Rehabilitation training method and device for stroke patient

Abstract

The embodiment of the invention provides a rehabilitation training method and device for a stroke patient, and is applied to the technical field of rehabilitation training. The method comprises the following steps: acquiring hobby data, mood data, physical sign data, illness state data, rehabilitation data and weather data at the current moment of a stroke patient; determining a virtual reality scene for rehabilitation training according to the hobby data, the mood data and the weather data; determining a personalized rehabilitation training task to be executed in the virtual reality scene according to the physical sign data, the illness state data and the rehabilitation data; and guiding the stroke patient to execute the personalized rehabilitation training task in the virtual reality scene by using a task guiding model corresponding to the personalized rehabilitation training task. In this way, personalized rehabilitation training tasks can be carried out according to actual conditions of the stroke patients, real-time guidance can be carried out when the tasks are executed, and the task completion degree is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation training, and in particular to a method and a device for rehabilitation training of stroke patients. Background Art

[0002] Stroke, also known as acute cerebrovascular accident (CVA), refers to an acute clinical syndrome of partial or complete brain function impairment caused by cerebrovascular disease. About 75% of patients have functional impairments of different types and degrees after a stroke, which makes them lose their basic ability to work and live. Good walking ability is one of the necessary abilities to carry out daily life activities and fully participate in community activities. Improving walking ability is one of the main goals of rehabilitation training for stroke patients. A suitable training environment is an important condition for improving the effect of rehabilitation training for walking ability.

[0003] At present, rehabilitation training for stroke patients is usually based on a virtual reality treadmill (VRT) to provide stroke patients with a close-to-reality environment, so that stroke patients can interact with objects and events in real time, and can simulate walking conditions in a real environment (including speed changes, uphill and downhill, disturbances, etc.). Under the protection of a weight belt, they can safely perform walking training that simulates daily life activities and participation in community activities. However, at present, this program can only assist stroke patients in carrying out general rehabilitation training tasks, and cannot carry out personalized rehabilitation training tasks based on the actual situation of stroke patients themselves. Summary of the invention

[0004] In a first aspect, an embodiment of the present invention provides a method for rehabilitation training of stroke patients, the method comprising:

[0005] Obtain the hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data of stroke patients;

[0006] Determine the virtual reality scene for rehabilitation training based on hobby data, mood data, and weather data;

[0007] Determine the personalized rehabilitation training tasks to be performed in the virtual reality scenario based on the physical sign data, disease data, and rehabilitation data;

[0008] The task guidance model corresponding to the personalized rehabilitation training tasks is used to guide stroke patients to perform personalized rehabilitation training tasks in a virtual reality scenario.

[0009] In some implementations of the first aspect, determining a personalized rehabilitation training task to be performed in a virtual reality scenario according to the physical sign data, the condition data, and the rehabilitation data includes:

[0010] According to the physical sign data, the condition data and the rehabilitation data, the corresponding rehabilitation training tasks are matched from the pre-built rehabilitation training task library, and the degree of fit between the matched rehabilitation training tasks and the virtual reality scene is calculated, and the rehabilitation training task with the greatest degree of fit is selected from the matched rehabilitation training tasks as the personalized rehabilitation training task to be performed in the virtual reality scene; or,

[0011] The virtual reality scene description data, physical sign data, disease condition data and rehabilitation data are input into the rehabilitation training large model, which processes the virtual reality scene description data, physical sign data, disease condition data and rehabilitation data to determine the personalized rehabilitation training tasks to be performed in the virtual reality scene.

[0012] In some implementations of the first aspect, the rehabilitation training large model is obtained by the following steps:

[0013] By using various data from the fields of virtual reality and rehabilitation training, the universal large model base is fine-tuned to obtain a large rehabilitation training model.

