Electrical stimulation control system and method

Abstract

The application relates to the technical field of medical rehabilitation, and provides an electrical stimulation control system and method, the system comprising: a communication-connected electromyographic signal acquisition module, an electrical stimulation module and a processor; the electromyographic signal acquisition module is used for collecting electromyographic signals of a healthy limb of a patient and inputting the electromyographic signals into the processor; the processor is used for determining a target joint angle for driving a diseased limb of the patient to move synchronously with the healthy limb of the patient according to the electromyographic signals; and according to the target joint angle, target electrical stimulation signal parameter values are obtained by inversely solving a preset stimulation-response model, and the target electrical stimulation signal parameter values are input into the electrical stimulation module; the electrical stimulation module is used for outputting electrical stimulation corresponding to the electrical stimulation signal parameter values to the diseased limb of the patient. The application realizes synchronous linkage of the healthy action and the electrical stimulation of the diseased limb by taking the target joint angle as an intermediate variable, and then realizes flexible adjustment of the electrical stimulation parameters with the change of the healthy action.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation technology, and in particular to an electrical stimulation control system and method. Background Technology

[0002] Mirror therapy, a classic intervention in the field of rehabilitation, works by constructing a virtual image of the healthy side of the body through a mirror. This allows patients to observe the activities of the healthy side and perceive them as movements of the affected side through visual illusion. Combined with synchronous imitation training on the affected side, this activates the neural remodeling mechanism of the brain's motor cortex, thereby promoting the recovery of motor function.

[0003] To achieve precise assistance in movement on the affected side, functional electrical stimulation (FES) technology has been widely integrated into mirror therapy. This technology captures movement signals from the healthy side via a motion recognition module, triggering the FES system to apply electrical stimulation to the target muscle groups on the affected side. This drives rhythmic muscle contraction, assisting the affected side in completing movements consistent with the healthy side, thus forming a closed-loop rehabilitation pathway.

[0004] Current mirror therapy generally employs on / off or threshold-triggered mechanisms, i.e., setting a fixed movement threshold, triggering the FES system to output electrical stimulation with fixed parameters when the movement on the healthy side reaches the threshold. However, fixed thresholds are difficult to adapt to the dynamic changes in the movements on the healthy side, resulting in significant deviations in the timing, amplitude, and trajectory of movements on the affected side compared to the healthy side, thus limiting the improvement of overall rehabilitation effects. Summary of the Invention

[0005] This invention provides an electrical stimulation control system to address the shortcomings of existing mirror therapy technologies, which generally employ on / off or threshold triggering mechanisms, failing to adapt to dynamic changes in the healthy side's movements and exhibiting insufficient matching between FES parameters and the muscle state of the affected side. This system achieves dynamic adaptation to changes in the healthy side's movements, improving the matching degree between electrical stimulation parameters and the movement needs of the affected side.

[0006] This invention provides an electrical stimulation control system, comprising: an electromyography signal acquisition module, an electrical stimulation module, and a processor connected in communication; The electromyography (EMG) signal acquisition module is used to acquire EMG signals from the patient's unaffected limb and input the EMG signals to the processor. The processor is configured to determine, based on the electromyographic signal, a target joint angle for driving the patient's affected limb to move synchronously with the patient's healthy limb; and, based on the target joint angle, obtain target electrical stimulation signal parameter values ​​by inversely solving a preset stimulus-response model, and input the target electrical stimulation signal parameter values ​​into the electrical stimulation module. The electrical stimulation module is used to output electrical stimulation corresponding to the target electrical stimulation signal parameter value to the patient's affected limb.

[0007] In some embodiments, the target joint angle includes a sequence of predicted joint angles that exhibit an increasing trend, and the processor includes a first model module; The first model module is used to input the temporal features formed by at least one temporal feature extracted from the electromyography signal, including the root mean square value, integral electromyography value, and waveform length, into a pre-trained target long short-term memory network model to obtain the predicted joint angle sequence output by the target long short-term memory network model. The target long short-term memory network model is trained in the following way: Collect electromyography timing signals and joint angle timing data of the patient's unaffected limb when performing the target action to form a training sample set; The electromyography test timing signals in the training sample set are used as training inputs, and the joint angle test timing data corresponding to the electromyography test timing signals are used as training labels. The initial long short-term memory network model is trained based on the training input and the training labels to obtain the trained target long short-term memory network model.

[0008] In some embodiments, the electrical stimulation control system further includes motion sensors; The motion sensor is used to collect the real joint angle sequence of the affected limb and input the real joint angle sequence into the first model module; The first model module is further configured to update the network weights of the target long short-term memory network model based on the loss function value between the actual joint angle sequence and the predicted joint angle sequence under at least one rehabilitation training round.

[0009] In some embodiments, the processor includes a second model module, which is configured to input the target joint angle into a pre-fitted stimulus-response model and perform inverse solving of the stimulus-response model to obtain a real-valued solution output by the stimulus-response model; and, When there is more than one real solution, the minimum solution within a preset monotonically increasing interval is determined as the target electrical stimulation signal parameter value; the preset monotonically increasing interval is obtained by differentiating the stimulus-response model; The stimulus-response model is obtained by fitting the model in the following way: Collect joint angle test data of the patient's affected limb under electrical stimulation signals corresponding to different electrical stimulation signal test parameter values, and combine the joint angle test data and the electrical stimulation signal test parameter values ​​corresponding to the joint angle test data into data points; Using the electrical stimulation signal parameters as independent variables and the joint angle as the dependent variable, a third-order polynomial fitting is performed on multiple sets of data points to obtain the stimulus-response model.

[0010] In some embodiments, the electrical stimulation signal parameter values ​​include one or more of stimulation current intensity, pulse width, or pulse frequency.

[0011] In some embodiments, the processor further includes a security threshold module; The safety threshold module is used to input the preset safety signal parameter threshold to the electrical stimulation module when the target electrical stimulation signal parameter value is detected to exceed the preset safety signal parameter threshold.

