Non-invasive brain-computer interface rehabilitation robot

Abstract

The invention discloses a non-invasive brain-computer interface rehabilitation robot, relates to the technical field of intelligent rehabilitation equipment, and solves the problems that an existing rehabilitation robot mainly adopts a set program to carry out repetitive rehabilitation training, a patient does not participate actively, and the passive training mode limits autonomous control and participation degree of the patient, so that the rehabilitation effect is poor. The rehabilitation curative effect is limited, and particularly, the effect in the aspect of nerve injury repair is poor. The method comprises the following steps: acquiring motion data of a patient; training an artificial intelligence model based on the historical motion data to obtain a motion recognition model; recognizing the motion data based on a motion recognition model to obtain a recognition result; performing rehabilitation training on the patient through a rehabilitation robot based on the recognition result; calculating a rehabilitation curative effect evaluation coefficient of the patient based on the myoelectricity characteristic data; and judging the recovery progress of the patient based on the recovery effect evaluation coefficient and a preset recovery effect evaluation threshold. The technical problem is solved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent rehabilitation equipment, and specifically is a non-invasive brain-computer interface rehabilitation robot. Background Art

[0002] Rehabilitation robots mainly passively drive patients to perform rehabilitation training, represented by Hocoma's Lokomat and Germany's LokoHelp robots; Hocoma's Lokomat provides a gait pattern that conforms to human physiological characteristics, provides training for specific tasks, and is based on continuous feedback and evaluation. It continuously increases the amount and intensity of training, improves patient participation, improves the efficiency of gait training, and greatly accelerates the patient's rehabilitation process; Germany's LokoHelp lower limb rehabilitation robot training system adjusts gait and posture according to the functional conditions of different patients, drives the patient's lower limbs to move on the treadmill, and makes large-scale gait training possible.

[0003] Existing rehabilitation robots mainly use set programs to perform repetitive rehabilitation training, and patients do not actively participate. This passive training method limits the patient's autonomous control and participation, resulting in limited rehabilitation efficacy, especially in the repair of nerve damage. At the same time, passive training will cause users to have a significant sense of pulling, affecting the use effect.

[0004] Therefore, the present invention proposes a non-invasive brain-computer interface rehabilitation robot to solve the above problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a non-invasive brain-computer interface rehabilitation robot, which is used to solve the problem that the existing rehabilitation robots mainly use set programs to perform repetitive rehabilitation training, and the patients do not actively participate. This passive training method limits the patients' autonomous control and participation, resulting in limited rehabilitation efficacy, especially poor effect in nerve damage repair. At the same time, passive training will cause the user to have an obvious pulling sensation, affecting the use effect.

[0006] To achieve the above-mentioned object, the first aspect of the present invention provides a non-invasive brain-computer interface rehabilitation robot, comprising: a data acquisition module, a data analysis module, a rehabilitation training module and a rehabilitation efficacy evaluation module;

[0007] Data acquisition module: used to obtain the patient's motion data; wherein the motion data includes: EEG characteristic data and myoelectric characteristic data;

[0008] Data analysis module: train the artificial intelligence model based on historical motion data to obtain the motion recognition model; identify the motion data based on the motion recognition model to obtain the recognition result;

[0009] Rehabilitation training module: Based on the recognition results, rehabilitation training is provided to patients through rehabilitation robots;

[0010] Rehabilitation efficacy evaluation module: The patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data; the patient's recovery progress is judged based on the rehabilitation efficacy evaluation coefficient and the preset rehabilitation efficacy evaluation threshold.

[0011] Preferably, the obtaining of the patient's motion data includes:

[0012] The EEG cap is used to collect EEG signals of the patient's brain activity area in real time when the patient is undergoing rehabilitation training with the rehabilitation robot;

[0013] The electromyographic signal of the affected limb is collected in real time through an electromyographic sensor attached to the skin surface of the motor muscle of the patient's affected limb;

[0014] By performing time domain analysis and frequency domain analysis on the EEG signal and the EMG signal, the EEG feature data corresponding to the EEG signal and the EMG feature data corresponding to the EMG signal are obtained.

