Rehabilitation training method and system based on mixed normal form brain-computer interface

Abstract

The invention relates to the technical field of brain-computer interfaces, and provides a rehabilitation training method and system based on a mixed normal form brain-computer interface. The method comprises the steps that electroencephalogram signals of the prefrontal lobe or the motor cortex of a patient are collected; performing cooperative decoding on the electroencephalogram signals according to a preset mixed normal form decoding model, and identifying signal features; if the electroencephalogram signal is a prefrontal lobe electroencephalogram signal, the signal feature is whether to imagine an intention intensity score of walking and movement, and if the electroencephalogram signal is a motor cortex electroencephalogram signal, the signal feature is a motor imagery type and an intention intensity score of movement; generating a lower limb rehabilitation training instruction matched with the rehabilitation stage of the patient according to the signal characteristics, wherein the training instruction at least comprises any one of walking, knee bending, leg lifting, left leg lifting and right leg lifting; the training instruction is transmitted to an execution module of the rehabilitation robot, and the rehabilitation robot is driven to complete corresponding training actions.

Description

Technical Field

[0001] This application relates to the field of brain-computer interface technology, and in particular to a rehabilitation training method and system based on a hybrid paradigm brain-computer interface. Background Technology

[0002] Existing technologies combining brain-computer interfaces with lower limb rehabilitation robots generally suffer from a bottleneck of "single rehabilitation training paradigms." Most systems employ only a single motor imagery (MI) paradigm or attention paradigm, failing to simultaneously address the dual needs of capturing patients' active motor intentions and assessing training focus. Some solutions claiming "multi-paradigm" approaches can only select a fixed paradigm before training, unable to dynamically adjust paradigm fusion strategies based on the patient's real-time state. This results in a lack of personalized adaptation in rehabilitation training, hindering effective patient motivation and particularly failing to meet the differentiated needs of stroke hemiplegic patients at different rehabilitation stages regarding training intensity and type. For example, traditional single-paradigm approaches cannot simultaneously quantify the intensity of motor intentions and attentional concentration, leading to a rigid training instruction generation mechanism that fails to accurately reflect the patient's true rehabilitation progress.

[0003] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0004] This application provides a rehabilitation training method and system based on a hybrid paradigm brain-computer interface, which aims to solve the following problems.

[0005] In a first aspect, embodiments of this application provide a rehabilitation training method based on a hybrid paradigm brain-computer interface, the method comprising: Collect electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex; The EEG signal is collaboratively decoded according to a preset hybrid paradigm decoding model to identify signal features; if the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined; if the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. Based on the signal characteristics, lower limb rehabilitation training instructions are generated to suit the patient's rehabilitation stage. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

[0006] In some embodiments, before performing collaborative decoding of the EEG signal according to a preset hybrid paradigm decoding model, the method further includes: validating the acquired EEG signal and removing invalid signals.

[0007] In some embodiments, the step of validating the acquired EEG signals and removing invalid signals includes: determining whether the electrodes of the EEG acquisition device have detached through a preset detachment detection mechanism; determining whether the impedance value of the EEG signal is within the effective range through a preset impedance detection mechanism; and determining and removing EEG signals that have detected electrode detachment or impedance values ​​exceeding the effective range as invalid signals.

[0008] In some embodiments, the step of collaboratively decoding the EEG signal according to a preset hybrid paradigm decoding model to identify signal features includes: if the EEG signal is a prefrontal EEG signal, determining whether the patient has generated a walking motion image through the hybrid paradigm decoding model, and outputting the corresponding motor intention intensity score; if the EEG signal is a motor cortex EEG signal, identifying the specific type of movement imagined by the patient through the hybrid paradigm decoding model, wherein the movement type includes at least one of raising the left leg and raising the right leg, and outputting the corresponding motor intention intensity score.

[0009] In some embodiments, generating lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal features includes: when it is determined that the patient has generated a motor imagery that meets preset requirements, determining gait parameters corresponding to the motor imagery according to the motor intention intensity score and a preset mapping relationship, wherein the gait parameters include at least one of stride length, stride frequency, and number of steps; generating training instructions that include the type of motor imagery and gait parameters; and generating a stationary instruction when it is not determined that the patient has generated a motor imagery that meets preset requirements.

[0010] In some embodiments, transmitting training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete corresponding training actions includes: transmitting the type of motor imagery and gait parameters contained in the training instructions to the execution module of the rehabilitation robot; the execution module controls the rehabilitation robot to perform walking, knee bending, leg raising, left leg raising or right leg raising actions according to the type of motor imagery and gait parameters, and adjusts the amplitude, frequency and number of actions according to the gait parameters.

[0011] In some embodiments, the acquisition of EEG signals from the patient's prefrontal cortex or motor cortex includes: acquiring EEG signals at preset points in the prefrontal cortex or motor cortex using a fully silicone flexible EEG acquisition device, utilizing a double-ear clip anti-interference structure and magnetically attached electrodes, wherein the preset points are 6-point simplified acquisition points.

[0012] In some embodiments, the method further includes: displaying the exercise intention intensity score, the action type corresponding to the training instruction, and the training completion status through a preset feedback interface, so as to achieve visual monitoring of the rehabilitation training process.

[0013] In some embodiments, after driving the rehabilitation robot to complete the corresponding training action, the method further includes: obtaining the corresponding motion intention quantification result and training completion feedback, and dynamically adjusting the fusion ratio of motion imagery and attention and the training difficulty information corresponding to the hybrid paradigm decoding model based on the motion intention quantification result and training completion feedback, thereby forming a personalized rehabilitation training closed loop.

[0014] Secondly, this application provides a rehabilitation training system based on a hybrid paradigm brain-computer interface, the system comprising: The signal acquisition unit is used to acquire electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex. The collaborative decoding unit is used to collaboratively decode the EEG signal according to a preset hybrid paradigm decoding model and identify signal features. If the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined. If the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. The training completion unit is used to generate lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal characteristics. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

[0015] This application utilizes a hybrid paradigm of motor imagery and attention for collaborative decoding, balancing active motor intention capture (MI paradigm) with training focus assessment (attention paradigm), thus addressing the limitation of single-paradigm functionality in existing technologies. Based on real-time EEG signal decoding results, the fusion ratio and training difficulty of the hybrid paradigm are dynamically adjusted, forming a closed-loop process of "acquisition-decoding-execution-feedback," adapting to the needs of patients at different rehabilitation stages and enhancing training targeting. Quantitative indicators such as motor intention intensity scores accurately reflect the patient's training status, and a real-time feedback mechanism motivates active patient participation, enhancing the effectiveness of rehabilitation training. Support for multi-site EEG acquisition, including the prefrontal cortex (simplified acquisition) and motor cortex (precise motion recognition), adapts to the signal acquisition needs of different rehabilitation stages, expanding the system's application scenarios.