[0012] In some embodiments, the electromyography signal acquisition module has at least one surface electromyography electrode; the electrical stimulation module includes at least one electrical stimulation electrode and at least one multichannel stimulator.

[0013] The present invention also provides an electrical stimulation control method, applied to a processor in the above-described electrical stimulation control system, the method comprising: Acquire electromyographic signals of the patient's unaffected limb from the electromyographic signal acquisition module in the electrical stimulation control system; Based on the electromyographic signals, a target joint angle is determined for driving the patient's affected limb to move synchronously with the patient's healthy limb; Based on the target joint angle, the target electrical stimulation signal parameter values ​​are obtained by inversely solving the preset stimulus-response model. The target electrical stimulation signal parameter value is input to the electrical stimulation module in the electrical stimulation control system to drive the electrical stimulation module to output the electrical stimulation corresponding to the target electrical stimulation signal parameter value to the patient's affected limb.

[0014] In some embodiments, the target joint angle includes an increasing sequence of predicted joint angles, and determining the target joint angle of the patient's affected limb based on the electromyographic signal includes: The root mean square value, integral electromyography value, and waveform length extracted from the electromyography signal are used to form a time-domain feature that is input into a pre-trained target long short-term memory network model to obtain the target joint angle output by the target long short-term memory network model. The target long short-term memory network model is trained in the following way: Collect electromyography timing signals and joint angle timing data of the patient's unaffected limb when performing the target action to form a training sample set; The electromyography test timing signals in the training sample set are used as training inputs, and the joint angle test timing data corresponding to the electromyography test timing signals are used as training labels. The initial long short-term memory network model is trained based on the training input and the training labels to obtain the trained target long short-term memory network model.

[0015] In some embodiments, determining the electrical stimulation signal parameter values ​​based on the target joint angle includes: The target joint angle is input into a pre-fitted stimulus-response model, and the stimulus-response model is solved inversely to obtain the real-valued solution output by the stimulus-response model; and, When there is more than one real solution, the minimum solution within a preset monotonically increasing interval is determined as the target electrical stimulation signal parameter value; the preset monotonically increasing interval is obtained by differentiating the stimulus-response model; The stimulus-response model is obtained by fitting the model in the following way: Collect joint angle test data of the patient's affected limb under electrical stimulation signals corresponding to different electrical stimulation signal test parameter values, and combine the joint angle test data and the electrical stimulation signal test parameter values ​​corresponding to the joint angle test data into data points; Using the electrical stimulation signal parameters as independent variables and the joint angle as the dependent variable, a third-order polynomial fitting is performed on multiple sets of data points to obtain the stimulus-response model.

[0016] The electrical stimulation control system provided by this invention captures electromyographic signals from the healthy limb in real time through an electromyographic signal acquisition module. Combined with the processor's precise prediction of the target joint angle of the affected limb and dynamic matching of electrical stimulation parameters, the electrical stimulation module outputs corresponding electrical stimulation, achieving synchronous linkage between healthy side movements and affected side electrical stimulation. This invention allows for flexible adjustment of electrical stimulation parameters according to changes in healthy side movements, improving the matching degree between electrical stimulation parameters and the movement needs of the affected side. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1This is a schematic diagram of the electrical stimulation control system provided by the present invention.

[0019] Figure 2 This is a flowchart illustrating the electrical stimulation control method provided by the present invention.

[0020] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] The electrical stimulation control system of this invention, as described in the embodiments, is as follows: Figure 1 As shown, it includes an electromyography signal acquisition module 10, an electrical stimulation module 20, and a processor 30 with communication connections.

[0023] The communication connection can be wired or wireless. For example, the modules can be connected via USB, RS-232 / 485 serial bus, etc., to ensure stable data transmission and low latency. Alternatively, wireless communication protocols such as Bluetooth, Wi-Fi, and ZigBee can be used, which allows for a more flexible physical layout of the system components, making it easier for patients to wear and use, reducing cable constraints, and improving the convenience of rehabilitation training.

[0024] In this embodiment, the electromyography (EMG) signal acquisition module is used to acquire EMG signals from the patient's unaffected limb and input the EMG signals to the processor; the processor is used to determine, based on the EMG signals, a target joint angle for driving the patient's affected limb to move synchronously with the patient's unaffected limb; and, based on the target joint angle, to obtain target electrical stimulation signal parameter values ​​by inversely solving a preset stimulus-response model, and input the electrical stimulation signal parameter values ​​to the electrical stimulation module; the electrical stimulation module is used to output electrical stimulation corresponding to the target electrical stimulation signal parameter values ​​to the patient's affected limb.

[0025] In practical implementation, the electromyography (EMG) signal acquisition module typically includes at least one or more surface EMG electrodes and corresponding signal conditioning and acquisition circuits. To ensure that the acquisition quality of the EMG signal is not affected by changes in skin condition and electrode contact state, the EMG signal acquisition module in this embodiment further includes a contact impedance dynamic monitoring and adaptive calibration unit. Before and during each training session, this unit periodically injects a weak, imperceptible high-frequency carrier signal into the skin through the electrode pair and measures the loop impedance of this signal in real time. Based on the measured loop impedance, the processor automatically adjusts the gain of the amplifier and the cutoff frequency parameters of the filter in the signal conditioning circuit. For example, when an increase in impedance is detected, the amplifier gain is appropriately increased to compensate for signal attenuation; when an increase in high-frequency noise interference is detected, the cutoff frequency of the filter is dynamically decreased. This adaptive calibration ensures the signal-to-noise ratio and amplitude stability of the EMG signal input to the processor. EMG signals refer to the bioelectrical signals generated when muscles contract under nerve control; they reflect the body's movement intentions and muscle activity levels. In this embodiment, by acquiring EMG signals from the unaffected limb, the patient's actively generated movement intentions can be captured in real time.

