Personalized motion scheme generation and adjustment method based on motion offset degree

Abstract

A personalized exercise scheme generation and adjustment method based on an action offset degree relates to the technical field of health management, and comprises the following steps: S1, collecting and classifying exercise videos, and labeling related information and timestamps corresponding to key frames; s2, performing health screening on the user, and determining the physical condition and exercise risk of the user; s3, performing a motion test on the user, collecting data and calculating a score; s4, generating a personalized exercise scheme according to the obtained data; s5, calculating the offset degree between the user action and the standard action of the motion video in real time; s6, giving corresponding exercise evaluation according to the motion offset degree of the user; s7, adding a motion compensation mechanism; and S8, dynamically adjusting the exercise scheme according to the deviation degree data of the user. According to the personalized exercise scheme generation and adjustment method based on the motion offset degree, the personalized exercise scheme can be generated according to the physical condition of the user, and dynamic adjustment is carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of health management, and particularly to a method for generating and adjusting a personalized exercise plan based on motion deviation degree. Background Art

[0002] With the development of intelligent health management and personalized exercise training, more and more users hope to obtain more accurate and personalized exercise plans. However, the existing technologies still face the following technical problems and defects in the process of generating personalized exercise plans: The existing exercise plan generation methods generally use standardized templates, ignoring the differences in exercise ability and health status of different body parts, and failing to design differentially according to the personalized needs of each part, resulting in overtraining of some parts and under-training of other parts, thus affecting the exercise effect and increasing the risk of exercise injuries. Especially for users in the recovery period, this problem may lead to secondary injuries. Therefore, it is necessary to generate personalized exercise plans according to the user's situation; Moreover, the existing technologies generally fail to deeply evaluate the motion deviation degree of users during exercise, especially during the execution of complex movements, and cannot detect and correct the motion deviation of users in a timely manner. The non-standard movements in the exercise plan cannot be corrected in a timely manner, increasing the risk of exercise injuries and affecting the exercise effect, especially in exercise training or rehabilitation that requires high-precision motion control. Therefore, it is necessary to detect the motion deviation degree of users and make dynamic adjustments in combination with the actual situation. Therefore, a method that can generate a personalized exercise plan and dynamically adjust the plan according to the motion deviation degree of users is needed, and the present invention solves this technical problem. Summary of the Invention

[0003] The present invention provides a method for generating and adjusting a personalized exercise plan based on motion deviation degree, which can generate a personalized exercise plan according to the user's physical condition and dynamically adjust the exercise plan according to the motion deviation degree during the user's exercise, thereby improving safety and ensuring the training effect.

[0004] A method for generating and adjusting a personalized exercise plan based on motion deviation degree includes the following steps: S1. Collect and classify exercise videos, and label the relevant information of the exercise videos and the timestamps corresponding to the key frames of the exercise videos; S2. Conduct a health screening on the user to determine the user's physical condition and exercise risk; S3. Collect the exercise test data of the user; S4. Generate a personalized exercise plan based on the obtained data. The user follows the exercise video according to the exercise plan and records it by taking pictures. Different exercise intensities and durations are set for different action types and different body parts of the user. S5. Calculate the deviation degree between the user's actions and the standard actions of the exercise video in real time. S6. Give corresponding exercise evaluations according to the size of the user's action deviation degree and provide action guidance to the user. S7. Add an exercise compensation mechanism. When the user's action deviation degree exceeds the preset threshold, dynamically adjust the exercise plan. S8. Dynamically adjust the exercise plan according to the user's deviation degree data.

[0005] Furthermore, the step S5 includes the following steps: S51. Calculate the deviation degree between the user's actions and each frame of the standard actions of the action video. S52. Calculate the deviation degree of each action. S53. Calculate the average value of the deviation degrees of all actions of the user during the entire exercise period. S54. Calculate the deviation degree of each part of the user's body.

