Personalized exoskeleton rehabilitation system and method based on conscious motor impairment patient classification

Abstract

The application discloses a personalized exoskeleton rehabilitation system and method based on consciousness and motor disorder patient grading, which comprises a comprehensive analysis platform, an electroencephalogram interface device, an electromyogram interface device and a modular exoskeleton combination. When applied, the patient wears the required device, connects the electroencephalogram signal collector through the electroencephalogram interface device, connects the electromyogram signal collector through the electromyogram interface device, sends the collected electroencephalogram signal and electromyogram signal to the comprehensive analysis platform for analysis, sends the control instruction to the controller according to the analysis result, and controls the modular exoskeleton combination to execute the corresponding personalized action combination. Through the integration of the comprehensive physiological signals of the patient and the dynamic evaluation mechanism, the accurate identification of the consciousness state and the motor ability of the patient is realized, the full-grade patient group from complete unconsciousness to basic autonomy can be covered, and the individual adaptability of the rehabilitation system is greatly improved. The application has good clinical popularization and productization potential.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of rehabilitation exoskeleton control, and particularly relates to a personalized exoskeleton rehabilitation system and method based on consciousness motor disorder patient grading. BACKGROUND

[0002] Consciousness motor disorder is a kind of nervous system disease involving abnormal cognitive-motor coupling mechanism, which is commonly seen in patients with stroke, brain trauma, Parkinson's disease, cerebral palsy, etc. Such patients often have different levels of consciousness disorder or executive function deficiency while their motor control ability is impaired. The rehabilitation process often involves motor stimulation such as turning over, lifting legs and hands, etc. The patient's motor function is enhanced by combining the body consciousness control exoskeleton. However, the existing motor rehabilitation system is mainly used for consciousness promotion or rehabilitation training for single consciousness motor disorder patients, driven by electromyography or posture information, and the consciousness level or cognitive participation of the patient cannot be monitored in real time, so the patient's cooperation degree and active participation willingness cannot be effectively judged, resulting in defects such as inadaptability of exoskeleton assistive device, difficulty in dynamically adjusting training intensity, insufficient consideration of patient's comprehensive physiological information, and insufficient human-machine cooperation control in the rehabilitation process of patients with different levels. Therefore, it is necessary to develop an exoskeleton rehabilitation system that integrates patient's comprehensive physiological information evaluation mechanism and has grading training and personalized intervention ability to meet the needs of consciousness motor disorder patients for high adaptability, precision and safety of intelligent auxiliary equipment.

[0003] Although the exoskeleton rehabilitation system has made progress in intention recognition and human-machine interaction, there are still problems such as lack of consciousness-motor coordination evaluation mechanism, insufficient individual adaptation, lack of effective human-machine closed-loop feedback mechanism and personalized rehabilitation training scheme when facing the special group of consciousness motor disorder patients. Most systems do not establish a grading classification mechanism, and the rehabilitation scheme cannot be dynamically adjusted according to the patient's ability level, resulting in unreasonable training intensity and the risk of excessive load or insufficient training. The existing technology generally uses a single signal source (such as EMG or EEG) for driving, and cannot integrate multiple physiological information for comprehensive evaluation, making it difficult to effectively identify the patient's consciousness level and willingness to participate, resulting in mismatch between training strategy and patient state. In addition, the current exoskeleton system is mostly open-loop control or preset strategy response, lacking the ability to automatically adjust control instructions according to the patient's immediate state, with feedback delay and adjustment lag, affecting the training effect and safety.

[0004] Therefore, there is an urgent need for an intelligent exoskeleton rehabilitation system for consciousness motor disorder patients, which integrates patient's comprehensive physiological information, has state grading recognition, closed-loop control and personalized training scheme. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a personalized exoskeleton rehabilitation system and method based on consciousness motor disorder patient grading, which solves the problem that the existing exoskeleton system does not establish a grading classification mechanism and the rehabilitation scheme cannot be dynamically adjusted according to the patient's ability level.