[0015] It should be noted that brain activity areas refer to specific parts of the brain that are related to specific functions or tasks. For example, the motor cortex is responsible for controlling body movements, and the visual cortex processes visual information. By collecting EEG signals from these areas through an EEG cap, changes in electrical activity related to the patient's motor intentions can be captured;

[0016] EEG signals reflect the electrical activity of brain neurons. Specific brain wave patterns correspond to different cognitive and motor intentions. By analyzing these patterns, the patient's motor intentions can be decoded, thereby achieving precise control of the rehabilitation robot.

[0017] Although the muscles of the affected limbs of paralyzed patients cannot contract voluntarily, some reflex activities and autonomic nervous system activities still exist; for example, when the patient tries to move the affected limb, some reflex muscle activities will be triggered, which can be captured by electromyographic sensors and used to assist in judging the patient's movement intention;

[0018] The EEG characteristic data include peak value, root mean square, zero crossing rate, signal amplitude, signal power, etc.; the myoelectric characteristic data include peak value, root mean square, zero crossing rate, signal amplitude, signal power, etc.

[0019] Preferably, the training of the artificial intelligence model based on historical motion data includes:

[0020] A1: Extract historical EEG feature data and historical EMG feature data from historical motion data;

[0021] A2: Obtaining the mapping relationship between historical EEG feature data and movement intention data and the mapping relationship between historical myoelectric feature data and movement intention data from the database; wherein the movement intention data includes: forward, backward, sprint and jog, etc.;

[0022] A3: Train an artificial intelligence model based on historical EEG feature data to obtain a movement intention recognition model; train an artificial intelligence model based on historical electromyography feature data to obtain a movement intention verification model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network;

[0023] A4: A motion recognition model is composed based on the motion intention recognition model and the motion intention verification model.

[0024] It should be noted that the historical motion data refers to the historical data of the collected motion data;

[0025] The database refers to existing public data sets on the mapping relationship between EEG feature data and movement intention data, as well as the mapping relationship between electromyography feature data and movement intention data, including: BCI Competition Datasets, PhysioNet and OpenBMI, etc.

[0026] Preferably, the training of an artificial intelligence model based on historical EEG feature data to obtain a movement intention recognition model includes:

[0027] Integrate historical EEG feature data as model input data, and integrate movement intention data corresponding to the historical EEG feature data as model output data;

[0028] The artificial intelligence model is trained based on the model input data and model output data to obtain a motion intention recognition model.

[0029] Preferably, the training of the artificial intelligence model based on the historical electromyographic feature data to obtain the movement intention verification model includes:

[0030] Integrate historical electromyographic feature data into standard input data, and integrate movement intention data corresponding to the historical electromyographic feature data into standard output data;

[0031] The artificial intelligence model is trained based on standard input data and standard output data to obtain a motion intention verification model.

[0032] Preferably, the identifying the motion data based on the motion identification model includes:

[0033] B1: Extracting real-time EEG feature data and EMG feature data;

[0034] B2: Input the EEG feature data into the movement intention recognition model to obtain the recognition result;

[0035] B3: Input the electromyographic feature data into the movement intention verification model to obtain the verification recognition result;

[0036] B4: Determine whether the verification recognition result is consistent with the recognition result; if yes, output the recognition result; if no, return to step B1 until the verification recognition result is consistent with the recognition result.

[0037] It should be noted that the recognition result, the verification recognition result and the motion intention data content attributes are consistent, that is, the recognition result and the verification recognition result both include: forward, backward, sprint and jog, etc.