[0026] For example, in wrist rehabilitation applications, the electromyography (EMG) signal acquisition module can place surface EMG electrodes on the extensor and flexor muscles of the patient's unaffected forearm, such as the radial carpi radialis flexor muscle and other major muscles related to wrist flexion and extension. The raw EMG signals acquired by the module are typically weak analog signals; therefore, the module can be equipped with amplifiers to amplify the signal amplitude and filters to remove noise. The processed signal is then converted into a digital signal by the module's internal analog-to-digital converter, and finally, via the aforementioned communication connection, the EMG signal sequence is input to the processor for further processing.

[0027] It should be noted that the processor is the core control module of the system, which can be a microprocessor, digital signal processor, field-programmable gate array, or a general-purpose computer or embedded system with computing capabilities.

[0028] In this embodiment, the processor first determines the target joint angle of the patient's affected limb based on the electromyographic (EMG) signals. Here, the target joint angle is a key intermediate variable, representing the expected angle of motion of the affected limb based on the movement intention of the healthy side. The processor can determine the target joint angle corresponding to the EMG signal using a prediction model that translates the EMG signal from the healthy limb to the joint motion angle of the affected limb.

[0029] Specifically, the processor first receives the electromyographic signal sequence from the electromyographic signal acquisition module and performs further preprocessing, such as feature extraction. Then, the processor uses a pre-trained prediction model to map the processed electromyographic signal sequence into continuous joint angle prediction values, i.e., target joint angles.

[0030] Furthermore, after determining the target joint angle that the affected limb needs to reach, the processor continues to calculate the level of electrical stimulation required to drive the affected limb to move to that target joint angle.

[0031] In one example, the processor stores a preset stimulus-response model. After determining the target joint angle of the patient's affected limb based on the electromyographic signal, the processor obtains the target electrical stimulation signal parameter value corresponding to the current electromyographic signal by performing inverse solution on the stimulus-response model.

[0032] In one example, the electrical stimulation control system may also include an angle measurement module that is connected in communication with the processor to measure and continuously feed back the actual joint angle of the patient's affected limb to the processor in real time.

[0033] The angle measurement module can be an inertial measurement unit strapped to the joint of the affected limb, which calculates the real-time angle of the joint by fusing data from its internal gyroscope and accelerometer. Alternatively, the angle measurement module can be an electronic goniometer fixed to the limb linkage.

[0034] Here, when the processor receives the electromyographic (EMG) signal from the EMG signal acquisition module, it calculates the target electrical stimulation signal parameter value corresponding to the current EMG signal by inversely solving the stimulus-response model. Then, it determines the patient's historical EMG signal data during this electrical stimulation control treatment and checks if there is a target historical EMG signal identical to the current EMG signal. If an identical target historical EMG signal exists, the error between the actual joint angle and the target joint angle corresponding to each target historical EMG signal is obtained. Then, the mean error of all errors is calculated. When the mean error is greater than a preset error threshold, a preset amplitude is added to the calculated target electrical stimulation signal parameter value corresponding to the current EMG signal to obtain the corrected target electrical stimulation signal parameter value, which is then input to the electrical stimulation module.

[0035] In one example, to improve the accuracy of error correction, the weight of the error corresponding to each target historical electromyography signal can be set according to the time difference between the signal time point corresponding to each target historical electromyography signal and the signal time point corresponding to the current electromyography signal. For example, the larger the time difference, the smaller the weight. Finally, the mean error corresponding to all errors is calculated according to the corresponding weight.

[0036] Furthermore, to overcome the lag in step correction based on historical error mean, this embodiment introduces a predictive error correction mechanism based on a Kalman filter. Specifically, a dynamic state-space model of the affected limb's movement is established internally by the processor, using the target joint angle as the system input and the actual joint angle fed back by the angle measurement module as the observed value. The Kalman filter fuses the joint angle predicted by the model with the actual joint angle measured by the sensor to estimate the system state and the prediction error covariance in real time. When calculating the electrical stimulation parameters for the next moment, the processor not only considers the target joint angle but also incorporates the tracking error predicted by the Kalman filter for the next moment to compensate for the calculated target electrical stimulation parameter value. For example, if the Kalman filter predicts a positive angle deviation in the next moment, the processor will slightly reduce the electrical stimulation parameter value in advance. This predictive, continuous closed-loop feedback control method enables the system to respond more smoothly and quickly to and suppress disturbances, improving the tracking accuracy and stability of the affected limb's movement.

[0037] In some embodiments, the target electrical stimulation signal parameter values ​​may include one or more of the following: stimulation current intensity, pulse frequency, and pulse width.

[0038] In practical applications, processors can employ various strategies to determine the target electrical stimulation signal parameter values. For example, the pulse frequency and pulse width can be set to fixed values, and then the varying stimulation current intensity value can be dynamically calculated and output based solely on the target joint angle.

[0039] In this embodiment, after receiving the target electrical stimulation signal parameter value from the processor, the electrical stimulation module generates electrical pulses with corresponding waveforms and intensities. These electrical pulses are applied through electrode pads attached to the patient's affected limb, thereby causing passive contraction of the muscles on the affected side and driving joint movement in the affected limb.

[0040] In some embodiments, the electrical stimulation module includes at least one electrical stimulation electrode and at least one multichannel stimulator.

[0041] Here, the electrical stimulation electrodes, similar to surface electromyography electrodes, are electrode pads adhered to the skin surface, used to transmit the electrical pulses generated by the stimulator to the target muscles of the affected limb. The multichannel stimulator has multiple independent output channels, capable of simultaneously outputting electrical stimulation signals with different parameters to multiple electrode pairs according to multiple sets of different instructions from the processor.

[0042] The electrical stimulation control system in this embodiment captures electromyographic signals from the healthy limb in real time through an electromyographic signal acquisition module. Combined with the processor's accurate prediction of the target joint angle of the affected limb and dynamic matching of electrical stimulation parameters, the electrical stimulation module outputs corresponding electrical stimulation, achieving synchronous linkage between healthy side movements and affected side electrical stimulation. This invention allows for flexible adjustment of electrical stimulation parameters according to changes in healthy side movements, improving the matching degree between electrical stimulation parameters and the movement needs of the affected side.

