Neuromuscular electrical stimulation rehabilitation control method and system based on electroencephalogram intention recognition

Abstract

The application discloses a neuromuscular electrical stimulation rehabilitation control method and system based on electroencephalogram intention recognition, relates to the technical field of intelligent rehabilitation control, and comprises the following steps: collecting electroencephalogram signal data of a user in a rehabilitation training process, processing the electroencephalogram signal data, constructing an electroencephalogram intention discrimination model to dynamically identify the motion intention category of the user at present, and outputting corresponding intention signals and confidence levels; according to the recognized intention signals and confidence levels and the current electromyography state, a neuromuscular electrical stimulation parameter optimization model is used to output a neuromuscular electrical stimulation parameter set, which is sent to a neuromuscular electrical stimulation device to control the electrical stimulation of a target muscle group. The method disclosed by the application realizes high-quality preprocessing of electroencephalogram signals, dynamic discrimination of intention, and self-adaptive optimization of individualized electrical stimulation parameters through the construction of an electroencephalogram intention real-time identification and neuromuscular electrical stimulation closed-loop optimization control method, and can accurately match the motion intention of the user with the electromyography state.

Description

Technical Field

[0001] This invention relates to the field of intelligent rehabilitation control technology, specifically to a neuromuscular electrical stimulation rehabilitation control method and system based on EEG intention recognition. Background Technology

[0002] With the continuous development of neuroscience and rehabilitation medicine, the application of neuromodulation technology driven by electroencephalogram (EEG) signals in the rehabilitation of motor dysfunction has attracted increasing attention. In recent years, brain-computer interface (BCI) technology has advanced rapidly, enabling real-time acquisition and intelligent decoding of EEG signals, providing a data foundation for motor intention recognition and neuromuscular modulation. Meanwhile, neuromuscular electrical stimulation (NMS), as an effective means of muscle function recovery, has been widely used in the rehabilitation training of patients with hemiplegia, paraplegia, and muscular atrophy. Current research focuses on how to deeply integrate EEG intention recognition with individualized electrical stimulation control to form a closed-loop, adaptive intelligent rehabilitation control strategy, thereby improving the accuracy and effectiveness of rehabilitation training.

[0003] Currently, existing methods for EEG intention recognition and neuromuscular electrical stimulation control still have many shortcomings. First, existing technologies mostly employ simple feature extraction and classification algorithms, lacking sufficient exploration of the dynamic temporal characteristics and nonlinear relationships in EEG signals. This limits the accuracy of intention recognition and makes it difficult to meet the real-time and stability requirements of complex rehabilitation scenarios. Second, most existing neuromuscular electrical stimulation strategies rely on preset parameters or fixed mapping relationships, lacking a deep linkage mechanism between the user's current intention confidence and electromyographic state. This makes it impossible to dynamically optimize stimulation parameters according to individual physiological states, easily leading to understimulation or overstimulation. Furthermore, existing systems mostly adopt open-loop or semi-open-loop control architectures, lacking real-time electromyographic feedback closed-loop optimization, making it difficult to achieve individualized adaptive control of the electrical stimulation process. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing EEG intention recognition and neuromuscular electrical stimulation control methods have insufficient intention recognition accuracy, lack dynamic optimization mechanisms for electrical stimulation parameters, lack real-time closed-loop adaptive control capabilities, and how to achieve individualized optimization control of neuromuscular electrical stimulation based on real-time EEG intention.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition, comprising: acquiring EEG signal data of the user in real time during rehabilitation training using a multi-channel EEG acquisition device; performing bandpass filtering, artifact removal, and adaptive noise suppression on the EEG signal data to output a purified EEG signal data stream; extracting features from the purified EEG signal data stream to output a movement intention feature vector; constructing an EEG intention discrimination model to dynamically identify the user's current movement intention category and outputting the corresponding intention signal and confidence level; and, based on the identified intention signal and confidence level and the current electromyographic state, using a neuromuscular electrical stimulation parameter optimization model to output a neuromuscular electrical stimulation parameter set, which is then sent to a neuromuscular electrical stimulation device to control the target muscle group for electrical stimulation.

[0007] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the bandpass filtering includes setting the passband range to 0.5 Hz to 45 Hz.

[0008] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the feature extraction includes extracting motion intention feature vectors from the purified EEG signal data stream by combining wavelet packet decomposition and power spectral density analysis.