[0038] Preferably, the patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data, including:

[0039] Obtain the duration of the patient's rehabilitation training, and extract the root mean square value and zero crossing rate from the electromyographic feature data;

[0040] The duration of rehabilitation training is marked as T, the root mean square value is marked as R, and the zero crossing rate is marked as L;

[0041] By formula:

[0042] P=α / e^[(T-ZT)^2 / ZT]+β×e^[(R-CR) / ln(CR)]+ɡ×e^[(L-CL) / ln(CL)];

[0043] The rehabilitation efficacy evaluation coefficient is calculated; where P is the rehabilitation efficacy evaluation coefficient, α, β, ɡ are proportional coefficients, and 0<α<1, 0<β<1, 0<ɡ<1, ZT is the optimal rehabilitation training duration, CR is the initial root mean square value, and CL is the initial zero crossing rate.

[0044] It should be noted that the optimal duration of rehabilitation training is set by experts in this field based on experience according to the actual condition of the patient; the initial root mean square value and initial zero crossing rate refer to the root mean square value and zero crossing rate of the electromyographic signal of the patient before rehabilitation training;

[0045] Training time that is too short may result in the patient's muscle endurance and overall physical fitness not being effectively improved; training time that is too long may lead to excessive muscle fatigue and increase the risk of injury. Long-term fatigue may also affect the muscle's ability to recover and regenerate.

[0046] The increase in the RMS value relative to the initial RMS value indicates that the intensity of muscle activation has increased, which is a manifestation of muscle strength recovery. In rehabilitation training, this means that the patient's muscle strength and activation ability are improving;

[0047] The increase in zero-crossing rate relative to the initial zero-crossing rate indicates that the frequency of muscle activation has increased, which is a manifestation of the recovery of neural control ability. In rehabilitation training, this shows that the patient's muscle activation frequency and coordination are improving.

[0048] Preferably, judging the patient's recovery progress based on the rehabilitation efficacy evaluation coefficient and a preset rehabilitation efficacy evaluation threshold includes:

[0049] Determine whether the rehabilitation efficacy evaluation coefficient is greater than or equal to the preset rehabilitation efficacy evaluation threshold; if yes, it means that the patient's rehabilitation efficacy is good; if no, it means that the patient's rehabilitation efficacy is poor.

[0050] It should be noted that the preset rehabilitation efficacy evaluation threshold is set by experts in this field based on experience according to the patient's initial condition.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] Existing rehabilitation robots mainly use set programs to carry out repetitive rehabilitation training, and patients are not actively involved. This passive training method limits the patient's autonomous control and participation, resulting in limited rehabilitation efficacy, especially poor effect in nerve injury repair. At the same time, passive training will cause users to have obvious pulling sensations, affecting the use effect. The present invention obtains the patient's EEG feature data and myoelectric feature data in real time through the data acquisition module, and uses the data analysis module to train the artificial intelligence model to obtain the movement intention recognition model and verification model, thereby realizing accurate recognition and verification of the patient's movement intention. The rehabilitation training module performs personalized rehabilitation training on the patient according to the recognition results, and the rehabilitation efficacy evaluation module calculates the rehabilitation efficacy evaluation coefficient and combines the preset threshold to judge the patient's recovery progress. This method not only improves the patient's active participation and rehabilitation effect, but also ensures the synchronization and coordination of human-computer interaction through real-time feedback and adaptive control algorithms, significantly improving the user's experience and the effect of rehabilitation training. The present invention solves the problem that the existing rehabilitation robot mainly uses set programs for repetitive passive training by introducing non-invasive brain-computer interface technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 A schematic diagram of a rehabilitation robot module according to an embodiment of the present invention;

[0055] Figure 2 Schematic diagram of the working process of the rehabilitation robot according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1-Figure 2 , the first aspect of the present invention provides a non-invasive brain-computer interface rehabilitation robot, including: a data acquisition module, a data analysis module, a rehabilitation training module and a rehabilitation efficacy evaluation module;

[0058] Data acquisition module: used to obtain the patient's motion data; wherein the motion data includes: EEG characteristic data and myoelectric characteristic data;

[0059] Data analysis module: train the artificial intelligence model based on historical motion data to obtain the motion recognition model; identify the motion data based on the motion recognition model to obtain the recognition result;

[0060] Rehabilitation training module: Based on the recognition results, rehabilitation training is provided to patients through rehabilitation robots;

[0061] Rehabilitation efficacy evaluation module: The patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data; the patient's recovery progress is judged based on the rehabilitation efficacy evaluation coefficient and the preset rehabilitation efficacy evaluation threshold.

