synchronously acquired at a preset first sampling frequency. Based on the electromyographic signal, the electromyographic envelope amplitude sequence is extracted using the sliding window method, and the electromyographic amplitude stability rate is obtained by calculating the variance of the amplitude difference between adjacent cycles. Based on the muscle strength signal, the muscle strength achievement rate is obtained by comparing the peak value with the preset training phase muscle strength target value. Based on the muscle displacement signal, the time delay and amplitude difference of the displacement waveforms of the two sides of the levator ani muscle are extracted using the cross-correlation function calculation method, and the muscle displacement coordination rate is calculated based on the time delay and amplitude difference. Based on the collected tissue impedance values, body pressure fluctuation values, and respiratory phase information, the muscle fatigue index was calculated using the normalized energy integral to recovery time ratio method.

7. The intelligent physiological training parameter optimization and adjustment method as described in claim 1, characterized in that, When the combined difference of the four indicators is less than a preset first threshold, the basic muscle function diagnosis result is output, including: The differences between the electromyographic amplitude stability rate, muscle strength achievement rate, muscle displacement coordination rate, and muscle fatigue index and their respective preset ideal target values ​​are calculated to obtain the amplitude stability difference, muscle strength achievement difference, displacement coordination difference, and fatigue index difference. The first weight coefficient, the second weight coefficient, the third weight coefficient, and the fourth weight coefficient are dynamically allocated according to the training stage parameters. The first weighted difference is obtained by multiplying the amplitude stability difference by the first weighting coefficient using the weighted Euclidean distance calculation method; the second weighted difference is obtained by multiplying the muscle strength achievement difference by the second weighting coefficient; the third weighted difference is obtained by multiplying the displacement coordination difference by the third weighting coefficient; and the fourth weighted difference is obtained by multiplying the fatigue index difference by the fourth weighting coefficient. The sum of the squares of the first, second, third, and fourth weighted differences is then taken as the square root to calculate the comprehensive difference of the four indicators. Determine whether the combined difference of the four indicators is less than a preset first threshold. If it is less than the preset first threshold, determine that the current user's combined measured state of electromyography, muscle strength, muscle displacement, and anti-fatigue performance is an ideal standard state. Determine that the parameter iterative calibration for muscle function analysis is complete. Based on the four indicators, the differences in each dimension, and the combined difference of the four indicators obtained from this iteration, output the basic muscle function diagnosis result.

8. The intelligent physiological training parameter optimization and adjustment method as described in claim 1, characterized in that, The specific steps of S4 include: In response to the diagnostic results of basic muscle function, a pre-stored training execution action sequence template is loaded; the training execution action sequence template includes a sequentially cyclical muscle endurance training phase, a power training phase, and a relaxation interval phase; The final pulse intensity and pulse frequency parameters determined at the time of completion of parameter iteration calibration are used as the basic parameters for output action. According to the training execution action time sequence template, the physiological training execution action sequence arranged on the time axis is generated in combination with the basic parameters, and the physiological training execution action is output. During the execution of physiological training actions, physiological feedback signals are collected at a preset second sampling frequency to establish a real-time physiological feedback buffer queue. Extract electromyography (EMG) signals and muscle displacement signals within the current execution cycle from the real-time physiological feedback buffer queue; calculate the real-time peak amplitude of EMG based on the EMG signals; and calculate the real-time muscle contraction displacement amplitude based on the muscle displacement signals. If the peak amplitude of the real-time electromyography exceeds the preset upper limit of the electromyography spasm threshold and the duration exceeds the preset spasm confirmation time window, it is determined that the user's muscles have abnormal spasm and a muscle spasm identifier is generated. If the real-time muscle contraction displacement amplitude is lower than the preset displacement insufficiency threshold and the number of consecutive occurrences exceeds the preset number threshold, then it is determined that the user has abnormal muscle strength deficiency and a muscle strength deficiency flag is generated. The identification of abnormal muscle conditions during training is accomplished by identifying muscle spasms and muscle weakness.

9. The intelligent physiological training parameter optimization and adjustment method as described in claim 8, characterized in that, After generating the muscle spasm identifier, an automatic motion intensity regression mechanism is triggered, including: In response to the muscle spasm indicator, the latest consecutive multiple waveform cycles of the electromyographic signal within the current execution action cycle are obtained, and the root mean square value of the electromyographic peak value of the latest consecutive multiple waveform cycles is calculated; Find the impedance dynamic upper limit threshold curve value corresponding to the current training time point from the dynamic safety threshold curve, and determine the smaller of the electromyography peak root mean square value and the impedance dynamic upper limit threshold curve value as the regression target reference value. A linear decreasing strategy is adopted to gradually reduce the current pulse intensity parameter according to a preset backtracking step size, and to re-acquire the electromyographic signal after each step of reduction. When the peak amplitude of the real-time electromyographic signal is lower than the preset spasm marker removal threshold, the backtracking is stopped and the muscle spasm marker is cleared.

10. The intelligent physiological training parameter optimization and adjustment method as described in claim 8, characterized in that, After generating the muscle weakness indicator, a progressive enhancement mechanism for movement intensity is triggered, including: In response to the muscle weakness indicator, the peak sequence of muscle displacement signal within multiple execution action cycles in the real-time physiological feedback buffer queue is extracted, the increasing trend slope of the peak sequence of muscle displacement signal is calculated, and it is determined whether the increasing trend slope is positive and exceeds a preset trend effective threshold. If the slope of the increasing trend is positive and exceeds the effective threshold of the trend, the pulse intensity parameter is gradually increased according to the preset incremental step size. After each increase, wait for a complete execution cycle and re-evaluate the peak value of the muscle displacement signal. If the peak value of the muscle displacement signal fails to reach the lower limit of the insufficient displacement threshold and the insufficient muscle strength indicator persists for multiple consecutive execution cycles, the current execution frequency will be adjusted up to the range of the optimal induction frequency for fast-twitch muscle fibers, and the muscle strength attainment status will be reassessed.

11. An intelligent physiological training parameter optimization and adjustment system, used to implement the intelligent physiological training parameter optimization and adjustment method according to any one of claims 1-10, characterized in that, include: The parameter acquisition module is used to acquire multi-source baseline physiological signal sequences and corresponding training phase parameters. The regulation configuration generation module identifies the user's skeletal muscle fiber type and constructs a multi-training target decoupled regulation model based on multi-source baseline physiological signal sequences and training phase parameters, generating muscle regulation configuration data. The functional diagnosis module initiates parameter iterative optimization based on muscle regulation configuration data, synchronously collects physiological feedback signals, calculates four indicators, and outputs basic muscle function diagnosis results when the comprehensive difference of the four indicators is less than the preset first threshold. The anomaly recognition module is used to receive the diagnostic results of basic muscle function, output physiological training execution actions according to the preset movement rhythm, continuously collect physiological feedback signals during the training execution process, and identify abnormal muscle symptoms. The abnormal adaptive adjustment module is used to dynamically adjust training parameters by triggering an automatic regression mechanism and a progressive enhancement mechanism for movement intensity after identifying abnormal muscle conditions.

Images