[0009] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the EEG intention discrimination model includes processing the motor intention feature vector through kernel mapping, constructing a multi-dimensional feature mapping space by combining dynamic temporal features and nonlinear activation functions, and outputting the corresponding intention signal and confidence level.

[0010] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the output of the corresponding intention signal and confidence level includes intention category discrimination based on the confidence level and a preset category threshold, and determining the final movement intention category by comparing the discrimination score of each movement intention category with the corresponding threshold.

[0011] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the neuromuscular electrical stimulation parameter optimization model includes the fusion of motor intention category, intention confidence and real-time electromyographic state information, and adopts an optimization algorithm that includes nonlinear mapping, integral feedback and normalization adjustment mechanism to output the amplitude, frequency and pulse width of the electrical stimulation current.

[0012] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition described in this invention, the output neuromuscular electrical stimulation parameter set includes real-time acquisition of electromyographic response feedback data during neuromuscular electrical stimulation, and the feature weights, normalization adjustment factors, and feedback gain coefficients in the neuromuscular electrical stimulation parameter optimization model are updated based on the feedback data using a dynamic adaptive adjustment mechanism.

[0013] Another objective of this invention is to provide a neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition. This system can extract features from the purified EEG signal data stream, output a motion intention feature vector, construct an EEG intention discrimination model to dynamically identify the user's current motion intention category, and output the corresponding intention signal and confidence level. This solves the problem of insufficient intention recognition accuracy in current EEG intention recognition and neuromuscular electrical stimulation control methods.

[0014] As a preferred embodiment of the neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition described in this invention, it includes: a real-time EEG signal acquisition and preprocessing module, a motor intention feature extraction and dynamic discrimination module, and a neuromuscular electrical stimulation parameter optimization and execution control module; the real-time EEG signal acquisition and preprocessing module is used to acquire the user's EEG signal data in real time during rehabilitation training through a multi-channel EEG acquisition device, and perform bandpass filtering, artifact removal, and adaptive noise suppression processing on the raw EEG signal to output a purified EEG signal data stream; the motor intention feature extraction and dynamic discrimination module is used to extract features from the purified EEG signal data stream, generate a motor intention feature vector, and dynamically identify the user's current motor intention category based on the EEG intention discrimination model, and output the intention signal and corresponding confidence level in real time; the neuromuscular electrical stimulation parameter optimization and execution control module is used to dynamically generate a set of stimulation current amplitude, frequency, and pulse width parameters based on the real-time identified motor intention signal, intention confidence level, and current electromyographic state, using a neuromuscular electrical stimulation parameter optimization model, and control the neuromuscular electrical stimulation device to drive the target muscle group to perform electrical stimulation.

[0015] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition.

[0017] The beneficial effects of this invention are as follows: The neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition provided by this invention effectively eliminates low-frequency drift, high-frequency noise, and non-brain-source interference through real-time acquisition and high-quality preprocessing of multi-channel EEG signals, significantly improving the signal-to-noise ratio and stability of motor intention features, and providing reliable data support for subsequent intention discrimination. Based on dynamic feature extraction and kernel mapping discrimination models, it achieves high-precision real-time recognition of the user's current motor intention category and confidence level, and can dynamically adjust the electrical stimulation control strategy according to the user's intention, enhancing the active interactivity and individual adaptability of rehabilitation training. By integrating the intention signal, confidence level, and electromyographic state into a neuromuscular electrical stimulation parameter optimization model, it can dynamically generate individualized electrical stimulation parameter sets, achieving precise matching of the user's physiological state and intention needs, avoiding the problems of insufficient or excessive stimulation caused by traditional fixed parameter stimulation. Overall, this invention achieves deep coupling and closed-loop optimization of EEG intention and electrical stimulation control, significantly improving the real-time performance, accuracy, and safety of rehabilitation training, and has superior technical effects compared to existing open-loop control or static parameter mapping methods, showing significant application prospects for improving the efficiency of neuromuscular functional rehabilitation and user experience. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 The first embodiment of the present invention provides an overall flowchart of a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition.

[0020] Figure 2 The first embodiment of the present invention provides a logical step diagram of a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition.

[0021] Figure 3 The following is an overall flowchart of a neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition, provided as a third embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figures 1-2 As an embodiment of the present invention, a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition is provided, comprising:

[0024] S1: Through a multi-channel EEG acquisition device, the EEG signal data of the user during the rehabilitation training process is collected in real time. The EEG signal data is processed by bandpass filtering, artifact removal and adaptive noise suppression, and the purified EEG signal data stream is output.