[0062] Obtain patient motion data, including:

[0063] The EEG cap is used to collect EEG signals of the patient's brain activity area in real time when the patient is undergoing rehabilitation training with the rehabilitation robot;

[0064] The electromyographic signal of the affected limb is collected in real time through an electromyographic sensor attached to the skin surface of the motor muscle of the patient's affected limb;

[0065] By performing time domain analysis and frequency domain analysis on the EEG signal and the EMG signal, the EEG feature data corresponding to the EEG signal and the EMG feature data corresponding to the EMG signal are obtained.

[0066] Training AI models based on historical sports data, including:

[0067] A1: Extract historical EEG feature data and historical EMG feature data from historical motion data;

[0068] A2: Obtaining the mapping relationship between historical EEG feature data and movement intention data and the mapping relationship between historical myoelectric feature data and movement intention data from the database; wherein the movement intention data includes: forward, backward, sprint and jog, etc.;

[0069] A3: Train an artificial intelligence model based on historical EEG feature data to obtain a movement intention recognition model; train an artificial intelligence model based on historical electromyography feature data to obtain a movement intention verification model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network;

[0070] A4: A motion recognition model is composed based on the motion intention recognition model and the motion intention verification model.

[0071] Based on the historical EEG feature data, the artificial intelligence model is trained to obtain the movement intention recognition model, including:

[0072] Integrate historical EEG feature data as model input data, and integrate movement intention data corresponding to the historical EEG feature data as model output data;

[0073] The artificial intelligence model is trained based on the model input data and model output data to obtain a motion intention recognition model.

[0074] The artificial intelligence model is trained based on historical electromyographic feature data to obtain a movement intention verification model, including:

[0075] Integrate historical electromyographic feature data into standard input data, and integrate movement intention data corresponding to the historical electromyographic feature data into standard output data;

[0076] The artificial intelligence model is trained based on standard input data and standard output data to obtain a motion intention verification model.

[0077] Identify motion data based on the motion recognition model, including:

[0078] B1: Extracting real-time EEG feature data and EMG feature data;

[0079] B2: Input the EEG feature data into the movement intention recognition model to obtain the recognition result;

[0080] B3: Input the electromyographic feature data into the movement intention verification model to obtain the verification recognition result;

[0081] B4: Determine whether the verification recognition result is consistent with the recognition result; if yes, output the recognition result; if no, return to step B1 until the verification recognition result is consistent with the recognition result.

[0082] The patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data, including:

[0083] Obtain the duration of the patient's rehabilitation training, and extract the root mean square value and zero crossing rate from the electromyographic feature data;

[0084] The duration of rehabilitation training is marked as T, the root mean square value is marked as R, and the zero crossing rate is marked as L;

[0085] By formula:

[0086] P=α / e^[(T-ZT)^2 / ZT]+β×e^[(R-CR) / ln(CR)]+ɡ×e^[(L-CL) / ln(CL)];

[0087] The rehabilitation efficacy evaluation coefficient is calculated; where P is the rehabilitation efficacy evaluation coefficient, α, β, ɡ are proportional coefficients, and 0<α<1, 0<β<1, 0<ɡ<1, ZT is the optimal rehabilitation training duration, CR is the initial root mean square value, and CL is the initial zero crossing rate.

[0088] The patient's recovery progress is judged based on the rehabilitation efficacy evaluation coefficient and the preset rehabilitation efficacy evaluation threshold, including:

[0089] Determine whether the rehabilitation efficacy evaluation coefficient is greater than or equal to the preset rehabilitation efficacy evaluation threshold; if yes, it means that the patient's rehabilitation efficacy is good; if no, it means that the patient's rehabilitation efficacy is poor.