[0016] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

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

[0018] Figure 1 This is a schematic flowchart illustrating the steps of a rehabilitation training method based on a hybrid paradigm brain-computer interface provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the lower limb rehabilitation robot provided in the embodiments of this application; Figure 4 This is a schematic diagram of a general hybrid paradigm provided in the embodiments of this application; Figure 5 This is a schematic diagram of a hybrid paradigm based on a prefrontal cortex brain-computer interface provided in an embodiment of this application; Figure 6 This is a schematic diagram of a hybrid paradigm based on a whole-brain brain-computer interface provided in an embodiment of this application; Figure 7 This is a schematic block diagram of a rehabilitation training system based on a hybrid paradigm brain-computer interface provided in one embodiment of this application; Figure 8 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0022] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0023] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] Existing technologies combining brain-computer interfaces with lower limb rehabilitation robots generally suffer from a bottleneck of "single rehabilitation training paradigms." Most systems employ only a single motor imagery (MI) paradigm or attention paradigm, failing to simultaneously address the dual needs of capturing patients' active motor intentions and assessing training focus. Some solutions claiming "multi-paradigm" approaches can only select a fixed paradigm before training, unable to dynamically adjust paradigm fusion strategies based on the patient's real-time state. This results in a lack of personalized adaptation in rehabilitation training, hindering effective patient motivation and particularly failing to meet the differentiated needs of stroke hemiplegic patients at different rehabilitation stages regarding training intensity and type. For example, traditional single-paradigm approaches cannot simultaneously quantify the intensity of motor intentions and attentional concentration, leading to a rigid training instruction generation mechanism that fails to accurately reflect the patient's true rehabilitation progress.

[0026] Therefore, a method is urgently needed to solve at least one of the above problems.

[0027] To solve the above problem, please refer to Figure 1 This application provides a rehabilitation training method based on a hybrid paradigm brain-computer interface, applied to computer equipment. The computer equipment can be deployed on a single server or a server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc. It should be noted that all information involved in the method provided in this application is extracted with the authorization of the relevant user and in accordance with relevant regulations, ensuring user data security.

[0028] The provided rehabilitation training method based on a hybrid paradigm brain-computer interface includes steps S101 to S103. Details are as follows: Step S101. Collect EEG signals from the patient's prefrontal cortex or motor cortex.

[0029] Specifically, this step provides a highly reliable, high signal-to-noise ratio data source for the entire rehabilitation training method. The core objective is to complete the acquisition of EEG signals in a compliant, low-interference manner that is suitable for patients at different rehabilitation stages. At the same time, invalid signals are eliminated through a dual-verification mechanism to ensure the accuracy of subsequent decoding from the source. It is compatible with both minimally invasive prefrontal cortex acquisition and whole-brain motor cortex acquisition modes, covering the full-cycle rehabilitation needs of stroke hemiplegic patients from early bed rest to later functional reconstruction.

[0030] Adopting such Figure 2 The fully silicone flexible EEG acquisition device shown features a dual-ear clip anti-interference design, magnetic female electrode, and a minimalist 6-point acquisition layout. It is also compatible with medical-grade whole-brain multi-channel EEG acquisition devices. The device has a sampling rate of no less than 250Hz, a signal bandwidth covering 0.5-30Hz (core frequency band of EEG characteristics), and built-in power frequency filtering and baseline drift suppression basic anti-interference modules to meet the signal accuracy requirements of medical rehabilitation scenarios.

[0031] The prefrontal lobe acquisition mode is designed for stroke patients in the early stages of hemiplegia, severe hemiplegia, and those who are bedridden and unable to cooperate with complex motor imagery. It adopts a minimalist 6-point prefrontal lobe electrode layout, which only collects prefrontal lobe EEG signals, greatly reducing the difficulty of wearing the device and the threshold for patient cooperation, making it suitable for early rehabilitation scenarios. The motor cortex acquisition mode is designed for stroke patients in the middle and late stages of hemiplegia who have some ability to coordinate motor imagery. It adopts the international 10-20 system standard electrode layout, focusing on covering core sites such as C3, C4, and Cz in the motor cortex of the brain, accurately capturing the neural electrical signal characteristics related to motor imagery, and supporting the recognition and decoding of multiple movement types.

[0032] During the acquisition process, dual-dimensional real-time verification is performed simultaneously to ensure that only valid signals enter the subsequent process: Electrode detachment detection: Real-time monitoring of the contact status between the electrode and the skin. When electrode detachment is detected, an immediate prompt is triggered, pausing the acquisition process until the electrode returns to normal contact; Contact impedance verification: Continuous monitoring of electrode-skin contact impedance. The impedance value is required to be controlled below 5kΩ. Signals exceeding the threshold will be marked and noise-removed, retaining only valid EEG signals that meet the impedance requirements.

[0033] All EEG signal acquisitions were conducted with prior written authorization from the patient and relevant guardians, strictly adhering to relevant regulations on medical data privacy protection. The acquired signals were used solely for real-time decoding and assessment during this rehabilitation training and were not stored or transmitted for any unauthorized purposes.

[0034] Step S102. Perform collaborative decoding on the EEG signal according to the preset hybrid paradigm decoding model to identify signal features; if the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not to imagine walking and movement; if the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of movement intention.

[0035] Specifically, this step is the core technical link of the solution. The core goal is to break through the technical bottleneck of a single paradigm. Through a collaborative decoding model that dynamically integrates motor imagery and attention, it can simultaneously complete the recognition of the patient's active motor intention and the assessment of training focus, and output standardized signal features. For the two types of input signals from the prefrontal cortex and the motor cortex, it outputs corresponding feature dimensions in a differentiated manner, providing accurate and real-time core basis for the subsequent generation of personalized instructions.

[0036] The hybrid paradigm decoding model adopts a decoding framework that dynamically integrates the motor imagery (MI) paradigm and the attention paradigm. Unlike existing solutions that fix a single paradigm or select a fixed paradigm before training, this model can dynamically adjust the feature weights and fusion ratio of the two paradigms based on the patient's real-time EEG characteristics. The motor imagery paradigm is responsible for capturing active motor intentions and identifying action-related neural features, while the attention paradigm is responsible for assessing the patient's level of focus on the training task. Through dual-paradigm collaborative verification, the misjudgment rate is significantly reduced and the decoding reliability is improved.

[0037] The model was pre-trained and validated using a full-cycle rehabilitation EEG dataset of stroke hemiplegic patients. It was specifically optimized for patients with different rehabilitation stages and different degrees of hemiplegia, and has good generalization ability. At the same time, it supports online incremental learning and can be personalized fine-tuned based on the historical training data of individual patients to continuously improve individual adaptability.

[0038] The effective EEG signal output by S101 is subjected to standardized preprocessing, including 50Hz power frequency notch filtering, 0.5-30Hz bandpass filtering, removal of EEG / EMG artifacts, baseline correction, and 2s sliding window (0.5s step) segmentation processing. The preprocessed standardized signal is then input into the decoding model.