[0025] Furthermore, by configuring a multi-channel EEG acquisition device, the EEG activity signals of users in rehabilitation training scenarios can be collected in real time. It is preferable to use a high sampling rate (not less than 512 Hz) and multi-channel (not less than 16 channels) EEG acquisition system, covering key brain regions related to motor intentions such as the prefrontal cortex, central area, and parietal lobe.

[0026] Raw EEG signals collected The signal suffers from motion artifacts, power frequency interference, and physiological noise, thus requiring multi-stage preprocessing. First, a bandpass filter (typically with a passband range of 0.5 Hz to 45 Hz) is used to clean the frequency domain of the signal, removing low-frequency drift and high-frequency noise.

[0027] Subsequently, the Independent Component Analysis (ICA) algorithm was applied to remove artifacts from the signal, separating and removing non-brain-derived components such as electrooculography (EOG) and electromyography (EMG).

[0028] Based on this, an adaptive noise suppression algorithm (such as a transform domain adaptive filter) is used to suppress the remaining noise in real time, and finally the purified EEG signal is output. .

[0029] To further enhance feature availability, Normalization and segmented windowing are performed, using a sliding window mechanism (window width...). Sliding step size The continuous signal is sliced ​​to provide a high-quality and stable data input source for subsequent feature extraction and intent discrimination models.

[0030] S2: Extract features from the purified EEG signal data stream, output a motion intention feature vector, construct an EEG intention discrimination model to dynamically identify the user's current motion intention category, and output the corresponding intention signal and confidence level.

[0031] Furthermore, using the purified EEG signal data, time-frequency domain analysis methods are employed for feature extraction, with wavelet packet decomposition and power spectral density methods being preferred for extracting multidimensional EEG intention feature vectors. Based on this, an EEG intention discrimination model is constructed using the extracted EEG features. A preferred discrimination mathematical model is designed based on nested kernel mapping and a temporal dynamic integral structure, with a discrimination score of [missing information]. Represented as:

[0032]

[0033] in, For a moment For category The calculated motion intent discrimination score, For the dimension of EEG feature vectors, For the first The coefficients of each characteristic, For a moment The extracted first Each EEG characteristic value is expressed in energy density. For the first The dynamic adjustment function corresponding to each feature The characteristic scaling factor is... To normalize the adjustment parameter, For the first The inhibitory factor for each feature is a positive real number. Indicates at time , No. EEG feature values ​​corresponding to each channel (or dimension).

[0034] like , indicating category The intent to discern is very strong; if It is in a critical discrimination state; if If so, it will not be classified as a category at present. .

[0035] It should be noted that the trained EEG intention classification model distinguishes the following four types of motor intention charts, as shown in Table 1.

[0036] Table 1. Sports Intent Chart

[0037]

[0038] The threshold should be determined based on a large number of EEG training samples to identify scores. The distribution is used to determine this. A common practice is to determine the mean score based on the positive examples in the training set. and standard deviation Set as:

[0039]

[0040] in It is a regulatory factor, often taken as This ensures that the judgment has a certain tolerance and avoids overfitting.

[0041] Finally, output After: If If so, it is classified as a category. The system can be configured according to each time interval. Calculate separately And compare their respective thresholds. Finally, the category that meets the conditions and has the highest score is selected as the judgment result.

[0042] During the rehabilitation training process, the system will use the current moment... Collected EEG feature vectors The input is used to construct the EEG intention discrimination model, and the values ​​of each category are dynamically calculated. Discrimination score The discrimination rule is set as follows: if If so, it is currently classified as a motion intention category. If multiple categories meet the criteria, then select the one that best meets the criteria. The largest category is used as the final judgment result.

[0043] The final output motion intention category signal The corresponding confidence information will serve as an important input for the generation and optimization of neuromuscular electrical stimulation parameters, driving the subsequent rehabilitation training control process.

[0044] S3: Based on the identified intention signal, confidence level, and current electromyographic state, the neuromuscular electrical stimulation parameter set is output using the neuromuscular electrical stimulation parameter optimization model and sent to the neuromuscular electrical stimulation device to control the target muscle group for electrical stimulation.

