Chronic lumbago rehabilitation method and system combined with electromyographic signal mode recognition

Abstract

The invention discloses a chronic lumbago rehabilitation method and system combined with electromyographic signal mode recognition, and relates to the technical field of electromyographic signal recognition. By collecting electromyographic data of lumbar and abdominal muscle groups in real time, extracting characteristics such as an integral electromyographic value and a median frequency, and calculating dynamic adjustment parameters based on a change rate in a continuous time window; a dynamic fatigue threshold interval is formed, and the real-time fatigue state of the patient is judged. According to the comparison relation between the instantaneous fatigue degree data and the dynamic threshold value, an adjusting coefficient used for adjusting the output power of the rehabilitation training equipment is generated, and the training intensity can be adjusted in a self-adaptive mode along with the muscle state. According to the method, excessive training or insufficient training can be effectively avoided, the safety, individuation and intelligence level of rehabilitation training can be improved, meanwhile, the real-time intervention burden of a rehabilitator can be relieved, the training efficiency and the rehabilitation effect can be improved, and the advantages of closed-loop dynamic regulation and control can be fully embodied.

Description

Technical Field

[0001] This invention relates to the field of electromyography (EMG) signal recognition technology, specifically to a method and system for chronic low back pain rehabilitation that combines EMG signal pattern recognition. Background Technology

[0002] Chronic low back pain is a common musculoskeletal disorder, and long-term pain and functional impairment severely affect patients' daily lives and mobility. Current rehabilitation methods typically rely on physical therapy, rehabilitation equipment, or biofeedback technology to help patients restore the function of their abdominal and lumbar muscle groups. Among these, surface electromyography (EMG), as an important physiological parameter reflecting muscle activity, has been widely used in rehabilitation training for assessing movement status and monitoring fatigue. However, existing methods largely focus on collecting and displaying static indicators such as EMG signal amplitude and frequency domain characteristics, making it difficult to effectively reflect the dynamic fatigue changes experienced by patients during training.

[0003] In existing technologies, rehabilitation training equipment typically relies on preset fixed thresholds or manually set training intensities for output control, failing to fully incorporate the dynamic changes in electromyographic (EMG) signals. When patients experience fatigue accumulation during training, the equipment cannot promptly identify and adjust its output power, easily leading to a mismatch between the rehabilitation training intensity and the patient's actual tolerance, thus affecting rehabilitation outcomes and even increasing the risk of secondary injury. Therefore, how to combine EMG signal pattern recognition methods to determine the patient's real-time fatigue state and dynamically adjust the output power of the rehabilitation training equipment based on this determination is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for chronic low back pain rehabilitation that combines electromyographic signal pattern recognition.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention discloses a method for rehabilitation of chronic low back pain that combines electromyographic signal pattern recognition, comprising the following steps:

[0007] Obtain raw electromyographic data of the target muscle groups in the waist and abdomen during the user's use of rehabilitation training equipment;

[0008] The raw electromyography (EMG) data is preprocessed, and a sliding time window mechanism is used to extract the integrated EMG values ​​and median frequency data of the raw EMG data within the current time window.

[0009] Based on the latest continuous The rate of change of integral electromyography values ​​and median frequency data within a time window is used to calculate dynamic adjustment parameters.

[0010] The dynamic adjustment parameter is added to the static baseline threshold determined based on the user's historical data to obtain the upper limit of the dynamic fatigue threshold range at the current moment. The upper limit is then multiplied by a preset scaling factor to obtain the lower limit of the dynamic fatigue threshold range.

[0011] Based on the integrated electromyography values ​​and median frequency data of the current time window, instantaneous fatigue data is generated by calculating using a fatigue state discrimination model.

[0012] The instantaneous fatigue data is compared with the dynamic fatigue threshold interval to determine the relative position of the instantaneous fatigue data with respect to the dynamic fatigue threshold interval, and a corresponding adjustment coefficient value for adjusting the output power of the rehabilitation training device is generated based on the relative position relationship.

[0013] The adjustment coefficient value is sent to the rehabilitation training device.

[0014] Secondly, this invention discloses a chronic low back pain rehabilitation system incorporating electromyographic signal pattern recognition, comprising:

[0015] The data acquisition module is used to acquire raw electromyographic data of the target muscle groups in the waist and abdomen during the user's use of rehabilitation training equipment;

[0016] The feature extraction module is used to preprocess the raw electromyography (EMG) data and use a sliding time window mechanism to extract the integrated EMG value and median frequency data of the raw EMG data within the current time window.

[0017] The dynamic parameter calculation module is used to calculate parameters based on the latest continuous parameters. The rate of change of integral electromyography values ​​and median frequency data within a time window is used to calculate dynamic adjustment parameters.

[0018] The threshold calculation module is used to add the dynamic adjustment parameter to the static baseline threshold determined based on the user's historical data to obtain the upper limit of the dynamic fatigue threshold range at the current moment, and multiply the upper limit by a preset proportional coefficient to obtain the lower limit of the dynamic fatigue threshold range.

[0019] The fatigue state discrimination module is used to calculate and generate instantaneous fatigue data based on the integrated electromyography value and median frequency data of the current time window through the fatigue state discrimination model.

[0020] The adjustment coefficient generation module is used to compare the instantaneous fatigue data with the dynamic fatigue threshold interval, determine the relative positional relationship of the instantaneous fatigue data with respect to the dynamic fatigue threshold interval, and generate a corresponding adjustment coefficient value for adjusting the output power of the rehabilitation training device based on the relative positional relationship.