[0014] In some implementations of the first aspect, a task guidance model corresponding to the personalized rehabilitation training task is used to guide a stroke patient to perform the personalized rehabilitation training task in a virtual reality scene, including:

[0015] During the execution of personalized rehabilitation training tasks, the joint activity signal flow and electromyographic signal flow of the stroke patient in the current action cycle are obtained;

[0016] Generate a joint real-time state model according to the joint activity signal flow, compare it with the standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task, and display the corrected image in the virtual reality scene based on the comparison result;

[0017] The electromyographic signal stream is segmented based on a sliding window, and the signal indicators corresponding to each segment are extracted. The signal indicators corresponding to each segment are input into the electromyographic signal stream prediction model in the task guidance model corresponding to the personalized rehabilitation training task. The electromyographic signal stream prediction model processes the signal indicators corresponding to each segment, and predicts the standard electromyographic signal stream in the next action cycle. According to the difference between the predicted standard electromyographic signal stream in the next action cycle and the actual electromyographic signal stream in the next action cycle, the muscles of the stroke patient are electrically stimulated.

[0018] In some implementations of the first aspect, the standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task is obtained by the following steps:

[0019] When healthy subjects perform personalized rehabilitation training tasks, the joint activity signal streams of the healthy subjects in each action cycle are obtained, and a standard joint real-time state model corresponding to the personalized rehabilitation training tasks is generated according to the joint activity signal streams in each action cycle.

[0020] In some implementations of the first aspect, the electromyographic signal flow prediction model in the task guidance model corresponding to the personalized rehabilitation training task is obtained by the following steps:

[0021] During the period when healthy subjects perform personalized rehabilitation training tasks, the electromyographic signal flow of the healthy subjects in each action cycle is obtained, the electromyographic signal flow is segmented based on the sliding window, and the signal indicators corresponding to each segment are extracted. The signal indicators corresponding to each segment are used as the sample body, and the electromyographic signal flow of the next action cycle is used as the sample label to generate samples. The data set including each sample is used to train the long short-term memory neural network to obtain the electromyographic signal flow prediction model corresponding to the personalized rehabilitation training task.

[0022] In some implementations of the first aspect, obtaining a joint activity signal flow and an electromyographic signal flow of a stroke patient in a current action cycle includes:

[0023] The inertial posture sensor deployed on the body of the stroke patient collects the inertial posture data of the stroke patient in the current motion cycle, thereby generating a joint activity signal stream of the stroke patient in the current motion cycle;

[0024] The surface electromyographic signal sensor deployed on the body of the stroke patient collects the electromyographic signal data of the stroke patient in the current movement cycle, thereby generating the electromyographic signal stream of the stroke patient in the current movement cycle.

[0025] In some implementable embodiments of the first aspect, inertial posture sensors are deployed on the outside of the shoulder, outside of the elbow, palm, front and back of the pelvis, outside of the thigh, and outside of the calf of the stroke patient; surface electromyography signal sensors are deployed on the deltoid muscle, triceps brachii, biceps brachii, wrist dorsiflexor muscle, erector spinae, gluteus maximus, gluteus medius, quadriceps femoris, biceps femoris, tibialis anterior, and gastrocnemius of the stroke patient.

[0026] In some implementations of the first aspect, the method further includes:

[0027] Calculate muscle activation and / or fatigue based on electromyographic signal data;

[0028] A real-time joint status model is displayed in a virtual reality scene, and the activation degree and / or fatigue degree of the corresponding muscles are displayed on the real-time joint status model.

[0029] In a second aspect, an embodiment of the present invention provides a rehabilitation training device for stroke patients, the device comprising:

[0030] The acquisition module is used to obtain the hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data of stroke patients;

[0031] A first determination module is used to determine a virtual reality scene for rehabilitation training based on the hobby data, mood data and weather data;

[0032] The second determination module is used to determine the personalized rehabilitation training task to be performed in the virtual reality scene according to the physical sign data, the condition data and the rehabilitation data;

[0033] The guidance module is used to guide stroke patients to perform personalized rehabilitation training tasks in a virtual reality scenario by using a task guidance model corresponding to the personalized rehabilitation training tasks.

[0034] In the embodiment of the present invention, personalized rehabilitation training tasks can be carried out according to the actual situation of the stroke patient himself, and real-time guidance can be provided during the execution of the task to improve the degree of task completion.