[0090] For example, suppose there is a patient who needs to use a non-invasive brain-computer interface rehabilitation robot for rehabilitation training; the details are as follows:

[0091] Basic information of the patient:

[0092] Initial condition: paralysis of lower limbs due to stroke;

[0093] Initial RMS value (CR): 50μV;

[0094] Initial zero crossing rate (CL): 10 times / second;

[0095] The patient stands or sits on the exoskeleton robot, wears an EEG cap and affixes an EMG sensor. After the device is powered on, the display terminal displays prompts and icons, and the patient starts and stops the device by looking at it with brain activity.

[0096] Data acquisition module;

[0097] 1. EEG signal collection:

[0098] Use an EEG cap to collect the patient's EEG signals in real time during rehabilitation training, focusing on the EEG activity of the motor cortex.

[0099] The collected EEG signals are analyzed in the time domain and frequency domain to obtain EEG characteristic data, including peak value, root mean square, zero crossing rate, signal amplitude and signal power.

[0100] 2. EMG signal collection:

[0101] Electromyographic sensors are attached to the quadriceps and gastrocnemius muscles of the patient's lower limbs to collect electromyographic signals in real time.

[0102] The collected electromyographic signals are analyzed in the time domain and frequency domain to obtain electromyographic characteristic data, including peak value, root mean square, zero crossing rate, signal amplitude and signal power.

[0103] Data analysis module;

[0104] 1. Training AI models:

[0105] Historical data extraction: extract historical EEG feature data and historical EMG feature data, as well as corresponding movement intention data (such as forward, backward, sprint and jog, etc.) from the database.

[0106] Model training:

[0107] Movement intention recognition model: The convolutional neural network (CNN) is trained using historical EEG feature data and corresponding movement intention data to obtain a movement intention recognition model.

[0108] Movement intention verification model: Use historical electromyographic feature data and corresponding movement intention data to train a deep belief network (DBN) to obtain a movement intention verification model.

[0109] Combined model: The motion intention recognition model and the motion intention verification model are combined into a motion recognition model.

[0110] 2. Real-time data recognition:

[0111] Data extraction: Extract features from the EEG feature data and EMG feature data collected in real time.

[0112] Model input:

[0113] The EEG feature data is input into the movement intention recognition model to obtain the recognition result.

[0114] The electromyographic feature data is input into the movement intention verification model to obtain the verification recognition result.

[0115] Result judgment: judge whether the recognition result and the verification recognition result are consistent. If they are consistent, output the recognition result; if they are inconsistent, re-extract data for judgment until the recognition result and the verification recognition result are consistent.

[0116] Rehabilitation training module;

[0117] Training duration: Patients underwent 45 minutes of rehabilitation training.

[0118] Recognition results: The movement intention recognition model identified the patient's movement intention as "forward", and the verification recognition model also confirmed this result.

[0119] The "forward" command is sent to the control terminal of the rehabilitation robot to control the robot to perform "forward" training.

[0120] Rehabilitation efficacy evaluation module;

[0121] 1. Data extraction:

[0122] Rehabilitation training time (T): 45 minutes;

[0123] RMS value (R): 60μV;

[0124] Zero crossing rate (L): 15 times / second;

[0125] 2. Calculate the rehabilitation efficacy evaluation coefficient:

[0126] Assumed optimal rehabilitation training time (ZT): 40 minutes;

[0127] Assume proportionality coefficients: α = 0.4, β = 0.3, ɡ = 0.3;

[0128] By formula:

[0129] P=α / e^[(T-ZT)^2 / ZT]+β×e^[(R-CR) / ln(CR)]+ɡ×e^[(L-CL) / ln(CL)];

[0130] Calculate the rehabilitation efficacy evaluation coefficient;

[0131] P=0.4 / e^[(45-40)^2 / 40]+0.3×e^[(60-50) / ln(50)]+0.3×e^[(15-10) / ln(10)]≈6.7;

[0132] 3. Evaluation results:

[0133] Assume that the preset rehabilitation efficacy assessment threshold is 6;

[0134] Since the rehabilitation efficacy evaluation coefficient (6.7) is greater than the preset rehabilitation efficacy evaluation threshold (6), it indicates that the patient's rehabilitation efficacy is good.

