Intelligent rehabilitation training auxiliary system

Abstract

The intelligent rehabilitation training auxiliary system comprises a remote cloud processing center and a plurality of sub rehabilitation systems, the remote cloud processing center is used for a rehabilitative or a doctor to manage and monitor the physical function level and the training condition of each patient and distribute a training scheme for each patient, and the sub-rehabilitation system is used for rehabilitation training of the patients and inputting of new actions by the rehabilitative; the system makes full use of artificial intelligence, edge computing and other technologies, gives full play to the advantages of deep learning, and achieves the purpose that a rehabilitation teacher checks patient data, formulates a training scheme and records training actions; evaluating and correcting rehabilitation actions of the patient, evaluating a physical state and monitoring an abnormal state; interaction with a patient is realized through a voice interaction technology and a human-computer interaction interface, the patient is stimulated and danced, and the rehabilitation enthusiasm of the patient is enhanced; the system integrates guidance and protection, provides real-time guidance and safety guarantee for rehabilitation training of patients in hospitals or at home, and has high practical value and wide application prospect.

Description

Technical Field

[0001] The present invention belongs to the field of medical rehabilitation and computer vision technology, and in particular relates to an intelligent rehabilitation training auxiliary system. Background Art

[0002] At present, my country's rehabilitation medical resources are in short supply, while the number of people who need rehabilitation is huge, and the rehabilitation product market has broad prospects. In particular, patients who need rehabilitation training at home after discharge are currently facing the following problems: (1) rehabilitation medical resources are in short supply, while the number of people who need rehabilitation is huge; (2) patients' dependence decreases after discharge and psychological problems need to be solved; (3) it is difficult to grasp the exercise standards and amount during self-rehabilitation.

[0003] The currently disclosed rehabilitation systems such as CN115036007A and CN115206484A, although they have the function of assisting rehabilitation training, can only train specific local muscle groups due to sensor limitations, and have great limitations. The active auxiliary mode disclosed in CN115463000A and CN115300325A is used to conduct rehabilitation training for patients. Due to the limitation of its mechanical structure, the patient's movement will be constrained to a certain extent, and only a small part of the rehabilitation plan can be implemented. The rehabilitation system disclosed in CN115148058A uses robot-assisted rehabilitation training equipment, which is relatively expensive and difficult for ordinary families to afford.

[0004] In summary, there are mainly the following types of products for assisting rehabilitation training on the market, which have the following problems:

[0005] (1) Body sensor collection device

[0006] It is difficult for non-professionals to wear these devices correctly and independently, and the direction or position of wearing may be deviated, making it impossible to meet the requirements of refined assessment.

[0007] (2) Active assist device

[0008] The mechanical structure restricts the patient's movement to a certain extent. The device is large and expensive, and is not highly targeted to different home environments, making it unsuitable for home or community use.

[0009] (3) Physician offline guidance device

[0010] This type of equipment cannot analyze the patient's recovery status in real time, and most public rehabilitation systems do not consider the difficulty of patients interacting with the system.

[0011] Therefore, for patients who need rehabilitation training at home after being discharged from hospital, it is of great practical significance to effectively solve the rehabilitation training needs of patients. Summary of the invention

[0012] In order to overcome the defects of the above-mentioned prior art, the purpose of the present invention is to provide an intelligent rehabilitation training assistance system, which utilizes technologies such as artificial intelligence and edge computing to give full play to the advantages of deep learning, so that rehabilitation therapists can view patient data, formulate training plans, and record training movements; evaluate and correct patients' rehabilitation movements, assess their physical condition, and monitor abnormal conditions; interact with patients through voice interaction technology and human-computer interaction interface; this system integrates guidance and protection, and can provide real-time guidance and safety guarantees for patients' rehabilitation training in hospitals or at home from multiple physiological and psychological aspects, and has high practical value and broad application prospects.