[0006] Furthermore, the step S51 includes the following steps: S511. Analyze the bone points of each frame of the action in the video and calculate the angle of each joint. S512. Represent one of the joints as k. For each joint k in the video, calculate the angle difference △ between the j-th frame and the (j - 1)-th frame of the video ; S513. According to the angle difference △ between the j-th frame and the (j - 1)-th frame , calculate the weight of joint k , and construct the adaptive weight formula as follows: = ; where is the smoothing factor. The larger the angle difference, the more important the joint is in the action change, and thus a higher weight is assigned; S514. Calculate the angle difference △ between each joint of the i-th frame of the user's video and the j-th frame of the action video , and the calculation formula is as follows: △ = | - |; where, is the angle of the k-th joint in the i-th frame of the user, is the included angle of the k-th joint in the j-th frame of the user; S515. According to the obtained adaptive weight , perform weighted averaging on the included angle difference △ of each joint. The specific formula is as follows: = ; S516. Convert to the action offset from 0 to x. The specific formula is as follows: =x * ; where x represents the set value, represents the possible maximum included angle difference.

[0007] Further, the step S52 includes the following steps: S521. There is a time difference between the user and the standard action video. Calculate the best matching path between the user frame and the video frame in the following way: D(i,j)= ; where w is the window size, D(i,j) represents the cumulative distance of the matching path between the 0-i frames of the user and the 0-j frames of the action video, and the path determined by min in the search is the matching relationship between the user frame and the video frame; S522. Calculate the action fluency: ; where, represents the absolute value of the joint angle change rate from the i-th frame to the (i + 1)-th frame, is the fluency factor; S523. Calculate the angle change rate of a single joint from the i-th frame to the (i + 1)-th frame: - | / △t; where, represents the angle of the j-th joint in the i-th frame, and △t represents the time interval between two frames; Then calculate the weighted average angle change rate of all joints between two frames , where, represents the weight of the j-th joint; S524. Assume there are N frames, the key frames are , and the key frame timestamps are , then the weight of each frame i is The calculation formula is as follows: = ; Among them, λ is the attenuation coefficient, which is used to control the influence of the time difference on the weight; S525. Establish an action comprehensive offset formula: =( )* ; Among them, represents the offset of each frame.

[0008] Furthermore, in step S54, the offsets of each part of the body are calculated by the following formula: = ; Among them, k is the body part number, j is the j-th frame, i is the i-th frame, represents the overall offset of the k-th body part, and m is the total number of frames of the video; After that, is converted into an action offset from 0 to x, and the specific formula is as follows: =x* ; Among them, represents the possible maximum value, and x represents the set value.

[0009] Furthermore, step S6 specifically includes the following steps: S61. Analyze the bone points of the captured picture to obtain the coordinates and scores of each key point of the user's body. After removing the head points, the average score of the other key points is used as the overall confidence. If the overall confidence is lower than the set value, it is prompted that no one is captured; S62. In the evaluation once every t seconds, if the action offset of the user within t seconds is greater than the set value, and the action offset of the user within t + 2 seconds is less than the set value, it means that the user stands still. If standing still continuously for multiple times, a voice prompt is given and the video playback is paused; S63. When doing a small action in the exercise video, if the number of times the action is not standard is greater than or equal to the parameter of "the action is unqualified if not up to standard in multiple evaluations" configured for this action, it is prompted that the current action is not standard as a whole; S64. When evaluating once every t seconds, when the overall offset of the user , then for the body part with the offset , an unqualified prompt is given, the movement essentials of the current body part are displayed on the page, and a voice broadcast is made, where ; S65. Obtain the average action offset in the previous t seconds. When it , corresponding prompts and broadcasts will be made.

[0010] Further, step S7 specifically includes the following steps: S71. After the original action video follow - up ends, replace the actions with an offset greater than or equal to the set value in the exercise plan with the action videos of similar actions. The similar actions are actions with the same action type, the same main exercise part but with a decreased intensity; S72. Extract the actions corresponding to the offset greater than or equal to b and less than a, and add warm - up exercises and cool - down exercises to enable the user to further follow - up.