[0039] For the prefrontal cortex input signal, the decoding output includes two core signal features: a binary classification result indicating whether walking was imagined, and a motion intention intensity score of 0-100. The specific process is as follows: Dual-paradigm feature extraction is achieved by simultaneously extracting prefrontal P300 event-related potential features and continuous visual attention-related EEG rhythm features corresponding to the attention paradigm, as well as prefrontal motor preparation potential (BP) features corresponding to the motor imagery paradigm, thus completing the simultaneous extraction of dual-dimensional features. The dynamic fusion and classification judgment dynamically adjusts the fusion ratio of the two paradigm features based on the patient's real-time attention feature weights, and outputs the judgment result of "imagine walking" or "not imagining walking" through a pre-trained binary classifier; when the patient's attention concentration is lower than the preset threshold, the weight of the motor imagination feature is automatically reduced to avoid invalid misjudgment. The intensity of motor intention quantification targets valid signals identified as "imaginary walking" and outputs a continuous intensity score of 0-100 based on dual-paradigm fusion features. The score is positively correlated with the intensity of the patient's active movement intention and training focus, and the quantitative assessment of training focus is completed simultaneously.

[0040] For the input signal from the motor cortex, two core signal features are decoded and output: a multi-classification result of the motor imagery type and a motor intention intensity score of 0-100. The specific process is as follows: Simultaneously extract motor cortical ERD / ERS features (mu / beta frequency band rhythm changes) corresponding to motor imagery paradigms to identify specific neural features of different motor imagery, and extract whole-brain attention network-related EEG features corresponding to attention paradigms to quantify patients' training focus. The dual-paradigm feature fusion ratio is dynamically adjusted based on real-time attention and focus. The motor imagery type is identified through a multi-classifier. The identified types include at least preset rehabilitation movements such as raising the left leg, raising the right leg, bending the knee, walking on flat ground, and raising the leg. At the same time, the validity is verified. Only when the identified imagery type is consistent with the specified movement of the current training task is it judged as "valid imagery"; otherwise, it is judged as "invalid imagery". Motor intention intensity quantification: For signals judged as "effective imagination", a motor intention intensity score of 0-100 is output based on dual-paradigm fusion features. The score is simultaneously associated with the neural activation level of motor imagination and training focus, realizing the synchronous quantification of motor intention and focus.

[0041] The model's entire decoding latency is controlled within 300ms to ensure the real-time nature of the decoding results and meet the closed-loop control requirements of rehabilitation training. If three consecutive sliding windows are judged as invalid imaginations, a focus prompt is triggered to guide the patient to concentrate on completing the training.

[0042] Step S103. Generate lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal characteristics. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. Transmit the training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

[0043] Specifically, this step is the implementation of the decoding results. The core objective is to generate personalized lower limb rehabilitation training instructions that are suitable for the patient's rehabilitation stage and are safe and controllable, based on the signal characteristics output by S102, and to complete the accurate mapping of "brainwave active intention - robot execution action". At the same time, a real-time feedback closed loop is constructed to dynamically adjust the training strategy, continuously encourage the patient to actively participate, and realize personalized rehabilitation training throughout the entire cycle.

[0044] The rehabilitation phase is pre-set and basic safety parameters are configured in advance based on the patient's clinical rehabilitation assessment results, dividing the rehabilitation into three phases: early bedridden period, mid-term recovery period, and late-term functional reconstruction period. Medical staff pre-set the training movement library, gait / movement parameter thresholds, and training difficulty levels for the corresponding phases. The core rules are as follows: Early bedridden period: Only simple single-joint movements such as knee flexion and leg raising are allowed, and the maximum values ​​of stride length, stride frequency and joint range of motion are strictly limited to meet the safe training needs of bedridden patients; Mid-term recovery period: Encourage compound movements such as walking on flat ground and alternating steps with both legs, gradually widen the parameter range, and focus on improving the patient's active motor control ability; Later functional reconstruction period: Introduce complex walking training with varying pace and stride to increase the difficulty of training and focus on the reconstruction of the patient's walking function.

[0045] For the "Whether to imagine walking + intensity score of movement intention" output by S102, execute the following instructions to generate rules: Invalid command generation: When the judgment result is "not imagined walking", or the motor intention intensity score is ≤60 points, a stationary command is directly generated and sent to the rehabilitation robot execution module. The robot remains stationary and does not perform any walking movements to avoid accidental triggering. Effective instruction generation: When the judgment result is "imaginary walking" and the motor intention intensity score is >60, training instructions for walking on flat ground are generated based on the score, supporting two parameter mapping modes pre-selected by medical staff: Linear mapping mode: Within the score range of 60-100, the stride length is adjusted to 60%~100% of the patient's maximum suitable stride length, and the step frequency is adjusted to 15~25 steps / minute. The higher the score, the larger the stride length and the faster the step frequency. At the same time, it matches the preset number of execution steps to generate a complete walking instruction. Gear mapping mode: Parameters are locked according to the grade level of the score. 60-70 points correspond to 100% stride length and 15 steps / minute; 70-80 points correspond to 100% stride length and 20 steps / minute; 80-100 points correspond to 100% stride length and 25 steps / minute, generating walking commands for the corresponding gear level.

[0046] For the "Motion Imagery Type + Motion Intent Intensity Score" output by S102, execute the following instructions to generate rules: Invalid command generation: When the judgment result is "invalid imagination" or the motion intention intensity score is ≤60, a stationary command is generated directly, and the robot maintains a stationary state; Effective instruction generation: When the judgment result is "effective imagination" and the motor intention intensity score is >60, first match the motor instruction that is completely consistent with the identified motor imagination type (raise left leg, raise right leg, bend knee, raise leg, walk on flat ground, etc.), and then adjust the execution parameters of the corresponding action (leg height, knee bend angle, stride / step frequency, action holding time, etc.) based on the motor intention intensity score. The higher the score, the higher the execution range and speed of the action, and all parameters do not exceed the safety threshold of the patient's current rehabilitation stage.

[0047] The generated training instructions are transmitted in real time to the execution module of the lower limb rehabilitation robot via low-latency communication protocols (Bluetooth 5.0, Ethernet, CAN bus), with the transmission latency controlled within 100ms to ensure the synchronization of the action execution with the patient's intention. After receiving the instructions, the robot execution module drives the servo motors, exoskeleton mechanism and other execution units to complete the corresponding training actions according to the instruction parameters. At the same time, it collects execution status data in real time, and immediately triggers an emergency stop when abnormalities such as excessive torque or jamming occur to ensure patient safety.

[0048] After the movement is completed, the quantitative results of the movement intention, the training completion status, and the movement execution parameters are fed back to the patient's training interface and the medical staff's management terminal in real time, allowing the patient to intuitively see the training effect corresponding to the active intention and encouraging the willingness to participate actively. The system records the entire process data of this training simultaneously, and dynamically adjusts the fusion ratio of the hybrid paradigm, the training difficulty, and the parameter mapping rules based on the training results to form a personalized rehabilitation training closed loop, continuously adapting to the training needs of patients at different rehabilitation stages and improving the rehabilitation effect of lower limb motor function.