[0035] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and other features, advantages and aspects of the embodiments of the present invention will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are provided for a better understanding of the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0037] Figure 1 A flowchart of a stroke patient rehabilitation training method provided by an embodiment of the present invention is shown;

[0038] Figure 2 A structural diagram of a rehabilitation training device for stroke patients provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not 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.

[0040] In addition, the term "and / or" in the present invention is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.

[0041] In order to solve the technical problems in the background technology, the embodiment of the present invention provides a method and device for rehabilitation training of stroke patients. Specifically, the hobby data, mood data, physical sign data, disease data, rehabilitation data and weather data of the stroke patient are obtained; the virtual reality scene for rehabilitation training is determined according to the hobby data, mood data and weather data; the personalized rehabilitation training task to be performed in the virtual reality scene is determined according to the physical sign data, disease data and rehabilitation data; the task guidance model corresponding to the personalized rehabilitation training task is used to guide the stroke patient to perform the personalized rehabilitation training task in the virtual reality scene.

[0042] In this way, personalized rehabilitation training tasks can be carried out according to the actual situation of stroke patients, and real-time guidance can be provided during the execution of the tasks to improve the completion rate of the tasks.

[0043] A method and device for rehabilitation training of stroke patients provided by the embodiments of the present invention will be described in detail below with reference to the accompanying drawings through specific embodiments.

[0044] Figure 1 A flowchart of a method for rehabilitation training of stroke patients provided by an embodiment of the present invention is shown. Figure 1 As shown, the stroke patient rehabilitation training method 100 may include the following steps:

[0045] S110, obtaining the stroke patient's hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data.

[0046] In some embodiments, the hobby data of stroke patients can be extracted by capturing their social platform data; the mood data of stroke patients can be obtained through smart wearable devices worn by stroke patients; and the vital sign data, condition data, and rehabilitation data of stroke patients can be obtained through the cloud server of the rehabilitation center.

[0047] S120, determining a virtual reality scene for rehabilitation training according to the hobby data, mood data, and weather data.

[0048] In some embodiments, corresponding virtual reality scene materials (such as objects, weather, light, sound, etc.) can be selected from a virtual reality scene material library according to hobby data, mood data, and weather data, and then a virtual reality scene can be constructed based on the selected virtual reality scene materials.

[0049] S130, determining a personalized rehabilitation training task to be performed in a virtual reality scenario based on the physical sign data, the disease condition data, and the rehabilitation data.

[0050] In some embodiments, corresponding rehabilitation training tasks can be matched from a pre-built rehabilitation training task library based on physical sign data, disease data, and rehabilitation data, and the degree of compatibility between the matched rehabilitation training tasks and the virtual reality scene is calculated. Then, the rehabilitation training task with the greatest degree of compatibility is selected from the matched rehabilitation training tasks as the personalized rehabilitation training task to be executed in the virtual reality scene.

[0051] In addition, the virtual reality scene description data, physical sign data, disease data and rehabilitation data can also be input into the rehabilitation training model, which processes the virtual reality scene description data, physical sign data, disease data and rehabilitation data to determine the personalized rehabilitation training tasks to be performed in the virtual reality scene.

[0052] It is worth noting that the above rehabilitation training model can be obtained through the following steps:

[0053] By using various data from the fields of virtual reality and rehabilitation training, the universal large model base is fine-tuned to obtain a large rehabilitation training model.

[0054] S140, using a task guidance model corresponding to the personalized rehabilitation training task to guide the stroke patient to perform the personalized rehabilitation training task in a virtual reality scenario.