[0043] It should be noted that each implementation method of this application can be freely combined, rearranged, or executed individually, and does not need to rely on or depend on a fixed execution order.

[0044] In some embodiments, the target joint angle includes a predicted joint angle sequence with an increasing trend. The processor includes a first model module, which is used to input the time-series features formed by at least one time-domain feature extracted from the electromyography signal, including the root mean square value, integral electromyography value, and waveform length, and at least one frequency-domain feature extracted from the average power frequency, median frequency, peak frequency, and power spectral density, into a pre-trained target long short-term memory network model to obtain the target joint angle output by the target long short-term memory network model.

[0045] It should be noted that the first model module can be a software program or firmware that performs a specific function inside the processor, or a specially configured hardware logic unit; there are no restrictions on this.

[0046] In this embodiment, the first model module acquires the raw electromyographic (EMG) signal input from the EMG signal acquisition module and extracts its temporal features. Here, temporal features refer to features that reflect the dynamic changes of the EMG signal over time. For example, within a sliding time window, features such as the root mean square (RMS), integral EMG value, and waveform length of the EMG signal can be calculated, and these feature values ​​can be arranged in chronological order to form a multidimensional temporal feature. The temporal feature is then input into the target long short-term memory (LSTM) network model.

[0047] Furthermore, to enable the model to perceive and adapt to muscle fatigue, this embodiment, while extracting time-domain features, also performs Fast Fourier Transform or Wavelet Transform on the electromyographic signals within the same sliding time window to extract their frequency-domain features, such as median frequency and average power frequency. It should be noted that during muscle fatigue, the power spectrum of the electromyographic signal shifts to the lower frequency region, resulting in changes in median frequency and average power frequency. The processor fuses the extracted time-domain and frequency-domain features to form a richer hybrid feature vector sequence, which is then input into the target long short-term memory network model. Since the input features also contain frequency-domain information directly related to fatigue, the trained target long short-term memory network model can not only predict movement intentions but also implicitly learn the impact of fatigue on motor ability, thereby outputting a predicted joint angle sequence that considers fatigue factors and is more consistent with physiological reality.

[0048] After receiving the temporal features corresponding to the electromyographic signals, the target long short-term memory network model outputs a predicted joint angle sequence, which is the target joint angle. Here, the target joint angle is not a single, discrete angle value, but rather includes a sequence of predicted joint angles that exhibit an increasing trend.

[0049] The target long short-term memory network model is trained in the following way: Collect electromyography timing signals and joint angle timing data of the patient's unaffected limb when performing the target action to form a training sample set; The electromyography test timing signals in the training sample set are used as training inputs, and the joint angle test timing data corresponding to the electromyography test timing signals are used as training labels. The initial long short-term memory network model is trained based on the training input and the training labels to obtain the trained target long short-term memory network model.

[0050] In some embodiments, electrodes of an electromyography (EMG) signal acquisition module can be placed on the healthy limb with temporal characteristics. At the same time, an inertial measurement unit is used to measure the joint angle temporal test data of the healthy limb when performing a target action, such as a uniform wrist flexion and extension movement. This yields multiple sets of paired EMG test temporal signals and joint angle temporal test data, which together constitute a training sample set.

[0051] Next, the training sample set is divided into a training set and a validation set. The electromyography (EMG) test timing signals in the training set are used as the training input, and the joint angle test timing data corresponding to their sampling times are used as the training labels.

[0052] Then, an initial long short-term memory (LSTM) network model with initial weight parameters is constructed, and optimization algorithms, such as the Adam optimizer and loss function, are used to continuously adjust the network weights through backpropagation to minimize the error between the predicted joint angle sequence output by the model and the true joint angle labels. Once the model reaches the preset accuracy or training number of iterations on the validation set, training is complete, and the resulting model is the trained target LSM network model.

[0053] The electrical stimulation control system in this embodiment processes the temporal relationship between electromyographic signals and joint angles through a long short-term memory network model, so as to drive the affected limb to smoothly and continuously reproduce the movement process of the healthy side.

[0054] In some embodiments, the electrical stimulation control system further includes motion sensors; The motion sensor is used to collect the real joint angle sequence of the affected limb and input the real joint angle sequence into the first model module; The first model module is further configured to update the network weights of the target long short-term memory network model based on the loss function value between the actual joint angle sequence and the predicted joint angle sequence under at least one rehabilitation training round.

[0055] It should be understood that the true joint angle sequence refers to the actual dynamic change in the joint angle of the affected limb over time under electrical stimulation. In this embodiment, the true joint angle sequence of the affected limb is collected in real time by setting a motion sensor, and the collected true joint angle sequence is input into the first model module.

[0056] In this embodiment, the first model module stores the actual joint angle sequence and the corresponding predicted joint angle sequence for each rehabilitation training round, and then calculates the difference between the two, which is the loss function value. When the first model module detects that the loss function value exceeds the preset loss threshold, it adjusts the model parameters of the target long short-term memory network model according to the preset model parameter adjustment rules.

[0057] In some embodiments, the processor includes a second model module, which is configured to input the target joint angle into a pre-fitted stimulus-response model and perform inverse solving of the stimulus-response model to obtain the real-valued solution output by the stimulus-response model; and .

[0058] When there is more than one real solution, the minimum solution within a preset monotonically increasing interval is determined as the target electrical stimulation signal parameter value; the preset monotonically increasing interval is obtained by differentiating the stimulus-response model; The stimulus-response model is obtained by fitting the model in the following way: Collect joint angle test data of the patient's affected limb under electrical stimulation signals corresponding to different electrical stimulation signal test parameter values, and combine the joint angle test data and the electrical stimulation signal test parameter values ​​corresponding to the joint angle test data into data points; Using the electrical stimulation signal parameters as independent variables and the joint angle as the dependent variable, a third-order polynomial fitting is performed on multiple sets of data points to obtain the stimulus-response model.