[0049] In some embodiments, before performing collaborative decoding of the EEG signal according to a preset hybrid paradigm decoding model, the method further includes: validating the acquired EEG signal and removing invalid signals.

[0050] This embodiment is a preprocessing step between S101 signal acquisition and S102 signal decoding. The core is to complete the validity verification and invalid signal removal before the EEG signal is input into the hybrid paradigm decoding model. This solves the technical problems of contact abnormalities, noise pollution, invalid data leading to decreased decoding accuracy, and robot motion mis-triggering in the original acquired signal. It ensures the input quality of the subsequent decoding step from the data source, reduces the interference of invalid data on the decoding results, and improves the operational stability and reliability of the entire system.

[0051] After completing the real-time acquisition of the patient's EEG signal and before inputting the signal into the preset hybrid paradigm decoding model, the validity verification process is started simultaneously. The verification process is synchronized with the signal acquisition, and a segmented mode with a 2-second sliding window and a 0.5-second step size is adopted to perform validity verification frame by frame for each segment of the acquired EEG signal.

[0052] Based on preset verification rules, the validity of each EEG signal is determined by a binary judgment. Signal segments that meet the invalid signal determination rules are directly removed from the dataset to be decoded and do not enter the subsequent decoding process; only signal segments that are determined to be valid are retained, and after standardization processing, they are input into the hybrid paradigm decoding model.

[0053] When three or more consecutive sliding window signals are determined to be invalid, the system immediately triggers a device wearing abnormality prompt through the terminal interface, reminding the patient or medical staff to check the wearing status of the EEG acquisition device and the electrode contact status, so as to troubleshoot the fault in time and ensure the continuity of the acquisition process.

[0054] In some embodiments, the step of validating the acquired EEG signals and removing invalid signals includes: determining whether the electrodes of the EEG acquisition device have detached through a preset detachment detection mechanism; determining whether the impedance value of the EEG signal is within the effective range through a preset impedance detection mechanism; and determining and removing EEG signals that have detected electrode detachment or impedance values ​​exceeding the effective range as invalid signals.

[0055] This embodiment is a concrete implementation scheme for validity verification. The core adopts a dual-dimensional joint verification mechanism of electrode detachment detection and contact impedance detection to solve the problems of missed detection, false judgment and insufficient detection accuracy of single verification methods. It accurately identifies signal failure scenarios from the hardware contact level, realizes accurate judgment and elimination of invalid signals, and at the same time takes into account the real-time performance and rigor of detection, which is suitable for the high reliability requirements of medical rehabilitation scenarios.

[0056] Electrode detachment detection is implemented using a pre-set detachment detection mechanism and a real-time conductivity pulse detection method. The main control unit of the EEG acquisition device sends a low-amplitude conductivity detection pulse to each acquisition electrode at a fixed frequency of 10Hz. By detecting the on / off state of the echo signal of the detection pulse, the physical contact state between the electrode and the patient's skin is determined. If no effective echo signal is returned from a certain electrode, it is directly determined that the electrode has detached.

[0057] The contact impedance detection is implemented using a constant current source excitation method through a preset impedance detection mechanism. During the signal acquisition process, the electrode-skin contact impedance of each electrode is continuously monitored. The preset effective impedance range is 0-5kΩ. If the real-time impedance value of a certain electrode exceeds the effective range of 5kΩ, the signal acquired by that electrode is determined to be an impedance abnormality signal.

[0058] The invalid signal determination and removal rules determine that if any core acquisition electrode is detected to be detached, or if the impedance values ​​of more than half of the acquisition electrodes are outside the valid range, the entire EEG signal acquired in the corresponding time period is directly determined as an invalid signal and completely removed from the dataset to be decoded. If only a single non-core electrode shows an impedance abnormality, the abnormal signal of that electrode is marked, and data completion is performed by interpolation of adjacent electrode signals. It is not directly determined as invalid for the entire segment, thus balancing signal integrity and verification rigor.

[0059] In some embodiments, the step of collaboratively decoding the EEG signal according to a preset hybrid paradigm decoding model to identify signal features includes: if the EEG signal is a prefrontal EEG signal, determining whether the patient has generated a walking motion image through the hybrid paradigm decoding model, and outputting the corresponding motor intention intensity score; if the EEG signal is a motor cortex EEG signal, identifying the specific type of movement imagined by the patient through the hybrid paradigm decoding model, wherein the movement type includes at least one of raising the left leg and raising the right leg, and outputting the corresponding motor intention intensity score.

[0060] This embodiment is a core refinement of step S102. The core is to set differentiated decoding targets and output dimensions for EEG signals from two different acquisition sources: the prefrontal cortex and the motor cortex. Through dynamic collaborative decoding of the motor imagery + attention hybrid paradigm, it can simultaneously complete the recognition of the patient's motor intention and the quantification of the intensity of the motor intention. This solves the pain point that a single paradigm cannot simultaneously take into account the capture of active motor intention and the assessment of training focus. At the same time, it adapts to the differentiated decoding needs of stroke hemiplegic patients at different rehabilitation stages.

[0061] The decoding of prefrontal EEG signals is implemented by simultaneously extracting two-dimensional features when the input EEG signal to the decoding model is a signal collected from the prefrontal cortex. This involves extracting the prefrontal preparatory potential features corresponding to the motor imagery paradigm, as well as the P300 event-related potentials and continuous visual attention-related EEG rhythm features corresponding to the attention paradigm. The fusion weights of the two features are dynamically adjusted according to the patient's real-time attention state. A pre-trained binary classifier is used to make a binary judgment on whether "walking motor imagery" is generated, and a continuous motor intention intensity score of 0-100 is output. This score simultaneously integrates the neural activation level of motor imagery and the patient's attention level during training, achieving simultaneous output of intention recognition and attention measurement.

[0062] The decoding of motor cortex EEG signals is implemented by simultaneously extracting ERD / ERS features of the motor cortex mu / beta bands corresponding to the motor imagery paradigm and EEG features of the whole-brain attention network corresponding to the attention paradigm when the input EEG signal to the decoding model is a signal collected from the motor cortex. The hybrid paradigm decoding model dynamically adjusts the fusion ratio of the dual paradigm features. The recognition of motor imagery types is completed through a pre-trained multi-classifier. The recognizable motor imagery types include at least raising the left leg, raising the right leg, bending the knee, walking on flat ground, and raising the leg. At the same time, the recognition results are matched and verified with the specified actions of the current training task. Only when the recognized imagery type matches the specified action is it judged as a valid motor imagery. The motor intention intensity score of 0-100 is output simultaneously. The score is positively correlated with the neural activation degree of the motor imagery and the training focus.

[0063] In some embodiments, generating lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal features includes: when it is determined that the patient has generated a motor imagery that meets preset requirements, determining gait parameters corresponding to the motor imagery according to the motor intention intensity score and a preset mapping relationship, wherein the gait parameters include at least one of stride length, stride frequency, and number of steps; generating training instructions that include the type of motor imagery and gait parameters; and generating a stationary instruction when it is not determined that the patient has generated a motor imagery that meets preset requirements.

