Bone fracture rehabilitation training plan intelligent recommendation method and system based on traditional Chinese medicine

Abstract

The invention relates to the technical field of information and data processing, and discloses an intelligent recommendation method and system for a bone fracture rehabilitation training plan based on traditional Chinese medicine, and the method comprises the steps: dividing image information according to labels, and obtaining skeleton data sets with different functions; taking the first position information and the second position information as a special information set; deleting the special data set from the skeleton data set to obtain a target data set; obtaining position information of the bones to be subjected to bone fracture rehabilitation training from the target data set, and selecting matched contents such as shadowboxing, eight-section brocade or five-poultry play from a pre-stored traditional Chinese medicine rehabilitation training database; deleting related actions of the second position information from an action data set corresponding to the content; guiding the trainee to perform bone fracture rehabilitation training through the visual terminal; the system comprises a first data set construction module, a second data set construction module and a target action output module. According to the bone fracture rehabilitation training system and method, biological feature recognition and personalized, precise and systematic bone fracture rehabilitation training are achieved.

Description

Technical Field

[0001] The present invention relates to the field of information and data processing technology, and in particular to a method and system for intelligently recommending a bone injury rehabilitation training plan based on traditional Chinese medicine. Background Art

[0002] Bone injury refers to the damage to bones caused by external impact, falls, sprains, etc. Common bone injuries include fractures, cracks, and dislocations. Fixed injuries generally require rehabilitation through physical therapy, exercise therapy, and auxiliary equipment. Rehabilitation can effectively shorten the recovery period and promote the healing of fixed injuries. However, the existing rehabilitation methods, in addition to the use of instruments, are carried out with the assistance of professional rehabilitation therapists. On the one hand, it is restricted by the environment, and on the other hand, the rehabilitation cost is relatively high.

[0003] Prior art 1, Chinese patent, application number: CN202410677434.9 discloses an intelligent rehabilitation training dynamic monitoring and personalized feedback system, including a user information management module, a biological signal acquisition module, a motion analysis and evaluation module, a personalized rehabilitation program generation module, a real-time feedback and interaction module, and a remote monitoring and communication module; although the Atrous spatiotemporal motion recognition method is used to identify the patient's motion situation, the strategy of comprehensive global context understanding, multi-scale time information integration, and improving the distinguishability of motion features is used to achieve efficient and accurate detection of motions in video images; and the improved dual-tower model is used to recommend a suitable rehabilitation training plan for each patient, and the existing rehabilitation training recommendation algorithm is improved by strengthening the interactive communication between the patient and the dual-tower model of therapy characteristics, generating a more powerful and comprehensive rehabilitation demand and therapy effect prediction model. However, it lacks the combination with relevant knowledge of traditional Chinese medicine, resulting in the need for further improvement in the intelligence of the system.

[0004] Prior art 2, Chinese patent, application number: CN 202410161370.7 discloses a method for formulating an intelligent training plan based on data collection and its application, including the following steps: collecting the patient's characteristic information and symptom information; generating polygons according to the patient's symptom information, evaluating the scores according to the results of the polygon tracing by the patient, and obtaining an initial score; formulating a rehabilitation plan, binding the rehabilitation plan with the characteristic information, symptom information and the initial score and storing it; after the patient has completed the training according to the rehabilitation plan, evaluating the scores according to the results of the polygon tracing by the patient, and obtaining the final score; when the final score is greater than the initial score, converting the rehabilitation plan into a shared rehabilitation plan; although by comparing the changes in the evaluation scores before and after the patient's use, a real and effective rehabilitation plan is collected, and the transformation from an ordinary rehabilitation plan to a shared rehabilitation plan is realized, and a rehabilitation plan suitable for the patient can be recommended from the shared rehabilitation plan. However, the process of obtaining the training plan is relatively complicated, which to a certain extent hinders the acquisition of the training plan.

[0005] Prior art three, Chinese patent, application number: 201910370527.6 discloses a tendon condition assessment method, device and storage medium based on deep learning, including: obtaining a first action video of the target user; extracting a first image from the video; extracting a second image based on the first image; inputting the first image and the second image into a convolutional neural network model; if the model recognizes that the tendon of the target part of the target user is damaged, obtaining adjacent frame images in the first action video; calculating the first similarity of adjacent frame images, obtaining the second similarity of adjacent frames of the preset non-damaged tendon action posture; determining the degree of tendon damage of the target user and the recommended rehabilitation training information corresponding to the degree of tendon damage according to the difference between the first similarity and the second similarity as the rehabilitation plan for the target user. Although it accurately assesses whether the user's tendon condition is damaged and provides more accurate rehabilitation training advice after the tendon is damaged. However, its function is relatively simple, and it only targets the rehabilitation plan for the tendon, resulting in a limited scope of application.

[0006] At present, the existing technologies 1, 2 and 3 have the problems that the rehabilitation plan acquisition method is relatively complicated, the efficiency is low, the principles of traditional Chinese medicine are not reflected, and the degree of intelligence needs to be further improved. Therefore, the present invention provides a method and system for intelligently recommending bone injury rehabilitation training plans based on traditional Chinese medicine. Summary of the invention

[0007] The main purpose of the present invention is to provide an intelligent recommendation method and system for bone injury rehabilitation training plans based on traditional Chinese medicine, so as to solve the problems in the prior art that the method of obtaining rehabilitation plans is relatively complicated, inefficient, does not reflect the principles of traditional Chinese medicine, and the degree of intelligence needs to be further improved.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for intelligently recommending a bone injury rehabilitation training plan based on traditional Chinese medicine, the method comprising:

[0010] Obtain image information of all bones of the target bone injury and mark the location information; use the function of the organ where the bone is located as a label, divide the image information according to the label, and obtain bone data sets with different functions;

[0011] Obtain the bone name and the type of bone injury through the input device, obtain the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set, and use the first position information and the second position information as a special information set; delete the special data set from the bone data set to obtain the target data set;

[0012] The position information of the bones to be subjected to bone injury rehabilitation training is obtained from the target data set, and matching Tai Chi, Ba Duan Jin or Wu Qin Xi content is selected from the pre-stored TCM rehabilitation training database; the related actions of the second position information are deleted from the action data set corresponding to the content; the trainee is guided to perform bone injury rehabilitation training through a visual terminal, and the cycle and time information of the bone injury rehabilitation training are prompted at the same time.

[0013] As a further improvement of the present invention, the process of obtaining the skeleton data sets of different functions includes:

[0014] Use medical imaging equipment to obtain image information of all bones of the bone injury target; perform denoising, enhancement and standardization on the acquired images; use edge detection to remove noise and artifacts in the images to obtain pre-processed images;

[0015] Use convolutional neural networks to automatically annotate bone images, automatically identify and annotate the location information of bones; functionally classify the annotated bones according to the function of the organ where the bones are located, generate a unique label for each functional classification, and associate the label with the corresponding bone image;

[0016] According to the generated functional labels, the annotated bone images are divided according to the labels to construct a tree-structured bone dataset for each different function. The root node is the bone dataset and the child nodes are the datasets of each functional classification; each dataset contains all the bone images under the same functional classification.

[0017] As a further improvement of the present invention, the process of obtaining the target data set includes:

[0018] Create an empty target dataset to store the data that meets the requirements filtered out from the original bone dataset; preprocess the bone dataset to unify the data format; extract key features from the bone dataset;

[0019] A hierarchical method is used to screen the data, and the data related to the special information set is gradually eliminated; the data directly related to the special information set is screened out according to the bone name and type of bone injury; the data indirectly related to the special information set is further screened, and the second position information of the bone associated with the bone injury is checked for duplicates;

[0020] Integrate the hierarchically screened data into the target dataset and verify the target dataset.

[0021] As a further improvement of the present invention, the process of gradually eliminating data related to the special information set includes:

[0022] Extract all recorded bone names from the bone data set and accurately match them with the bone names in the special information set; the successfully matched records enter the next level of screening; for the successfully matched records, further extract the bone injury type and match it with the bone injury type in the special information set; the successfully matched records enter the candidate list of the target data set;

[0023] For the records in the candidate list, extract the second position information of the bone associated with the bone injury; match the extracted second position information with the relevant position information in the special information set; the successfully matched records are further entered into the candidate list of the target data set;

[0024] Perform a duplicate check on the records in the candidate list and calculate the duplicate check rate; based on the duplicate check results, remove records with a duplication rate exceeding 30%.