[0013] In order to achieve the above object, the technical solution of the present invention is:

[0014] An intelligent rehabilitation training auxiliary system, comprising a remote cloud processing center 1 and a plurality of sub-rehabilitation systems 2;

[0015] The remote cloud processing center 1 is used by a rehabilitation therapist or doctor to manage and monitor the physical function level and training status of each patient and assign a training plan to each patient; the remote cloud processing center 1 includes a first processor 101, a first power module 102 for supplying power to the first processor 101, a manual processing module 103 for realizing signal interaction with the first processor 101, a first information exchange module 104 and a first storage module 105;

[0016] The first processor 101 is used to analyze the patient's physical function level, training movement specifications and training plan completion status, use single-view and multi-view clustering algorithms to classify patients in multiple dimensions, and automatically assign new training plans to multiple patients; the first current module 102 is used to provide a stable power supply voltage to the first processor 101; the manual processing module 103 is used for rehabilitation therapists and doctors to view and process patient training plans and communicate with patients, and pass information to the first processor 101 and the first information exchange module 104; the first information exchange module 104 is used to exchange information with multiple sub-rehabilitation systems 2, including sending each patient's training plan, training movements and medical advice, and receiving multiple patients' physical function levels, training posture evaluations, and training plan completion status; the first storage module 105 is used to store information of multiple patients.

[0017] The sub-rehabilitation system 2 is used for patient rehabilitation training and rehabilitation therapists to record new movements, including a second processor 201, a second power supply module 202 for powering the second processor 201, a human-computer interaction module 203 for information interaction with the second processor 201, a voice interaction module 204, a health monitoring module 205, an identity information collection module 206, a video stream collection module 207, a posture analysis module 208, a fall monitoring module 209, a second information exchange module 210 and a second storage module 211;

[0018] The first processor 101 and the second processor 201 are computers.

[0019] The second current module 202 is used to provide a stable power supply voltage to the second processor 201 .

[0020] The human-computer interaction module 203 is used for the patient or the rehabilitation therapist to manually control the subsystem.

[0021] The voice interaction module 204 is used for the patient to voice control the subsystem during training, and the system gives the patient voice feedback.

[0022] The health monitoring module 205 is used to measure various body functions of the patient and send the monitored physiological data to the second processor 201 .

[0023] The identity information collection module 206 collects and identifies the identity information of the patient or the rehabilitation therapist.

[0024] The video stream acquisition module 207 is a camera, which is used to collect the patient's video stream information in real time during the patient's rehabilitation training and send it to the first processor 201, and for the rehabilitation therapist to enter new rehabilitation movements.

[0025] The posture analysis module 208 establishes a skeleton point model based on the video stream information collected by the video stream acquisition module 207, and then compares the patient's skeleton point information with the standard action skeleton point information entered in advance by the rehabilitation therapist: the DTW algorithm is used to align the two time series of inconsistent lengths of the patient's action and the standard action provided by the rehabilitation therapist, and then the difference in the angle between the human skeleton vectors formed by the standard action and the patient's action in certain key frames specified by the rehabilitation therapist is analyzed to judge the similarity between the two, and further obtain evaluation data.

[0026] The specific analysis method of the posture analysis module 208 is:

[0027] (1)Accumulated distance according to the DTW algorithm

[0028] γ(i, j) = d(q i , c j )+min{γ(i-1,j-1),γ(i-1,j),γ(i,j-1)}

[0029] In the formula, d(q i , c j ) is the distance between the points at positions i and j of the two time series, using the Euclidean distance, γ(i-1, j), γ(i, j-1) and (i-1, j-1) are the minimum cumulative distances from the previous row, left column and upper left corner to the current position (i, j), respectively;

[0030] Starting from the lower right corner of the matrix, backtracking along the path of the minimum cumulative distance, finding the best matching point pair sequence, finding the shortest path between the patient's action skeleton point information sequence and the standard action skeleton point information sequence, thereby aligning the two sequences and matching the patient's action with the specified standard action frame by frame;

[0031] (2) Read the pre-set key frame of the action and the patient action frame matched in the previous step, and calculate the difference δθ between the stick figure angle of the key part of the patient action and the standard action. If |δθ|<θ threshold , then it is determined as the action standard, otherwise it is determined as non-standard, and the obtained evaluation data is sent to the processor 201;

[0032] The fall monitoring module 209 analyzes the patient's physical characteristics based on the information collected by the video stream acquisition module 207 and determines whether the patient has any dangerous characteristics of falling. Specifically, it determines whether the patient has fallen by analyzing whether the key point data of the human body meets the three characteristics of falling. If the patient falls, the fall monitoring module 209 immediately calls for help from the patient's family or doctor, and sends the fall event to the second processor 201 to stop training.