[0064] This embodiment is the core implementation scheme of step S103. The core is to generate differentiated rehabilitation training instructions based on the signal characteristics output by S102, establish a standardized mapping relationship between the intensity of movement intention and robot motion parameters, solve the problems of rigid training instruction generation, inability to accurately match the patient's real-time movement intention, and insufficient safety redundancy in traditional solutions, and realize personalized, safe and accurate generation of training instructions to adapt to the training needs of patients at different rehabilitation stages.

[0065] The two-dimensional criteria for determining effective motor imagery are: first, the decoding model determines that the patient has generated motor imagery that is completely matched with the current training task (i.e., walking in the prefrontal cortex mode and the designated action imagery consistent with the training task in the motor cortex mode); second, the corresponding motor intention intensity score is >60 points. Only when both conditions are met simultaneously is it determined to be motor imagery that meets the preset requirements.

[0066] When a motor imagery is determined to meet preset requirements, gait parameters are determined based on the real-time output score of the motor intention intensity, according to a pre-selected mapping relationship (linear mapping or step-by-step mapping) by medical staff. Gait parameters include at least stride length, cadence, and number of steps: In the linear mapping mode, within a score range of 60-100, stride length is set to 60%-100% of the patient's maximum stride length suitable for their current rehabilitation stage, cadence is set to 15-25 steps / minute, and the number of steps is preset by medical staff or dynamically adjusted according to the training task. The score shows a positive linear correlation with stride length and cadence. In the step-by-step mapping mode, the 60-100 score range is divided into three consecutive steps: 60-70 points correspond to 100% stride length and 15 steps / minute, 70-80 points correspond to 100% stride length and 20 steps / minute, and 80-100 points correspond to 100% stride length and 25 steps / minute. The system synchronizes and matches the preset number of execution steps; finally, it generates a complete lower limb rehabilitation training instruction that includes the type of motor imagery, gait parameters, execution duration, and safety threshold.

[0067] Static command generation rules: When no motor imagery that meets the preset requirements is detected, including any of the following situations: no effective motor imagery is detected, the imagined action does not match the training task, or the motor intention intensity score is ≤60 points, a static command is generated directly. The command clearly requires the rehabilitation robot to maintain the current static state and not to perform any incremental power output actions to avoid safety risks caused by accidental triggering.

[0068] In some embodiments, transmitting training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete corresponding training actions includes: transmitting the type of motor imagery and gait parameters contained in the training instructions to the execution module of the rehabilitation robot; the execution module controls the rehabilitation robot to perform walking, knee bending, leg raising, left leg raising or right leg raising actions according to the type of motor imagery and gait parameters, and adjusts the amplitude, frequency and number of actions according to the gait parameters.

[0069] This embodiment is the implementation step of S103. The core is to clarify the transmission specifications of training instructions and the control logic of the robot execution module, so as to achieve low-latency and high-precision mapping from EEG active intention to robot actual action. At the same time, it refines the full-dimensional parameter control of action execution, solves the problems of asynchronous action execution with patient intention, insufficient action control precision and poor adaptability in traditional solutions, and ensures the safety, accuracy and synchronization of rehabilitation training actions.

[0070] Using medical-grade low-latency wired / wireless communication protocols (including CAN bus, industrial Ethernet, and Bluetooth 5.0), the generated training instructions, including the type of motor imagery (action type), gait parameters, and safety thresholds, are transmitted completely and in real time to the main control execution module of the lower limb rehabilitation robot. The end-to-end transmission latency is controlled within 100ms, ensuring the synchronization between action execution and the patient's movement intention. At the same time, a cyclic redundancy check code is added during the instruction transmission process to verify the integrity of the transmitted data and avoid action execution errors caused by data packet loss or errors.

[0071] like Figure 3 After receiving the training instruction, the execution module of the rehabilitation robot shown first performs a safety check, verifying whether the gait parameters and movement amplitude in the instruction are within the preset safety threshold range for the patient's current rehabilitation stage. If the check fails, execution is refused and a parameter abnormality prompt is triggered. If the check passes, the corresponding servo motor and exoskeleton execution mechanism are driven to complete the corresponding action according to the action type in the instruction. Executable actions include walking, knee flexion, leg lifting, left leg lifting, and right leg lifting. At the same time, the robot strictly controls the details of the action according to the gait parameters: adjusting the single step length when walking and the joint range of motion when lifting the leg / bending the knee according to the stride length parameter; adjusting the execution frequency of the action according to the step frequency parameter; and adjusting the number of repetitions of the action according to the number of steps parameter, so as to achieve precise and controllable control of the movement amplitude, frequency, and number of repetitions in all dimensions.

[0072] If the execution module receives a stop command, it immediately cuts off the incremental control of power output, locks the robot's current posture, and maintains a stationary state; if a stop command is received while the robot is performing an action, it executes a smooth deceleration and stop procedure, completing smooth braking within 300ms to avoid sudden stops that could cause limb injury to the patient.

[0073] In some embodiments, the acquisition of EEG signals from the patient's prefrontal cortex or motor cortex includes: acquiring EEG signals at preset points in the prefrontal cortex or motor cortex using a fully silicone flexible EEG acquisition device, utilizing a double-ear clip anti-interference structure and magnetically attached electrodes, wherein the preset points are 6-point simplified acquisition points.

[0074] like Figure 2 As shown, by adopting a dedicated all-silicone flexible EEG acquisition device, and through a double-ear clip anti-interference structure, magnetic female electrode, and a 6-point minimalist acquisition layout, it solves the problems of traditional EEG acquisition devices, such as complicated wearing, poor anti-interference ability, high patient cooperation threshold, and poor long-term wearing comfort. It is especially suitable for early rehabilitation scenarios of stroke hemiplegic patients who are bedridden, have limited limb movement, and have weak cooperation ability, while ensuring the stability of the acquired signal and a high signal-to-noise ratio.

[0075] It adopts a dedicated all-silicone flexible EEG acquisition device. The core configuration of the device includes a skin-friendly and bendable all-silicone flexible base, a double-ear clip-on common-mode anti-interference structure, a magnetic female electrode, and a front-mounted low-noise amplification and filtering circuit. The device has no rigid pressure structure and is suitable for patients who are bedridden for a long time or undergo long-term training.

[0076] The preset acquisition points adopt a minimalist layout of 6 points. For the prefrontal cortex acquisition mode, the 6 points are evenly distributed in the core feature acquisition areas such as FP1 and FP2 in the prefrontal cortex. There is no need to wear complicated whole-brain electrodes. Signal acquisition can be completed simply by placing them on the skin of the forehead, which greatly reduces the difficulty of wearing and the threshold for patient cooperation. For the motor cortex acquisition mode, the 6 points focus on covering the core motor imagery feature acquisition sites such as C3, C4, and Cz in the motor cortex, balancing acquisition accuracy and wearing convenience.