[0006] The present application solves the above technical problems by adopting the following technical solutions:

[0007] The personalized exoskeleton rehabilitation system based on consciousness motor disorder patient grading comprises a comprehensive analysis platform, an electroencephalogram interface device, an electromyogram interface device, and a modular exoskeleton combination, wherein the electroencephalogram interface device and the modular exoskeleton combination are wearable devices; the modular exoskeleton combination comprises detachably connected upper limb exoskeletons, hip joint exoskeletons, and knee joint exoskeletons, and the upper limb exoskeletons and the hip joint exoskeletons are each provided with a controller; the electromyogram interface device is arranged at a position close to the arm of the upper limb exoskeleton; the comprehensive analysis platform is connected to the controller through a wired or wireless connection; in application, the patient wears the required devices, connects the electroencephalogram signal collector through the electroencephalogram interface device, connects the electromyogram signal collector through the electromyogram interface device, sends the collected electroencephalogram signals and electromyogram signals to the comprehensive analysis platform for analysis, sends control instructions to the controller according to the analysis results, and controls the modular exoskeleton combination to perform corresponding personalized action combinations.

[0008] The comprehensive analysis platform comprises a workstation, a display screen, and an electrical cabinet, wherein the workstation processes the received collected signals and issues instructions; when the modular exoskeleton combination is fixedly used, the modular exoskeleton combination is connected to the comprehensive analysis platform through a bus; and when the modular exoskeleton combination is used in a mobile manner, the modular exoskeleton combination is connected to the comprehensive analysis platform in a wireless manner.

[0009] The workstation comprises a data processing module and an instruction output module; the data processing module comprises a data preprocessing and feature extraction model, a machine learning model, and a consciousness-motor function grading evaluation prediction model; the data preprocessing and feature extraction model filters and denoises the received physiological signals of the patient, adopts a time domain and frequency domain feature extraction method, and acquires representative features; the preprocessed features are input to the machine learning model, which is trained and optimized through historical data to generate the consciousness-motor function grading evaluation prediction model of the patient function grading.

[0010] According to the consciousness level and motor ability of the patient, the machine learning model divides the patient into different function levels and gives an evaluation result for each level.

[0011] The system further comprises a walking assistance device for assisting walking movement during rehabilitation training.

[0012] The patient motion intention and consciousness state evaluation method comprises the following steps:

[0013] Step 1, physiological signal acquisition and data fusion: real-time acquisition of patient's multi-source signals including electroencephalogram signal, electromyogram signal, and inertial sensor data, and synchronous processing; synthesizing an initial patient state model through data fusion algorithm according to the acquired multi-source signal data, inputting the initial patient state model into the consciousness-motor function classification evaluation prediction model to obtain the patient's functional classification results including I level complete unconsciousness, II level partial consciousness, III level conscious but movement limited, and IV level basic self-determination;

[0014] Step 2, motor intention evaluation and dynamic adjustment, specifically including:

[0015] Step 2.1, extracting the root mean square amplitude of electromyogram signal and the event-related desynchronization index ERD of motor cortex region of electroencephalogram signal;

[0016] Step 2.2, setting intention recognition threshold according to the patient functional classification results obtained in step 1: for I-II level patients, the ERD threshold is 30% of the maximum ERD value and the RMS threshold is 20% of the maximum voluntary contraction; for III level patients, the ERD threshold is 50% of the maximum ERD value and the RMS threshold is 40% of the maximum voluntary contraction; for IV level patients, the ERD threshold is 70% of the maximum ERD value and the RMS threshold is 60% of the maximum voluntary contraction;

[0017] Step 2.3, when ERD and exceed their respective thresholds at the same time, it is determined that there is a motor intention, and the joint angular velocity data provided by the inertial sensor is used to judge the type of motor intention: angular velocity < 10° / s is static support intention, 10° / s ≤ angular velocity < 50° / s is slow movement intention, and angular velocity ≥ 50° / s is fast movement intention;

[0018] Step 3, dynamic motion control strategy optimization: the system dynamically adjusts the control strategy of the exoskeleton according to the type of motor intention identified in step 2, adjusts the motion direction, speed and intensity of the exoskeleton to adapt to the current state of the patient;

[0019] Step 4, real-time feedback mechanism and driving parameter optimization, specifically including:

[0020] Step 4.1, real-time monitoring of patient's joint range of motion and movement duration through inertial sensor, and calculating muscle fatigue index through electromyogram signal;

[0021] Step 4.2, adjusting the assistive torque and driving stiffness of the exoskeleton according to the muscle fatigue index to optimize the driving parameters;

[0022] Step 4.3: Send the optimized drive parameters to the exoskeleton controller to adjust the exercise intensity in real time and prevent over-fatigue or under-training.