[0025] As a further improvement of the present invention, the process of matching the extracted second location information with the relevant location information in the special information set includes:

[0026] Gaussian pyramids are constructed for the images in the image information and the standard anatomical position information respectively to detect the features of the image at different scales; multiple scale space layers are generated through Gaussian blur and downsampling operations; on the basis of the Gaussian pyramid, the Gaussian difference image of two adjacent layers of images is calculated to detect extreme points in the image, that is, potential key points;

[0027] In the Gaussian difference image, by comparing the values ​​of each pixel with the 26 neighboring pixels around it, the extreme points are detected, which are potential key points; the detected extreme points are located; the edge response is eliminated by calculating the eigenvalues ​​of the Hessian matrix to retain stable key points; the gradient direction and amplitude are calculated for the pixels around each key point; in the local area around the key point, a gradient direction histogram of 36 directions is constructed, and the gradient amplitude of each direction is counted; the direction with the highest peak in the histogram is selected as the main direction of the key point, and a key point with direction is generated;

[0028] The local image area around the key point is divided into 16 sub-areas. In each sub-area, the gradient direction histogram in 8 directions is calculated to generate a 128-dimensional feature descriptor; the descriptor is normalized; the Euclidean distance between the feature descriptors of the image information and the image in the standard anatomical position information is calculated; for each feature descriptor, the feature descriptor with the smallest Euclidean distance to it is found in the other image as a matching pair;

[0029] A Euclidean distance threshold is set. Only when the Euclidean distance between two feature descriptors is less than the threshold, they are considered to be a valid matching pair; all matching pairs are screened and matching results with Euclidean distance greater than the set threshold are eliminated.

[0030] As a further improvement of the present invention, the process of selecting matching Tai Chi, Ba Duan Jin or Wu Qin Xi content includes:

[0031] Based on the action data set in the TCM rehabilitation training database, Tai Chi, Ba Duan Jin, and Wu Qin Xi movements are classified according to the skeletal parts and functions involved; an index is created for each classified movement, and the index contains the key features of the movement;

[0032] The extracted skeletal position information is matched with the action index in the TCM rehabilitation training database, and the matching degree between the skeletal position information and the action index is calculated using cosine similarity;

[0033] Set a matching threshold, filter out movements with matching degrees higher than the threshold, and select a movement from Tai Chi, Ba Duan Jin, or Wu Qin Xi.

[0034] As a further improvement of the present invention, the process of using cosine similarity to calculate the matching degree between the skeleton position information and the action index includes:

[0035] Extract the key features of each action from the TCM rehabilitation training database, including the involved bone parts, movement angles and strengths; convert the standardized bone position information into a feature vector, where each bone position information is represented as a multidimensional vector, where each dimension corresponds to a specific bone part or movement parameter; convert the key features in each action index into a feature vector, which will be used to match the bone position information;

[0036] For each bone position information vector, calculate the cosine similarity between it and each action index vector; cosine similarity is a method to measure the similarity of the directions of two vectors;

[0037] The matching degree between the skeleton position information and the action index is determined according to the calculated cosine similarity. The higher the matching degree, the higher the similarity between the skeleton position information and the action index.

[0038] As a further improvement of the present invention, the process of determining the matching degree between the skeleton position information and the action index includes:

[0039] The filtered action index vectors are sorted in descending order according to their cosine similarity with the skeleton position information vector; from the sorted list, the action index vectors with the highest similarity are selected in turn for matching; after each match, the matching result is updated;

[0040] In each matching process, the hierarchy of the current matching result is evaluated and the continuity index of the matching result is calculated; if the hierarchy of the current matching result is poor, the weight needs to be readjusted;

[0041] Determine the matching degree between the skeletal position information and the motion index; output the final matching result for use in traditional Chinese medicine rehabilitation training; the output of the matching result includes key information of the matching degree, the matching motion index, and the matching skeletal position information.

[0042] As a further improvement of the present invention, the process of calculating the continuity index of the matching result includes:

[0043] Extract global feature vectors from the skeleton position information and action index. The global feature vectors include the global features of the overall motion trend, posture change and speed change. Calculate the cosine similarity between global feature vectors, calculate the Euclidean distance between global feature vectors, and calculate the global continuity index by combining the cosine similarity and the Euclidean distance.

[0044] Extract local feature vectors from the skeletal position information and action index, the local feature vectors include the local characteristics of the motion trajectory of the local joint and the activation state of the local muscle; calculate the Pearson correlation coefficient between the local feature vectors, calculate the mutual information between the local feature vectors, and calculate the local continuity index by combining the correlation analysis and the mutual information;

[0045] Detail feature vectors are extracted from the skeleton position information and action index. The detail feature vectors include detail features of small motion changes and subtle posture adjustments. The entropy value of the detail feature vectors is calculated, and the KL divergence between the detail feature vectors is calculated. The detail continuity index is calculated by combining entropy analysis and KL divergence.

[0046] To achieve the above object, the present invention also provides the following technical solutions:

[0047] An intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine is applied to the intelligent recommendation enrichment of bone injury rehabilitation training plan based on traditional Chinese medicine. The intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine comprises:

[0048] The first data set construction module is used to obtain the image information of all bones of the bone injury target and mark the position information; according to the function of the organ where the bone is located, the image information is divided according to the label to obtain bone data sets with different functions;

[0049] The second data set construction module is used to obtain the bone name and the bone injury type of the bone injury through the input device, obtain the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set, and use the first position information and the second position information as a special information set; delete the special data set from the bone data set to obtain the target data set;

[0050] The target action output module is used to obtain the position information of the bones to be subjected to bone injury rehabilitation training from the target data set, select matching Tai Chi, Ba Duan Jin or Wu Qin Xi content from the pre-stored traditional Chinese medicine rehabilitation training database; delete the relevant actions of the second position information from the action data set corresponding to the content; guide the trainee to perform bone injury rehabilitation training through the visual terminal, and at the same time prompt the cycle and time information of the bone injury rehabilitation training.

[0051] The present invention acquires and labels bone image information. By acquiring the image information of all bones of the bone injury target, the morphology and position of the bones can be intuitively understood, providing basic data for analysis and processing; labeling the position information of the bones helps to accurately identify and locate the specific part of the bone injury, providing an accurate target for subsequent rehabilitation training; the labels are divided according to the function of the organ position where the bones are located (for example, ribs are marked as chest bones), making the data set more structured and orderly, and convenient for classification and retrieval. Significance: Through detailed image and position information, bone injuries can be diagnosed more accurately, providing a scientific basis for rehabilitation training; the division of functional labels makes the data set more refined, which helps to improve the efficiency and accuracy of data processing. The bone injury information is acquired and the data set is processed. The bone name and type of the bone injury are acquired through the input device, and the specific part and nature of the bone injury can be quickly located; the first position information of the bone injury and the second position information of the associated bone are extracted from the bone data set to form a special information set, providing key data for rehabilitation training; the special data set is deleted to obtain the target data set, making the data set purer, which is convenient for rehabilitation training recommendation, and at the same time, it is avoided that the given training content affects the bone where the bone injury is located and the associated bones, and also for providing a personalized rehabilitation training plan. Significance: By extracting and processing the position information, the bone injury site can be accurately located, providing a precise target for rehabilitation training; deleting data unrelated to the bone injury makes the target data set purer, which helps to improve the pertinence and effectiveness of rehabilitation training. Select rehabilitation training content and guide the training, select matching Tai Chi, Ba Duan Jin or Wu Qin Xi from the pre-stored TCM rehabilitation training database to ensure the scientificity and effectiveness of rehabilitation training; delete the actions related to the second position information to avoid secondary damage to the bone injury site during rehabilitation training; guide the trainee to conduct bone injury rehabilitation training through the visual terminal, provide intuitive training guidance, and prompt the training cycle and time to ensure the systematicness and continuity of the training. Significance: Select appropriate rehabilitation training content according to the specific situation of the bone injury, realize personalized rehabilitation training, and improve the rehabilitation effect; by deleting the actions related to the bone injury site, avoid secondary damage to the bone injury site during rehabilitation training, realize the identification of biological characteristics, and ensure the safety of the rehabilitation process; through visual guidance and time prompts, ensure the systematicness and continuity of rehabilitation training, which helps to improve the efficiency and effect of rehabilitation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic diagram of the steps of an embodiment of a method for intelligently recommending a bone injury rehabilitation training plan based on traditional Chinese medicine according to the present invention;

[0053] Figure 2 A schematic flow chart of steps for obtaining bone data sets of different functions according to an embodiment of the intelligent recommendation method for bone injury rehabilitation training plan based on traditional Chinese medicine of the present invention;

[0054] Figure 3 A schematic flow chart of steps for obtaining a target data set in an embodiment of an intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine of the present invention;

[0055] Figure 4 A schematic flow chart of steps for selecting matching Tai Chi, Ba Duan Jin or Wu Qin Xi and other contents according to an embodiment of an intelligent recommendation method for bone injury rehabilitation training plan based on traditional Chinese medicine of the present invention;

[0056] Figure 5 This is a functional module diagram of an embodiment of an intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine of the present invention;

[0057] Figure 6 It is a structural schematic diagram of an embodiment of an electronic device of the present invention;

[0058] Figure 7 It is a schematic structural diagram of an embodiment of the storage medium of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0060] The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0061] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0062] like Figure 1 As shown, this embodiment provides an embodiment of a method for intelligently recommending a bone injury rehabilitation training program based on traditional Chinese medicine. In this embodiment, the method for intelligently recommending a bone injury rehabilitation training program based on traditional Chinese medicine specifically includes the following steps:

[0063] Step S1: Obtain image information of all bones of the bone injury target and mark the position information; use the function of the organ position where the bone is located as a label, divide the image information according to the label, and obtain bone data sets with different functions;

[0064] Step S2: obtaining the bone name and the type of bone injury through the input device, obtaining the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set, and taking the first position information and the second position information as a special information set; deleting the special data set from the bone data set to obtain the target data set;

[0065] Step S3: Obtain the position information of the bones to be subjected to bone injury rehabilitation training from the target data set, select matching Tai Chi, Ba Duan Jin or Wu Qin Xi and other contents from the pre-stored TCM rehabilitation training database; delete the relevant actions of the second position information from the action data set corresponding to the content; guide the trainee to perform bone injury rehabilitation training through the visual terminal, and prompt information such as the cycle and time of bone injury rehabilitation training.