[0055] In some embodiments, during the execution of the personalized rehabilitation training task, the joint activity signal flow and the electromyographic signal flow of the stroke patient in the current action cycle can be obtained. A joint real-time state model is generated according to the joint activity signal flow, and it is compared with the standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task. Based on the comparison result, the corrected image is displayed in the virtual reality scene to guide the stroke patient to complete the personalized rehabilitation training task. At the same time, based on the sliding window, the electromyographic signal stream is segmented, and the signal indicators corresponding to each segment are extracted (such as time domain features such as mean, standard deviation, root mean square, maximum value and minimum value; frequency domain features such as spectrum energy and main frequency component), and the signal indicators corresponding to each segment are input into the electromyographic signal stream prediction model in the task guidance model corresponding to the personalized rehabilitation training task, which processes the signal indicators corresponding to each segment, predicts the standard electromyographic signal stream in the next action cycle, and calculates the difference between the predicted standard electromyographic signal stream in the next action cycle and the actual electromyographic signal stream in the next action cycle. If the difference is greater than or equal to the preset threshold, the preset numerical interval where the difference is located is determined, and according to the stimulation parameters corresponding to the preset numerical interval, the muscles of the stroke patient are electrically stimulated in a subsequent action cycle to guide the stroke patient to complete the personalized rehabilitation training task. Obviously, here the stroke patient is first prompted to make active corrections according to the correction image, and then passive corrections are made through electrical stimulation, so that the user can be effectively guided to complete the personalized rehabilitation training task through double correction.

[0056] As an example, here, the inertial posture data of the stroke patient in the current motion cycle can be collected by an inertial posture sensor deployed on the body of the stroke patient, and the inertial posture data can be filtered using a Butterworth high-pass filter and a moving average filter to generate a joint activity signal stream of the stroke patient in the current motion cycle; the electromyography signal data of the stroke patient in the current motion cycle can be collected by a surface electromyography signal sensor deployed on the body of the stroke patient, and the electromyography signal data can be filtered using a Butterworth high-pass filter and a band-stop filter to generate an electromyography signal stream of the stroke patient in the current motion cycle.

[0057] Exemplarily, inertial posture sensors are deployed on the outside of the shoulder, outside of the elbow, palm, front and back of the pelvis, outside of the thigh, and outside of the calf of the stroke patient; surface electromyographic signal sensors are deployed on the deltoid, triceps, biceps, wrist dorsiflexor, erector spinae, gluteus maximus, gluteus medius, quadriceps, biceps femoris, tibialis anterior, and gastrocnemius of the stroke patient. Optionally, the joint activity signal streams described herein are the joint activity signal streams corresponding to the joints at the above positions; the electromyographic signal streams described herein are the electromyographic signal streams corresponding to the above muscles.

[0058] The difference equation of the Butterworth high-pass filter can be:

[0059] y[n]=b0·x[n]+b1·x[n―1]+b2·x[n―2]―a1·y[n―1]―a2·y[n―2] (1)

[0060] Where x[n] is the input data; y[n] is the output data; b0, b1, b2, a1, a2 are the coefficients of the Butterworth high-pass filter, which can be calculated using the Butterworth high-pass filter design formula.

[0061] The working principle of the moving average filter can be expressed by the following formula:

[0062]

[0063] Where x[n] is the input data; y[n] is the output data; and N is the window size of the moving average filter.

[0064] The difference equation for the band-stop filter can be expressed as:

[0065]

[0066] In the formula, x[n] is the input data, y[n] is the output data, and f c is the frequency to be removed, f s is the sampling frequency of the data.

[0067] It is worth noting that the standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task can be obtained through the following steps:

[0068] When healthy subjects perform personalized rehabilitation training tasks, the joint activity signal streams of the healthy subjects in each action cycle are obtained, and a standard joint real-time state model corresponding to the personalized rehabilitation training tasks is generated according to the joint activity signal streams in each action cycle.

[0069] The electromyographic signal flow prediction model in the task guidance model corresponding to the personalized rehabilitation training task can be obtained by the following steps:

[0070] During the period when healthy subjects perform personalized rehabilitation training tasks, the electromyographic signal flow of healthy subjects in each action cycle is obtained, the electromyographic signal flow is segmented based on a sliding window, the signal index corresponding to each segment is extracted, the signal index corresponding to each segment is used as the sample body, and the electromyographic signal flow of the next action cycle is used as the sample label, so as to generate samples, and the long short-term memory neural network is trained using the data set including each sample to obtain the electromyographic signal flow prediction model corresponding to the personalized rehabilitation training task. Optionally, the data set is divided into a training set (70%), a validation set (15%) and a test set (15%), the training set is used to train the long short-term memory neural network, and then the validation set is used to evaluate the trained long short-term memory neural network to verify its accuracy and effect on the electromyographic signal flow prediction task, and the test set is used to test the trained long short-term memory neural network to observe its prediction performance on new data, and finally the trained long short-term memory neural network that has passed the verification and test is used as the electromyographic signal flow prediction model.