[0077] When wearing the device, the magnetic female electrode quickly completes the tool-free connection between the electrode and the flexible substrate, and the device is fixed by the double ear clip structure. At the same time, common-mode suppression of power frequency interference is achieved, improving the device's resistance to environmental interference. After wearing, the acquisition device is started, the sampling rate is set to no less than 250Hz, the signal bandwidth covers 0.5-30Hz (core frequency band of EEG characteristics), and continuous real-time acquisition is started. The patient's EEG signal is collected from preset points in the prefrontal cortex or motor cortex. During the acquisition process, the pre-filter circuit simultaneously completes baseline drift suppression and basic noise filtering to ensure the signal-to-noise ratio of the acquired signal.

[0078] In some embodiments, the method further includes: displaying the exercise intention intensity score, the action type corresponding to the training instruction, and the training completion status through a preset feedback interface, so as to achieve visual monitoring of the rehabilitation training process.

[0079] Through a pre-set dual-end feedback interface, the entire rehabilitation training process can be visualized and monitored in real time. This solves the problems of traditional methods where patients cannot intuitively perceive their own training effects and medical staff cannot monitor the training process in real time. Visual positive feedback effectively motivates patients to actively participate in training, while providing full data support for medical staff to adjust rehabilitation plans.

[0080] By pre-setting two sets of visual feedback interfaces—a training interaction interface for patients and a management and monitoring interface for medical staff—both are deployed on the touch display device of the brain-computer interface terminal. The two interfaces share the same data source and are synchronized in real time, each adapting to the needs of different users.

[0081] During training, the patient's interface displays three core pieces of information in real time: first, the patient's current exercise intention intensity score, visually presented using a combination of numbers and a dynamic progress bar, with positive visual cues provided when the score reaches the target; second, the type of movement corresponding to the current training instruction and the progress of the training task; and third, the completion status of this training session and the historical achievement rate. This intuitive visual feedback allows patients to perceive the training effect corresponding to their active exercise intentions in real time, generating positive motivation and increasing their willingness and cooperation in actively participating in training.

[0082] The interface for medical staff synchronously displays all patient training data, including real-time EEG signal waveforms, curves showing changes in the intensity of motor intentions, a complete record of training instructions issued, robot action execution status, training task completion rate, and abnormal event records, enabling full-process visual monitoring of the rehabilitation training process. It also allows medical staff to view the patient's historical training data and adjust training parameters, rehabilitation tasks, and model configurations online based on real-time monitoring data and historical data, without interrupting the training process.

[0083] The interface has a built-in anomaly warning module. When situations such as electrode detachment, abnormal signal quality, robot execution failure, or persistent failure of the patient's movement intention are detected, a prominent graded anomaly prompt will immediately pop up on the corresponding interface to remind relevant personnel to handle the situation in a timely manner and ensure the safety of the training process.

[0084] In some embodiments, after driving the rehabilitation robot to complete the corresponding training action, the method further includes: obtaining the corresponding motion intention quantification result and training completion feedback, and dynamically adjusting the fusion ratio of motion imagery and attention and the training difficulty information corresponding to the hybrid paradigm decoding model based on the motion intention quantification result and training completion feedback, thereby forming a personalized rehabilitation training closed loop.

[0085] This embodiment is the core closed-loop component of the entire rehabilitation training method. The core is to dynamically adjust the dual-paradigm fusion ratio and training difficulty of the hybrid paradigm decoding model based on training feedback data after the robot completes the training movements, forming a personalized rehabilitation closed loop of "training-feedback-adjustment-optimization". This solves the problem that traditional programs have fixed training strategies and cannot dynamically adapt to the differentiated needs of patients at different rehabilitation stages, enabling continuous iterative optimization of rehabilitation programs and continuously improving the rehabilitation effect of lower limb motor function.

[0086] After driving the rehabilitation robot to complete the corresponding training movements, the system automatically retrieves the full closed-loop feedback data of this training, which includes: the quantitative results of motor intention (mean, peak, and achievement rate of the intensity score of motor intention throughout the process), the training completion status (action execution completion rate, task completion degree, and abnormal situation records), the temporal data of the patient's attention and focus during the training process, and the judgment accuracy and misjudgment rate data of the hybrid paradigm decoding model, which serve as the core basis for dynamic optimization.

[0087] Based on the feedback data from this training, the model parameters are optimized. If the patient's attention span remains below the preset threshold during training and the decoding model has a high misjudgment rate, the weight of the attention paradigm feature in the hybrid paradigm decoding model is automatically increased, while the weight of the motor imagery paradigm is reduced. This strengthens the verification of the patient's training focus and reduces invalid misjudgments. If the patient's attention span is stable and meets the standard, but the sensitivity of motor imagery recognition is insufficient and the ability to capture subtle intentions is poor, the weight of the motor imagery paradigm feature is automatically increased to optimize the model's ability to recognize subtle motor intentions and achieve personalized dynamic adaptation of the dual paradigm fusion ratio.

[0088] Based on the quantitative results of motor intention and training completion, the training difficulty is adaptively adjusted. If the patient's average motor intention intensity score is consistently above 80 points for three or more consecutive training sessions, the task completion rate is 100%, and there are no abnormalities, the training difficulty is automatically increased. This includes raising the safe upper limit of gait parameters, increasing the complexity of movements, shortening the execution interval of single movements, and increasing multi-movement combination training tasks. If the patient's motor intention intensity achievement rate is below 30% for three or more consecutive training sessions, and the task is difficult to complete, the training difficulty is automatically reduced. This includes lowering the threshold of gait parameters, simplifying movement types, extending the training interval, and reducing the duration of a single training session, ensuring that the training difficulty is always accurately matched with the patient's current rehabilitation ability.

[0089] The adjusted dual-paradigm fusion ratio and training difficulty parameters are updated in real time to the hybrid paradigm decoding model and training task configuration library, and directly applied to the next rehabilitation training. At the same time, feedback data from each training session is continuously collected, and the above optimization process is repeated to form a closed loop of personalized rehabilitation training throughout the entire process. This continuously adapts to the training needs of stroke hemiplegic patients at different rehabilitation stages, and achieves efficient recovery of lower limb motor function.

[0090] In some embodiments, the workflow of a general motion visualization + attention hybrid paradigm is as follows: Figure 4As shown, EEG signals from the patient's frontal lobe are collected using an EEG acquisition device. Simultaneously, the validity of the EEG signals is verified through a combination of detachment and impedance detection, eliminating invalid signals. Subsequently, the hybrid paradigm decoding module of the brain-computer interface terminal performs collaborative decoding of the signals, identifying signal characteristics and quantifying the intensity of motor intention. Based on the decoding results and the quantified motor intention value, lower limb rehabilitation training instructions (such as walking on flat ground, knee flexion, and leg raising) are generated to suit the patient's rehabilitation stage. These instructions are transmitted to the lower limb rehabilitation robot execution module, driving the robot to complete the corresponding training movements. Simultaneously, the quantified motor intention results and training completion status are fed back to the patient and medical staff in real time, dynamically adjusting the fusion ratio of the hybrid paradigm and the training difficulty to form a personalized rehabilitation training closed loop. This adapts to the training needs of stroke hemiplegic patients at different rehabilitation stages, ultimately achieving efficient recovery of lower limb motor function.