[0021] A control signal output module is used to send the adjustment coefficient value to the rehabilitation training device.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. By extracting the integral electromyography value and the rate of change of the median frequency, dynamic adjustment parameters are calculated, and combined with the user's historical data to form a dynamic fatigue threshold range. Then, based on the comparison relationship between this threshold range and real-time fatigue data, a coefficient for driving the power adjustment of the device is generated, thereby realizing the personalized adaptive control of the rehabilitation training device on the patient's fatigue state and improving the safety and effectiveness of rehabilitation training.

[0024] 2. It can dynamically adjust thresholds according to the individual differences of different patients, realize personalized rehabilitation plans, improve training efficiency and rehabilitation effect, while reducing the real-time intervention burden of rehabilitation therapists and improving the intelligence and automation level of training. Attached Figure Description

[0025] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0026] Figure 1 This is a flowchart of the method of the present invention;

[0027] Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation

[0028] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0029] Application Overview

[0030] Existing rehabilitation training methods generally rely on fixed training intensity or the therapist's subjective judgment, which makes it difficult to reflect the patient's muscle fatigue state in real time. This can easily lead to undertraining or overtraining, affecting rehabilitation outcomes and posing potential injury risks.

[0031] To address the aforementioned issues, this invention proposes a closed-loop adaptive rehabilitation training method. This method involves real-time acquisition of electromyographic (EMG) signals from the target muscle groups in the patient's lumbar and abdominal regions during training. A sliding time window is used to extract features such as integrated EMG values ​​and median frequency. The rate of change within the continuous time window is analyzed to calculate dynamic adjustment parameters. Combined with the patient's historical data, a dynamic fatigue threshold range is generated, and an instantaneous fatigue level data is generated using a fatigue state discrimination model. The system generates an adjustment coefficient for the output power of the training equipment based on the relative positional relationship between the instantaneous fatigue level and the threshold range, achieving real-time adaptive control of training intensity.

[0032] This invention, through dynamic monitoring and intelligent adjustment, keeps rehabilitation training within a safe and effective range, avoiding the risk of muscle injury, improving training efficiency and rehabilitation effects, reducing the real-time intervention burden on rehabilitation therapists, and significantly enhancing the intelligence and personalization of rehabilitation training.

[0033] like Figure 1 As shown, this application proposes a rehabilitation method for chronic low back pain that combines electromyographic signal pattern recognition, including the following steps:

[0034] Obtain raw electromyographic data of the target muscle groups in the waist and abdomen during the user's use of rehabilitation training equipment;

[0035] The raw electromyography (EMG) data is preprocessed using a sliding time window mechanism to extract the integrated EMG value and median frequency data within the current time window. Preprocessing includes signal amplification, filtering, and noise reduction. Filtering can employ a bandpass filter to remove high-frequency noise and low-frequency motion artifacts. The sliding time window mechanism can use a fixed-length window, such as 500 ms, or an adaptive-length window. The integrated EMG value is obtained by integrating the rectified EMG signal within the sliding time window; rectification can be full-wave rectification or half-wave rectification. The median frequency data is obtained by performing a Fast Fourier Transform on the EMG signal within the sliding time window and calculating the median frequency of its power spectrum.

[0036] Based on the latest continuous The dynamic adjustment parameters are calculated based on the rate of change of integrated electromyography (EMG) values ​​and median frequency data within a time window. The dynamic adjustment parameters can be calculated by weighted summation of the rate of change of integrated EMG values ​​and median frequency data, with the weights determined based on correlation analysis of historical data.

[0037] The dynamic adjustment parameter is added to the static baseline threshold determined based on the user's historical data to obtain the upper limit of the dynamic fatigue threshold range at the current moment; the static baseline threshold can be calculated by statistically analyzing the median and standard deviation of fatigue data in the user's historical training data.

[0038] Multiply the upper limit value by a preset scaling factor to obtain the lower limit value of the dynamic fatigue threshold range;

[0039] Based on the integrated electromyography (EMG) values ​​and median frequency data within the current time window, instantaneous fatigue data is generated using a fatigue state discrimination model. This model can be weighted and fused using frequency-domain fatigue factors and time-domain compensation factors. The adjustment coefficient values ​​can be determined using different mapping rules based on the relative positional relationship between the instantaneous fatigue data and the dynamic fatigue threshold interval.

[0040] The instantaneous fatigue data is compared with the dynamic fatigue threshold interval to determine the relative position of the instantaneous fatigue data with respect to the dynamic fatigue threshold interval, and a corresponding adjustment coefficient value for adjusting the output power of the rehabilitation training device is generated based on the relative position relationship.

[0041] The adjustment coefficient value is sent to the rehabilitation training device.

[0042] This method, by monitoring the dynamic changes in electromyography (EMG) signals in real time and combining this with historical user data to establish personalized dynamic fatigue threshold ranges, can more accurately reflect changes in the patient's fatigue state during training. Compared with existing technologies, this method can dynamically adjust the output power of rehabilitation training equipment according to the patient's real-time fatigue state, avoiding the problem of mismatch between training intensity and patient tolerance caused by fixed threshold control, thereby improving the safety and effectiveness of rehabilitation training. Through comprehensive analysis of integrated EMG values ​​and median frequency data, this method can more comprehensively assess muscle fatigue state, providing a more precise control basis for rehabilitation training.