[0071] In summary, in the embodiments of the present invention, personalized rehabilitation training tasks can be carried out according to the actual situation of the stroke patient himself, and real-time guidance can be provided during the execution of the task to improve the degree of task completion.

[0072] In addition, in order to facilitate intuitive display of the rehabilitation training effect, the stroke patient rehabilitation training method 100 may also include:

[0073] Calculate muscle activation and / or fatigue based on electromyographic signal data;

[0074] A real-time joint status model is displayed in a virtual reality scene, and the activation degree and / or fatigue degree of the corresponding muscles are displayed on the real-time joint status model.

[0075] The following describes in detail the stroke patient rehabilitation training method 100 in conjunction with a specific application scenario, as shown below:

[0076] In this application scenario, there are processing modules, inertial posture sensors, surface electromyography signal sensors (which can both collect electromyography signal data and perform functional electrical stimulation), VR display devices, force platforms, universal treadmills, weight-reducing suspensions, data collection suits and other components.

[0077] Before rehabilitation training, stroke patients, with the help of relevant staff, wear data acquisition suits equipped with inertial sensors and surface electromyography signal sensors and enter the omnidirectional treadmill. Relevant staff will lock and link the weight-reducing suspension protection belt on the data acquisition suit to the weight-reducing suspension device above the omnidirectional treadmill.

[0078] After the rehabilitation training begins, the processing module first obtains the stroke patient's hobby data, mood data, physical sign data, disease data, rehabilitation data and current weather data, and then determines the virtual reality scene for rehabilitation training based on the hobby data, mood data and weather data, and determines the personalized rehabilitation training tasks to be performed in the virtual reality scene based on the physical sign data, disease data and rehabilitation data, and then displays the virtual reality scene through a VR display device, adjusts the speed, slope and stability of the universal treadmill based on the virtual reality scene and the personalized rehabilitation training tasks, and then uses the task guidance model corresponding to the personalized rehabilitation training tasks to guide the stroke patients to perform the personalized rehabilitation training tasks in the virtual reality scene.

[0079] During the execution of personalized rehabilitation training tasks, the processing module collects the inertial posture data of stroke patients in the current action cycle through inertial posture sensors, and uses Butterworth high-pass filters and moving average filters to filter the inertial posture data, thereby generating the joint activity signal stream of stroke patients in the current action cycle; the surface electromyography signal sensor collects the electromyography signal data of stroke patients in the current action cycle, and uses Butterworth high-pass filters and band-stop filters to filter the electromyography signal data, thereby generating the electromyography signal stream of stroke patients in the current action cycle. A joint real-time state model is generated according to the joint activity signal stream, and it is compared with the standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task. Based on the comparison results, the VR display device displays the corrected image in the virtual reality scene to guide the stroke patients to complete the personalized rehabilitation training task. At the same time, the EMG signal stream is segmented based on the sliding window, and the signal indicators corresponding to each segment are extracted. The signal indicators corresponding to each segment are input into the EMG signal stream prediction model in the task guidance model corresponding to the personalized rehabilitation training task. The signal indicators corresponding to each segment are processed by the EMG signal stream prediction model to predict the standard EMG signal stream in the next action cycle, and the difference between the predicted standard EMG signal stream in the next action cycle and the actual EMG signal stream in the next action cycle is calculated. If the difference is greater than or equal to the preset threshold, the preset numerical interval of the difference is determined, and according to the stimulation parameters corresponding to the preset numerical interval, the muscles of the stroke patient are electrically stimulated through the surface EMG signal sensor in a subsequent action cycle to guide the stroke patient to complete the personalized rehabilitation training task.