[0023] A personalized training method for human-machine closed-loop control based on patients with consciousness-motor disorders includes the following steps:

[0024] Step a: Develop a personalized, graded rehabilitation training plan in advance;

[0025] Step b: Real-time assessment of the patient's state of consciousness is achieved by collecting electroencephalogram (EEG) signals, and basic limb responsiveness is monitored by combining EMG signals to select an initial rehabilitation training program.

[0026] Step c: Apply corresponding movements to the patient according to the preliminary rehabilitation training plan. At the same time, use the patient's movement intention and consciousness status assessment method to collect the patient's real-time physiological signals and provide assessment feedback.

[0027] Step d: Based on the feedback of the patient's movement status and physiological response, automatically optimize the output parameters of the exoskeleton controller. Specific optimization methods include:

[0028] Step d1, Movement Quality Assessment: Calculate the patient's movement completion indicators, including the joint range of motion achievement rate, using inertial sensor data. Trajectory deviation ;

[0029] Step d2, Training Progress Evaluation: Count the cumulative number of training cycles. Each completed set of preset actions is counted as one cycle. When the cumulative number of training cycles increases by 10 cycles, a parameter optimization evaluation is triggered.

[0030] Step d3, Parameter optimization decision:

[0031] If within 10 consecutive cycles The average value is ≥0.90 and <5cm indicates good patient recovery, reducing the need for assistance;

[0032] If 0.75≤ <0.90 or 5cm≤ <10cm indicates that the patient is in a stable training period, and the current parameters should be kept unchanged;

[0033] like <0.75 or ≥10cm indicates insufficient assistive force; increase assistive force.

[0034] Step d4, Output Parameter Update: The optimized assist force, speed parameters, and force parameters are sent to the exoskeleton controller. The exoskeleton controller adjusts the motor drive current and position control gain according to the optimized parameters.

[0035] The patient's state of consciousness included complete unconsciousness and consciousness with impaired motor function.

[0036] The graded rehabilitation training program includes a passive guidance mode, a semi-active assisted mode, and an active self-management mode. Patients who are completely unconscious use the passive guidance mode for rehabilitation training. Patients who are conscious but have motor impairments initially choose the semi-active assisted mode for rehabilitation training. As the training progresses, the program automatically switches to the active self-management mode based on feedback information.

[0037] The rehabilitation process in passive guidance mode is as follows:

[0038] By collecting electroencephalogram (EEG) signals to assess the patient's state of consciousness in real time, combined with electromyography (EMG) signals to monitor the patient's basic limb reaction ability, and through a passive guidance mode to perform precise limb movements, promote joint mobility and muscle training, and adjust the intensity of exercise in a timely manner through a feedback mechanism.

[0039] The rehabilitation process using a semi-active assisted mode is as follows:

[0040] The lower limb exoskeleton training unit is matched according to the patient's level, and walking aids are provided according to the patient's rehabilitation stage. Training is carried out according to the training plan, and the patient's electromyographic response is monitored in real time. The driving force and movement pattern of the exoskeleton are adjusted. As the patient's recovery progresses, the assistance is gradually reduced to enhance the patient's independent movement ability. When the training feedback meets the target, it automatically switches to active and autonomous mode for rehabilitation training.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. This invention integrates comprehensive physiological signals of patients with dynamic assessment mechanisms to achieve accurate identification of patients' state of consciousness and motor ability. It can cover all levels of patients, from complete unconsciousness to basic autonomy, and greatly improve the individual adaptability of the rehabilitation system.

[0043] 2. The rehabilitation system of this invention can match the exoskeleton rehabilitation module according to the patient's comprehensive physiological information assessment results and dynamically adjust the exoskeleton assistance mode, so that the training process is highly coordinated with the patient's recovery rhythm. Through multi-level training modules such as passive guidance, semi-active assistance, and active self-management, it ensures that the intervention intensity always matches the patient's current ability, significantly improving rehabilitation efficiency and safety.

[0044] 3. This invention constructs a closed-loop control path of perception-evaluation-execution-feedback, which can adjust control parameters in real time to respond to changes in the patient's state. While ensuring the stability and accuracy of training, the system achieves real-time response and intervention to states such as physiological load and movement abnormalities.

[0045] 4. This system adopts a modular design and supports the selection of functional units such as upper limb, hip joint, and knee joint according to the needs of different patients, meeting the rehabilitation needs of different limb parts. It is also compatible with various training postures such as lying down, sitting, standing, and walking, and is suitable for hospitals, rehabilitation centers and home environments. It has good clinical promotion and productization potential. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the overall structure of the personalized exoskeleton rehabilitation system described in this invention.