[0043] The calculation of dynamically adjusted parameters is the core of realizing the dynamic fatigue threshold range. Its purpose is to transform the abstract trend of electromyographic signal changes into a quantifiable parameter for dynamically adjusting the threshold. This calculation process mainly includes three stages: data standardization, weighted fusion, and mapping. The following is a detailed explanation of this process:

[0044] The integral electromyography values ​​( The rate of change and median frequency of ) The rate of change of ) is normalized to obtain the first standardized parameter. Second standardized parameters The purpose of this step is to eliminate the differences in dimensions and orders of magnitude of the original feature data, transforming it into dimensionless pure numerical values, which facilitates subsequent weighted comparisons and fusion.

[0045] The integral electromyography value was analyzed using a normalization function. The rate of change and median frequency of ) The rate of change of the data is normalized. The normalization function can be a simple linear scaling function or a scaling function determined based on the statistical characteristics of historical data (such as maximum, minimum, mean, and standard deviation).

[0046] After obtaining the standardized parameters, a weighted summation operation is performed on them to generate a preliminary, comprehensive adjustment parameter. This step reflects the relative importance of different features in fatigue assessment.

[0047] To simulate the nonlinear physiological process of muscle fatigue from accumulation to burst, and to map the initial adjustment parameters to an output range more suitable for control decisions, the initial adjustment parameters are... Input the preset mapping function for final processing.

[0048] This preset mapping function has a key characteristic: its first derivative is greater in the middle range of the input range than in the ranges at both ends. This means that the function is more sensitive to changes in input values ​​within the middle range.

[0049] When initial adjustment parameters When the system is in the middle range (corresponding to a muscle fatigue trend that is forming but has not yet reached the critical point), the high sensitivity of the mapping function allows the dynamic adjustment parameters of the final output to change significantly, thereby driving the dynamic threshold range to tighten rapidly, putting the system into a highly "alert" state and preparing for the upcoming intervention.

[0050] When initial adjustment parameters When at either end (corresponding to good muscle condition or extreme fatigue), the low sensitivity of the function makes the output change smoothly, avoiding system misjudgment or oscillation caused by small fluctuations in parameters in stages where no intervention is needed or where forced intervention is required.

[0051] Finally, the value output by this mapping function is used as the dynamic adjustment parameter and subsequently used to calculate the upper limit of the dynamic fatigue threshold range.

[0052] By introducing a preset mapping function to nonlinearly transform the initial adjustment parameters, the relationship between changes in electromyography (EMG) signals and fatigue state can be reflected more accurately. The high slope design of the mapping function near the midpoint allows the system to produce a more pronounced response when the EMG signal is in a critical range of change, thereby improving the sensitivity of fatigue state discrimination. Compared with existing technologies that directly use linear weighting results, this scheme better adapts to the complex relationship between EMG signals and fatigue state through nonlinear transformation, solving the problem of insufficient discrimination accuracy caused by linear processing in existing technologies.

[0053] The calculation of initial adjustment parameters is a crucial step in achieving personalized weight allocation. Its core idea lies in quantitatively assessing the indicative strength of different electromyographic characteristics to fatigue states based on the user's historical data, and dynamically allocating weights accordingly. This overcomes the limitation of fixed weights in adapting to individual differences. The process specifically includes the following steps:

[0054] The system first extracts valid information from the stored user historical data and constructs two data sequences:

[0055] Obtain from user historical data The rate of change of median frequency data for each historical time window constitutes the first sequence; this sequence reflects the changing trend of muscle spectral characteristics in historical periods.

[0056] Obtain from user historical data The instantaneous fatigue data corresponding to each historical time window constitute the second sequence; this sequence represents the muscle fatigue state finally determined by the system in the historical stage.

[0057] To objectively measure the predictive power of median frequency variation on fatigue state, the Pearson correlation coefficient of the two sequences was calculated. The Pearson correlation coefficient is a classic statistical indicator for measuring the degree and direction of linear correlation between two variables, and its value ranges from [value missing]. ; This indicates a positive correlation, meaning that a decrease in median frequency is associated with an increase in fatigue, which is consistent with the physiological manifestations of muscle fatigue. The larger the absolute value, the stronger the linear relationship between the median frequency change trend and the fatigue state, and the higher the reliability of this feature in judging fatigue.

[0058] To convert the correlation coefficient into a value that can be directly used as a weighting coefficient, it needs to be normalized.

[0059] First, take the absolute value. To eliminate directionality while preserving relation strength, a linear transformation is then used to map it to... The standardized correlation coefficient is obtained from the interval. .

[0060] A simple mapping method is:

[0061] ,at this time Already in If further adjustments to the distribution are needed within the interval, linear scaling can be used for further processing.

[0062] The normalized correlation coefficient Directly assign to the second normalized parameter Second weighting coefficient This weighting coefficient is used as the second standardized parameter in the weighted fusion formula. The reliability of features characterized by historical data is directly used as the weighting basis for current data fusion; if historical data shows that the median frequency is highly correlated with fatigue, then the median frequency feature is given a higher weight in the current calculation.

[0063] Accordingly, based on the second weighting coefficient Calculate the first standardized parameter First weighting coefficient ,make .