[0080] In addition, the processing module calculates the activation and / or fatigue of the muscle based on the electromyographic signal data, displays the real-time state model of the joint in a virtual reality scene through a VR display device, and displays the activation and / or fatigue of the corresponding muscle on the real-time state model of the joint. The medial, lateral and anterior loads of the joint are calculated based on the joint force information collected by the force platform set on the universal treadmill, so as to synchronously change the weight loss of the weight loss suspension device. The voice signal of the stroke patient is collected through the VR display device and the voice collection and output device is synchronized, and feedback is output.

[0081] It is worth noting that, considering the personal rehabilitation situation of stroke patients, they may not be able to cope with personalized rehabilitation training tasks at the beginning. The number of continuous electrical stimulations can be counted here. If the number of continuous electrical stimulations is greater than the specified number, the preset threshold corresponding to the above difference will be lowered to avoid the stroke patients' decreased enthusiasm due to continuous electrical stimulation.

[0082] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0083] The above is an introduction to a method embodiment. The following is a further explanation of the solution of the present invention through an apparatus embodiment.

[0084] Figure 2 FIG. 4 shows a structural diagram of a rehabilitation training device for stroke patients provided by an embodiment of the present invention. Figure 2 As shown, the stroke patient rehabilitation training device 200 may include:

[0085] The acquisition module 210 is used to acquire the hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data of the stroke patient.

[0086] The first determination module 220 is used to determine a virtual reality scene for rehabilitation training according to the hobby data, mood data and weather data.

[0087] The second determination module 230 is used to determine the personalized rehabilitation training task to be performed in the virtual reality scene according to the physical sign data, the disease condition data and the rehabilitation data.

[0088] The guidance module 240 is used to guide the stroke patient to perform the personalized rehabilitation training task in a virtual reality scene by using the task guidance model corresponding to the personalized rehabilitation training task.

[0089] Understandably, Figure 2 Each module / unit in the stroke patient rehabilitation training device 200 shown has the function of realizing Figure 1 The functions of the various steps in the stroke patient rehabilitation training method 100 shown and the ability to achieve their corresponding technical effects are not described in detail here for the sake of brevity.

[0090] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and the present invention is not limited here.

[0091] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for rehabilitation training of stroke patients, characterized in that: The method comprises: Obtain the hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data of stroke patients; determining a virtual reality scene for rehabilitation training according to the hobby data, the mood data, and the weather data; Determining, according to the physical sign data, the condition data, and the rehabilitation data, a personalized rehabilitation training task to be performed in the virtual reality scene; The task guidance model corresponding to the personalized rehabilitation training task is used to guide the stroke patient to perform the personalized rehabilitation training task in the virtual reality scene.

2. The method according to claim 1, characterized in that The step of determining, according to the physical sign data, the condition data, and the rehabilitation data, a personalized rehabilitation training task to be performed in the virtual reality scene includes: According to the physical sign data, the condition data and the rehabilitation data, a corresponding rehabilitation training task is matched from a pre-built rehabilitation training task library, and for the matched rehabilitation training task, a degree of fit between the matched rehabilitation training task and the virtual reality scene is calculated, and a rehabilitation training task with the greatest degree of fit is selected from the matched rehabilitation training tasks as the personalized rehabilitation training task to be performed in the virtual reality scene; or, The virtual reality scene description data, the vital sign data, the disease condition data and the rehabilitation data are input into a large rehabilitation training model, which processes the virtual reality scene description data, the vital sign data, the disease condition data and the rehabilitation data to determine the personalized rehabilitation training tasks to be performed in the virtual reality scene.

3. The method according to claim 2, characterized in that The rehabilitation training large model is obtained by the following steps: By using various data from the fields of virtual reality and rehabilitation training, the universal large model base is fine-tuned to obtain a large rehabilitation training model.