[0047] Figure 2 This invention relates to a consciousness-motor function grading assessment module.

[0048] Figure 3 This is a diagram of the human-machine closed-loop control system of the present invention.

[0049] Figure 4 This invention provides a personalized rehabilitation training program.

[0050] Figure 5 This is a schematic diagram illustrating the rehabilitation application of the present invention for patients with different levels of consciousness and motor impairment.

[0051] The labels in the diagram are: 1-Comprehensive analysis platform; 2-Electroencephalogram (EEG) interface device; 3-Upper limb exoskeleton; 4-Electromyogram (EMG) interface device; 5-Hip joint exoskeleton; 6-Knee joint exoskeleton; 7-Walking aid. Detailed Implementation

[0052] The structure and working process of the present invention will be further described below with reference to the accompanying drawings.

[0053] The purpose of this invention is to address the problems in existing exoskeleton rehabilitation systems, such as the lack of a consciousness-motor coordination assessment mechanism, insufficient personalized adaptation, and the absence of an effective human-machine closed-loop feedback mechanism and personalized rehabilitation training programs. Most systems lack a grading and classification mechanism, and rehabilitation programs cannot be dynamically adjusted according to the patient's ability level. This invention provides a personalized exoskeleton rehabilitation system and its training method based on the grading of patients with consciousness-motor disorders, enabling graded training for patients with different levels of consciousness-motor disorders. The system features personalized combinations to meet the rehabilitation needs of different limb parts, while being compatible with various training postures such as lying down, sitting, standing, and walking.

[0054] A personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders includes a comprehensive analysis platform, an EEG interface device, an EMG interface device, and a modular exoskeleton assembly. The EEG interface device and the modular exoskeleton assembly are wearable devices. The modular exoskeleton assembly includes a detachably connected upper limb exoskeleton, a hip exoskeleton, and a knee exoskeleton. The upper limb and hip exoskeletons are equipped with controllers, and the EMG interface device is positioned close to the arm on the upper limb exoskeleton. The comprehensive analysis platform and controllers are connected via wired or wireless means. In use, the patient wears the required devices, connects the EEG interface device to an EEG signal collector, and the EMG interface device to an EMG signal collector. The acquired EEG and EMG signals are sent to the comprehensive analysis platform for analysis. Based on the analysis results, control commands are sent to the controllers to control the modular exoskeleton assembly to execute corresponding personalized movement combinations.

[0055] Specific embodiments, such as Figures 1 to 5 As shown:

[0056] A personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders comprises a comprehensive analysis platform 1, an electroencephalogram (EEG) interface device 2, an upper limb exoskeleton 3, an electromyography (EMG) interface device 4, a hip exoskeleton 5, a knee exoskeleton 6, and a walking aid 7. The specific structure of the personalized exoskeleton system is as follows: Figure 1 As shown. The integrated analysis platform 1 consists of a high-performance workstation, an LCD screen, a caster wheeled mobile platform, and an electrical control cabinet. It is connected to the upper limb exoskeleton 3 controller via a bus to achieve stable data transmission. It can also wirelessly interact with the lower limb exoskeleton modules (including the hip exoskeleton 5 and the knee exoskeleton 6) and the EEG interface device 2. When the modular exoskeleton assembly is in fixed use, it is connected to the integrated analysis platform 1 via a bus. When the modular exoskeleton assembly is in mobile use, it is connected to the integrated analysis platform 1 wirelessly.

[0057] This device can comprehensively assess various physiological information of the patient, effectively identify the patient's level of consciousness and willingness to participate, and match appropriate training strategies accordingly. The upper limb exoskeleton 3 consists of a control box, joint motors, Bowden cable drive mechanism, rigid frame, and upper limb flexible straps. Simultaneously, the electromyography (EMG) interface device 4 is integrated into the upper limb flexible straps for real-time monitoring of the patient's EMG signals. The hip exoskeleton 5 mainly consists of a waist belt, joint motors, inertial sensors, controller, joint linkages, and thigh flexible straps. The knee exoskeleton 6 mainly consists of Bowden cables, calf braces, muscle interface device, and foot braces.

[0058] The workstation includes a data processing module and an instruction output module. The data processing module includes a data preprocessing and feature extraction model, a machine learning model, and a consciousness-motor function grading assessment prediction model. The data preprocessing and feature extraction model filters and denoises the received patient physiological signals and uses time-domain and frequency-domain feature extraction methods to obtain representative features. The preprocessed features are input into the machine learning model, which is trained and optimized using historical data to generate a consciousness-motor function grading assessment prediction model for patient functional grading.