[0066] Preferably, step S1 of this embodiment obtains and labels the bone image information. By obtaining the image information of all bones of the bone injury target, the morphology and position of the bones can be intuitively understood, and accurate recognition of biological characteristics can be achieved, providing basic data for analysis and processing; labeling the position information of the bones helps to accurately identify and locate the specific parts of the bone injury, providing accurate targets for subsequent rehabilitation training; and the labels are divided according to the functions of the organ positions where the bones are located (such as ribs, labeled as chest bones), making the data set more structured and orderly, and easy to classify and retrieve. Significance: Through detailed images and position information, bone injuries can be diagnosed more accurately, providing a scientific basis for rehabilitation training; the division of functional labels makes the data set more refined, which helps to improve the efficiency and accuracy of data processing. Step S2 obtains bone injury information and processes the data set. The name and type of the bone of the bone injury are obtained through the input device, so that the specific part and nature of the bone injury can be quickly located; the first position information of the bone injury and the second position information of the associated bone are extracted from the bone data set to form a special information set to provide key data for rehabilitation training; the special data set is deleted to obtain the target data set, so that the data set is purer, which is convenient for rehabilitation training recommendations. At the same time, it is avoided that the given training content affects the bone where the bone injury is located and the associated bones, and also in order to provide a personalized rehabilitation training plan. Significance: By extracting and processing the position information, the bone injury site can be accurately located to provide a precise target for rehabilitation training; deleting data unrelated to the bone injury makes the target data set purer, which helps to improve the pertinence and effectiveness of rehabilitation training. Step S3 selects rehabilitation training content and guides the training, selects matching Tai Chi, Ba Duan Jin or Wu Qin Xi and other content from the pre-stored TCM rehabilitation training database, and makes the content into animation for display; ensures the scientificity and effectiveness of rehabilitation training; deletes the actions related to the second position information to avoid secondary injury to the bone injury part during rehabilitation training; guides the trainee to conduct bone injury rehabilitation training through the visual terminal, provides intuitive training guidance, and prompts the training cycle and time to ensure the systematicness and continuity of the training. Significance: Select appropriate rehabilitation training content according to the specific situation of bone injury, realize personalized rehabilitation training, and improve rehabilitation effect; by deleting the actions related to the bone injury part, avoid secondary injury to the bone injury part during rehabilitation training, and ensure the safety of the rehabilitation process; through visual guidance and time prompts, ensure the systematicness and continuity of rehabilitation training, which helps to improve rehabilitation efficiency and effect.

[0067] In summary, this embodiment achieves personalized, precise and systematic bone injury rehabilitation training through detailed image information acquisition, accurate location information annotation, scientific rehabilitation training content selection and visual training guidance. It not only improves the scientificity and effectiveness of rehabilitation training, but also ensures the safety and continuity of the rehabilitation process, which has important clinical application value and social significance.

[0068] Tai Chi, Ba Duan Jin and Wu Qin Xi in this embodiment are all traditional Chinese health-preserving methods. Through specific movements and breathing methods, they can effectively regulate qi and blood and promote physical health. Tai Chi is known for its slow and gentle movements and deep breathing, which can effectively regulate qi and blood; Tai Chi's movements are slow and gentle, emphasizing the relaxation of the body and the coordination of breathing; through slow movements, blood circulation can be promoted and the flow of qi and blood can be enhanced; Tai Chi emphasizes "qi sinking to Dantian", that is, through deep breathing, the breath sinks to the Dantian area, which helps to enhance the circulation and regulation of qi and blood; Tai Chi can reconcile yin and yang, balance qi and blood, enhance the body's immunity and resistance, and has a significant effect on regulating qi and blood. Ba Duan Jin is an ancient fitness method with simple and easy-to-learn movements, suitable for people of all ages, and can effectively regulate qi and blood; the movement design of Ba Duan Jin is scientific and reasonable, and each movement has a specific conditioning effect. For example, "choose health-preserving movements suitable for your own qi and blood conditions with both hands, which can be adjusted according to personal constitution and qi and blood conditions. The following is a detailed description of how to choose Tai Chi, Ba Duan Jin, and Wu Qin Xi movements according to different qi and blood: People with insufficient qi and blood usually show symptoms such as pale complexion, fatigue, and easy fatigue. It is suitable to choose some movements that can replenish qi and nourish blood. Tai Chi can choose movements such as "Cloud Hands" and "Single Whip", which can promote the flow of qi and blood and enhance physical strength and energy. Ba Duan Jin can choose movements such as "Two Hands Climbing the Feet to Strengthen the Kidneys and Waist" and "Shaking the Head and Wagging the Tail to Remove Heart Fire", which can help to replenish the kidneys and replenish qi and strengthen qi and blood. Wu Qin Xi can choose movements such as "Bear Play" and "Deer Play", which can enhance physical strength and endurance and help to replenish qi and nourish blood. People with qi and blood stasis usually show symptoms such as local pain, swelling, and poor blood circulation. It is suitable to choose some movements that can activate blood circulation and remove blood stasis. Tai Chi can choose movements such as "Grasping the Bird's Tail" and "Slanting Flying Style", which can promote blood circulation and relieve stasis. Stagnation. For Ba Duan Jin, you can choose movements such as "Drawing the bow left and right like shooting an eagle" and "Looking back after five kinds of fatigue and seven kinds of injuries". These movements help to stretch the muscles and bones and promote the circulation of qi and blood. For Wu Qin Xi, you can choose movements such as "Tiger Play" and "Ape Play". These movements can enhance the flexibility of the body and blood circulation and help to remove blood stasis. Examples of movements that are not suitable for people with bone injuries or joint problems: Tai Chi, Single Whip: This movement involves single-leg support and a large body twist, which may put pressure on the knee and hip joints and is not suitable for people with knee or hip problems. Oblique Flying Style: This movement requires a large range of motion of the shoulder joint, which may put pressure on the shoulder joint, and is not suitable for people with shoulder problems. Ba Duan Jin: Climbing the feet with both hands to strengthen the kidney and waist: This movement requires bending and stretching, which may put pressure on the lumbar spine and knee joints, and is not suitable for people with lumbar spine or knee problems. Draw the bow left and right like shooting an eagle: This movement requires a large range of motion of the shoulder and elbow joints, which may put pressure on the shoulder and elbow joints, and is not suitable for people with shoulder or elbow problems.Five Animal Exercises: Tiger Exercises require greater twisting and stretching of the body, which may put pressure on the lumbar spine and knee joints, and are not suitable for people with lumbar spine or knee joint problems; Monkey Exercises require a greater range of motion for the shoulder and elbow joints, which may put pressure on the shoulder and elbow joints, and are not suitable for people with shoulder or elbow joint problems.

[0069] Furthermore, if Figure 2 As shown, the process of obtaining the skeleton data sets of different functions in step S1 specifically includes the following steps:

[0070] Step S11: using medical imaging equipment to obtain image information of all bones of the bone injury target; performing denoising, enhancement and standardization processing on the acquired image; using edge detection to remove noise and artifacts in the image to obtain a preprocessed image;

[0071] The acquisition process of the preprocessed image includes the following steps:

[0072] Extract image features through convolution operation:

[0073]

[0074] Where: I c (x, y) is the value of the feature map output by the convolutional layer at position (x, y), W i,j is the weight of the convolution kernel at position (i, j), b is the bias term, and k is the size of the convolution kernel;

[0075] Introducing nonlinear characteristics to enhance the expressiveness of the model;

[0076] I a (x,y)=ReLU(I c (x,y))=max(0,I c (x,y)

[0077] Where: I a (x, y) is the value of the feature map at position (x, y) after nonlinear activation, and ReLU is the rectified linear unit activation function; I c (x,y) The value of the feature map output by the convolutional layer at position (x,y);

[0078] The pooling layer reduces the dimension of the feature map, reduces the amount of computation, and retains important features;

[0079] I p (x,y)=maxpool(I a (x,y))=max i,j∈pool_size I a (x+i,y+j)

[0080] Where: Ip (x, y) is the value of the feature map output by the pooling layer at position (x, y), maxpool is the maximum pooling operation, and pool_size is the size of the pooling window;

[0081] The fully connected layer maps the feature map to the final edge detection result;

[0082] I fc (x,y)=W fc I p (x,y)+b fc

[0083] Where: I fc (x, y) is the value of the feature map output by the fully connected layer at position (x, y), W fc is the weight matrix of the fully connected layer, b fc is the bias term of the fully connected layer;

[0084] Edge detection results, generate the final edge detection image.