[0135] If the judgment result is that the rehabilitation effect is poor, appropriate measures will be taken to improve it, such as:

[0136] Adjust the intensity and time of training, appropriately extend or shorten the training time, find the optimal time for rehabilitation training, and appropriately adjust the training intensity according to the patient's physical fitness and endurance;

[0137] Increase the diversity of training by introducing different types of training tasks, such as walking at different speeds, going up and down stairs, etc., to increase the patient's interest and participation, and develop a personalized training plan to ensure that the training content matches the patient's actual situation;

[0138] Provide psychological support and motivation, help patients build confidence and overcome psychological barriers through psychological counseling, and establish a reward mechanism to encourage patients to actively participate in rehabilitation training and improve their enthusiasm and initiative in training;

[0139] Through these measures, the rehabilitation effect of patients can be effectively improved and the effect of rehabilitation training can be enhanced.

[0140] Through the above examples, it can be seen that the present invention has significant benefits in evaluating and improving the rehabilitation efficacy of patients through the non-invasive brain-computer interface rehabilitation robot; first, by real-time acquisition and analysis of the patient's EEG and EMG signals, it can accurately identify the patient's movement intentions and provide personalized rehabilitation training, thereby improving the patient's active participation and training effect; secondly, by calculating the rehabilitation efficacy evaluation coefficient, the system can scientifically evaluate the patient's rehabilitation progress, promptly discover the situation of poor rehabilitation efficacy, and take corresponding improvement measures, such as adjusting training intensity and time, increasing training diversity, providing psychological support and motivation, strengthening rehabilitation evaluation and technical optimization, etc.; this comprehensive rehabilitation training method not only improves the patient's rehabilitation effect, but also provides rehabilitation doctors and therapists with a scientific basis for evaluation and adjustment, and promotes the efficiency and personalized development of rehabilitation training.

[0141] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0142] Working principle of the present invention:

[0143] The present invention obtains the patient's motion data; trains an artificial intelligence model based on historical motion data to obtain a motion recognition model; recognizes the motion data based on the motion recognition model to obtain a recognition result; performs rehabilitation training on the patient through a rehabilitation robot based on the recognition result; calculates the patient's rehabilitation efficacy evaluation coefficient based on electromyographic characteristic data; and judges the patient's recovery progress based on the rehabilitation efficacy evaluation coefficient and a preset rehabilitation efficacy evaluation threshold. This solves the technical problem that existing rehabilitation robots mainly use set programs to perform repetitive rehabilitation training, and patients do not actively participate. This passive training method limits the patient's autonomous control and participation, resulting in limited rehabilitation efficacy, especially poor results in nerve injury repair.

[0144] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A non-invasive brain-computer interface rehabilitation robot, characterized in that: Data collection module, data analysis module, rehabilitation training module and rehabilitation efficacy evaluation module; Data acquisition module: used to obtain the patient's motion data; wherein the motion data includes: EEG characteristic data and myoelectric characteristic data; Data analysis module: train the artificial intelligence model based on historical motion data to obtain the motion recognition model; identify the motion data based on the motion recognition model to obtain the recognition result; Rehabilitation training module: Based on the recognition results, rehabilitation training is provided to patients through rehabilitation robots; Rehabilitation efficacy evaluation module: The patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data; the patient's recovery progress is judged based on the rehabilitation efficacy evaluation coefficient and the preset rehabilitation efficacy evaluation threshold.