[0033] The specific analysis method of the fall monitoring module 209 is:

[0034] (1) Extract the average value of the y coordinates of the key points of the human body in each frame in the sliding window

[0035]

[0036] (2) Extract the average y coordinates of the 12 points in the head of each frame in the sliding window

[0037] Use the least squares method to find the sliding window The slope v of (1≤t≤25);

[0038] (3) Extract the variance x_var and range x_range = x_max-x_min of the x coordinate of each frame in the sliding window;

[0039] (4) Fall judgment: If the following conditions are met simultaneously in the selected sliding window: a. the average value of the y coordinates of the key points of the human body for more than 10 frames is higher than the specified threshold; b. the slope v, i.e., the velocity, is higher than the specified threshold; c. the variance of the x coordinates for more than 15 frames is higher than the specified threshold; d. the range of the x coordinates for more than 5 frames is higher than the specified threshold; then it is judged that the patient has fallen.

[0040] The second information exchange module 210 is used to exchange information with the remote cloud processing center 1 .

[0041] The second storage module 211 is used to store information of multiple patients.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] (1) The present invention ensures the richness, scientificity, and pertinence of system actions and the effectiveness of patient training through the mutual cooperation between the remote cloud processing center 1 and multiple sub-rehabilitation systems 2.

[0044] (2) The present invention provides a health monitoring module 205 and a fall monitoring module 209 to effectively monitor the patient's physical condition during exercise and promptly alarm when the patient is in danger, thereby ensuring the patient's safety during exercise.

[0045] (3) The sub-rehabilitation system 2 of the present invention effectively solves the problem of needing to wear sensory input devices during rehabilitation training through the second processor 201, the human-computer interaction module 203, the voice interaction module 204, the video stream acquisition module 207, and the posture analysis module 208, thereby greatly reducing the cost of use while improving the convenience of rehabilitation training for patients.

[0046] (4) The sub-rehabilitation system 2 of the present invention realizes the analysis and feedback of the patient's training movements through the posture analysis module 208 and the voice interaction module 204, and encourages the patient, thereby improving communication efficiency, improving the efficiency and quality of rehabilitation training, and enhancing the patient's enthusiasm for exercise.

[0047] (5) The posture analysis module 208 of the present invention uses the DTW algorithm to align the two time series of inconsistent lengths, the patient's movements and the standard movements provided by the rehabilitation therapist, and then judges the similarity between the two by analyzing the difference in the angles formed by the human skeleton vectors formed by the standard movements and the patient's movements in certain key frames specified by the rehabilitation therapist. The obtained evaluation data is more precise and accurate.

[0048] (6) The fall monitoring module 209 of the present invention uses the coordinates, slope, variance and range of each frame to analyze and determine whether the patient has fallen, which has the characteristics of comprehensive consideration of various factors and accurate analysis.

[0049] (7) The remote cloud processing center 1 of the present invention can provide the rehabilitation therapist with the patient's recent information. The rehabilitation therapist can record new training movements and set rehabilitation training plans through the remote cloud processing center 1, which effectively ensures the patient's scientific rehabilitation training.

[0050] In summary, the system of the present invention makes full use of technologies such as artificial intelligence and edge computing, and gives full play to the advantages of deep learning, so that rehabilitation therapists can view patient data, formulate training plans, and record training movements; evaluate and correct patient rehabilitation movements, assess physical conditions, and monitor abnormal conditions; interact with patients through voice interaction technology and human-computer interaction interface, motivate and inspire patients, and enhance their enthusiasm for rehabilitation. This system integrates guidance and protection, provides real-time guidance and safety guarantees for patients in hospital or at home rehabilitation training, and has high practical value and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the edge computing system structure of the present invention.

[0052] Figure 2 It is a schematic diagram of the structure of the first processor of the present invention.

[0053] Figure 3 It is a schematic diagram of the structure of the second processor of the present invention.

[0054] Figure 4 It is a flow chart of an implementation method of the present invention.

[0055] Figure 5 A flow chart for evaluating and correcting the actions of the present invention.

[0056] Figure 6 This is a flow chart of fall detection according to the present invention. DETAILED DESCRIPTION

[0057] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0058] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0059] Reference Figure 1 , an intelligent rehabilitation training auxiliary system, including a remote cloud processing center 1 and multiple sub-rehabilitation systems 2.

[0060] Reference Figure 2The remote cloud processing center 1 is set up in the rehabilitation therapist's home, hospital, rehabilitation center, etc., and is used by the rehabilitation therapist or doctor to manage and monitor the physical function level and training status of each patient and assign training plans to each patient; the remote cloud processing center 1 includes a first processor 101, a first power supply module 102 for supplying power to it, an artificial processing module 103 for realizing signal interaction with the first processor 101, a first information exchange module 104 and a first storage module 105.