[0059] The specific implementation of the data preprocessing and feature extraction model is as follows:

[0060] The EEG signals were denoised using 0.5-45Hz bandpass filtering and independent component analysis to extract the motor cortex region. , , The event-related desynchronization ERD exponent of the electrode is calculated using the following formula:

[0061]

[0062] in, The baseline power spectral density, This represents the power spectral density during the motion phase.

[0063] The electromyography (EMG) signals were subjected to 20-450Hz bandpass filtering, full-wave rectification, and 50ms window RMS smoothing to extract the root mean square amplitude. The calculation formula is:

[0064]

[0065] in, This represents the number of sampling points within the window.

[0066] The machine learning model employs an ensemble learning approach combining random forest and Bayesian inference. Training data comprises 7500 samples from 150 patients, divided into training, validation, and test sets in a 7:1.5:1.5 ratio. The random forest contains 500 decision trees with a maximum depth of 12, using 5-fold cross-validation. The model outputs the patient's functional level (Level I: Completely unconscious; Level II: Partially conscious; Level III: Conscious but with limited movement; Level IV: Basically autonomous), achieving a classification accuracy of 91.2% on the test set. The model utilizes an incremental learning mechanism, updating parameters every 100 newly added samples, retaining 70% of the original decision trees and adding 30% based on the new data, thus dynamically adapting to the patient's rehabilitation progress.

[0067] The consciousness-motor function grading assessment and prediction model is based on collected multi-source signals (such as EEG signals, EMG signals, inertial sensor data, etc.) and utilizes machine learning models (such as random forest, Bayesian inference, etc.) for state classification and dynamic assessment. The specific steps are as follows:

[0068] 1) Data Acquisition: Physiological signals from patients are collected in real time using devices such as electroencephalography (EEG), electromyography (EMG), and inertial measurement unit (IMU). Each signal is connected to a comprehensive analysis platform via a hardware interface to ensure the real-time nature and accuracy of the data.

[0069] 2) Data preprocessing and feature extraction: The collected signals are filtered and denoised, and representative features are obtained by using time-domain and frequency-domain feature extraction methods (such as FFT, power spectrum analysis, etc.).

[0070] 3) Machine learning training and model optimization: Input the preprocessed features into the machine learning model (such as random forest, Bayesian inference, etc.), train and optimize it with historical data, and generate a predictive model for patient functional classification.

[0071] 4) Functional level classification: Based on the patient's level of consciousness and motor ability, the machine learning model classifies the patient into different functional levels (such as Level I completely unconscious - Level II partially conscious - Level III conscious but motor limited - Level IV basically autonomous) and provides the assessment results for each level to guide the selection of subsequent personalized rehabilitation plans.

[0072] A personalized training method based on human-machine closed-loop control for patients with consciousness-motor disorders dynamically selects and adjusts training modes according to the patient's physiological state and functional assessment results, in order to achieve precise rehabilitation intervention for different patients with consciousness-motor disorders. The specific process is as follows: Mode selection: Based on the patient's level of consciousness and motor ability, the system automatically selects the most suitable training mode. This includes the following steps:

[0073] Step a: Develop a personalized, graded rehabilitation training plan in advance;

[0074] Step b: Real-time assessment of the patient's state of consciousness is achieved by collecting electroencephalogram (EEG) signals, and basic limb responsiveness is monitored by combining EMG signals to select an initial rehabilitation training program.

[0075] Step c: Apply corresponding movements to the patient according to the preliminary rehabilitation training plan. Simultaneously, use the patient's motor intention and consciousness state assessment method to collect the patient's real-time physiological signals and provide assessment feedback. The patient's motor intention and consciousness state assessment method includes the following steps:

[0076] Step 1: Physiological Signal Acquisition and Data Fusion: Multi-source physiological signals, including EEG, EMG, and IMU data, are acquired in real-time from the patient using devices such as electroencephalography (EEG), electromyography (EMG), and inertial measurement unit (IMU). These signals are processed synchronously, and each signal is connected to a comprehensive analysis platform via a hardware interface to ensure real-time data accuracy. Based on the acquired multi-source signal data, a comprehensive initial patient state model is synthesized using a data fusion algorithm. This patient state model is then input into a consciousness-motor function grading assessment and prediction model to obtain the patient's functional grading results (Grade I: Completely unconscious; Grade II: Partially conscious; Grade III: Conscious but with limited movement; Grade IV: Basically autonomous).