2. A non-invasive brain-computer interface rehabilitation robot according to claim 1, characterized in that: The step of obtaining the patient's motion data comprises: The EEG cap is used to collect EEG signals of the patient's brain activity area in real time when the patient is undergoing rehabilitation training with the rehabilitation robot; The electromyographic signal of the affected limb is collected in real time through an electromyographic sensor attached to the skin surface of the motor muscle of the patient's affected limb; By performing time domain analysis and frequency domain analysis on the EEG signal and the EMG signal, the EEG feature data corresponding to the EEG signal and the EMG feature data corresponding to the EMG signal are obtained.

3. A non-invasive brain-computer interface rehabilitation robot according to claim 1, characterized in that: The training of the artificial intelligence model based on historical motion data includes: A1: Extract historical EEG feature data and historical EMG feature data from historical motion data; A2: by obtaining the mapping relationship between historical EEG feature data and movement intention data and the mapping relationship between historical EMG feature data and movement intention data from the database; A3: Train an artificial intelligence model based on historical EEG feature data to obtain a movement intention recognition model; train an artificial intelligence model based on historical electromyography feature data to obtain a movement intention verification model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network; A4: A motion recognition model is composed based on the motion intention recognition model and the motion intention verification model.

4. A non-invasive brain-computer interface rehabilitation robot according to claim 3, characterized in that: The method of training an artificial intelligence model based on historical EEG feature data to obtain a movement intention recognition model includes: Integrate historical EEG feature data as model input data, and integrate movement intention data corresponding to the historical EEG feature data as model output data; The artificial intelligence model is trained based on the model input data and model output data to obtain a motion intention recognition model.

5. The non-invasive brain-computer interface rehabilitation robot according to claim 3, characterized in that: The artificial intelligence model is trained based on the historical electromyographic feature data to obtain the movement intention verification model, including: Integrate historical electromyographic feature data into standard input data, and integrate movement intention data corresponding to the historical electromyographic feature data into standard output data; The artificial intelligence model is trained based on standard input data and standard output data to obtain a motion intention verification model.

6. The non-invasive brain-computer interface rehabilitation robot according to claim 1, characterized in that: The identifying the motion data based on the motion identification model includes: B1: Extracting real-time EEG feature data and EMG feature data; B2: Input the EEG feature data into the movement intention recognition model to obtain the recognition result; B3: Input the electromyographic feature data into the movement intention verification model to obtain the verification recognition result; B4: Determine whether the verification recognition result is consistent with the recognition result; if yes, output the recognition result; if not, return to step B1 until the verification recognition result is consistent with the recognition result.

7. The non-invasive brain-computer interface rehabilitation robot according to claim 1, characterized in that: The patient's rehabilitation efficacy evaluation coefficient is calculated based on the electromyographic characteristic data, including: Obtain the duration of the patient's rehabilitation training, and extract the root mean square value and zero crossing rate from the electromyographic feature data; The duration of rehabilitation training is marked as T, the root mean square value is marked as R, and the zero crossing rate is marked as L; By formula: P=α / e^[(T-ZT)^2 / ZT]+β×e^[(R-CR) / ln(CR)]+ɡ×e^[(L-CL) / ln(CL)]; The rehabilitation efficacy evaluation coefficient is calculated; where P is the rehabilitation efficacy evaluation coefficient, α, β, ɡ are proportional coefficients, and 0<α<1, 0<β<1, 0<ɡ<1, ZT is the optimal rehabilitation training duration, CR is the initial root mean square value, and CL is the initial zero crossing rate.

8. The non-invasive brain-computer interface rehabilitation robot according to claim 1, characterized in that: The step of judging the patient's recovery progress based on the rehabilitation efficacy evaluation coefficient and the preset rehabilitation efficacy evaluation threshold comprises: Determine whether the rehabilitation efficacy evaluation coefficient is greater than or equal to the preset rehabilitation efficacy evaluation threshold; if yes, it means that the patient's rehabilitation efficacy is good; if no, it means that the patient's rehabilitation efficacy is poor.

Images