[0061] The first processor 101 is used to analyze the patient's physical function level, training movement standardization and training plan completion status, and use single-view and multi-view clustering algorithms to classify patients in multiple dimensions, such as training enthusiasm level, physical function level, etc., so that rehabilitation therapists can give priority to patients who should be given more attention, and it is convenient for rehabilitation therapists to uniformly set training plans for patients with the same pathological level or cause of disease, and automatically assign new training plans to multiple patients; the first current module 102 is used to provide a stable power supply voltage to the first processor 101 to ensure the normal operation of the remote cloud processing center 1; the manual processing module 103 is used for rehabilitation therapists and doctors to view and process patient training plans and communicate with patients, and transmit information to the first processor 101 and the first information exchange module 104; the first information exchange module 104 is used to exchange information with multiple sub-rehabilitation systems 2, including sending each patient's training plan, training movement and doctor's advice, and receiving multiple patients' physical function levels, training posture evaluations, and training plan completion status; the first storage module 105 is used to store information of multiple patients.

[0062] Reference Figure 3 The sub-rehabilitation system 2 is set at the patient's home, hospital or rehabilitation center, and is used for patient rehabilitation training or rehabilitation therapists to enter new movements. It includes a second processor 201, a second power supply module 202 for powering it, a human-computer interaction module 203 that interacts with the second processor 201, a voice interaction module 204, a health monitoring module 205, an identity information collection module 206, a video stream collection module 207, a posture analysis module 208, a fall monitoring module 209, a second information exchange module 210 and a second storage module 211; the second processor 201 is used for information communication and analysis and calculation between several modules.

[0063] The first processor 101 and the second processor 201 are computers, and the computers are equipped with a keyboard, a mouse, a display and a speaker.

[0064] The second current module 202 is used to provide a stable power supply voltage to the second processor 201 to ensure the normal operation of the remote processing center 1.

[0065] The human-computer interaction module 203 is used for the patient to manually control the subsystem. In this example, the human-computer interaction module uses a touch display screen.

[0066] The voice interaction module 204 is used for the patient to voice control the subsystem during training, and the system gives the patient voice feedback.

[0067] The health monitoring module 205 is used to measure various body functions of the patient, such as a digital blood pressure monitor, and send the monitored physiological data to the second processor 201 .

[0068] The identity information collection module 206 collects and identifies the patient's identity information; in this example, it includes two methods: face login and card swipe login.

[0069] The video stream acquisition module 207 is a camera, which is used to collect the patient's video stream information in real time during the patient's rehabilitation training and send it to the first processor 201.

[0070] The posture analysis module 208 establishes a skeleton point model based on the video stream information collected by the video stream acquisition module 207, and then compares the patient's skeleton point information with the standard action skeleton point information entered in advance by the rehabilitation therapist: the DTW algorithm is used to align the two time series of inconsistent lengths of the patient's action and the standard action provided by the rehabilitation therapist, and then the difference in the angle between the human skeleton vectors formed by the standard action and the patient's action in certain key frames specified by the rehabilitation therapist is analyzed to judge the similarity between the two, and further obtain evaluation data.

[0071] The specific analysis method of the posture analysis module 208 is:

[0072] (1)Accumulated distance according to the DTW algorithm

[0073] γ(i, j) = d(q i , c j )+min{γ(i-1,j-1),γ(i-1,j),γ(i,j-1)}

[0074] In the formula, d(q i , c j ) is the distance between the points at positions i and j of the two time series, using the Euclidean distance, γ(i-1, j), γ(i, j-1) and (i-1, j-1) are the minimum cumulative distances from the previous row, left column and upper left corner to the current position (i, j), respectively;

[0075] Starting from the lower right corner of the matrix, backtracking along the path of the minimum cumulative distance, finding the best matching point pair sequence, finding the shortest path between the patient's action skeleton point information sequence and the standard action skeleton point information sequence, thereby aligning the two sequences and matching the patient's action with the specified standard action frame by frame;

[0076] (2) Read the pre-set key frame of the action and the patient action frame matched in the previous step, and calculate the difference δθ between the stick figure angle of the key part of the patient action and the standard action. If |δθ|<θ threshold , then it is determined as the action standard, otherwise it is determined as non-standard, and the obtained evaluation data is sent to the processor 201;

[0077] The fall monitoring module 209 analyzes the patient's physical characteristics based on the information collected by the video stream acquisition module 207 and determines whether the patient has any dangerous characteristics of falling. Specifically, it determines whether the patient has fallen by analyzing whether the key point data of the human body meets the three characteristics of falling. If the patient falls, the fall monitoring module 209 immediately calls for help from the patient's family or doctor, and sends the fall event to the second processor 201 to stop training.