4. The method according to claim 1, characterized in that: The step of using the task guidance model corresponding to the personalized rehabilitation training task to guide the stroke patient to perform the personalized rehabilitation training task in the virtual reality scene includes: During the execution of the personalized rehabilitation training task, obtaining the joint activity signal flow and the myoelectric signal flow of the stroke patient in the current action cycle; generating a joint real-time state model according to the joint activity signal stream, comparing the model with a standard joint real-time state model in a task guidance model corresponding to the personalized rehabilitation training task, and displaying a corrected image in the virtual reality scene based on the comparison result; The electromyographic signal stream is segmented based on a sliding window, and the signal index corresponding to each segment is extracted. The signal index corresponding to each segment is input into the electromyographic signal stream prediction model in the task guidance model corresponding to the personalized rehabilitation training task, and the electromyographic signal stream prediction model processes the signal index corresponding to each segment to predict the standard electromyographic signal stream in the next action cycle. According to the difference between the predicted standard electromyographic signal stream in the next action cycle and the actual electromyographic signal stream in the next action cycle, the muscles of the stroke patient are electrically stimulated.

5. The method according to claim 4, characterized in that The standard joint real-time state model in the task guidance model corresponding to the personalized rehabilitation training task is obtained by the following steps: While healthy subjects perform the personalized rehabilitation training task, the joint activity signal stream of the healthy subjects in each action cycle is obtained, and a standard joint real-time state model corresponding to the personalized rehabilitation training task is generated according to the joint activity signal stream in each action cycle.

6. The method according to claim 4, characterized in that The myoelectric signal flow prediction model in the task guidance model corresponding to the personalized rehabilitation training task is obtained by the following steps: During the period when healthy subjects perform the personalized rehabilitation training task, the electromyographic signal flow of the healthy subjects in each action cycle is obtained, the electromyographic signal flow is segmented based on a sliding window, and the signal indicators corresponding to each segment are extracted. The signal indicators corresponding to each segment are used as the sample body, and the electromyographic signal flow of the next action cycle is used as the sample label to generate samples, and the long short-term memory neural network is trained using the data set including each sample to obtain the electromyographic signal flow prediction model corresponding to the personalized rehabilitation training task.

7. The method according to claim 4, characterized in that The obtaining of the joint activity signal flow and the myoelectric signal flow of the stroke patient in the current action cycle includes: collecting inertial posture data of the stroke patient in a current motion cycle by using an inertial posture sensor deployed on the body of the stroke patient, thereby generating a joint motion signal stream of the stroke patient in the current motion cycle; The surface electromyographic signal sensor deployed on the body of the stroke patient collects the electromyographic signal data of the stroke patient in the current action cycle, thereby generating the electromyographic signal stream of the stroke patient in the current action cycle.

8. The method according to claim 7, characterized in that The inertial posture sensor is deployed on the outside of the shoulder, outside of the elbow, palm, front and back of the pelvis, outside of the thigh, and outside of the calf of the stroke patient; the surface electromyography signal sensor is deployed on the deltoid muscle, triceps brachii, biceps brachii, wrist dorsiflexor muscle, erector spinae, gluteus maximus, gluteus medius, quadriceps femoris, biceps femoris, tibialis anterior, and gastrocnemius of the stroke patient.

9. The method according to claim 4, characterized in that The method further comprises: Calculating the activation degree and / or fatigue degree of the muscle according to the electromyographic signal data; The real-time state model of the joint is displayed in the virtual reality scene, and the activation degree and / or fatigue degree of the corresponding muscle is displayed on the real-time state model of the joint.

10. A rehabilitation training device for stroke patients, characterized in that: The device comprises: The acquisition module is used to obtain the hobby data, mood data, physical sign data, condition data, rehabilitation data and current weather data of stroke patients; A first determination module, configured to determine a virtual reality scene for rehabilitation training according to the hobby data, the mood data and the weather data; A second determination module is used to determine the personalized rehabilitation training task to be performed in the virtual reality scene according to the physical sign data, the condition data and the rehabilitation data; The guidance module is used to guide the stroke patient to perform the personalized rehabilitation training task in the virtual reality scene by using the task guidance model corresponding to the personalized rehabilitation training task.

Images