[0011] Further, step S8 specifically includes the following steps: S81. Dynamic adjustment based on each part of the body: When the offset of a certain part is less than b for consecutive days, the intensity of this part is increased until the maximum intensity when generating the next plan; when the offset is greater than or equal to a, a new plan is immediately generated and the intensity of this part is decreased until the low intensity; when the offset is greater than or equal to b for consecutive days, the intensity of this part is decreased when generating the next plan; in other cases, the current intensity is maintained; S82. Dynamic adjustment based on the action type: When the offset of a certain action type is less than b for consecutive days, the intensity of this action type is increased until the maximum intensity when generating the next plan; when the offset is greater than or equal to a, a new plan is immediately generated and the intensity of this action type is decreased until the low intensity; when the offset is greater than or equal to b for consecutive days, the intensity of this action type is decreased when generating the next plan; in other cases, the current intensity is maintained; S83. Dynamic adjustment based on the whole: When the overall offset is less than b for consecutive days, the overall intensity is increased until the maximum intensity when generating the next plan; when the offset is greater than or equal to a, a new plan is immediately generated and the overall intensity is decreased until the low intensity; when the offset is greater than or equal to b for consecutive days, the overall intensity is decreased when generating the next plan; in other cases, the current intensity is maintained.

[0012] Further, step S3 specifically includes the following steps: S31. Obtain the comprehensive physical fitness index of the user - Physical function test index - Physical form test index - ; S32. Calculate the above - mentioned indexes through the comprehensive rating score calculation formula: ; α + β+γ = 1; Wherein, 、 、 Represents the sum of the standardized values of each index, and α, β, γ are weight coefficients.

[0013] The technical effects of the present invention are as follows: (1) In this solution, the physical condition and exercise risk of the user are determined through health screening and exercise tests, and a personalized exercise plan is generated, thereby reducing the risk of the user being injured during exercise. Moreover, this solution can also detect the movement deviation of the user when the user exercises according to the exercise plan, and give corresponding guidance to the user according to the actual situation. It can also dynamically adjust the exercise plan according to the compensation mechanism and the deviation situation within consecutive days, thereby improving the safety of the user during exercise and enhancing the exercise effect; (2) This solution can perform bone point analysis, calculate the angle of each joint, and compare the user's movement with the movement in the movement video, thereby calculating the difference between the user's movement and the standard movement in the movement video, and further calculating the deviation. Through the above method, not only can the user timely understand whether their movement is standard, but also the risk of exercise injury can be reduced, thereby improving safety; (3) This solution adopts a compensation mechanism. When the movement deviation of the user exceeds the threshold, the exercise plan can be automatically adjusted, and the exercise plan can also be dynamically adjusted according to the deviation situation of the user within a period of time, thereby improving the exercise effect on the premise of ensuring the safety of the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is the overall flowchart of the present invention.

[0015] Figure 2 is the four - quadrant diagram of the user's exercise risk of the present invention.

[0016] Figure 3 is the flowchart of the deviation calculation in the present invention.

[0017] Figure 4 is the flowchart of the exercise evaluation and guidance for the user in the present invention.

[0018] Figure 5 is the flowchart of the compensation mechanism in the present invention.

[0019] Figure 6 is the flowchart of the dynamic adjustment of the exercise plan in the present invention.

[0020] Figure 7 are the evaluation indicators and their weights for adults and the elderly aged 20 - 79 in the present invention.

[0021] Figure 8 are the comprehensive physical fitness rating scores for adults and the elderly aged 20 - 79 in the present invention.

[0022] Figure 9This is the comparison table of physical fitness assessment and exercise goals in the present invention.

[0023] Figure 10 This is the comparison table of exercise test item ratings and exercise intensities in the present invention. Detailed implementation manners

[0024] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with specific embodiments and the accompanying drawings.

[0025] Refer to Figures 1 - 6 , a personalized exercise plan generation and adjustment method based on action deviation degree, comprising the following steps: S1. Collect and classify exercise videos, and label the relevant information of the exercise videos and the timestamps corresponding to the key frames of the exercise videos; S2. Conduct a health screening on the user to determine the user's physical condition and exercise risks; S3. Collect the user's exercise test data; S4. Generate a personalized exercise plan according to the obtained data. The user follows the exercise video according to the exercise plan and records it by shooting. Different exercise intensities and durations are set for different action types and different body parts of the user; S5. Calculate the deviation degree between the user's actions and the standard actions of the exercise video in real time; S6. Give corresponding exercise evaluations according to the magnitude of the user's action deviation degree, and guide the user's actions; S7. Add an exercise compensation mechanism to dynamically adjust the exercise plan when the user's action deviation degree exceeds a preset threshold; S8. Dynamically adjust the exercise plan according to the user's deviation degree data.