[0078] The specific analysis method of the fall monitoring module 209 is:

[0079] 1) Extract the average value of the y coordinates of the key points of the human body in each frame in the sliding window

[0080]

[0081] 2) Extract the average y coordinates of the 12 points in the head of each frame in the sliding window

[0082] Use the least squares method to find the sliding window The slope v of (1≤t≤25).

[0083] 3) Extract the variance x_var and x_range = x_max - x_min of the x coordinate of each frame in the sliding window.

[0084] 4) Fall judgment: If the following conditions are met in the selected sliding window at the same time: a. the average value of the y coordinates of the key points of the human body for more than 10 frames is higher than the specified threshold; b. the slope v, that is, the speed, is higher than the specified threshold; c. the variance of the x coordinates for more than 15 frames is higher than the specified threshold; d. the range of the x coordinates for more than 5 frames is higher than the specified threshold; then it is judged that the patient has fallen.

[0085] The second information exchange module 210 is used to exchange information with the remote cloud processing center 1 .

[0086] The second storage module 211 is used to store information of multiple patients.

[0087] Reference Figure 4-6 As shown, the working principle of the system of the present invention, taking the stroke training program as an example, includes the following steps:

[0088] (1) The rehabilitation therapist enters all rehabilitation training movements in the rehabilitation sub-system 2 according to the system instructions;

[0089] (2) The patient logs in to the rehabilitation sub-system 2 through face recognition or card recognition;

[0090] (3) After the patient enters the activity area, the system displays the training plan on the touch screen, and the patient watches the standard action video and completes the corresponding action training according to the voice prompts. After the posture analysis module 208 receives the patient training video stream collected by the video stream acquisition module 207, it uses the DTW algorithm to align the two time series of inconsistent lengths of the patient's action and the standard action provided by the rehabilitation therapist, and then judges the similarity between the two by analyzing the difference in the angle between the human skeleton vectors formed by the standard action and the patient's action in certain key frames specified by the rehabilitation therapist, and further obtains evaluation data and sends it to the processor 201;

[0091] (4) During the training process, the fall monitoring module 209 is used to monitor whether the patient has fallen or other dangerous behaviors during the training, and judge whether the patient has fallen by analyzing whether the key point data of the human body meets the three characteristics of falling. When the patient falls, the module immediately calls for help from the patient's family or doctor, and sends the fall event to the processor 201 to stop the training;

[0092] (5) After the training, the patient uses the health monitoring module 205 to measure and detect physiological data such as his / her body function, heart rate and blood pressure, and sends the physiological data to the processor 201;

[0093] (6) After the training is completed, the second processor 201 sends the patient data to the remote cloud processing center 1 through the second information transmission module 211. The remote cloud processing center 1 analyzes the training movement standardization of the individual patient, the completion of the training plan, etc., and designs and allocates a new training plan for the patient according to the algorithm;

[0094] (7) After the training, the rehabilitation therapist checks the patient's training data, physical function and the new training plan designed by the system in the remote cloud processing center 1. After the rehabilitation therapist reviews and approves or modifies the new training plan, the new training plan is sent to the sub-rehabilitation system 2. The cycle repeats until the patient is completely recovered.

Claims

1. An intelligent rehabilitation training auxiliary system, comprising a remote cloud processing center (1) and a plurality of sub-rehabilitation systems (2); characterized in that: The remote cloud processing center (1) is used by a rehabilitation therapist or doctor to manage and monitor the physical function level and training status of each patient and to assign a training plan to each patient; the remote cloud processing center (1) comprises a first processor (101), a first power module (102) for supplying power to the first processor (101), a manual processing module (103) for realizing signal interaction with the first processor (101), a first information exchange module (104) and a first storage module (105); The first processor (101) is used to analyze the patient's physical function level, training movement standardization and training plan completion status, use single-view and multi-view clustering algorithms to classify patients in multiple dimensions, and automatically assign new training plans to multiple patients; the first current module (102) is used to provide a stable power supply voltage to the first processor (101); the manual processing module (103) is used for rehabilitation therapists and doctors to view and process patient training plans and communicate with patients, and transmit information to the first processor (101) and the first information exchange module (104); the first information exchange module (104) is used to exchange information with multiple rehabilitation sub-systems (2), including sending each patient's training plan and training movement and medical advice, and receiving multiple patients' physical function levels, training posture evaluation status, and training plan completion status; the first storage module (105) is used to store information of multiple patients. The sub-rehabilitation system (2) is used for patient rehabilitation training and for a rehabilitation therapist to record new movements, and comprises a second processor (201), a second power supply module (202) for supplying power to the second processor, a human-computer interaction module (203) for information interaction with the second processor (201), a voice interaction module (204), a health monitoring module (205), an identity information collection module (206), a video stream collection module (207), a posture analysis module (208), a fall monitoring module (209), a second information exchange module (210) and a second storage module (211); The first processor (101) and the second processor (201) are computers.

2. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The second current module (202) is used to provide a stable power supply voltage to the second processor (201).

3. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The human-computer interaction module (203) is used for the patient or the rehabilitation therapist to manually control the subsystem.

4. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The voice interaction module (204) is used for the patient to voice control the subsystem during training, and the system gives the patient voice feedback.

5. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The health monitoring module (205) is used to measure various body functions of the patient and send the monitored physiological data to the second processor (201).

6. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The identity information collection module (206) collects and identifies the identity information of the patient or the rehabilitation therapist.

7. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The video stream acquisition module (207) is a camera, which is used to collect the patient's video stream information in real time during the patient's rehabilitation training and send it to the first processor (201), and for the rehabilitation therapist to enter new rehabilitation movements.

8. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The posture analysis module (208) establishes a skeleton point model based on the video stream information collected by the video stream collection module (207), and then compares the patient's skeleton point information with the standard action skeleton point information entered in advance by the rehabilitation therapist: the DTW algorithm is used to align the two time series of inconsistent lengths, namely, the patient's action and the standard action provided by the rehabilitation therapist, and then the similarity between the two is judged by analyzing the difference in the angle formed by the human skeleton vectors formed by the standard action and the patient's action in certain key frames specified by the rehabilitation therapist, so as to further obtain evaluation data; The specific analysis method of the posture analysis module (208) is: (1)Accumulated distance according to the DTW algorithm γ(i,j)=d(q i ,c j )+min{y(i-1,j-1),γ(i-1,j),γ(i,j-1)} In the formula, d(q i , c j ) is the distance between the points at positions i and j of the two time series, using the Euclidean distance, γ(i-1, j), γ(i, j-1) and (i-1, j-1) are the minimum cumulative distances from the previous row, left column and upper left corner to the current position (i, j), respectively; Starting from the lower right corner of the matrix, backtracking along the path of minimum cumulative distance, finding the best matching point pair sequence, finding the shortest path between the patient's action skeleton point information sequence and the standard action skeleton point information sequence, thereby aligning the two sequences and matching the patient's action with the specified standard action frame by frame; (2) Read the pre-set key frame of the action and the patient action frame matched in the previous step, and calculate the difference in the stick figure angle between the patient's action and the key part of the standard action δθ, if |δθ|<θ thresh old , then it is judged as the action standard, otherwise it is judged as non-standard, and the obtained evaluation data is sent to the processor (201).

9. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The fall monitoring module (209) analyzes the patient's physical characteristics based on the information collected by the video stream collection module (207) and determines whether the patient has fall danger characteristics, specifically by analyzing whether the key point data of the human body meets the three characteristics of fall to determine whether the patient has fallen. If the patient falls, the fall monitoring module (209) immediately calls for help from the patient's family or doctor, and sends the fall event to the second processor (201) to stop training; The specific analysis method of the fall monitoring module (209) is: (1) Extract the average value of the y coordinates of the key points of the human body in each frame in the sliding window (2) Extract the average y coordinates of the 12 points in the head of each frame in the sliding window Use the least squares method to find the sliding window The slope v of (1≤t≤25); (3) Extract the variance x_var and range x_range = x_max-x_min of the x coordinate of each frame in the sliding window; (4) Fall judgment, if in the selected sliding window, the following conditions are met at the same time: a. The average value of the y coordinates of the key points of the human body for more than 10 frames is higher than the specified threshold; b. The slope v, that is, the speed, is higher than the specified threshold; c. The variance of the x-coordinates of more than 15 frames is higher than the specified threshold; d. The range of the x-coordinates of more than 5 frames is higher than the specified threshold; then it is determined that the patient has fallen.

10. The intelligent rehabilitation training auxiliary system according to claim 1, characterized in that: The second information exchange module (210) is used to exchange information with the remote cloud processing center (1); the second storage module (211) is used to store information of multiple patients.

Images