[0045] Furthermore, based on the real-time identified intent signals and confidence levels, combined with the currently collected electromyographic data, a neuromuscular electrical stimulation parameter optimization model is employed;

[0046]

[0047] in, For a moment The amplitude of the electrical stimulation current, The frequency of electrical stimulation. For electrical stimulation pulse width, The total dimension of the input features. For the first The weight of the dimensional feature on the amplitude adjustment, For the first Dimensional input feature values, select ,in Assess the confidence level of the current intent category. For the first Real-time electromyographic signals of the channel, It is a dynamic activation function. As a normalization adjustment factor, The suppression coefficient, The scaling factor is the reference frequency. To optimize the integration window width for frequency, Indicates dynamic feedback weights. To account for the error between the expected and actual electromyography (EMG) values, The angle representing the rate of change of electromyography. This is the pulse width modulation gain coefficient. Indicates composite feature input, For pulse width mapping coefficients, As the pulse width normalization regulator, Indicates at time , No. The fused signal value of each input feature is typically the product of the intention confidence and the current electromyographic signal.

[0048] The range is Typical example This corresponds to stimuli ranging from low to high intensity. The range is Typical example Low-frequency stimulation is used for relaxation, while high-frequency stimulation is used for contraction; The range is Typical example The wider the pulse width, the stronger the stimulation depth.

[0049] This optimization model integrates intent category, intent confidence, and electromyographic state. Through a nonlinear mapping function, integral feedback, and normalization adjustment mechanism, it dynamically adjusts electrical stimulation parameters to ensure that the output parameters match the user's current movement intent and physiological state in real time. The optimization model supports adaptive parameter adjustment, dynamically optimizing the electrical stimulation control strategy based on individual user differences and real-time state changes. The final generated electrical stimulation parameter set is sent to the neuromuscular electrical stimulation device in real time via an interface, driving the target muscle group to produce a movement output highly matched to the user's intent.

[0050] It should be noted that during the execution of neuromuscular electrical stimulation, electromyographic response feedback data is collected in real time, including the dynamic change trend of electromyographic signals of the target muscle group after stimulation response, as well as the difference information between the output and the expected motor intention.

[0051] The collected feedback data is input into the neuromuscular electrical stimulation parameter optimization model. Through a dynamic adaptive adjustment mechanism, the feature weights, normalization adjustment factors, and feedback gain coefficients in the model are updated in real time, thereby achieving online optimization and adaptive learning of the model parameters.

[0052] Through closed-loop feedback optimization, the system can continuously optimize the output of electrical stimulation parameters according to changes in the user's current state, improve the accuracy, stability and individualized adaptability of stimulation control, further enhance the effect of rehabilitation training, reduce the risk of stimulation side effects, and promote the intelligence and efficiency of the neuromuscular function recovery process.

[0053] Example 2, one embodiment of the present invention, provides a neuromuscular electrical stimulation rehabilitation control method based on EEG intention recognition. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0054] First, an experiment was conducted based on a "closed-loop rehabilitation system for neuromuscular electrical stimulation driven by EEG intention recognition" to verify the effectiveness and optimization of the invented method in rehabilitation training. Six patients with upper limb motor dysfunction, aged 35 to 55 years, were selected as experimental subjects. All were in the rehabilitation training stage and possessed the ability to express voluntary motor intentions. The experimental equipment included a 16-channel EEG acquisition system (sampling rate of 512 Hz), neuromuscular electrical stimulation devices, a surface electromyography (EMG) acquisition system, and a customized rehabilitation training platform.

[0055] During the experimental preparation phase, each subject was first fitted with an EEG acquisition cap, using a 16-channel layout covering the prefrontal, central, and parietal lobes, with a focus on acquiring EEG signals related to upper limb motor intentions. During EEG acquisition, real-time bandpass filtering (0.5–45 Hz) was performed, and independent component analysis (ICA) was used to remove artifacts from electrooculography (EOG) and electromyography (EMG). Further adaptive filtering in the transform domain was then applied to suppress residual noise, resulting in a purified EEG signal stream.

[0056] Subsequently, the purified EEG signals were sliced ​​using a sliding window mechanism (window width 1 second, step size 0.2 seconds) and input into the feature extraction module. In the feature extraction stage, wavelet packet decomposition and power spectral density analysis were used to extract EEG feature vectors, which were then input into a trained EEG intention discrimination model based on nested kernel mapping and temporal dynamic integration. The model outputs four categories of motor intentions (clenched fist, flexed elbow, raised arm, and rest) and their corresponding confidence scores in real time, serving as input to the neuromuscular electrical stimulation parameter optimization model.