[0064] Finally, the first weighting coefficient was calculated. Second weighting coefficient The initial adjustment parameters are obtained by performing a weighted summation according to the following formula. :

[0065] .

[0066] This technical solution establishes a correlation analysis between median frequency changes and fatigue levels, dynamically adjusting the weight allocation of integrated electromyography (EMG) values ​​and median frequency. This makes the initial adjustment parameter calculations more aligned with the physiological characteristics of individual users. Compared to existing technologies that use fixed weights, this solution automatically optimizes parameter weights based on the user's actual training data, improving the accuracy of fatigue state identification. Specifically, when the correlation between a user's median frequency changes and fatigue levels is high, the system automatically increases the weight of the median frequency parameter; conversely, it increases the weight of the integrated EMG value, thereby achieving adaptive adjustment to individual differences among users.

[0067] The preset mapping function is the core component for implementing nonlinear transformation; its function is to adjust the initial parameters obtained from the preliminary fusion. ( The normalized value is mapped to a final adjustment parameter suitable for controlling the dynamic fatigue threshold range. This function can simulate the nonlinear physiological response during muscle fatigue.

[0068] The specific expression for the preset mapping function is:

[0069] .

[0070] The initial adjustment parameter after normalization has a value range of [0, 1]. It is a dimensionless scalar obtained by weighted fusion of the standardized parameters of integral electromyography value and median frequency, followed by linear scaling, and comprehensively represents the intensity of muscle fatigue trend calculated in real time.

[0071] This is the output of the mapping function, i.e., the final adjustment parameter. The physical meaning of this output value is that it will be directly used to calculate the upper limit of the dynamic fatigue threshold range (or as its adjustment amount). Its value is precisely limited to the range... Inside.

[0072] This is the lower limit of the dynamic fatigue threshold range, usually set based on the user's historical baseline data or a safe conservative value, ensuring that the output will not fall below a safe threshold.

[0073] This is the upper limit of the dynamic fatigue threshold range, usually set based on the user's maximum tolerance level or a preset limit value, ensuring that the output does not exceed a reasonable range.

[0074] The final adjustment parameters are defined. The maximum possible range of variation.

[0075] The core part of the function formula is essentially the standard hyperbolic tangent function. A variant of the standard. The output range of the function is After this transformation, the output range is linearly mapped to... This makes it very convenient to output the function via... Scaling, and through Perform a translation, and finally input Mapped to Interval.

[0076] It is a preset sensitivity coefficient, which is a positive real number and is a key parameter that controls the shape and response characteristics of the control function.

[0077] when When the value is large (e.g.) ), the function in The area becomes very steep, enter Tiny changes can lead to changes in output exist and There is a sharp jump between them. This is suitable for scenarios that require a quick and decisive response to fatigue trends.

[0078] when When the value is small (e.g.) The function curve is relatively flat, and the input... Significant changes are needed to affect the output. This represents a significant change. It is suitable for scenarios requiring a smooth, stable transition, avoiding system output oscillations caused by minor signal fluctuations.

[0079] Therefore, by configuring parameters It allows for flexible adjustment of the system's dynamic response characteristics to adapt to the physiological characteristics of different users and different rehabilitation stages and strategies.

[0080] For input offset, the input will be... Subtracting 0.5 means moving the function's sensitivity center from 0 to 0.5. This ensures that when initially adjusting the parameters... When the value is in the middle of its range, the rate of change of the mapping function is the greatest, which perfectly matches the design requirement of "most sensitive in the middle range".

[0081] Therefore, this technical solution solves the nonlinear transformation problem between dynamically adjusted parameters and fatigue threshold range by introducing a mapping function with a specific mathematical form. The result of linear fusion is nonlinearly transformed, and its output is strictly limited to a pre-defined, reasonable range. Furthermore, its response sensitivity can be determined by parameters. This allows for flexible adjustments. It enables the system to better simulate the nonlinear physiological process of muscle fatigue accumulation and exhibits greater sensitivity at key decision points (moderate fatigue trends), thus allowing for more timely and precise regulatory decisions.

[0082] This application also proposes that the integrated electromyographic value is obtained by integrating the rectified electromyographic signal within the sliding time window; and the median frequency data is obtained by performing a fast Fourier transform on the electromyographic signal within the sliding time window and calculating the median frequency of its power spectrum.

[0083] The calculation of integrated electromyographic (EMG) values ​​can be performed by preprocessing the raw EMG signal using full-wave or half-wave rectification. The rectified signal is then used to accumulate time-domain energy through an analog integrator circuit or a digital integration algorithm. As a preferred implementation, digital integration can be performed by accumulating discrete sampling points using the trapezoidal or rectangular method. The window width can be set to 100-500 ms to accommodate the activity characteristics of different muscle groups. For obtaining median frequency data, the signal within the window needs to be windowed to reduce spectral leakage; the Hanning or Hamming window can be used as the preferred window function. The number of points in the Fast Fourier Transform (FFT) should be selected based on the sampling rate; typical values ​​are 256 or 512 points. The median frequency of the power spectrum refers to the frequency point that bisects the total area of ​​the power spectrum, and can be accurately calculated using an accumulation algorithm.