[0091] A hybrid paradigm of motor imagery and attention based on prefrontal brain-computer interface, such as Figure 5 As shown, the overall process is similar to the general process described above. The details of the decoding results and motion commands need to be clarified here: The game page only requires the patient to imagine walking. The hybrid paradigm decoding module will output two indicators: "whether walking was imagined" and "motor intention intensity score." If it is determined that the patient imagined walking, the gait parameters (stride length, cadence, and mapping relationship as follows) are adjusted according to the motor intention intensity score, and a walking command is issued. The robot walks a specified number of steps with the specified cadence and stride length (the number of steps is adjustable). If it is determined that the patient did not imagine walking, a stationary command is issued, and the robot remains stationary. The gait parameters are adjusted according to a linear mapping relationship, as shown in the table below: The gait parameters are adjusted according to the gear position as shown in the table below: Lower limb rehabilitation training and methods based on a hybrid paradigm of motor imagery and attention using whole-brain brain-computer interfaces include: because whole-brain brain-computer interfaces can collect EEG signals from the motor cortex, they can identify different types of movements, such as... Figure 6 As shown, the overall process is similar to the general process described above. Here, the decoding results and details of the movement instructions need to be clarified: The game page can ask the patient to imagine specific actions such as raising the left leg or the right leg. The hybrid paradigm decoding module will output two indicators: "type of movement imagination" and "intensity score of movement intention". If it is determined that the patient has imagined the specified action, the gait parameters are adjusted according to the intensity score of movement intention and the instruction for the specified action (such as raising the left leg) is issued. The robot executes the action with the specified parameters. If it is determined that the patient has not performed movement imagination or the imagined movement is not the specified type, a stationary instruction is issued and the robot remains stationary.

[0092] Please see Figure 7 As shown, Figure 7This is a schematic diagram of the structure of a rehabilitation training system 200 based on a hybrid paradigm brain-computer interface provided in this application embodiment. The rehabilitation training system 200 based on a hybrid paradigm brain-computer interface is used to execute the steps of the rehabilitation training methods based on a hybrid paradigm brain-computer interface shown in the above embodiments. The rehabilitation training system 200 based on a hybrid paradigm brain-computer interface can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0093] like Figure 7 As shown, the rehabilitation training system 200 based on a hybrid paradigm brain-computer interface includes: The signal acquisition unit 201 is used to acquire electroencephalogram (EEG) signals from the patient's frontal lobe or motor cortex. The collaborative decoding unit 202 is used to collaboratively decode the EEG signal according to a preset hybrid paradigm decoding model and identify signal features; if the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined; if the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. The training completion unit 203 is used to generate lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal characteristics. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

[0094] In some embodiments, before performing collaborative decoding of the EEG signal according to a preset hybrid paradigm decoding model, the method further includes: validating the acquired EEG signal and removing invalid signals.

[0095] In some embodiments, the step of validating the acquired EEG signals and removing invalid signals includes: determining whether the electrodes of the EEG acquisition device have detached through a preset detachment detection mechanism; determining whether the impedance value of the EEG signal is within the effective range through a preset impedance detection mechanism; and determining and removing EEG signals that have detected electrode detachment or impedance values ​​exceeding the effective range as invalid signals.

[0096] In some embodiments, the step of collaboratively decoding the EEG signal according to a preset hybrid paradigm decoding model to identify signal features includes: if the EEG signal is a prefrontal EEG signal, determining whether the patient has generated a walking motion image through the hybrid paradigm decoding model, and outputting the corresponding motor intention intensity score; if the EEG signal is a motor cortex EEG signal, identifying the specific type of movement imagined by the patient through the hybrid paradigm decoding model, wherein the movement type includes at least one of raising the left leg and raising the right leg, and outputting the corresponding motor intention intensity score.

[0097] In some embodiments, generating lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal features includes: when it is determined that the patient has generated a motor imagery that meets preset requirements, determining gait parameters corresponding to the motor imagery according to the motor intention intensity score and a preset mapping relationship, wherein the gait parameters include at least one of stride length, stride frequency, and number of steps; generating training instructions that include the type of motor imagery and gait parameters; and generating a stationary instruction when it is not determined that the patient has generated a motor imagery that meets preset requirements.

[0098] In some embodiments, transmitting training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete corresponding training actions includes: transmitting the type of motor imagery and gait parameters contained in the training instructions to the execution module of the rehabilitation robot; the execution module controls the rehabilitation robot to perform walking, knee bending, leg raising, left leg raising or right leg raising actions according to the type of motor imagery and gait parameters, and adjusts the amplitude, frequency and number of actions according to the gait parameters.

[0099] In some embodiments, the acquisition of EEG signals from the patient's prefrontal cortex or motor cortex includes: acquiring EEG signals at preset points in the prefrontal cortex or motor cortex using a fully silicone flexible EEG acquisition device, utilizing a double-ear clip anti-interference structure and magnetically attached electrodes, wherein the preset points are 6-point simplified acquisition points.

[0100] In some embodiments, the method further includes: displaying the exercise intention intensity score, the action type corresponding to the training instruction, and the training completion status through a preset feedback interface, so as to achieve visual monitoring of the rehabilitation training process.

[0101] In some embodiments, after driving the rehabilitation robot to complete the corresponding training action, the method further includes: obtaining the corresponding motion intention quantification result and training completion feedback, and dynamically adjusting the fusion ratio of motion imagery and attention and the training difficulty information corresponding to the hybrid paradigm decoding model based on the motion intention quantification result and training completion feedback, thereby forming a personalized rehabilitation training closed loop.

[0102] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the rehabilitation training system and its modules based on the hybrid paradigm brain-computer interface described above can be found in the corresponding embodiments of the rehabilitation training method based on the hybrid paradigm brain-computer interface, and will not be repeated here.

[0103] The aforementioned rehabilitation training method based on a hybrid paradigm brain-computer interface can be implemented as a computer program, which can be used in, for example... Figure 7 It runs on the device shown.

[0104] Please see Figure 8 , Figure 8 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0105] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any rehabilitation training method based on a hybrid paradigm brain-computer interface.

[0106] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0107] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any rehabilitation training method based on a hybrid paradigm brain-computer interface.

[0108] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0109] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0110] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Collect electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex; The EEG signal is collaboratively decoded according to a preset hybrid paradigm decoding model to identify signal features; if the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined; if the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. Based on the signal characteristics, lower limb rehabilitation training instructions are generated to suit the patient's rehabilitation stage. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

[0111] In some embodiments, before performing collaborative decoding of the EEG signal according to a preset hybrid paradigm decoding model, the method further includes: validating the acquired EEG signal and removing invalid signals.

[0112] In some embodiments, the step of validating the acquired EEG signals and removing invalid signals includes: determining whether the electrodes of the EEG acquisition device have detached through a preset detachment detection mechanism; determining whether the impedance value of the EEG signal is within the effective range through a preset impedance detection mechanism; and determining and removing EEG signals that have detected electrode detachment or impedance values ​​exceeding the effective range as invalid signals.