[0077] Step 2, Motor Intent Assessment and Dynamic Adjustment: By analyzing electromyographic signals and combining them with motion data provided by inertial sensors, the patient's motor intent is inferred and assessed in conjunction with the state of consciousness reflected by electroencephalogram (EEG) signals.

[0078] Step 3: Optimization of dynamic motion control strategy: Based on real-time assessment results, the system dynamically adjusts the exoskeleton's control strategy, adjusting the exoskeleton's movement direction, speed, and force to adapt to the patient's current state;

[0079] Step 4, Real-time Feedback Mechanism: By monitoring the patient's movement status and fatigue level in real time, feedback is provided to perceive the patient's fatigue level and optimize the exoskeleton's driving parameters to adjust the movement intensity.

[0080] Step d: Automatically optimize the output parameters of the exoskeleton controller based on the feedback of the patient's movement status and physiological response.

[0081] In this embodiment, the specific implementation method of step 2, motion intent assessment, is as follows:

[0082] The system extracts the median frequency of the electromyographic signal every 10 seconds using a fast Fourier transform. Calculate the muscle fatigue index ,according to Adjust the auxiliary torque of the exoskeleton and driving stiffness :when <1.15 o'clock, , When 1.15≤ When <1.35, , ;when When ≥1.35, , The optimized parameters are sent to the exoskeleton controller via the communication interface. The exoskeleton controller uses 2-second linear interpolation to achieve a smooth transition and adjusts the motor drive current and PID controller gain accordingly. The auxiliary torque is limited to the range of 5-50 Nm, the drive stiffness is limited to the range of 20-200 Nm / rad, and the adjustment range in a single instance does not exceed 20%.

[0083] Step d: Automatically optimize the output parameters of the exoskeleton controller based on the feedback of the patient's movement status and physiological response.

[0084] The specific implementation method for step d, which automatically optimizes the output parameters of the exoskeleton controller, is as follows:

[0085] The system calculates the joint range of motion achievement rate using inertial sensors. and trajectory deviation ,in For joint angle, These are the spatial coordinates. The system counts the cumulative number of training cycles. A parameter optimization evaluation is triggered every 10 training cycles, calculating the parameters from the past 10 cycles. and .

[0086] Parameters will be adjusted based on the evaluation results:

[0087] when and At that time, the auxiliary force gain coefficient ;

[0088] when or hour, ;

[0089] when or hour, .

[0090] Synchronous calculation of speed parameters and force parameters The optimized parameters are sent to the exoskeleton controller, which then... Adjust motor drive current ,according to and Adjust the position control gain and speed control gain.

[0091] System security settings: Single adjustment range not exceeding ±15%. The system is limited to the range of 0.15-0.95. If the system requires additional assistance after three consecutive assessments, the training will be paused and medical staff will be notified to intervene.

[0092] In the above scheme, a comprehensive initial patient state model is synthesized based on the collected multi-source signal data through a data fusion algorithm. The data fusion algorithm mentioned is a conventional algorithm in this field. Technicians can choose the appropriate algorithm as needed to implement it. It is not an innovative part of this scheme and has no impact on the implementation results of this scheme. Therefore, the specific implementation process of the data fusion algorithm will not be described in detail.

[0093] The patient's state of consciousness included complete unconsciousness and consciousness with impaired motor function.

[0094] The graded rehabilitation training program includes a passive guidance mode, a semi-active assisted mode, and an active self-management mode. Patients who are completely unconscious use the passive guidance mode for rehabilitation training. Patients who are conscious but have motor impairments initially choose the semi-active assisted mode for rehabilitation training. As the training progresses, the program automatically switches to the active self-management mode based on feedback information.

[0095] The passive-guided mode is suitable for patients with severe impairment of consciousness, with the exoskeleton providing passive movement throughout the entire process; the specific rehabilitation process is as follows:

[0096] By collecting electroencephalogram (EEG) signals to assess the patient's state of consciousness in real time, combined with electromyography (EMG) signals to monitor the patient's basic limb responsiveness, and using a passive guidance mode to perform precise limb movements, promoting joint mobility and muscle training, and adjusting the exercise intensity in a timely manner through a feedback mechanism, the patient can effectively avoid problems such as muscle atrophy and joint stiffness caused by prolonged bed rest. The feedback mechanism also ensures the safety and comfort of the training process by adjusting the exercise intensity in a timely manner.