[0057] During the electrical stimulation optimization phase, the model integrates the intent category, confidence level, and current electromyographic state to dynamically generate the electrical stimulation current amplitude A(t), frequency f(t), and pulse width d(t), which then drives the neuromuscular electrical stimulation device in real time via an interface. The target muscle groups for electrical stimulation are the flexor muscles of the affected forearm and the biceps brachii muscle group. The stimulation frequency range is 10-100 Hz, the current amplitude is 5-100 Hz, the current amplitude is 580 mA, and the pulse width is 50-400 μs.

[0058] Simultaneously, electromyographic response data of the target muscle group is collected in real time and input into the optimization model for online adjustment, achieving closed-loop adaptive control. The experiment was designed for continuous training for 30 minutes, during which subjects were guided to actively attempt three types of movement intentions multiple times. The system responded in real time, adjusting the stimulation parameters to dynamically drive muscle movement.

[0059] Table 2 Experimental Data

[0060]

[0061] As shown in Table 2, the "Neuromuscular Electrical Stimulation Rehabilitation Control Method Based on EEG Intent Recognition" described in this invention demonstrated significant advantages in its application to six experimental subjects. Firstly, the accuracy rate of intent category discrimination was generally above 88%, reaching a maximum of 93%, indicating that the EEG intent discrimination model possesses excellent dynamic recognition capabilities. Compared to traditional methods based on linear discrimination or fixed thresholds, it significantly improves the accuracy and real-time performance of intent recognition, ensuring that the electrical stimulation control process can more accurately match the user's intent.

[0062] Secondly, the mean confidence score remained stable above 0.80, demonstrating the model's good adaptability to different subjects in actual rehabilitation training. The confidence score output provided a stable basis for subsequent optimization of stimulation parameters, supporting real-time dynamic parameter adjustment. Regarding electrical stimulation parameters, the optimized model dynamically adjusted the current amplitude, frequency, and pulse width according to the individual subject's condition. Throughout the training process, the electrical stimulation parameters remained stably distributed within a safe and effective range, without any abnormally high or low values, ensuring the individualized adaptability and safety of the electrical stimulation.

[0063] More importantly, the average improvement rate of electromyographic response was approximately 34%, compared to traditional methods with fixed stimulation parameters (which typically show an improvement rate in the 20%–25% range), demonstrating the significant advantage of the closed-loop optimization mechanism in dynamically matching the user's physiological state. Subjective comfort scores were generally above 4 points, further validating the method's good adaptability in terms of user experience, which helps improve adherence and active participation in rehabilitation training.

[0064] In summary, the results of the embodiments fully demonstrate that the present invention is superior to the prior art in terms of EEG intention discrimination accuracy, neuromuscular electrical stimulation parameter optimization capability, closed-loop feedback effect, and user experience. It reflects the innovation and novelty of the technical solution in dynamic adaptive control and individualized rehabilitation training, and has good potential for clinical application and promotion.

[0065] Example 3, referring to Figure 3As an embodiment of the present invention, a neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition is provided, including a real-time EEG signal acquisition and preprocessing module, a motor intention feature extraction and dynamic discrimination module, and a neuromuscular electrical stimulation parameter optimization and execution control module.

[0066] The module for real-time acquisition and preprocessing of EEG signals is used to acquire EEG signal data of users during rehabilitation training in real time through a multi-channel EEG acquisition device. It performs bandpass filtering, artifact removal, and adaptive noise suppression on the raw EEG signals and outputs a purified EEG signal data stream. The module for motor intention feature extraction and dynamic discrimination is used to extract features from the purified EEG signal data stream, generate a motor intention feature vector, and dynamically identify the user's current motor intention category based on the EEG intention discrimination model. The module for neuromuscular electrical stimulation parameter optimization and execution control is used to dynamically generate a set of stimulation current amplitude, frequency, and pulse width parameters based on the real-time identified motor intention signal, intention confidence, and current electromyographic state, using a neuromuscular electrical stimulation parameter optimization model, and control the neuromuscular electrical stimulation device to drive the target muscle group to perform electrical stimulation.