[0113] In some embodiments, the step of collaboratively decoding the EEG signal according to a preset hybrid paradigm decoding model to identify signal features includes: if the EEG signal is a prefrontal EEG signal, determining whether the patient has generated a walking motion image through the hybrid paradigm decoding model, and outputting the corresponding motor intention intensity score; if the EEG signal is a motor cortex EEG signal, identifying the specific type of movement imagined by the patient through the hybrid paradigm decoding model, wherein the movement type includes at least one of raising the left leg and raising the right leg, and outputting the corresponding motor intention intensity score.

[0114] In some embodiments, generating lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal features includes: when it is determined that the patient has generated a motor imagery that meets preset requirements, determining gait parameters corresponding to the motor imagery according to the motor intention intensity score and a preset mapping relationship, wherein the gait parameters include at least one of stride length, stride frequency, and number of steps; generating training instructions that include the type of motor imagery and gait parameters; and generating a stationary instruction when it is not determined that the patient has generated a motor imagery that meets preset requirements.

[0115] In some embodiments, transmitting training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete corresponding training actions includes: transmitting the type of motor imagery and gait parameters contained in the training instructions to the execution module of the rehabilitation robot; the execution module controls the rehabilitation robot to perform walking, knee bending, leg raising, left leg raising or right leg raising actions according to the type of motor imagery and gait parameters, and adjusts the amplitude, frequency and number of actions according to the gait parameters.

[0116] In some embodiments, the acquisition of EEG signals from the patient's prefrontal cortex or motor cortex includes: acquiring EEG signals at preset points in the prefrontal cortex or motor cortex using a fully silicone flexible EEG acquisition device, utilizing a double-ear clip anti-interference structure and magnetically attached electrodes, wherein the preset points are 6-point simplified acquisition points.

[0117] In some embodiments, the method further includes: displaying the exercise intention intensity score, the action type corresponding to the training instruction, and the training completion status through a preset feedback interface, so as to achieve visual monitoring of the rehabilitation training process.

[0118] In some embodiments, after driving the rehabilitation robot to complete the corresponding training action, the method further includes: obtaining the corresponding motion intention quantification result and training completion feedback, and dynamically adjusting the fusion ratio of motion imagery and attention and the training difficulty information corresponding to the hybrid paradigm decoding model based on the motion intention quantification result and training completion feedback, thereby forming a personalized rehabilitation training closed loop.

[0119] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the rehabilitation training method based on a hybrid paradigm brain-computer interface as provided in any embodiment of this application.

[0120] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A rehabilitation training method based on a hybrid paradigm brain-computer interface, characterized in that, include: Collect electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex; The EEG signals are collaboratively decoded according to a preset hybrid paradigm decoding model to identify signal features; If the EEG signal is a prefrontal cortex EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined; if the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. Based on the signal characteristics, lower limb rehabilitation training instructions are generated to suit the patient's rehabilitation stage. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

2. The method according to claim 1, characterized in that, Before performing collaborative decoding of the EEG signal according to the preset hybrid paradigm decoding model, the method further includes: The collected EEG signals are validated, and invalid signals are removed.

3. The method according to claim 2, characterized in that, The validity verification of the acquired EEG signals and the removal of invalid signals includes: The system uses a pre-set detachment detection mechanism to determine whether the electrodes of the EEG acquisition device have detached. The system uses a preset impedance detection mechanism to determine whether the impedance value of the EEG signal is within the effective range. EEG signals that detect electrode detachment or impedance values ​​exceeding the effective range are identified as invalid signals and discarded.

4. The method according to claim 1, characterized in that, The step of collaboratively decoding the EEG signal according to a preset hybrid paradigm decoding model to identify signal features includes: If the EEG signal is a prefrontal EEG signal, the hybrid paradigm decoding model is used to determine whether the patient has a walking motion image and outputs the corresponding motion intention intensity score. If the EEG signal is a motor cortex EEG signal, the specific type of movement imagined by the patient is identified by the hybrid paradigm decoding model. The type of movement includes at least one of raising the left leg and raising the right leg, and the corresponding movement intention intensity score is output.

5. The method according to claim 4, characterized in that, The step of generating lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal features includes: When it is determined that the patient has generated a motor imagery that meets the preset requirements, the gait parameters corresponding to the motor imagery are determined according to the motor intention intensity score and a preset mapping relationship. The gait parameters include at least one of stride length, cadence, and number of steps. Generate training instructions that include the motor imagery type and gait parameters; when no motor imagery meeting the preset requirements is detected from the patient, generate a stationary instruction.

6. The method according to claim 1, characterized in that, The step of transmitting training instructions to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions includes: The training instructions, which include the type of motor imagery and gait parameters, are transmitted to the execution module of the rehabilitation robot. The execution module controls the rehabilitation robot to perform walking, knee bending, leg lifting, left leg lifting, or right leg lifting actions according to the type of motor imagery and gait parameters, and adjusts the amplitude, frequency, and number of movements according to the gait parameters.

7. The method according to claim 1, characterized in that, The acquisition of electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex includes: Using a flexible silicone EEG acquisition device, and utilizing a double ear clip anti-interference structure and magnetically attached electrodes, EEG signals are acquired at preset points in the prefrontal cortex or motor cortex. The preset points are a simplified 6-point acquisition point system.

8. The method according to claim 5, characterized in that, The method further includes: The preset feedback interface displays the intensity score of the exercise intention, the type of movement corresponding to the training instruction, and the training completion status, thereby enabling visual monitoring of the rehabilitation training process.

9. The method according to claim 1, characterized in that, After driving the rehabilitation robot to complete the corresponding training action, the method further includes: Obtain the corresponding quantitative results of motor intention and feedback on training completion. Based on the quantitative results of motor intention and feedback on training completion, dynamically adjust the fusion ratio of motor imagery and attention and the training difficulty information of the hybrid paradigm decoding model to form a personalized rehabilitation training closed loop.

10. A rehabilitation training system based on a hybrid paradigm brain-computer interface, characterized in that, The method applied to any one of claims 1-9 includes: The signal acquisition unit is used to acquire electroencephalogram (EEG) signals from the patient's prefrontal cortex or motor cortex. The collaborative decoding unit is used to collaboratively decode the EEG signal according to a preset hybrid paradigm decoding model and identify signal features. If the EEG signal is a prefrontal EEG signal, the signal feature is the intensity score of whether or not the intention to walk and move is imagined. If the EEG signal is a motor cortex EEG signal, the signal feature is the type of motor imagination and the intensity score of the intention to move. The training completion unit is used to generate lower limb rehabilitation training instructions adapted to the patient's rehabilitation stage based on the signal characteristics. The training instructions include at least one of walking, knee flexion, leg raising, raising the left leg, or raising the right leg. The training instructions are transmitted to the execution module of the rehabilitation robot to drive the rehabilitation robot to complete the corresponding training actions.

Images