[0097] The semi-active assist mode is suitable for patients who have regained some level of consciousness, with the exoskeleton providing appropriate assistance; the specific rehabilitation process is as follows:

[0098] The lower limb exoskeleton training unit is matched according to the patient's level, and walking aids are provided according to the patient's rehabilitation stage. Training is carried out according to the training plan, and the patient's electromyographic response is monitored in real time. The driving force and movement pattern of the exoskeleton are adjusted. As the patient's recovery progresses, the assistance force is gradually reduced to enhance the patient's independent movement ability. When the training feedback meets the target, it automatically switches to active and autonomous mode for rehabilitation training.

[0099] The active, autonomous mode is suitable for patients with strong independent movement capabilities, with the exoskeleton providing only minimal support. In each training mode, the system monitors the patient's physiological state in real time (such as electromyography, electroencephalography, and motor responses), automatically adjusting the intensity and control strategies through a feedback mechanism. In this mode, the exoskeleton not only provides necessary assistance but also monitors the patient's electromyographic responses in real time, adjusting the exoskeleton's driving force and movement patterns to ensure the patient's coordination and gait stability. As the patient's recovery progresses, the system can gradually reduce the level of assistance, enhancing the patient's autonomous movement capabilities. This tiered training mode not only improves rehabilitation efficiency but also optimizes feedback in real time during training, effectively avoiding overtraining or sports injuries, and maximizing patient compliance and rehabilitation outcomes.

[0100] Training Process Optimization: By combining the patient's recovery progress and training feedback, the system adjusts the training intensity to avoid overtraining or sports injuries while ensuring rehabilitation efficiency. Feedback data during training is continuously updated to help the system optimize future training strategies. Furthermore, the exoskeleton designed for this system has a degree of scalability, allowing for the matching of different exoskeleton components according to the patient's specific needs. The upper limb exoskeleton is used for arm rehabilitation, while the lower limb exoskeleton is used for hip and knee joint rehabilitation. This personalized selection of training modes and exoskeleton combinations ensures that patients undergo rehabilitation training at an appropriate intensity, thereby improving training effectiveness.

[0101] In summary, the key points of this invention are summarized as follows:

[0102] 1. This invention adopts a modular rehabilitation exoskeleton structure design, which supports the selection of functional units such as upper limb, knee joint, and hip joint according to the needs of different patients, meets the rehabilitation needs of different limb parts, and is compatible with multiple training postures such as lying down, sitting, standing, and walking.

[0103] 2. The exoskeleton controller of the present invention also supports multiple rehabilitation training modes, including passive guidance mode, semi-active assistance mode and active autonomous mode, and can dynamically switch training modes according to changes in the patient's condition.

[0104] 3. This invention integrates multiple signal sources such as electroencephalography (EEG), electromyography (EMG), and inertial measurement unit (IMU) to construct a real-time dynamic human state perception system. Through machine learning algorithms, it accurately identifies the patient's level of consciousness and motor function, realizes the classification of individual functional status, and provides intelligent decision-making basis for subsequent personalized rehabilitation strategies.

[0105] 4. This invention constructs a complete human-machine closed-loop control path: physiological information acquisition—intent recognition—exoskeleton execution—sensory feedback—control parameter optimization. Through the neural pathway sensing feedback mechanism, the system can sense the patient's movement state and physiological response in real time, automatically optimize the output parameters of the exoskeleton controller, and achieve safe and precise closed-loop training control.

[0106] 5. This invention targets patients with different levels of consciousness and motor impairment (such as complete unconsciousness, partial consciousness, consciousness but limited motor function, and basic autonomy). The system automatically matches exoskeleton training functional units with rehabilitation training modes to achieve personalized rehabilitation training.

[0107] It should be understood that this solution is not limited to the specific embodiments described above. Devices and structures not described in detail herein should be understood as being implemented in a manner common to the art. Any person skilled in the art can make many possible variations and modifications to this solution, or modify it into equivalent embodiments, without departing from the scope of this solution, using the methods and techniques disclosed above. This does not affect the substantive content of this solution. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this solution, without departing from its scope, still fall within the protection scope of this solution.