[0059] In this embodiment, the second model module is equipped with a pre-fitted stimulus-response model, which can calculate the target electrical stimulation signal parameter value that drives the affected limb to achieve the target joint angle.

[0060] In practical applications, electrical stimulation electrodes can be attached to the patient's affected limb, and an angle measuring device can be installed on the corresponding joint, such as the wrist joint, to obtain joint angle test data in real time.

[0061] Subsequently, the system controls the electrical stimulation electrodes to output a series of electrical stimulation signals with different parameter values ​​according to a preset program. For example, using the electrical stimulation current intensity as a variable parameter, starting from 0 mA, the current intensity is gradually increased in a fixed step until it reaches the upper limit that is comfortable and safe for the patient. At each current intensity value, the stimulation is maintained for a short period of time to allow the joint angle to reach a stable value. At this time, the angle measuring device records the joint angle value of the stable state as joint angle test data.

[0062] Through the above process, multiple pairs of data points are collected, consisting of electrical stimulation signal test parameter values ​​and their corresponding joint angle test data. After obtaining these multiple sets of data points, the electrical stimulation signal parameters are set as independent variables, and the joint angles as dependent variables. A third-order polynomial is used for fitting, such as y=ax. 3 +bx 2 +cx+d, where y is the joint angle and x is the electrical stimulation signal parameter.

[0063] It should be noted that the response of muscles to electrical stimulation is typically nonlinear. In the low-intensity stimulation region, muscle contraction is not significant; as the intensity increases, the change in joint angle accelerates; when the stimulation intensity reaches a certain level, muscle contraction tends to saturate, and the change in joint angle slows down again. Therefore, the use of third-order polynomial fitting in this embodiment can improve control accuracy while ensuring a relatively small computational load.

[0064] Furthermore, to address the time-varying nature of muscle response due to fatigue, this embodiment further improves the stimulus-response model by establishing a model that incorporates fatigue factors. Specifically, when fitting the stimulus-response model, in addition to collecting electrical stimulation parameters and joint angle data, a fatigue index F is also recorded simultaneously. This fatigue index F can be calculated based on the cumulative electrical stimulation energy applied to the affected limb or the frequency domain features extracted from the electromyographic signal of the healthy side. Therefore, the fitted result is no longer a two-dimensional curve of y = f(x), but a three-dimensional surface model of y = f(x, F), where y is the joint angle, x is the electrical stimulation signal parameter, and F is the fatigue index. During the reverse solution process, the processor calculates the current fatigue index in real time and then solves the equation to obtain the target electrical stimulation parameter x required under the current fatigue state. This allows the stimulus-response model to dynamically adapt to changes in muscle state. When the muscle is fatigued, i.e., the F value increases, the model automatically outputs a higher electrical stimulation parameter value to achieve the same target joint angle, thereby achieving precise compensation for fatigue effects.

[0065] The electrical stimulation control system in this embodiment uses a stimulus-response model to convert the target joint angle of the affected side into a matching target electrical stimulation signal parameter value, thereby achieving precise matching between muscle contraction and joint movement angle on the affected side, and improving the synchronization and accuracy of the movements on the affected side and the healthy side during rehabilitation training.

[0066] In some embodiments, the processor further includes a security threshold module; The safety threshold module is used to input the preset safety signal parameter threshold to the electrical stimulation module when the target electrical stimulation signal parameter value is detected to exceed the preset safety signal parameter threshold.

[0067] In this embodiment, before rehabilitation training begins, one or more safety signal parameter thresholds can be preset according to the patient's specific condition, such as skin sensitivity, pain threshold, and muscle tolerance. For example, the maximum permissible stimulation current intensity and maximum pulse width can be set, and these safety signal parameter thresholds can be stored in the safety threshold module.

[0068] During real-time system operation, before sending the electrical stimulation signal parameter values ​​to the electrical stimulation module, the processor first sends the electrical stimulation signal parameter values ​​to the safety threshold module for checking. The safety threshold module continuously monitors whether the electrical stimulation signal parameter values ​​exceed the preset safety signal parameter threshold. If the detected electrical stimulation signal parameter value does not exceed the preset safety signal parameter threshold, the electrical stimulation signal parameter value is allowed to pass and is normally input to the electrical stimulation module. If the detected electrical stimulation signal parameter value exceeds the preset safety signal parameter threshold, the safety threshold module will perform a limiting operation, that is, use the preset safety signal parameter threshold as the output and input it to the electrical stimulation module.

[0069] The electrical stimulation control system in this embodiment, by introducing a safety threshold module, effectively ensures the safety and comfort of the patient's rehabilitation diagnosis without affecting the control accuracy within the normal treatment range.

[0070] This invention also proposes an electrical stimulation control method, applied to the processor in the electrical stimulation control system proposed in the above embodiments, with reference to... Figure 2 As shown, the method includes steps 210, 220 and 230.

[0071] Step 210: Acquire the electromyographic signals of the patient's unaffected limbs collected by the electromyographic signal acquisition module in the electrical stimulation control system.

[0072] In practical implementation, an electromyography (EMG) signal acquisition module typically includes at least one or more surface EMG electrodes and corresponding signal conditioning and acquisition circuitry. EMG signals refer to the bioelectrical signals generated when muscles contract under nerve control; they reflect a person's motor intentions and muscle activity levels. In this embodiment, by acquiring EMG signals from the unaffected limb, the patient's actively generated motor intentions can be captured in real time.

[0073] For example, in applications targeting the upper limbs, such as wrist rehabilitation, the electromyography (EMG) signal acquisition module can place surface EMG electrodes on the extensor and flexor muscles of the patient's unaffected forearm, such as the flexor carpi radialis, flexor carpi ulnaris, flexor digitorum superficialis, and extensor carpi radialis longus—the main muscles related to wrist flexion and extension movements. The raw EMG signals acquired by the module are typically weak analog signals; therefore, the module can be equipped with amplifiers to amplify the signal amplitude and filters to remove noise. The processed signals are then converted into digital signals by the module's internal analog-to-digital converter and finally input to the processor for further processing via the aforementioned communication connection.