[0085] I edge (x,y)=sigmoid(I fc (x,y)

[0086] Where: I edge (x, y) is the pixel value of the final edge detection image at position (x, y), and sigmoid is the sigmoid activation function, which is used to map the output value to the range [0, 1]. The formula and steps together constitute a complex edge detection process based on deep learning, which combines the powerful feature extraction capability of convolutional neural networks and the advantages of traditional edge detection technology to help extract more accurate and clear bone edge information from the image.

[0087] Step S12: Use a convolutional neural network to automatically annotate the bone image, automatically identify and annotate the location information of the bones; functionally classify the annotated bones according to the function of the organ where the bones are located. For example, annotate the ribs as chest bones, annotate the femur as lower limb bones, etc.; generate a unique label for each functional classification, and associate the label with the corresponding bone image;

[0088] Step S13: According to the generated functional labels, the annotated bone images are divided according to the labels, and a tree-structured bone data set of different functions is constructed, where the root node is the bone data set and the child nodes are data sets of each functional classification; each data set contains all bone images under the same functional classification;

[0089] For example: root node: bone dataset; child node 1: chest bone dataset; grandchild node 1.1: rib dataset; grandchild node 1.2: sternum dataset; child node 2: lower limb bone dataset; grandchild node 2.1: femur dataset; grandchild node 2.2: tibia dataset; child node 3: upper limb bone dataset; grandchild node 3.1: humerus dataset; grandchild node 3.2: radius dataset; child node 4: spinal bone dataset; grandchild node 4.1: cervical vertebrae dataset; grandchild node 4.2: lumbar vertebrae dataset.

[0090] Preferably, the image preprocessing of step S11 of this embodiment improves the quality of the image through denoising and image enhancement technology, and reduces the impact of noise and artifacts on subsequent analysis; the standardization process ensures the consistency of the image under different devices and conditions, which is convenient for subsequent unified analysis and processing; the edge detection helps to clearly identify the outline of the bone, remove irrelevant background information, and make the bone image clearer and more accurate. Significance: The preprocessing step provides high-quality image data for bone recognition and classification, ensuring the accuracy and reliability of the analysis. Step S12 automatic labeling and functional classification, using convolutional neural network (CNN) for automatic labeling, greatly improves the efficiency and accuracy of labeling, and reduces the error of manual labeling; classification according to the function of the organ position where the bone is located, makes the bone data set more structured and orderly, which is convenient for subsequent functional analysis and application. Significance: Automatic labeling and functional classification make the bone data set more refined and professional, and provide more accurate and targeted data support for medical research and clinical applications. Step S13 data set construction, constructing a tree-structured bone data set, making data management and retrieval more efficient and convenient; dividing the data set according to the functional label, so that the bone images under the same functional classification are stored centrally, which is convenient for targeted research and application. Significance: The structured management of data sets makes the use of bone data more efficient and flexible, provides strong data support for medical research and clinical applications, and helps promote the development and progress of related fields.

[0091] In summary, the construction of the skeleton data set of the present embodiment not only realizes efficient and accurate processing and classification in technology, but also provides powerful data support in practical applications, which is helpful to promote the development of medical research and clinical applications. The technical effect and significance of each step complement each other and jointly construct a high-quality and efficient skeleton data set. Through the above-mentioned hierarchical process, the skeleton data set of each different function is obtained; not only the hierarchy and scalability of the data set are guaranteed, but also the purity and efficiency of the data set are improved through duplicate checking and optimization processing. The data set finally constructed provides scientific, accurate and reliable basic data for subsequent bone injury rehabilitation training.

[0092] Furthermore, if Figure 3As shown, the process of obtaining the target data set in step S2 specifically includes the following steps:

[0093] Step S21: Create an empty target data set to store the data that meets the requirements screened from the original bone data set; preprocess the bone data set to unify the data format; extract key features from the bone data set, such as the specific location of the bones, functional labels, etc.;

[0094] Step S22: using a hierarchical method to filter data, gradually eliminating data related to the special information set; filtering out data directly related to the special information set according to the bone name and type of bone injury; further filtering out data indirectly related to the special information set, such as the second position information of the bone associated with the bone injury; and performing a duplicate check operation;

[0095] Step S23: Integrate the hierarchically screened data into the target data set and verify the target data set.

[0096] Preferably, in step S21 of this embodiment, data preprocessing and feature extraction provide a storage space for the screened data to ensure orderly management and efficient use of data; by standardizing the bone data set, the problem of inconsistent data format is eliminated to ensure the comparability and operability of data; extract key features such as the specific position and function label of the bone to provide basic data support for data screening and analysis. The significance achieved: by unifying the data format, it is ensured that data from different sources can be analyzed and processed on the same platform, improving the efficiency and accuracy of data processing; extracting key features helps to understand the bone data more deeply, providing strong support for subsequent screening and analysis, and ensuring that the screened data is representative and practical. Step S22 hierarchical data screening, through a hierarchical screening method, gradually eliminates data that does not meet the requirements to ensure the purity of the final target data set; according to the bone name and bone injury type of bone injury, screen out data directly related to the special information set to ensure the pertinence of the data; further screen the indirect related data such as the second position information of the bone associated with the bone injury to ensure the comprehensiveness and integrity of the data; through the duplicate checking operation, it is ensured that there is no duplicate data in the target data set to improve the quality and reliability of the data. Significance achieved: Through the hierarchical screening method, the purity of the final target data set is guaranteed to avoid interference from irrelevant data; data directly and indirectly related to the special information set are screened out to ensure the pertinence and comprehensiveness of the data, providing high-quality data support for subsequent research and analysis; through duplicate checking operations, the quality of the target data set is guaranteed to avoid the impact of duplicate data on the analysis results. Step S23 Data integration and verification Integrate the data after hierarchical screening into the target data set to ensure the integrity and consistency of the data; verify the target data set to ensure the accuracy and reliability of the data. Significance achieved: Through data integration, the integrity of the target data set is guaranteed to avoid the impact of missing data on the analysis results; through data verification, the accuracy and reliability of the target data set are guaranteed to provide reliable data support for analysis.

[0097] In summary, the target dataset construction process of this embodiment not only ensures the standardization, purity and pertinence of the data, but also improves the quality and reliability of the data. The ultimate goal is to provide high-quality and reliable data support for analysis and promote scientific research and application development in related fields.

[0098] Furthermore, the process of gradually eliminating data related to the special information set in step S22 specifically includes the following steps:

[0099] Step S221: extract all recorded bone names from the skeletal data set, and accurately match them with the bone names in the special information set; the successfully matched records enter the next level of screening; for the successfully matched records, further extract the bone injury type, and match it with the bone injury type in the special information set; the successfully matched records enter the candidate list of the target data set;

[0100] Step S222: extracting the second position information of the bone associated with the bone injury from the records in the candidate list; matching the extracted second position information with the relevant position information in the special information set; and the successfully matched records are further entered into the candidate list of the target data set;

[0101] Step S223: perform a duplication check on the records in the candidate list and calculate the duplication rate; based on the duplication check results, remove the records with a duplication rate exceeding 30%.

[0102] Preferably, step S221 of this embodiment ensures that each record in the data set is completely consistent with the bone name in the special information set, avoiding mismatching due to name differences; further screening out records that match the bone injury type in the special information set, improving the relevance and pertinence of the data. Significance achieved: By accurately matching the bone name and the bone injury type, it is ensured that the records in the target data set are highly relevant to the special information set, reducing the interference of irrelevant data; making the target data set more focused on the specific bone injury type, providing a high-quality data basis for subsequent analysis and research. Step S222 further refines the screening conditions, and ensures the accuracy of the data by matching the second position information of the bone associated with the bone injury; by matching with the relevant position information in the special information set, further screening out records that meet the conditions. Significance achieved: By introducing the matching of the second position information, the screening process is made more detailed, ensuring that the records in the target data set are also highly consistent with the special information set in terms of position information; making the target data set not only pertinent in terms of bone injury type, but also highly accurate in terms of position information, providing more valuable data for subsequent clinical applications and research. Step S223 calculates the duplication rate, identifies and removes records with a duplication rate of more than 30%, thereby ensuring the uniqueness and reliability of the data set; by removing records with a high duplication rate, redundant information in the data set is reduced, and the quality of the data is improved. Significance achieved: Through the duplication check operation, it is ensured that each record in the target data set is unique, avoiding the impact of duplicate data on the analysis results; removing records with a high duplication rate reduces the noise in the data set, making the target data set purer and more reliable, and providing a solid foundation for subsequent analysis and application.