[0084] Therefore, this technical solution solves the problem in existing technologies where a single feature cannot comprehensively reflect muscle state by extracting features in both the time and frequency domains. Integrated electromyography (EMG) values ​​effectively characterize muscle activation levels, while median frequency changes sensitively reflect muscle fatigue. The synergistic calculation of these two features provides more reliable input data for subsequent fatigue assessment. Compared to traditional methods that only use time-domain amplitude analysis, this solution supplements the electromyography signal with electrophysiological changes through frequency-domain features, resulting in higher timeliness and accuracy in fatigue monitoring.

[0085] Determining the static baseline threshold is a fundamental step in achieving personalized fatigue assessment. Its core purpose is to objectively quantify the user's typical fatigue level and fluctuation characteristics from their historical data, thereby establishing a benchmark reference value that aligns with the user's specific physiological state. This threshold is not fixed but dynamically updated as the user's recovery progresses, providing a stable computational foundation for subsequent dynamic adjustments. The specific determination process is as follows:

[0086] First, retrieve from the database the data completed by the user within the most recent historical training period (e.g., the past four weeks). Fatigue data from each effective training session.

[0087] A training session must meet preset quality standards (such as training duration exceeding the minimum requirement, performance of movements meeting the standard, and good signal quality) to be marked as "valid" in order to ensure that the data used for calculation can truly reflect the user's fatigue state.

[0088] Fatigue data is not a single value, but a collection of instantaneous fatigue data corresponding to all time windows within each effective training session. This collection fully records the evolution of a user's fatigue level during a complete training session and contains rich statistical information.

[0089] Statistical analysis was performed on all acquired historical fatigue data, and two key statistical features were extracted:

[0090] One is the intermediate fatigue standard. Calculate the median of all historical fatigue data. The median, rather than the arithmetic mean, is chosen as the benchmark because it is not sensitive to extreme values ​​and can better resist interference from single abnormal training sessions (such as those caused by extremely good or bad performance), thus more robustly representing the user's most typical fatigue level.

[0091] Secondly, the range of fatigue fluctuation. Calculate the standard deviation of all historical fatigue data. The standard deviation quantifies how user historical fatigue data revolves around its intermediate fatigue baseline. The degree of dispersion reflects the stability and consistency of the user's fatigue state. A large value indicates that the user's fatigue level fluctuates greatly, while a small value indicates that the state is stable.

[0092] Based on the intermediate fatigue level benchmark and the fatigue fluctuation range, the static baseline threshold is calculated using the following formula. :

[0093] ;

[0094] The physical meaning of this formula is that it sets the threshold at an intermediate fatigue baseline. Within a certain range above. Added This constitutes a safe buffer zone or fault tolerance range.

[0095] This is a preset linear coefficient, a dimensionless positive real number. It is a key parameter for adjusting whether the system is conservative or not.

[0096] From clinical experience, The value of is related to the balance between the desired specificity and sensitivity. For example, This roughly means setting the threshold higher than the historical fatigue level of about 84% (assuming the data is normally distributed).

[0097] From a technical perspective, The larger the value is set, the better. The higher the value, the earlier the intervention is triggered, resulting in greater safety but potentially slightly less precision; The smaller the value is set, The closer It is more in line with the user's own situation, but the fault tolerance is reduced. The preferred range is usually between 0.5 and 2.0, and can be determined through trial optimization based on clinical application scenarios.

[0098] Therefore, this technical solution determines the static baseline threshold by combining the median and standard deviation of the user's historical training data, which can more accurately reflect the user's individualized fatigue characteristics. Compared with existing methods that use fixed thresholds or simple averages, this solution has better robustness and adaptability. Intermediate fatigue baseline. It can represent the typical fatigue level of users, while the range of fatigue fluctuations This reflects the normal range of variation in the user's fatigue state. By introducing a linear coefficient... Furthermore, the threshold sensitivity can be adjusted based on individual user differences. This statistically based threshold determination method can effectively reduce the impact of outlier data and improve the accuracy of fatigue state identification.

[0099] The fatigue state discrimination model is the core algorithm for real-time assessment of muscle state. Its purpose is to simulate the judgment logic of experienced rehabilitation therapists: focusing not only on the absolute intensity of muscle activity but also on its changing trends over time and the synergistic and compensatory relationships between different features. Through multi-feature, multi-rule fusion decision-making, the model ultimately outputs comprehensive instantaneous fatigue data, providing accurate input for subsequent dynamic adjustments. Its specific calculation process is as follows:

[0100] First, monitor the median frequency. The change in frequency is the most classic characteristic of muscle fatigue in the frequency domain; the difference between the median frequency data of the current time window and the median frequency data of the previous time window (or the average of the previous multiple time windows) is calculated and denoted as the first difference. . Typically negative, this indicates a shift of frequency components towards lower frequencies; The absolute value of the difference between the first preset threshold value and the threshold value The comparison is performed. This threshold is used to filter out irrelevant changes caused by minor signal fluctuations or noise.

[0101] like The absolute value is less than the first preset difference threshold. If the frequency domain change is not significant, the frequency domain fatigue factor can be set to 0 or a very small base value.

[0102] Conversely if The absolute value is greater than or equal to the first preset difference threshold. If this is the case, then significant muscle fatigue frequency domain characteristics are considered to have appeared. At this point, a frequency domain characteristic similar to... Frequency domain fatigue factor data that are proportional to the absolute value .For example, , This is a proportionality coefficient. The larger this value, the more severe the left shift of the spectrum and the higher the fatigue component.