[0026] Refer to Figure 1 , for step S1, the following steps are mainly adopted in this embodiment: S11. Classify all the collected exercise videos into a warm-up exercise library, a main exercise library, and a finishing exercise library; S12. Label each action video: exercise part, equipment used, exercise type (aerobic exercise, balance exercise, flexibility exercise, resistance exercise, etc.), which diseases prohibit doing this action, which diseases this action is beneficial to, exercise intensity (METs), and configure the parameter of "if the action fails to meet the standard multiple times, the whole action is unqualified", where "multiple times" can be a specific number of times, and the specific number of times can be determined by the staff; S13. Mark the timestamps corresponding to the key frames of the exercise video for calculation when calculating the action deviation degree, where the key frame refers to the frame when an action reaches the standard level and the different frames in the video; The method for key frame extraction is as follows: Use the cv2 package in python to obtain each frame image of the video and its corresponding video time point. Analyze the skeletal data of each frame image using the skeletal analysis model movenet_thunder, calculate the overall confidence level, remove the frames with too low confidence level, and compare the frame differences (such as the absdiff method in cv2) starting from the first frame for the remaining frames. Select the frames with large differences. If the differences are small for several consecutive seconds, select one frame per second (for example, select the first frame image, and compare the second, third, fourth, etc. frames with the first frame respectively. If the differences of the second and third frames are small, do not select them. If the difference of the fourth frame is large, select it, and then compare the fifth, sixth, seventh, eighth, etc. frames with the fourth frame until the end. If the fourth frame is the second second of the video, the fifth, sixth, and seventh frames are the third second, and the eighth frame is the fourth second, but the differences between the fifth, sixth, and seventh frames and the fourth frame are all very small, then select the seventh frame, and continue to compare the seventh frame with the subsequent frames). After the program processing is completed, it is necessary to manually check the video and add the frames with the standard action degree to the selected list.

[0027] The confidence level mentioned above can be calculated in the following way: The camera captures the user's movement process, analyzes the skeletal points of the camera frame image through the model, obtains the coordinates and scores of each key point of the user's body, and after removing the head points, the average value of the scores of other points is used as the overall confidence level. If the result of the overall confidence level is lower than the set value (the threshold set by the staff according to the actual situation), it is prompted that there is no one in the camera.

[0028] The above content is the content of the data preparation part in the left half of the appendix Figure 1

[0029] See Figure 1 、 Figure 2 For step S2, this embodiment is specifically implemented in the following way: The health screening specifically involves understanding the user's current diseases through a questionnaire survey, then confirming whether the user has cardiovascular, metabolic, or liver diseases or their symptoms and signs, and whether they have exercise habits. Then, draw a four-quadrant diagram of the user's exercise risk according to the results. It can be learned from the four-quadrant diagram that when the user does not have cardiovascular, metabolic, or liver diseases or their symptoms and signs and has no exercise habits, it is default that the user's exercise test is unqualified and should start from low intensity (<3 METs); if they have exercise habits, it is default that the exercise test result is good and start exercising from medium intensity (3 - 6 METs); if they have cardiovascular, metabolic, or liver diseases or their symptoms and signs, they cannot use the system, and it is recommended that the user seek medical attention in time.

[0030] The four-quadrant diagram can refer to the appendix Figure 2 ​Draw according to whether the user has cardiovascular, metabolic or liver diseases or their symptoms and signs, and whether they have exercise habits, and give relevant suggestions in the corresponding quadrants. The above content is attached Figure 1 The content of the first point in the right half part.