[0103] In summary, this embodiment gradually eliminates data related to special information sets, which not only improves the accuracy and pertinence of the data set, but also enhances the precision and reliability of the data set, ensuring the high quality of the final target data set and providing a solid foundation for subsequent clinical research, data analysis and application.

[0104] Furthermore, the process of matching the extracted second location information with the relevant location information in the special information set in step S222 specifically includes the following steps:

[0105] Step S2221: constructing Gaussian pyramids for the images in the image information and the standard anatomical position information respectively, for detecting the features of the images at different scales; generating multiple scale space layers through Gaussian blur and downsampling operations; on the basis of the Gaussian pyramid, calculating the Gaussian difference image of two adjacent layers of images, for detecting extreme points in the image, i.e., potential key points;

[0106] Step S2222: In the Gaussian difference image, by comparing the values ​​of each pixel point with the 26 neighboring pixel points around it, extreme points are detected, and extreme points are potential key points; the detected extreme points are located; the edge response is eliminated by calculating the eigenvalues ​​of the Hessian matrix to retain stable key points; the gradient direction and amplitude are calculated for the pixels around each key point; in the local area around the key point, a gradient direction histogram of 36 directions is constructed, and the gradient amplitude of each direction is counted; the direction with the highest peak in the histogram is selected as the main direction of the key point, and a key point with direction is generated;

[0107] Step S2223: Divide the local image area around the key point into 16 sub-areas, calculate the gradient direction histogram in 8 directions in each sub-area, and generate a 128-dimensional feature descriptor; normalize the descriptor; calculate the Euclidean distance between the feature descriptors of the image information and the image in the standard anatomical position information; for each feature descriptor, find the feature descriptor with the smallest Euclidean distance in the other image as a matching pair;

[0108] Step S2224: Set a Euclidean distance threshold. Only when the Euclidean distance between two feature descriptors is less than the threshold, they are considered to be a valid matching pair; screen all matching pairs and eliminate matching results whose Euclidean distance is greater than the set threshold.

[0109] Preferably, step S2221 of this embodiment constructs a Gaussian pyramid and detects extreme points. Through Gaussian blur and downsampling operations, multiple scale space layers are generated, which can detect image features at different scales and enhance the robustness of features; Gaussian difference images of two adjacent layers of images are calculated to detect extreme points in the image, that is, potential key points; and significant feature points in the image can be effectively extracted. Significance: It can detect image features at different scales, is suitable for image matching of different resolutions and viewing angles, and improves the accuracy and robustness of matching; by detecting extreme points, key feature points in the image can be accurately extracted, providing a basis for subsequent feature matching. Step S2222 detects extreme points and generates key points, detects extreme points by comparing the values ​​of each pixel point with the 26 neighboring pixel points around it, and ensures the stability and accuracy of the key points; eliminates edge responses by calculating the eigenvalues ​​of the Hessian matrix, retains stable key points, and improves the reliability of key points; in the local area around the key point, a gradient direction histogram of 36 directions is constructed, and the gradient amplitude of each direction is counted to generate a key point with direction. Significance: By eliminating edge responses and retaining stable key points, the accuracy of matching is improved; key points with directions are generated to enhance the description ability of features and make matching more accurate. Step S2223 generates feature descriptors and matches, divides the local image area around the key points into 16 sub-areas, calculates the gradient direction histograms in 8 directions in each sub-area, generates 128-dimensional feature descriptors, and enhances the description ability of features; calculates the Euclidean distance between the feature descriptors of the image in the image information and the standard anatomical position information, and finds matching pairs. Significance: By generating 128-dimensional feature descriptors, the description ability of features is enhanced and the accuracy of matching is improved; by calculating the Euclidean distance, matching pairs are found, providing a basis for subsequent matching screening. Step S2224 sets a threshold and screens matching pairs, sets a Euclidean distance threshold, and only when the Euclidean distance between two feature descriptors is less than the threshold, it is considered to be a valid matching pair, which improves the reliability of matching; screens all matching pairs, and eliminates matching results with Euclidean distances greater than the set threshold to ensure the accuracy of matching results. Significance: By setting the threshold, the reliability of the matching results is improved and the validity of the matching results is guaranteed; by screening the matching pairs and eliminating the matching results that do not meet the conditions, the matching accuracy is improved.

[0110] In summary, this embodiment can effectively extract key feature points in an image, generate high-dimensional feature descriptors, perform matching by calculating the Euclidean distance, and filter matching pairs by setting a threshold, ultimately achieving high-precision and high-reliability image matching. This process has important application value in the fields of medical image analysis, computer vision, and the like.

[0111] Furthermore, if Figure 4As shown, the process of selecting matching Tai Chi, Ba Duan Jin or Wu Qin Xi in step S3 specifically includes the following steps:

[0112] Step S31: According to the action data set in the traditional Chinese medicine rehabilitation training database, Tai Chi, Ba Duan Jin, Wu Qin Xi and other actions are classified according to the bone parts and functions involved; an index is established for each classified action, and the index contains key features of the action (such as the bones involved, movement angle, intensity, etc.);

[0113] Step S32: matching the extracted skeletal position information with the action index in the traditional Chinese medicine rehabilitation training database, and using cosine similarity to calculate the matching degree between the skeletal position information and the action index;

[0114] Step S33: setting a matching degree threshold, filtering out movements with matching degrees higher than the threshold, and selecting a movement from Tai Chi, Ba Duan Jin or Wu Qin Xi.

[0115] Preferably, the classification and index construction of step S31 of this embodiment can greatly improve the retrieval efficiency of the subsequent matching step by classifying the actions according to the skeletal parts and functions and establishing an index containing key features; standardize the action data set to ensure the uniformity of the data format from different sources, which is convenient for systematic management and application. The significance achieved: through scientific classification and index construction, it is ensured that each action can be accurately retrieved in the database, laying the foundation for matching work; integrating the action data in the traditional Chinese medicine rehabilitation training database to form a systematic resource library, which is convenient for the subsequent rehabilitation training program design and optimization. Step S32 feature matching, using algorithms such as cosine similarity, can accurately calculate the matching degree between the skeletal position information and the action index, and ensure the accuracy of the matching results; the cosine similarity algorithm has a high computational efficiency and can complete the matching calculation of a large amount of data in a short time. The significance achieved: through accurate feature matching, it can provide personalized action selection for each trainee, ensure the pertinence and effectiveness of the training program; use data-driven matching algorithms to ensure the scientificity and objectivity of the rehabilitation training program, and avoid subjective judgment deviations. Step S33: Threshold screening. By setting the matching threshold, the actions that best meet the needs of the trainee can be screened out to optimize the training plan. Threshold screening realizes the automated decision-making process, reduces manual intervention, and improves decision-making efficiency and consistency. Significance achieved: Through threshold screening, it is ensured that the recommended actions are both in line with scientific requirements and highly personalized, which improves the rehabilitation effect of the trainee. Threshold screening is part of systematic management, which ensures the scientificity and systematicness of the entire rehabilitation training process and improves the overall quality of rehabilitation training.

[0116] In summary, this embodiment can efficiently and accurately select the most suitable Tai Chi, Ba Duan Jin or Wu Qin Xi movements for the trainee from the TCM rehabilitation training database, ensuring the scientificity, personalization and systematicness of the rehabilitation training program. This not only improves the effect of rehabilitation training, but also provides strong support for the modernization and intelligence of TCM rehabilitation training.

[0117] Furthermore, the process of using cosine similarity to calculate the matching degree between the skeleton position information and the action index in step S32 specifically includes the following steps:

[0118] Step S321: extracting key features of each action from the TCM rehabilitation training database, including the involved bone parts, movement angles and strengths, etc.; converting the standardized bone position information into feature vectors, where each bone position information is represented as a multidimensional vector, where each dimension corresponds to a specific bone part or movement parameter; converting the key features in each action index into feature vectors, which will be used to match the bone position information;

[0119] Step S322: For each skeleton position information vector, calculate the cosine similarity between it and each action index vector; cosine similarity is a method for measuring the similarity of two vector directions, and the calculation formula is:

[0120]

[0121] In the formula, A i and B i are the i-th component of the skeleton position information vector and the action index vector, respectively, w i is the weight of the i-th component, which is used to adjust the importance of different components, and n is the total dimension of the vector;

[0122] Step S323: Determine the matching degree between the skeleton position information and the action index according to the calculated cosine similarity. The higher the matching degree, the higher the similarity between the skeleton position information and the action index.