[0103] To verify the judgment of frequency domain indices and capture certain special fatigue patterns (such as muscle compensation by increasing recruitment), the model simultaneously monitors integral electromyography values. The change is as follows:

[0104] Calculate the difference between the integrated electromyography (EMG) value of the current time window and the integrated EMG value of the previous time window (or the average of the previous multiple time windows), and record it as the second difference. . A positive value usually indicates increased muscle activation.

[0105] Will The corresponding second preset difference threshold Comparison:

[0106] like Less than If the time-domain change is not significant, the time-domain compensation factor can be set to 0.

[0107] Fold, if Greater than or equal to Then a matching one is generated. Proportional time-domain compensation factor data .For example, , This is the proportionality coefficient. This phenomenon ( Significantly decreased A significant increase in output usually indicates that muscles are trying to maintain output by recruiting more motor units, and is strong evidence that fatigue is occurring and developing.

[0108] get and The model then performs a weighted fusion calculation on the two data points to output the final, comprehensive instantaneous fatigue data, Fatigue_instant.

[0109]

[0110] in, and These are the weighting coefficients, and typically the sum of the two is 1. The weights can be assigned as fixed values ​​based on prior knowledge of the reliability of the features.

[0111] The entire calculation process closely integrates the physiological manifestations of muscle fatigue (leftward shift of the spectrum, increased motor unit recruitment), giving the model's output clear physical and physiological meaning, rather than just a black-box mathematical result. The model focuses on the amount of change (difference, rate of change) of feature values ​​over time, rather than their absolute values. This reduces interference from factors such as baseline differences between individuals and electrode placement, enhancing sensitivity to the fatigue process. The model does not rely on a single indicator to make decisions. Only when the frequency domain indicator ( Significant decrease) and time-domain indicators ( A high fatigue level output is only generated when there is a significant increase (or a significant change) in fatigue levels. This cross-validation mechanism greatly reduces the false alarm rate and improves the reliability of the judgment.

[0112] The generation of adjustment coefficient values ​​is a crucial decision-making step connecting state recognition and device execution. Its core function is to transform the calculated instantaneous fatigue data, which characterizes the physiological state, into a clear control command that can be directly executed by the rehabilitation training device. This process employs a zoned control strategy based on the relative position of the instantaneous fatigue value within the dynamic threshold range, balancing the effectiveness, safety, and smoothness of training. The specific generation process is as follows:

[0113] When the system determines that the instantaneous fatigue data is less than the lower limit of the dynamic fatigue threshold range, it indicates that the user's current muscle condition is good, the fatigue level is low, and the user is in a "safe zone" where training can be carried out safely and effectively.

[0114] In this state, the user can receive full training stimulation to promote rehabilitation. Therefore, there is no need to provide additional assistance or reduce the load on the rehabilitation training equipment.

[0115] At this point, the instruction is to set the adjustment factor value to 100%. This instruction tells the rehabilitation training device to maintain its current output parameters. For example, for a device that provides resistance, it means maintaining the current resistance value; for a device that provides assistance, it means maintaining the current level of assistance. This decision aims to fully utilize the user's existing muscle strength level and maximize training effectiveness.

[0116] When the system determines that the instantaneous fatigue data is within the dynamic fatigue threshold range, it indicates that the user's muscles have entered a state of moderate fatigue, and are in the "adjustment zone" (or "working zone"), which requires close monitoring and intervention. Decisions in this zone are the most critical, requiring precise and smooth control.

[0117] At this point, the equipment output needs to be adjusted proportionally according to the severity of fatigue. The higher the degree of fatigue, the more assistance should be provided (or the less resistance should be) to achieve "real-time matching of load and capacity" and stabilize the user's muscle fatigue level within the ideal range.

[0118] The instruction is to query a preset control mapping table based on the difference between instantaneous fatigue data and a target threshold (e.g., the median of the dynamic range) to obtain an adjustment coefficient value between 0% and 100%.

[0119] This mapping table defines the relationship between "fatigue deviation value" and "adjustment coefficient". The mapping table can be a predefined lookup table or a discretized implementation of a simple mathematical function (such as a linear function or an exponential function).

[0120] The adjustment range of the device output is proportional to the degree to which the fatigue level deviates from the ideal value, making the control process very smooth, avoiding sudden changes in output, providing users with a gradual and comfortable training experience, while accurately maintaining the training load at the optimal level.

[0121] When the system determines that the instantaneous fatigue data is greater than or equal to the upper limit of the dynamic fatigue threshold range, it indicates that the user's muscles have reached or are close to their current capacity limit and are in a "protected area" with a high risk of injury.

[0122] In this situation, safety becomes the primary objective, thus requiring mandatory and forceful intervention to rapidly and significantly reduce training load in order to protect users from overtraining and sports injuries.

[0123] The instruction is to set the adjustment factor to a preset value less than 50%. This value is typically set at a low level, such as 30% or 20%, based on clinical safety guidelines. This means the device output will immediately drop to 30% or 20% of the normal level.

[0124] This decision is a safety overwrite mechanism. It goes beyond proportional adjustment, issuing a strong protective command. For example, for a resistance device, it will instantly reduce the resistance to a very low level; for a power assist device, it will provide significant assistance to share most of the load. Simultaneously, this operation is usually accompanied by audible and visual alarms to alert the user or prompt them to pause and rest.