[0074] Step 220: Based on the electromyographic signals, determine the target joint angle for driving the patient's affected limb to move synchronously with the patient's healthy limb.

[0075] It should be noted that the processor is the core control module of the system, which can be a microprocessor, digital signal processor, field-programmable gate array, or a general-purpose computer or embedded system with computing capabilities.

[0076] In this embodiment, the processor first determines the target joint angle of the patient's affected limb based on electromyographic signals. Here, the target joint angle is a key intermediate variable, representing the expected angle of motion of the affected limb based on the movement intention of the healthy side. The processor performs this task by establishing and applying a predictive model from the electromyographic signals of the healthy limb to the joint motion angle of the affected limb.

[0077] Specifically, the processor first receives the electromyographic signal sequence from the electromyographic signal acquisition module and performs further preprocessing, such as feature extraction. Then, the processor uses a pre-trained prediction model to map the processed electromyographic signal sequence into continuous joint angle prediction values, i.e., target joint angles.

[0078] Step 230: Based on the target joint angle, the target electrical stimulation signal parameter values ​​are obtained by inversely solving the preset stimulus-response model.

[0079] After determining the target joint angle that the affected limb needs to reach, the processor continues to calculate the level of electrical stimulation required to drive the affected limb to move to that target joint angle.

[0080] In one example, the processor stores a preset stimulus-response model. After determining the target joint angle of the patient's affected limb based on the electromyographic signal, the processor obtains the electrical stimulation signal parameter value corresponding to the current electromyographic signal by performing inverse solution on the stimulus-response model.

[0081] In one example, the electrical stimulation control system may also include an angle measurement module that is connected in communication with the processor to measure and continuously feed back the actual joint angle of the patient's affected limb to the processor in real time.

[0082] The angle measurement module can be an inertial measurement unit strapped to the joint of the affected limb, which calculates the real-time angle of the joint by fusing data from its internal gyroscope and accelerometer. Alternatively, the angle measurement module can be an electronic goniometer fixed to the limb linkage.

[0083] Here, when the processor receives the electromyographic (EMG) signal from the EMG signal acquisition module, it calculates the target electrical stimulation signal parameter value corresponding to the current EMG signal by inversely solving the stimulus-response model. Then, it determines the patient's historical EMG signal data during this electrical stimulation control treatment and checks if there is a target historical EMG signal identical to the current EMG signal. If an identical target historical EMG signal exists, the error between the actual joint angle and the target joint angle corresponding to each target historical EMG signal is obtained. Then, the mean error of all errors is calculated. When the mean error is greater than a preset error threshold, a preset amplitude is added to the calculated target electrical stimulation signal parameter value corresponding to the current EMG signal to obtain the corrected target electrical stimulation signal parameter value, which is then input to the electrical stimulation module.

[0084] In one example, to improve the accuracy of error correction, the weight of the error corresponding to each target historical electromyography signal can be set according to the time difference between the signal time point corresponding to each target historical electromyography signal and the signal time point corresponding to the current electromyography signal. For example, the larger the time difference, the smaller the weight. Finally, the mean error corresponding to all errors is calculated according to the corresponding weight.

[0085] Furthermore, to overcome the lag in step correction based on historical error mean, this embodiment introduces a predictive error correction mechanism based on a Kalman filter. Specifically, a dynamic state-space model of the affected limb's movement is established internally by the processor, using the target joint angle as the system input and the actual joint angle fed back by the angle measurement module as the observed value. The Kalman filter fuses the joint angle predicted by the model with the actual joint angle measured by the sensor to estimate the system state and the prediction error covariance in real time. When calculating the electrical stimulation parameters for the next moment, the processor not only considers the target joint angle but also incorporates the tracking error predicted by the Kalman filter for the next moment to compensate for the calculated target electrical stimulation parameter value. For example, if the Kalman filter predicts a positive angle deviation in the next moment, the processor will slightly reduce the electrical stimulation parameter value in advance. This predictive, continuous closed-loop feedback control method enables the system to respond more smoothly and quickly to and suppress disturbances, improving the tracking accuracy and stability of the affected limb's movement.

[0086] Step 240: Input the target electrical stimulation signal parameter value into the electrical stimulation module in the electrical stimulation control system to drive the electrical stimulation module to output electrical stimulation corresponding to the target electrical stimulation signal parameter value to the patient's affected limb.

[0087] After receiving the electrical stimulation signal parameter values ​​from the processor, the electrical stimulation module generates electrical pulses of corresponding waveforms and intensities. These electrical pulses are applied through electrode pads attached to the patient's affected limb, thereby causing passive contraction of the muscles on the affected side and driving joint movement in the affected limb.

[0088] The electrical stimulation control method in this embodiment captures electromyographic signals from the healthy limb in real time through an electromyographic signal acquisition module. Combined with the processor's accurate prediction of the target joint angle of the affected limb and dynamic matching of electrical stimulation parameters, the electrical stimulation module outputs corresponding electrical stimulation, achieving synchronous linkage between healthy side movements and affected side electrical stimulation. This invention allows for flexible adjustment of electrical stimulation parameters according to changes in healthy side movements, improving the matching degree between electrical stimulation parameters and the movement needs of the affected side.

[0089] In some embodiments, the target joint angle includes an increasing sequence of predicted joint angles, and determining the target joint angle of the patient's affected limb based on the electromyographic signal includes: The time sequence formed by at least one time-domain feature extracted from the electromyography signal, including the root mean square value, integral electromyography value, and waveform length, and at least one frequency-domain feature extracted from the average power frequency, median frequency, peak frequency, and power spectral density, is input into a pre-trained target long short-term memory network model to obtain the target joint angle output by the target long short-term memory network model.