[0123] Preferably, in step S321 of this embodiment, feature extraction and vectorization are used to extract the key features of each action from the TCM rehabilitation training database, including the involved bone parts, movement angles and strengths, etc., which is helpful to capture the core elements of the action; the bone position information is standardized to ensure the uniformity of different data sources; the standardized bone position information is converted into a feature vector, and each bone position information is represented as a multidimensional vector, so that the data can be mathematically operated and compared. Significance: It provides the necessary data basis for similarity calculation; it ensures that different actions and bone position information are compared on the same dimension, which improves the accuracy of matching. Step S322 cosine similarity calculation, for each bone position information vector, calculates the cosine similarity between it and each action index vector; cosine similarity is a method to measure the similarity of the directions of two vectors, which can effectively evaluate the similarity of two vectors; by introducing weights, the importance of different components is adjusted, so that the calculation results are more in line with actual needs. Significance: Through cosine similarity calculation, the similarity between bone position information and action index can be quantified, providing a basis for action recommendation; weight adjustment enables the system to make personalized action recommendations according to the characteristics and needs of different users. Step S323: Matching degree screening and action recommendation: Set a matching degree threshold to screen out actions with a matching degree higher than the threshold, which helps to filter out irrelevant actions and improve the accuracy of recommendations; select the action with the highest matching degree from the screened actions as the final recommended action, ensuring that the recommended action is most consistent with the user's actual needs; adjust the matching degree threshold based on user feedback so that the system can continuously optimize the recommendation effect. Significance: Through matching degree screening and action selection, it is ensured that the recommended action is highly matched with the user's bone position information, which improves the accuracy of the recommendation; adjust the matching degree threshold based on user feedback so that the system can continuously adapt to user needs and improve user experience; through personalized action recommendations, the rehabilitation training needs of different users are met and the rehabilitation effect is improved.

[0124] In summary, this embodiment can extract key features from the traditional Chinese medicine rehabilitation training database, convert the skeletal position information and action index into feature vectors, and evaluate the similarity through cosine similarity calculation. According to the matching screening and user feedback, the most suitable action is recommended; not only the accuracy and personalization of the action recommendation are improved, but also the system is continuously optimized through user feedback, which improves the overall user experience and rehabilitation effect. Through the above hierarchical process, this embodiment can effectively use cosine similarity to calculate the matching degree between the skeletal position information and the action index, ensuring that the selected action is highly matched with the actual needs of the user.

[0125] Furthermore, the process of determining the matching degree between the skeleton position information and the action index in step S323 specifically includes the following steps:

[0126] Step S3231: sorting the filtered action index vectors in descending order according to their cosine similarity with the skeleton position information vectors; selecting the action index vectors with the highest similarity from the sorted list for matching; and updating the matching results after each matching;

[0127] Step S3232: During each matching process, the hierarchy of the current matching result is evaluated and the continuity index of the matching result is calculated; if the hierarchy of the current matching result is poor, the weight needs to be readjusted;

[0128] Step S3233: Determine the matching degree between the skeletal position information and the motion index; output the final matching result for use in traditional Chinese medicine rehabilitation training; the output of the matching result includes key information such as the matching degree, the matching motion index, and the matching skeletal position information.

[0129] Preferably, in step S3231 of this embodiment, similarity sorting and preliminary matching are performed by sorting the action index vectors in descending order according to the cosine similarity to ensure that the action index vectors with the highest similarity are matched first; the action index vectors with the highest similarity are selected in turn for matching, and the matching results are updated to ensure the efficiency and accuracy of the matching process. The significance achieved: through sorting and preliminary matching, the action indexes most similar to the skeletal position information are quickly screened out to reduce the complexity of matching; the action indexes with the highest similarity are matched first to ensure the accuracy of the preliminary matching results, providing a basis for hierarchical evaluation and optimization. Step S3232 hierarchical evaluation and weight adjustment: by calculating the continuity index of the matching result, the hierarchy of the current matching result is evaluated to ensure the structural stability and continuity of the matching result; according to the hierarchical evaluation result, the weight is dynamically adjusted to optimize the matching strategy to ensure the hierarchy and accuracy of the matching result. The significance achieved: through hierarchical evaluation and weight adjustment, the hierarchical structure of the matching result is guaranteed to be stable, and the matching result is prevented from being broken or discontinuous; the weight is dynamically adjusted to optimize the matching strategy to improve the accuracy and reliability of the matching result, meeting the actual needs of traditional Chinese medicine rehabilitation training. Step S3233: Determination and output of the final matching result. According to the hierarchical matching and optimized results, the final matching degree between the skeletal position information and the action index is determined to reflect the similarity between the two. The final matching result is output, including key information such as matching degree, matched action index, and matched skeletal position information, for use in TCM rehabilitation training. Significance achieved: Through the determination and output of the final matching result, accurate and reliable matching information is provided for TCM rehabilitation training, supporting the scientificity and effectiveness of the training process. The output matching result can be directly applied to TCM rehabilitation training to help patients perform correct action training and improve rehabilitation effects.

[0130] In summary, the matching process of the present embodiment not only improves the efficiency and accuracy of matching, but also ensures the hierarchy and stability of the matching results. The matching results of the final output can provide scientific and reliable support for Chinese medicine rehabilitation training, help patients to carry out effective rehabilitation training, and achieve the expected rehabilitation effect. Through the above-mentioned hierarchical matching process, the present embodiment can effectively improve the matching degree between the skeletal position information and the action index, ensure the accuracy and hierarchy of the matching results, and meet the needs of applications such as Chinese medicine rehabilitation training.

[0131] Furthermore, the process of calculating the continuity index of the matching result in step S3232 specifically includes the following steps:

[0132] Step S32321: extracting a global feature vector from the skeleton position information and the action index, the global feature vector usually including global features such as the overall motion trend, posture change and speed change; calculating the cosine similarity between the global feature vectors, calculating the Euclidean distance between the global feature vectors, and combining the cosine similarity and the Euclidean distance to calculate the global continuity index;

[0133] Cosine similarity is used to measure the similarity between two vector directions. The formula is as follows:

[0134]

[0135] Where A and B are two global eigenvectors, · represents the vector dot product, ||A|| and ||B|| represent the vector modulus;

[0136] Euclidean distance is used to measure the actual distance between two vectors in space. The formula is as follows:

[0137]

[0138] In the formula, A i and B i is the i-th component of vectors A and B, and n is the dimension of the vector;

[0139] Combining cosine similarity and Euclidean distance, the global continuity index is calculated using the weighted average method. The formula is as follows:

[0140] Global Continuity=α·Cosine Similarity+β·Euclidean Distance

[0141] In the formula, α and β are weight coefficients used to balance the influence of the two measurement methods;

[0142] Step S32322: extracting local feature vectors from the skeletal position information and the action index, the local feature vectors including local features such as the motion trajectory of the local joints and the activation state of the local muscles; calculating the Pearson correlation coefficient between the local feature vectors, calculating the mutual information between the local feature vectors, and combining the correlation analysis and the mutual information to calculate the local continuity index;

[0143] Among them, the Pearson correlation coefficient is used to measure the linear correlation between two variables. The formula is as follows:

[0144]

[0145] In the formula, X and Y are two local eigenvectors, and is the mean of two local eigenvectors;

[0146] Mutual information is used to measure the degree of nonlinear correlation between two variables. The formula is as follows:

[0147]

[0148] Where p(x,y) is the joint probability distribution, p(x) and p(y) are the marginal probability distributions;

[0149] The local continuity index is calculated using the weighted average method, and the formula is as follows:

[0150] Local Continuity=γ·Pearson Correlation+δ·Mutual Information

[0151] In the formula, γ and δ are weight coefficients used to balance the influence of the two measurement methods;

[0152] Step S32323: extracting detail feature vectors from the skeleton position information and the action index, the detail feature vectors including detail features such as slight motion changes and slight posture adjustments; calculating the entropy of the detail feature vectors, calculating the KL divergence between the detail feature vectors, and combining the entropy analysis and the KL divergence to calculate the detail continuity index;

[0153] Entropy is used to measure the degree of confusion of information, and the formula is as follows:

[0154]

[0155] In the formula, p(x i ) is the probability distribution of the i-th element in the feature vector X;

[0156] KL divergence is used to measure the difference between two probability distributions. The formula is as follows:

[0157]

[0158] Where P and Q are the probability distributions of two detail feature vectors;

[0159] The detail continuity index is calculated using the weighted average method, and the formula is as follows:

[0160] Detail Continuity=∈·Entropy+ζ·KL Divergence

[0161] In the formula, ∈ and ζ are weight coefficients used to balance the influence of the two measurement methods;

[0162] Through the above detailed steps, the continuity of the matching results can be comprehensively evaluated at the global, local and detail levels to ensure the hierarchy and consistency of the matching results.