[0125] This technical solution achieves precise control of the output power of rehabilitation training equipment by establishing a dynamic fatigue threshold range and a multi-level adjustment mechanism. When muscles are detected to be in a low-fatigue state, the equipment maintains full power output to ensure training intensity; when fatigue enters a transitional range, gradual adjustment is used to avoid sudden power changes; and when the fatigue threshold is reached, a protective power reduction is automatically triggered. This graded adjustment strategy based on real-time fatigue status effectively solves the problem of mismatch between training intensity and patient tolerance caused by traditional fixed threshold control, reducing the risk of excessive muscle fatigue while ensuring training effectiveness.

[0126] This application also proposes that the target threshold be set as the arithmetic mean of the upper and lower limits of the dynamic fatigue threshold interval.

[0127] Specifically, the arithmetic mean is calculated by adding the upper limit and the lower limit and then dividing by two to obtain the target threshold located at the midpoint of the interval.

[0128] This technical solution addresses the asymmetric adjustment problem caused by using fixed or unilateral thresholds in existing technologies by introducing the arithmetic mean as the target threshold. Within the dynamic fatigue threshold range, the symmetrical adjustment mechanism based on the target threshold can more accurately reflect the real-time fatigue state changes of patients. Compared with existing technologies, this solution improves the accuracy and response speed of the output power adjustment of rehabilitation training equipment, enabling the equipment output to better match the patient's actual tolerance. Furthermore, by using the arithmetic mean, a parameter that is easy to calculate and has clear physical meaning, as the target threshold, the system's computational complexity is reduced while ensuring adjustment accuracy.

[0129] like Figure 2As shown, this application also proposes a chronic low back pain rehabilitation system that combines electromyographic signal pattern recognition. The system includes a data acquisition module, a feature extraction module, a dynamic parameter calculation module, a threshold calculation module, a fatigue state discrimination module, a regulation coefficient generation module, and a control signal output module.

[0130] The data acquisition module can use a surface electromyography (EMG) sensor array, which is attached to the skin surface of the target muscle group in the waist and abdomen through medical-grade conductive gel to collect raw EMG signals at a set sampling frequency.

[0131] The feature extraction module includes a digital signal processor to preprocess the raw electromyographic signals and implements a sliding time window mechanism through a programmable logic device. The window length is preferably 500ms and the overlap rate is 50%.

[0132] The dynamic parameter calculation module uses an embedded microprocessor to generate dynamically adjusted parameters by real-time monitoring of the rate of change of integral electromyography values ​​and median frequency data, combined with correlation analysis of historical data.

[0133] The threshold calculation module has a built-in non-volatile memory to store users' historical training data and establishes personalized static baseline thresholds through statistical analysis methods.

[0134] The fatigue state discrimination module integrates a machine learning accelerator and runs a lightweight neural network model trained on clinical data to achieve millisecond-level fatigue state assessment.

[0135] The adjustment coefficient generation module includes lookup table logic circuitry, supporting dynamic threshold interval comparison and fuzzy control algorithms. The control signal output module adopts an industrial-grade CAN bus interface to achieve real-time data interaction with the rehabilitation training equipment.

[0136] This system addresses the issue of delayed output power adjustment in existing rehabilitation training equipment through multi-module collaborative operation. Electromyography (EMG) signals acquired by the data acquisition module are processed by the feature extraction module, and then the dynamic parameter calculation module and threshold calculation module jointly construct a dynamic fatigue threshold range reflecting the individual's real-time state. The instantaneous fatigue data generated by the fatigue state discrimination module is compared in real-time with the dynamic threshold range. Based on this, the adjustment coefficient generation module generates precise control commands, ultimately achieving closed-loop adjustment of the rehabilitation training equipment's output power through the control signal output module. Compared to existing technologies, this system achieves a complete automated process from EMG signal acquisition to equipment control, avoiding response delays caused by manual intervention. It can adjust training intensity promptly based on dynamic changes in the patient's muscle state, ensuring rehabilitation effectiveness while reducing the risk of sports injuries. Each module is implemented using embedded hardware, featuring fast response speed, high calculation accuracy, and good system stability.

[0137] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A rehabilitation method for chronic low back pain combining electromyographic signal pattern recognition, characterized by: Includes the following steps: Obtain raw electromyographic data of the target muscle groups in the waist and abdomen during the user's use of rehabilitation training equipment; The raw electromyography (EMG) data is preprocessed, and a sliding time window mechanism is used to extract the integrated EMG values ​​and median frequency data of the raw EMG data within the current time window. Based on the latest continuous The rate of change of integral electromyography values ​​and median frequency data within a time window is used to calculate dynamic adjustment parameters. The dynamic adjustment parameter is added to the static baseline threshold determined based on the user's historical data to obtain the upper limit of the dynamic fatigue threshold range at the current moment. The upper limit is then multiplied by a preset scaling factor to obtain the lower limit of the dynamic fatigue threshold range. Based on the integrated electromyography values ​​and median frequency data of the current time window, instantaneous fatigue data is generated by calculating using a fatigue state discrimination model. The instantaneous fatigue data is compared with the dynamic fatigue threshold interval to determine the relative position of the instantaneous fatigue data with respect to the dynamic fatigue threshold interval, and a corresponding adjustment coefficient value for adjusting the output power of the rehabilitation training device is generated based on the relative position relationship. The adjustment coefficient value is sent to the rehabilitation training device.

2. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The calculation process for the dynamically adjusted parameters is as follows: The rate of change of integral electromyography values ​​and median frequency data were normalized to obtain the first standardized parameter and the second standardized parameter. The first standardized parameter and the second standardized parameter are weighted and summed to obtain the initial adjustment parameter; The initial adjustment parameters are input into a preset mapping function for processing, and the final adjustment parameters are output as the dynamic adjustment parameters. Wherein, the value of the first derivative of the preset mapping function in the middle interval of the input range is greater than the value in the intervals at both ends of the input range.

3. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 2, characterized in that: The calculation process for the initial adjustment parameters is as follows: Obtain from user historical data The median frequency data change rate of each historical time window constitutes the first sequence; Obtain from user historical data The instantaneous fatigue data corresponding to each historical time window constitute the second sequence; Calculate the Pearson correlation coefficient between the first sequence and the second sequence. ; Regarding the Pearson correlation coefficient Perform normalization to make it fall into The standardized correlation coefficient is obtained from the interval. ; Standardized correlation coefficient Assigned to the second normalized parameter Second weighting coefficient ; According to the second weighting coefficient Calculate the first standardized parameter First weighting coefficient ,make ; The initial adjustment parameters are calculated using the following formula. : 。 4. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The formula for the preset mapping function is: ； in, This is the initial adjustment parameter after normalization, and its value ranges from 0 to 1; The preset sensitivity coefficient, This represents the lower limit of the dynamic fatigue threshold range. This represents the upper limit of the dynamic fatigue threshold range.

5. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The integrated electromyographic value is obtained by integrating the rectified electromyographic signal within the sliding time window; the median frequency data is obtained by performing a fast Fourier transform on the electromyographic signal within the sliding time window and calculating the median frequency of its power spectrum.

6. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The process for determining the static baseline threshold is as follows: Obtain the data completed by the user during historical training periods. The fatigue data in each effective training session is a collection of instantaneous fatigue data corresponding to all time windows in each training session. Calculate the median value of the fatigue data to obtain the intermediate fatigue baseline. ; Calculate the standard deviation of the fatigue data to obtain the fatigue fluctuation range. ; Based on the intermediate fatigue level benchmark and the fatigue fluctuation range, the static baseline threshold is calculated using the following formula. : ； in, These are the preset linear coefficients.

7. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The process by which the fatigue state discrimination model generates instantaneous fatigue data is as follows: Calculate the first difference between the median frequency data of the current time window and the median frequency data of the previous one or more time windows. When the first difference exceeds the first preset difference threshold, generate frequency domain fatigue factor data that is proportional to the first difference. Calculate the second difference between the integrated electromyography (EMG) value of the current time window and the integrated EMG values ​​of the previous one or more time windows. When the second difference exceeds a second preset difference threshold, generate time-domain compensation factor data that is proportional to the second difference. Based on the rate of change of the integrated electromyography value within a continuous time window, time-domain compensation factor data is generated; The frequency domain fatigue factor data and the time domain compensation factor data are weighted and fused to calculate the instantaneous fatigue data.

8. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 1, characterized in that: The process of generating the adjustment coefficient value includes: When the relative positional relationship indicates that the instantaneous fatigue data is less than the lower limit of the dynamic fatigue threshold range, the adjustment coefficient value is set to 100%. When the relative positional relationship indicates that the instantaneous fatigue data is within the dynamic fatigue threshold range, the preset control mapping table is queried according to the difference between the instantaneous fatigue data and the target threshold, and an adjustment coefficient value between 0% and 100% is obtained according to the control mapping table. When the relative positional relationship indicates that the instantaneous fatigue data is greater than or equal to the upper limit of the dynamic fatigue threshold range, the adjustment coefficient value is set to a preset value of less than 50%.

9. The chronic low back pain rehabilitation method combining electromyographic signal pattern recognition according to claim 8, characterized in that: The target threshold is the arithmetic mean of the upper and lower limits of the dynamic fatigue threshold interval.

10. A chronic low back pain rehabilitation system combining electromyographic signal pattern recognition, characterized in that: include: The data acquisition module is used to acquire raw electromyographic data of the target muscle groups in the waist and abdomen during the user's use of rehabilitation training equipment; The feature extraction module is used to preprocess the raw electromyography (EMG) data and use a sliding time window mechanism to extract the integrated EMG value and median frequency data of the raw EMG data within the current time window. The dynamic parameter calculation module is used to calculate parameters based on the latest continuous parameters. The rate of change of integral electromyography values ​​and median frequency data within a time window is used to calculate dynamic adjustment parameters. The threshold calculation module is used to add the dynamic adjustment parameter to the static baseline threshold determined based on the user's historical data to obtain the upper limit of the dynamic fatigue threshold range at the current moment, and multiply the upper limit by a preset proportional coefficient to obtain the lower limit of the dynamic fatigue threshold range. The fatigue state discrimination module is used to calculate and generate instantaneous fatigue data based on the integrated electromyography value and median frequency data of the current time window through the fatigue state discrimination model. The adjustment coefficient generation module is used to compare the instantaneous fatigue data with the dynamic fatigue threshold interval, determine the relative positional relationship of the instantaneous fatigue data with respect to the dynamic fatigue threshold interval, and generate a corresponding adjustment coefficient value for adjusting the output power of the rehabilitation training device based on the relative positional relationship. A control signal output module is used to send the adjustment coefficient value to the rehabilitation training device.

Images