[0090] In this embodiment, the first model module acquires the raw electromyographic (EMG) signal input from the EMG signal acquisition module and extracts its temporal features. Here, temporal features refer to features that reflect the dynamic changes of the EMG signal over time. For example, within a sliding time window, features such as the root mean square (RMS), integral EMG value, and waveform length of the EMG signal can be calculated, and these feature values ​​can be arranged in chronological order to form a multidimensional temporal feature. Then, the temporal features are input into the target long short-term memory (LSTM) network model. Furthermore, to enable the model to perceive and adapt to muscle fatigue, this embodiment, while extracting time-domain features, also performs Fast Fourier Transform or Wavelet Transform on the electromyographic signals within the same sliding time window to extract their frequency-domain features, such as median frequency and average power frequency. It should be noted that during muscle fatigue, the power spectrum of the electromyographic signal shifts to the lower frequency region, resulting in changes in median frequency and average power frequency. The processor fuses the extracted time-domain and frequency-domain features to form a richer hybrid feature vector sequence, which is then input into the target long short-term memory network model. Since the input features also contain frequency-domain information directly related to fatigue, the trained target long short-term memory network model can not only predict movement intentions but also implicitly learn the impact of fatigue on motor ability, thereby outputting a predicted joint angle sequence that considers fatigue factors and is more consistent with physiological reality.

[0091] After receiving the temporal features corresponding to the electromyographic signals, the target long short-term memory network model outputs a predicted joint angle sequence, which is the target joint angle. Here, the target joint angle is not a single, discrete angle value, but rather includes a sequence of predicted joint angles that exhibit an increasing trend.

[0092] The target long short-term memory network model is trained in the following way: Electromyography timing signals and joint angle timing data of the unaffected limbs of patients and test subjects when performing the target action are collected to form a training sample set; The electromyography test timing signals in the training sample set are used as training inputs, and the joint angle test timing data corresponding to the electromyography test timing signals are used as training labels. The initial long short-term memory network model is trained based on the training input and the training labels to obtain the trained target long short-term memory network model.

[0093] In some embodiments, electrodes of an electromyography (EMG) signal acquisition module can be placed on the patient's unaffected limb. At the same time, an inertial measurement unit is used to measure the joint angle timing test data of the unaffected limb when performing a target action, such as uniform wrist flexion and extension. This yields multiple sets of paired EMG timing signals and joint angle timing test data, which together constitute a training sample set.

[0094] Next, the training sample set is divided into a training set and a validation set. The electromyography (EMG) test timing signals in the training set are used as the training input, and the joint angle test timing data corresponding to their sampling times are used as the training labels.

[0095] Then, an initial long short-term memory (LSTM) network model with initial weight parameters is constructed, and optimization algorithms, such as the Adam optimizer and loss function, are used to continuously adjust the network weights through backpropagation to minimize the error between the predicted joint angle sequence output by the model and the true joint angle labels. Once the model reaches the preset accuracy or training number of iterations on the validation set, training is complete, and the resulting model is the trained target LSM network model.

[0096] The electrical stimulation control method in this embodiment uses a long short-term memory network model to process the temporal relationship between electromyographic signals and joint angles, so as to drive the affected limb to smoothly and continuously reproduce the movement process of the healthy side.

[0097] In some embodiments, determining the electrical stimulation signal parameter values ​​based on the target joint angle includes: The target joint angle is input into a pre-fitted stimulus-response model. The stimulus-response model is then solved in reverse to obtain the real-valued solution output by the stimulus-response model; and, When there is more than one real solution, the minimum solution within a preset monotonically increasing interval is determined as the target electrical stimulation signal parameter value; the preset monotonically increasing interval is obtained by differentiating the stimulus-response model; The stimulus-response model is obtained by fitting the model in the following way: Collect joint angle test data of the patient's affected limb under electrical stimulation signals corresponding to different electrical stimulation signal test parameter values, and combine the joint angle test data and the electrical stimulation signal test parameter values ​​corresponding to the joint angle test data into data points; Using the electrical stimulation signal parameters as independent variables and the joint angle as the dependent variable, a third-order polynomial fitting is performed on multiple sets of data points to obtain the stimulus-response model.

[0098] In this embodiment, the second model module is equipped with a pre-fitted response-stimulus inverse model, which can calculate the electrical stimulation signal parameter values ​​that drive the affected limb to achieve the target joint angle.

[0099] In practical applications, electrical stimulation electrodes can be attached to the affected limb of the test subject, and an angle measuring device can be installed on the corresponding joint, such as the wrist joint, to obtain joint angle test data in real time.

[0100] Subsequently, the system controls the electrical stimulation electrodes to output a series of electrical stimulation signals with different parameter values ​​according to a preset program. For example, using the electrical stimulation current intensity as a variable parameter, starting from 0 mA, the current intensity is gradually increased in a fixed step until it reaches the upper limit that feels comfortable and safe for the tester. At each current intensity value, the stimulation is maintained for a short period of time to allow the joint angle to reach a stable value. At this point, the angle measuring device records the joint angle value at this stable state as joint angle test data.

[0101] Through the above process, multiple pairs of data points are collected, consisting of electrical stimulation signal test parameter values ​​and their corresponding joint angle test data. After obtaining these multiple sets of data points, the electrical stimulation signal parameters are set as independent variables, and the joint angles as dependent variables. A third-order polynomial is used for fitting, such as y=ax. 3 +bx 2 +cx+d, where y is the joint angle and x is the electrical stimulation signal parameter.

[0102] It should be noted that the response of muscles to electrical stimulation is typically nonlinear. In the low-intensity stimulation region, muscle contraction is not significant; as the intensity increases, the change in joint angle accelerates; when the stimulation intensity reaches a certain level, muscle contraction tends to saturate, and the change in joint angle slows down again. Therefore, the use of third-order polynomial fitting in this embodiment can improve control accuracy while ensuring a relatively small computational load.