[0163] Preferably, the extraction of global feature vectors and calculation of continuity index in step S32321 of this embodiment can capture the overall dynamic characteristics of the action by extracting global features such as overall motion trends, posture changes and speed changes; measure the similarity of the directions of two vectors, which is suitable for evaluating the overall consistency of the action; measure the actual distance between the two vectors in space, which is suitable for evaluating the overall difference of the action; and comprehensively evaluate the continuity at the global level by taking into account the direction similarity and spatial distance through the weighted average method. The significance achieved: ensure the consistency of the matching results in overall motion trends and posture changes, and improve the reliability and accuracy of the matching results; through the extraction and analysis of global features, the overall dynamic characteristics of the action can be better understood and described, providing basic data for subsequent analysis. Step S32322 extracts local feature vectors and calculates continuity indexes. By extracting local features such as the motion trajectory of local joints and the activation state of local muscles, the detailed dynamic characteristics of the action can be captured; the linear correlation between two variables is measured, which is suitable for evaluating the linear consistency of local joint motion; the nonlinear correlation between two variables is measured, which is suitable for evaluating the nonlinear consistency of local muscle activation state; through the weighted average method, linear and nonlinear correlations are comprehensively considered to comprehensively evaluate the continuity at the local level. The significance achieved: ensure the consistency of the matching results in local joint motion and muscle activation state, and improve the precision and accuracy of the matching results; through the extraction and analysis of local features, the detailed dynamic characteristics of the action can be better understood and described, and detailed data can be provided for subsequent analysis. Step S32323 extracts the detail feature vector and calculates the continuity index. By extracting detail features such as small motion changes and subtle posture adjustments, the microscopic dynamic characteristics of the action can be captured; the degree of information confusion is measured, which is suitable for evaluating the complexity and diversity of detail features; the difference between two probability distributions is measured, which is suitable for evaluating the distribution consistency of detail features; through the weighted average method, the complexity of information and the consistency of distribution are comprehensively considered to comprehensively evaluate the continuity of the detail level. The significance achieved: ensure the consistency of the matching results in small motion changes and subtle posture adjustments, and improve the precision and accuracy of the matching results; through the extraction and analysis of detail features, the microscopic dynamic characteristics of the action can be better understood and described, and more detailed data can be provided for subsequent analysis.

[0164] In summary, this embodiment ensures the hierarchy and consistency of the matching results by comprehensively evaluating the continuity of the matching results at the global, local and detail levels, thereby improving the reliability and accuracy of the matching results. Each step together builds a multi-level and multi-dimensional continuity evaluation system, providing comprehensive technical support for action matching and analysis.

[0165] Furthermore, the process of deleting the action related to the second location information from the action data set corresponding to the content in step S3 specifically includes the following steps:

[0166] Step S34: extracting the second position information of the bone associated with the bone injury from the special information set, and clarifying the bone parts that need to be excluded; constructing an exclusion list based on the second position information, listing all related actions that need to be deleted from the action data set;

[0167] Step S35: traverse the action data set, filter out actions involving the skeletal parts in the exclusion list according to the skeletal parts in the exclusion list, and mark them as actions to be deleted; calculate the similarity between the actions to be deleted, and evaluate the similarity between the actions; and evaluate whether the duplication rate meets the standard according to the similarity calculation result;

[0168] Step S36: based on the duplicate checking result, delete the actions involving the second position information from the action data set to ensure that the deleted action data set no longer contains related actions; update the action data set to ensure that the remaining actions do not involve the second position information.

[0169] Preferably, step S34 of this embodiment extracts the second position information of the bone associated with the bone injury, accurately extracts the second position information of the bone part associated with the bone injury from the special information set; constructs an exclusion list based on the extracted second position information, and clarifies the relevant actions that need to be deleted from the action data set. The significance achieved: ensure that the bone parts that need to be excluded are accurately identified and located; provide a clear goal for the step, that is, clarify which actions need to be deleted from the data set. Step S35 screens and marks the actions to be deleted, traverses the action data set, screens out the actions involving the bone parts in the exclusion list, and marks them as actions to be deleted; calculates the similarity between the actions to be deleted, and evaluates the similarity between the actions; and evaluates whether the duplication rate meets the standard based on the similarity calculation results. The significance achieved: ensure that only actions involving specific bone parts are marked as to be deleted; evaluate the similarity between the actions through similarity calculation, and ensure that the deleted actions will not affect the overall structure and diversity of the data set; ensure that the duplication rate meets the standard to avoid accidental deletion or omission. Step S36 deletes and updates the action data set, deletes the actions involving the second position information from the action data set based on the duplication result; updates the action data set to ensure that the remaining actions do not involve the second position information. Significance achieved: Ensure that the action dataset no longer contains actions related to specific skeletal parts, maintain the purity of the dataset; the updated dataset remains consistent, ensuring the accuracy and reliability of subsequent analysis and application.

[0170] In summary, this embodiment achieves accurate screening and purification of the action data set, ensuring that the actions in the data set do not involve specific skeletal parts, thereby improving the quality and applicability of the data set. It is of great significance for action analysis, model training and application, and ensures the accuracy and reliability of the data set.

[0171] like Figure 5 As shown, this embodiment also provides an embodiment of an intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine. In this embodiment, the intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine is applied to the intelligent recommendation method for bone injury rehabilitation training plan based on traditional Chinese medicine in the above embodiment. The intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine includes a first data set construction module 1, a second data set construction module 2, and a target action output module 3 which are electrically connected in sequence;

[0172] Among them, the first data set construction module 1 is used to obtain the image information of all bones of the bone injury target and mark the position information; according to the function of the organ position where the bone is located, the image information is divided according to the label to obtain the bone data set of different functions; the second data set construction module 2 is used to obtain the bone name and bone injury type of the bone injury through the input device, obtain the first position information of the bone injury, and the second position information of the bone associated with the bone injury from the bone data set, and use the first position information and the second position information as a special information set; delete the special data set from the bone data set to obtain the target data set; the target action output module 3 is used to obtain the position information of the bones to be subjected to bone injury rehabilitation training from the target data set, select matching Tai Chi, Ba Duan Jin or Wu Qin Xi and other contents from the pre-stored traditional Chinese medicine rehabilitation training database; delete the relevant actions of the second position information from the action data set corresponding to the content; guide the trainee to perform bone injury rehabilitation training through the visual terminal, and prompt information such as the cycle and time of bone injury rehabilitation training.

[0173] Preferably, the first data set construction module 1 of this embodiment obtains the image information of all bones of the bone injury target and marks the position information; according to the function of the organ position of the bone as a label, the image information is divided according to the label to obtain bone data sets with different functions. The significance achieved: by marking and labeling, the standardization and structuring of image information are ensured, which is convenient for data processing and analysis; by classifying bones according to function, it is helpful to formulate personalized rehabilitation training plans for bones with different functions. The second data set construction module 2 obtains the bone name and bone injury type of the bone injury through the input device; obtains the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set to form a special information set; deletes the special data set from the bone data set to obtain the target data set. The significance achieved: by obtaining the specific location information of the bone injury, the pertinence and effectiveness of the rehabilitation training are ensured; by deleting the special data set related to the bone injury, the purity of the target data set is ensured to avoid interference factors. The target action output module 3 obtains the position information of the bone to be trained for bone injury rehabilitation from the target data set; selects matching Tai Chi, Ba Duan Jin or Wu Qin Xi and other contents from the pre-stored TCM rehabilitation training database; deletes the relevant actions of the second position information from the action data set corresponding to the content; guides the trainee to perform bone injury rehabilitation training through the visual terminal, and prompts the cycle and time of bone injury rehabilitation training and other information. The significance achieved: According to the specific location and type of bone injury, recommend the most suitable TCM rehabilitation training content to improve the rehabilitation effect; delete unsuitable actions to ensure the safety and effectiveness of training; provide intuitive training guidance through the visual terminal to help the trainee perform rehabilitation training correctly and improve the efficiency and effect of training.

[0174] In summary, this embodiment provides accurate rehabilitation training programs according to the specific conditions of bone injuries to improve the rehabilitation effect; ensures the scientificity and reliability of the training programs through data standardization and purification; provides intuitive training guidance through visual terminals to improve user experience and enhance training compliance. This embodiment provides a scientific, efficient, and safe solution for bone injury rehabilitation training through intelligent data processing and personalized recommendations, which has important clinical application value and social significance.

[0175] like Figure 6 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0176] The memory 42 stores program instructions for implementing the intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to any of the above embodiments.

[0177] The processor 41 is used to execute the program instructions stored in the memory 42 to make intelligent recommendations for bone injury rehabilitation training programs based on traditional Chinese medicine.