Claims

1. A personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders, characterized in that: The system comprises a comprehensive analysis platform, an EEG interface device, an EMG interface device, and a modular exoskeleton assembly. The EEG interface device and the modular exoskeleton assembly are wearable devices. The modular exoskeleton assembly includes a detachably connected upper limb exoskeleton, a hip exoskeleton, and a knee exoskeleton. Both the upper limb and hip exoskeletons are equipped with controllers. The EMG interface device is positioned close to the arm on the upper limb exoskeleton. The comprehensive analysis platform and controllers are connected via wired or wireless means. In use, the patient wears the required devices and connects the EEG interface device to an EEG signal acquisition device and the EMG interface device to an EMG signal acquisition device. The acquired EEG and EMG signals are sent to the comprehensive analysis platform for analysis. Based on the analysis results, control commands are sent to the controllers to control the modular exoskeleton assembly to execute corresponding personalized movement combinations, constructing a closed-loop control path of perception-assessment-execution-feedback. The system's assessment of the patient's motor intention and state of consciousness, and its closed-loop control method, includes the following steps: Step 1, Physiological Signal Acquisition and Data Fusion: Real-time acquisition of multi-source signals from the patient, including EEG signals, EMG signals, and inertial sensor data, and simultaneous processing; Based on the acquired multi-source signal data, a comprehensive initial patient state model is synthesized through a data fusion algorithm. The initial patient state model is then input into the consciousness-motor function grading assessment and prediction model to obtain the patient's functional grading results, including Level I completely unconscious, Level II partially conscious, Level III conscious but with limited movement, and Level IV basically autonomous. Step 2, assessment and dynamic adjustment of exercise intention, specifically includes: Step 2.1: Extract the root mean square amplitude of the electromyographic signal. Event-related desynchronization index (ERD) is associated with the motor cortex region of electroencephalogram (EEG) signals. Step 2.2: Based on the patient functional classification results obtained in Step 1, set the intent recognition thresholds: For patients in grades I-II, the ERD threshold is 30% of the maximum ERD value and the RMS threshold is 20% of the maximum spontaneous contraction; for patients in grade III, the ERD threshold is 50% of the maximum ERD value and the RMS threshold is 40% of the maximum spontaneous contraction; for patients in grade IV, the ERD threshold is 70% of the maximum ERD value and the RMS threshold is 60% of the maximum spontaneous contraction. Step 2.3, when ERD and Simultaneously, when the respective thresholds are exceeded, it is determined that there is a motion intention. The type of motion intention is determined by combining the joint angular velocity data provided by the inertial sensor: angular velocity <10° / s indicates static support intention, 10° / s ≤ angular velocity <50° / s indicates slow motion intention, and angular velocity ≥50° / s indicates fast motion intention. Step 3: Dynamic motion control strategy optimization: Based on the type of motion intent identified in Step 2, the system dynamically adjusts the exoskeleton's control strategy, adjusting the exoskeleton's motion direction, speed, and force to adapt to the patient's current state. Step 4, Real-time feedback mechanism and driver parameter optimization, specifically including: Step 4.1: Monitor the patient's joint range of motion and duration of movement in real time using inertial sensors, and calculate the muscle fatigue index using electromyography signals. ; Step 4.2: Based on the muscle fatigue index Adjust the auxiliary torque and driving stiffness of the exoskeleton to optimize the driving parameters; Step 4.3: Send the optimized drive parameters to the exoskeleton controller to adjust the exercise intensity in real time and prevent over-fatigue or under-training.

2. The personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders according to claim 1, characterized in that: The integrated analysis platform includes a workstation, a display screen, and an electrical cabinet. The workstation processes the received acquisition signals and issues commands. When the modular exoskeleton assembly is in fixed use, it is connected to the integrated analysis platform via a bus. When the modular exoskeleton assembly is in mobile use, it is connected to the integrated analysis platform wirelessly.

3. The personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders according to claim 2, characterized in that: The workstation includes a data processing module and an instruction output module. The data processing module includes a data preprocessing and feature extraction model, a machine learning model, and a consciousness-motor function grading assessment and prediction model. The data preprocessing and feature extraction model filters and denoises the received patient physiological signals and uses time-domain and frequency-domain feature extraction methods to obtain representative features. The preprocessed features are input into a machine learning model, which is then trained and optimized using historical data to generate a predictive model for assessing and evaluating the consciousness-motor function of patients.

4. The personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders according to claim 3, characterized in that: Based on the patient's level of consciousness and motor ability, the machine learning model classifies the patient into different functional levels and provides an assessment result for each level.

5. The personalized exoskeleton rehabilitation system based on the classification of patients with consciousness and motor disorders according to claim 4, characterized in that: It also includes walking aids to assist walking during rehabilitation training.

Images