[0103] Furthermore, to address the time-varying nature of muscle response due to fatigue, this embodiment further improves the stimulus-response model by establishing a model that incorporates fatigue factors. Specifically, when fitting the stimulus-response model, in addition to collecting electrical stimulation parameters and joint angle data, a fatigue index F is also recorded simultaneously. This fatigue index F can be calculated based on the cumulative electrical stimulation energy applied to the affected limb or the frequency domain features extracted from the electromyographic signal of the healthy side. Therefore, the fitted result is no longer a two-dimensional curve of y = f(x), but a three-dimensional surface model of y = f(x, F), where y is the joint angle, x is the electrical stimulation signal parameter, and F is the fatigue index. During the reverse solution process, the processor calculates the current fatigue index in real time and then solves the equation to obtain the target electrical stimulation parameter x required under the current fatigue state. This allows the stimulus-response model to dynamically adapt to changes in muscle state. When the muscle is fatigued, i.e., the F value increases, the model automatically outputs a higher electrical stimulation parameter value to achieve the same target joint angle, thereby achieving precise compensation for fatigue effects.

[0104] The electrical stimulation control method in this embodiment uses a response-stimulus inverse model to convert the target joint angle of the affected side into a matching electrical stimulation signal parameter value, thereby achieving precise matching between muscle contraction and joint movement angle on the affected side, and improving the synchronization and accuracy of the movements on the affected side and the healthy side during rehabilitation training.

[0105] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device includes a processor 310, a communications interface 320, a memory 330, and a communication bus 340. The processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logic instructions stored in the memory 330 to execute an electrical stimulation control method, which includes: Acquire electromyographic signals of the patient's unaffected limb from the electromyographic signal acquisition module in the electrical stimulation control system; Based on the electromyographic signals, a target joint angle is determined for driving the patient's affected limb to move synchronously with the patient's healthy limb; Based on the target joint angle, the target electrical stimulation signal parameter values ​​are obtained by inversely solving the preset stimulus-response model. The target electrical stimulation signal parameter value is input to the electrical stimulation module in the electrical stimulation control system to drive the electrical stimulation module to output the electrical stimulation corresponding to the target electrical stimulation signal parameter value to the patient's affected limb.

[0106] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and stored in a computer-readable storage medium when sold or used as independent products. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, is embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., each of which can store program code.

[0107] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the devices can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in each of the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. An electrical stimulation control system, characterized in that, include: The electromyography signal acquisition module, electrical stimulation module, and processor are connected via communication. The electromyography (EMG) signal acquisition module is used to acquire EMG signals from the patient's unaffected limb and input the EMG signals to the processor. The processor is configured to determine, based on the electromyographic signals, a target joint angle for driving the patient's affected limb to move synchronously with the patient's healthy limb. Furthermore, based on the target joint angle, the target electrical stimulation signal parameter value is obtained by inversely solving the preset stimulus-response model, and the target electrical stimulation signal parameter value is input to the electrical stimulation module; The electrical stimulation module is used to output electrical stimulation corresponding to the target electrical stimulation signal parameter value to the patient's affected limb; The target joint angle includes a predicted joint angle sequence with an increasing trend. The processor includes a first model module. The first model module is used to input the time-series features formed by at least one time-domain feature extracted from the electromyography signal, including the root mean square value, integral electromyography value, and waveform length, and at least one frequency-domain feature extracted from the average power frequency, median frequency, peak frequency, and power spectral density, into a pre-trained target long short-term memory network model to obtain the predicted joint angle sequence output by the target long short-term memory network model. The processor includes a second model module, which is used to input the target joint angle into a pre-fitted stimulus-response model and perform inverse solving of the stimulus-response model to obtain the real-valued solution output by the stimulus-response model; and, When there is more than one real solution, the minimum solution within a preset monotonically increasing interval is determined as the target electrical stimulation signal parameter value; the preset monotonically increasing interval is obtained by differentiating the stimulus-response model; The stimulus-response model is obtained by fitting the model in the following way: Collect joint angle test data of the patient's affected limb under electrical stimulation signals corresponding to different electrical stimulation signal test parameter values, and combine the joint angle test data and the electrical stimulation signal test parameter values ​​corresponding to the joint angle test data into data points; Using the electrical stimulation signal parameters as independent variables and the joint angle as the dependent variable, a third-order polynomial fitting is performed on multiple sets of data points to obtain the stimulus-response model.

2. The electrical stimulation control system according to claim 1, characterized in that, The target long short-term memory network model was trained in the following way: Collect electromyography timing signals and joint angle timing data of the patient's unaffected limb when performing the target action to form a training sample set; The electromyography test timing signals in the training sample set are used as training inputs, and the joint angle test timing data corresponding to the electromyography test timing signals are used as training labels. The initial long short-term memory network model is trained based on the training input and the training labels to obtain the trained target long short-term memory network model.

3. The electrical stimulation control system according to claim 2, characterized in that, The electrical stimulation control system also includes motion sensors; The motion sensor is used to collect the real joint angle sequence of the affected limb and input the real joint angle sequence into the first model module; The first model module is further configured to update the network weights of the target long short-term memory network model based on the loss function value between the actual joint angle sequence and the predicted joint angle sequence under at least one rehabilitation training round.

4. The electrical stimulation control system according to claim 1, characterized in that, The target electrical stimulation signal parameter values ​​include one or more of the following: stimulation current intensity, pulse width, or pulse frequency.

5. The electrical stimulation control system according to claim 1, characterized in that, The processor also includes a security threshold module; The safety threshold module is used to input the preset safety signal parameter threshold to the electrical stimulation module when the target electrical stimulation signal parameter value is detected to exceed the preset safety signal parameter threshold.

6. The system according to claim 1, characterized in that, The electromyography (EMG) signal acquisition module has at least one surface EMG electrode; the electrical stimulation module includes at least one electrical stimulation electrode and at least one multichannel stimulator.

Images