[0031] For step S4, this embodiment is implemented in the following manner: S41. Determine the exercise frequency: Determine the exercise frequency according to the user's current exercise stage and user dynamic data. The default frequency in the adaptation stage is 3 times / week, increase by one time every two weeks in the improvement stage until it reaches 5 times / week, and the default frequency in the maintenance stage is 5 times / week. It can be adjusted according to the user's feedback; S42. Determine the exercise intensity and duration of each type of action, each part of the body, and the whole body: Total duration of a single exercise: The default duration in the adaptation stage (four weeks in this embodiment) is 30 minutes, and the duration of each exercise will be extended by 5 minutes per week until it reaches 50 minutes at most; it can also be adjusted according to the user's feedback; in the improvement stage (4 months) and the maintenance stage (always maintained), it will be changed according to the user's dynamic data; Duration of each part of the body: Make an average distribution; Duration distribution of each type of action: The default distribution of aerobic, flexibility, balance, resistance and other exercise types is equal, each accounting for 25%. When a detection item in the exercise detection is unqualified, the recommended exercise type corresponding to it will increase by 10%. Aerobic and resistance exercises will not appear in the same exercise. (For example: the total duration is 30 minutes, and the aerobic exercise becomes 35% when the 30 - second sit - to - stand test is unqualified. Do aerobic exercise on Monday, aerobic 35%, flexibility 25%, balance 25%, aerobic 12 minutes, flexibility 9 minutes, balance 9 minutes; do resistance exercise on Wednesday, resistance 25%, flexibility 25%, balance 25%, resistance 10 minutes, flexibility 10 minutes, balance 10 minutes); Overall exercise intensity: The default exercise intensity is confirmed by health screening and exercise tests, and is dynamically adjusted according to the user's data after exercise; Intensity of each part of the body and each type of action: By default, it is the same as the overall intensity and is adjusted according to the user's dynamic data.

[0032] S43. Action selection: First, filter out the contraindicated exercises for the disease from the video library according to the user's current disease; Warm - up exercise: The user needs to do warm - up exercises to stretch the body and prevent injury during the main exercise; Just select a few exercises from the warm - up exercise library that can move each part of the body, and the total duration is 5 minutes; Main exercise: Screen the exercise list according to the intensity and duration of each type of action, and the intensity and duration of each part of the body; Cool-down exercise: Through various movements, gradually relax the muscles and adjust the heart rate, breathing, blood pressure, etc. to normal levels; Select a few from the cool-down exercise library that can relax various parts of the body, with a total duration of 5 - 10 minutes; S44. Exercise progression The exercise progression is divided into three stages: adaptation stage, improvement stage, and maintenance stage; Adaptation stage: The adaptation stage defaults to 4 weeks, with the exercise intensity and frequency remaining unchanged. Extend the duration of each exercise by 5 minutes per week, and the duration per session, exercise frequency, exercise intensity, and stage duration can be adjusted according to user feedback; Improvement stage: This stage defaults to 4 - 8 months; Increase the exercise frequency and duration every 2 weeks, but at most 5 times a week and at most 50 minutes per session; When reaching a certain level, strengthen the intensity of each exercise, reduce the number of weekly exercises and the duration per session until each type of exercise reaches the high-intensity limit of 5 times a week and 50 minutes per session. Adjustments need to be made according to user feedback; Maintenance stage: After reaching the maximum exercise volume in the improvement stage, it is the maintenance stage; Users can make adaptive adjustments based on the maximum exercise volume to confirm the exercise volume that they will maintain throughout the year.

[0033] Thus, through the above methods, select a suitable exercise plan according to the user's physical condition to achieve the generation of a personalized exercise plan. The above content is Figure 1 The content of the third point on the right half regarding the generation of a personalized exercise plan in

[0034] Furthermore, the step S5 includes the following steps: S51. Calculate the offset of each frame of the user's movement from the standard movement of the movement video; S52. Calculate the offset of each movement; S53. Calculate the average value of the offsets of all movements during the entire exercise period of the user; S54. Calculate the offsets of various parts of the user's body.