[0178] The processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip having signal processing capabilities. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0179] Further, Figure 7 The schematic diagram of the structure of the storage medium of an embodiment of the present application is that the storage medium 5 of the embodiment of the present application stores program instructions 51 that can implement all the above methods, wherein the program instructions 51 can be stored in the above storage medium in the form of a biometric recognition application software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0180] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0181] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of a biometric recognition application software functional unit. The above is only an implementation method of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of the present invention, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

[0182] The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims

1. An intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine, characterized in that: The intelligent recommendation method for bone injury rehabilitation training plan based on traditional Chinese medicine includes: Obtain image information of all bones of the target bone injury and mark the location information; use the function of the organ where the bone is located as a label, divide the image information according to the label, and obtain bone data sets with different functions; Obtain the bone name and the type of bone injury through the input device, obtain the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set, and use the first position information and the second position information as a special information set; delete the special data set from the bone data set to obtain the target data set; The position information of the bones to be subjected to bone injury rehabilitation training is obtained from the target data set, and matching Tai Chi, Ba Duan Jin or Wu Qin Xi content is selected from the pre-stored TCM rehabilitation training database; the related actions of the second position information are deleted from the action data set corresponding to the content; the trainee is guided to perform bone injury rehabilitation training through a visual terminal, and the cycle and time information of the bone injury rehabilitation training are prompted at the same time.

2. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 1 is characterized in that: The process of obtaining bone data sets of different functions includes: Use medical imaging equipment to obtain image information of all bones of the bone injury target; perform denoising, enhancement and standardization on the acquired images; use edge detection to remove noise and artifacts in the images to obtain pre-processed images; Use convolutional neural networks to automatically annotate bone images, automatically identify and annotate the location information of bones; functionally classify the annotated bones according to the function of the organ where the bones are located, generate a unique label for each functional classification, and associate the label with the corresponding bone image; According to the generated functional labels, the annotated bone images are divided according to the labels to construct a tree-structured bone dataset for each different function. The root node is the bone dataset and the child nodes are the datasets of each functional classification; each dataset contains all the bone images under the same functional classification.

3. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 1 is characterized in that: The process of obtaining the target data set includes: Create an empty target dataset to store the data that meets the requirements filtered out from the original bone dataset; preprocess the bone dataset to unify the data format; extract key features from the bone dataset; A hierarchical method is used to screen the data, and the data related to the special information set is gradually eliminated; the data directly related to the special information set is screened out according to the bone name and type of bone injury; the data indirectly related to the special information set is further screened, and the second position information of the bone associated with the bone injury is checked for duplicates; Integrate the hierarchically screened data into the target dataset and verify the target dataset.

4. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 3 is characterized in that: The process of gradually eliminating data related to a particular set of information, including: Extract all recorded bone names from the bone data set and accurately match them with the bone names in the special information set; the successfully matched records enter the next level of screening; for the successfully matched records, further extract the bone injury type and match it with the bone injury type in the special information set; the successfully matched records enter the candidate list of the target data set; For the records in the candidate list, extract the second position information of the bone associated with the bone injury; match the extracted second position information with the relevant position information in the special information set; the successfully matched records are further entered into the candidate list of the target data set; Perform a duplicate check on the records in the candidate list and calculate the duplicate check rate; based on the duplicate check results, remove records with a duplication rate exceeding 30%.

5. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 4 is characterized in that: The process of matching the extracted second location information with the relevant location information in the special information set includes: Gaussian pyramids are constructed for the images in the image information and the standard anatomical position information respectively to detect the features of the image at different scales; multiple scale space layers are generated through Gaussian blur and downsampling operations; on the basis of the Gaussian pyramid, the Gaussian difference image of two adjacent layers of images is calculated to detect extreme points in the image, that is, potential key points; In the Gaussian difference image, by comparing the values ​​of each pixel with the 26 neighboring pixels around it, the extreme points are detected, which are potential key points; the detected extreme points are located; the edge response is eliminated by calculating the eigenvalues ​​of the Hessian matrix to retain stable key points; the gradient direction and amplitude are calculated for the pixels around each key point; in the local area around the key point, a gradient direction histogram of 36 directions is constructed, and the gradient amplitude of each direction is counted; the direction with the highest peak in the histogram is selected as the main direction of the key point, and a key point with direction is generated; The local image area around the key point is divided into 16 sub-areas. In each sub-area, the gradient direction histogram in 8 directions is calculated to generate a 128-dimensional feature descriptor; the descriptor is normalized; the Euclidean distance between the feature descriptors of the image information and the image in the standard anatomical position information is calculated; for each feature descriptor, the feature descriptor with the smallest Euclidean distance to it is found in the other image as a matching pair; A Euclidean distance threshold is set. Only when the Euclidean distance between two feature descriptors is less than the threshold, they are considered to be a valid matching pair; all matching pairs are screened and matching results with Euclidean distance greater than the set threshold are eliminated.

6. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 1 is characterized in that: The process of selecting matching Tai Chi, Ba Duan Jin or Wu Qin Xi content includes: Based on the action data set in the TCM rehabilitation training database, Tai Chi, Ba Duan Jin, and Wu Qin Xi movements are classified according to the skeletal parts and functions involved; an index is created for each classified movement, and the index contains the key features of the movement; The extracted skeletal position information is matched with the action index in the TCM rehabilitation training database, and the matching degree between the skeletal position information and the action index is calculated using cosine similarity; Set a matching threshold, filter out movements with matching degrees higher than the threshold, and select a movement from Tai Chi, Ba Duan Jin, or Wu Qin Xi.

7. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 6 is characterized in that: The process of using cosine similarity to calculate the matching degree between the skeleton position information and the action index includes: Extract the key features of each action from the TCM rehabilitation training database, including the involved bone parts, movement angles and strengths; convert the standardized bone position information into a feature vector, where each bone position information is represented as a multidimensional vector, where each dimension corresponds to a specific bone part or movement parameter; convert the key features in each action index into a feature vector, which will be used to match the bone position information; For each bone position information vector, calculate the cosine similarity between it and each action index vector; cosine similarity is a method to measure the similarity of the directions of two vectors; The matching degree between the skeleton position information and the action index is determined according to the calculated cosine similarity. The higher the matching degree, the higher the similarity between the skeleton position information and the action index.

8. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 7 is characterized in that: The process of determining the matching degree between the bone position information and the action index includes: The filtered action index vectors are sorted in descending order according to their cosine similarity with the skeleton position information vector; from the sorted list, the action index vectors with the highest similarity are selected in turn for matching; after each match, the matching result is updated; In each matching process, the hierarchy of the current matching result is evaluated and the continuity index of the matching result is calculated; if the hierarchy of the current matching result is poor, the weight needs to be readjusted; Determine the matching degree between the skeletal position information and the motion index; output the final matching result for use in traditional Chinese medicine rehabilitation training; the output of the matching result includes key information of the matching degree, the matching motion index, and the matching skeletal position information.

9. The intelligent recommendation method for bone injury rehabilitation training program based on traditional Chinese medicine according to claim 8 is characterized in that: The process of calculating the continuity index of the matching results includes: Extract global feature vectors from the skeleton position information and action index. The global feature vectors include the global features of the overall motion trend, posture change and speed change. Calculate the cosine similarity between global feature vectors, calculate the Euclidean distance between global feature vectors, and calculate the global continuity index by combining the cosine similarity and the Euclidean distance. Extract local feature vectors from the skeletal position information and action index, the local feature vectors include the local characteristics of the motion trajectory of the local joint and the activation state of the local muscle; calculate the Pearson correlation coefficient between the local feature vectors, calculate the mutual information between the local feature vectors, and calculate the local continuity index by combining the correlation analysis and the mutual information; Detail feature vectors are extracted from the skeleton position information and action index. The detail feature vectors include detail features of small motion changes and subtle posture adjustments. The entropy value of the detail feature vectors is calculated, and the KL divergence between the detail feature vectors is calculated. The detail continuity index is calculated by combining entropy analysis and KL divergence.

10. An intelligent recommendation system for bone injury rehabilitation training program based on traditional Chinese medicine, which is applied to the intelligent recommendation system for bone injury rehabilitation training program based on traditional Chinese medicine as claimed in any one of claims 1 to 9, characterized in that: The intelligent recommendation system for bone injury rehabilitation training plan based on traditional Chinese medicine includes: The first data set construction module is used to obtain the image information of all bones of the bone injury target and mark the position information; according to the function of the organ where the bone is located, the image information is divided according to the label to obtain bone data sets with different functions; The second data set construction module is used to obtain the bone name and the bone injury type of the bone injury through the input device, obtain the first position information of the bone injury and the second position information of the bone associated with the bone injury from the bone data set, and use the first position information and the second position information as a special information set; delete the special data set from the bone data set to obtain the target data set; The target action output module is used to obtain the position information of the bones to be subjected to bone injury rehabilitation training from the target data set, select matching Tai Chi, Ba Duan Jin or Wu Qin Xi content from the pre-stored traditional Chinese medicine rehabilitation training database; delete the relevant actions of the second position information from the action data set corresponding to the content; guide the trainee to perform bone injury rehabilitation training through the visual terminal, and at the same time prompt the cycle and time information of the bone injury rehabilitation training.

Images