[0067] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be 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 methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0069] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0070] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition, characterized in that: It includes a module for real-time acquisition and preprocessing of electroencephalogram (EEG) signals, a module for extracting and dynamically judging motor intention features, and a module for optimizing and controlling neuromuscular electrical stimulation parameters. The real-time acquisition and preprocessing module for EEG signals is used to acquire EEG signal data of users during rehabilitation training in real time through a multi-channel EEG acquisition device, perform bandpass filtering, artifact removal and adaptive noise suppression on the raw EEG signals, and output the purified EEG signal data stream. The motion intention feature extraction and dynamic discrimination module is used to extract features from the purified EEG signal data stream, generate motion intention feature vectors, and dynamically identify the user's current motion intention category based on the EEG intention discrimination model, and output the intention signal and corresponding confidence level in real time. The purified EEG signal data was processed using time-frequency domain analysis methods for feature extraction, specifically wavelet packet decomposition and power spectral density methods, to extract multidimensional EEG intention feature vectors. Based on the extracted EEG features, an EEG intention discrimination model is constructed; Discrimination score Represented as: in, For a moment For category The calculated motion intent discrimination score, For the dimension of EEG feature vectors, For the first The coefficients of each characteristic, For a moment The extracted first Each EEG characteristic value is expressed in energy density. For the first The dynamic adjustment function corresponding to each feature For normalized adjustment parameters, For the first The inhibitory factor for each feature is a positive real number. Indicates at time , No. The EEG characteristic values ​​corresponding to each channel; The threshold should be determined based on a large number of EEG training samples to identify scores. The distribution is used to determine the mean score based on the positive examples in the training set; and standard deviation Set as: in It is a regulatory factor, often taken as This ensures that the judgment has a certain tolerance and avoids overfitting; Output After: If Then it is classified as a category. The system can be configured according to each time period. Calculate separately And compare their respective thresholds Finally, the category that meets the conditions and has the highest score is selected as the judgment result; During the rehabilitation training process, the system will use the current moment... Collected EEG feature vectors The input is used to construct the EEG intention discrimination model, and the values ​​of each category are dynamically calculated. Discrimination score The discrimination rule is set as follows: if If so, it is currently classified as a motion intention category. If multiple categories meet the criteria, then select the one that best meets the criteria. The largest category is used as the final judgment result; The final output motion intention category signal The corresponding confidence information will serve as an important input for the generation and optimization of neuromuscular electrical stimulation parameters, driving the subsequent rehabilitation training control process; The neuromuscular electrical stimulation parameter optimization and execution control module is used to dynamically generate a set of stimulation current amplitude, frequency and pulse width parameters based on the real-time identified motor intention signal, intention confidence and current electromyographic state, using a neuromuscular electrical stimulation parameter optimization model, and control the neuromuscular electrical stimulation device to drive the target muscle group to perform electrical stimulation.

2. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that, include: The system uses a multi-channel EEG acquisition device to collect EEG signal data from users in real time during rehabilitation training. The EEG signal data is then processed with bandpass filtering, artifact removal, and adaptive noise suppression to output a purified EEG signal data stream. Feature extraction is performed on the purified EEG signal data stream to output a motion intention feature vector. An EEG intention discrimination model is constructed to dynamically identify the user's current motion intention category and output the corresponding intention signal and confidence level. Based on the identified intent signal, confidence level, and current electromyographic state, a neuromuscular electrical stimulation parameter set is output using a neuromuscular electrical stimulation parameter optimization model and sent to the neuromuscular electrical stimulation device to control the target muscle group for electrical stimulation.

3. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The bandpass filter includes setting the passband range to 0.5 Hz to 45 Hz.

4. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The feature extraction includes extracting motion intention feature vectors from the purified EEG signal data stream using a combination of wavelet packet decomposition and power spectral density analysis.

5. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The EEG intention discrimination model includes processing the motion intention feature vector with kernel mapping, constructing a multi-dimensional feature mapping space by combining dynamic temporal features and nonlinear activation functions, and outputting the corresponding intention signal and confidence level.

6. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The output corresponding intention signal and confidence level include intention category discrimination based on the confidence level and preset category threshold, and the final motion intention category is determined by comparing the discrimination score of each motion intention category with the corresponding threshold.

7. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The neuromuscular electrical stimulation parameter optimization model integrates motor intention category, intention confidence, and real-time electromyographic state information. It employs an optimization algorithm that includes nonlinear mapping, integral feedback, and normalization adjustment mechanisms to output the amplitude, frequency, and pulse width of the electrical stimulation current.

8. The neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in claim 1, characterized in that: The output neuromuscular electrical stimulation parameter set includes real-time acquisition of electromyographic response feedback data during neuromuscular electrical stimulation, and the feature weights, normalization adjustment factors, and feedback gain coefficients in the neuromuscular electrical stimulation parameter optimization model are updated based on the feedback data using a dynamic adaptive adjustment mechanism.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in any one of claims 2 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the neuromuscular electrical stimulation rehabilitation control system based on EEG intention recognition as described in any one of claims 2 to 8.

Images