[0035] The above content in this embodiment is implemented through bone point analysis and the dynamic time warping (DTW) algorithm. The above content is Figure 1 The content of the fourth point on the right half regarding the calculation of offsets in

[0036] Furthermore, the step S51 includes the following steps: S511. Use the model move net_thunder to perform bone point analysis on each frame of the movement in the video and calculate the angle of each joint; S512. Represent one of the joints as k. For each joint k in the video, calculate the angle difference △ between the jth frame and the (j - 1)th frame of the video ; S513. Calculate the weight of joint k according to the angular difference Δ between the j-th frame and the (j - 1)-th frame , and construct the adaptive weight formula as follows: = ; where is the smoothing factor. The larger the angular difference, the more important the joint is in the action change, so a higher weight is assigned. Through the above method, the dynamic weight of each joint in this frame can be calculated according to the angular difference of each joint; S514. Calculate the angular difference Δ between each joint in the i-th frame of the user video (i.e., the video of the user's actions captured by the camera) and the j-th frame of the action video , and the calculation formula is as follows: Δ = | - |; where is the angle of the k-th joint in the i-th frame of the user, is the angle of the k-th joint in the j-th frame of the user; S515. According to the obtained adaptive weight of each joint in this frame , perform weighted average on the angular difference Δ of each joint (i.e., the angular difference between the user and the video) to calculate the offset degree. The specific formula is as follows: = ; S516. Convert to the action offset degree from 0 to x. The specific formula is as follows: = x * ; where x represents the set value, represents the possible maximum angular difference.

[0037] In this embodiment, x is 100, that is, convert to the action offset degree from 0 to 100. Since the maximum angle of a normal joint is 180, so the maximum value ofis 180.

[0038] The above content is Figure 3 the content of calculating the offset degree of one frame in

[0039] Furthermore, the step S52 includes the following steps: S521. There is a difference between the user and the standard action video time. In this embodiment, the dynamic time warping (DTW) algorithm is combined with windowed delay matching and continuous frame matching to calculate the best matching path between the user frame and the video frame: D(i,j)= ; Among them, w is the window size, D(i,j) represents the cumulative distance of the matching path between the user's 0-i frames and the action video's 0-j frames. The path determined by min during the search is the matching relationship between the user frame and the video frame; When , D(i,j) = ; Conversely, D(i,j)= , that is, transfer to the adjacent frame. For example Continue the search and calculation, where w is the window size, such as set to 5, which can be adjusted as needed; This embodiment uses the fastdtw framework of python to implement, and the Sakoe-Chiba window is used for the delay.

[0040] Video frame list: j_list = [1, 2, 3, 4], User frame list: i_list = [1, 2, 3, 4], Calculate: distance, path = fastdtw(j_list,i_list,dist=custom_distance), custom_distance is the method of calculating the value, distance is the D(i,j) value, the minimum cumulative distance.

[0041] path is the index matching relationship between j_list and i_list.

[0042] S522. Calculate the action fluency: ; Among them, represents the absolute value of the joint angle change rate from the i-th frame to the i+1-th frame, is the fluency factor; S523. Calculate the angle change rate of a single joint from the i-th frame to the i+1-th frame to achieve the calculation of the absolute value of the frame angle change rate: - | / △t; Among them, represents the angle of the jth joint in the i-th frame, and △t represents the time interval between two frames; Then calculate the weighted average angle change rate of all joints between the two frames ,in, represents the weight of the jth joint; S524, assuming there are N frames, the key frame is , the key frame timestamp is , then the weight of each frame i is The calculation formula is as follows: = ; Among them, λ is the attenuation coefficient, which is used to control the influence of time difference on weight. λ>0, so as to realize dynamic adjustment of frame weight. S525. Establish the action comprehensive deviation formula to calculate the comprehensive deviation: =( )* ; in, Indicates the offset of each frame.

[0043] Furthermore, in step S54, the deviation of each body part is calculated by the following formula: = ; Where k is the body part number, j is the jth frame, i is the i-th frame, represents the overall deviation of the kth body part, and m is the total number of video frames; Afterwards, Converted to action offset from 0 to 100, the specific formula is as follows: =x* ; in, Indicates the maximum possible value.

[0044] The above content is Figure 3 The content of calculating the offset of an action.

[0045] See also Figure 4 Further, the step S6 specifically includes the following steps: S61, the system previews the exercise and announces the essentials of the action. After that, the user starts exercising. The system performs a skeleton point analysis on the image captured by the camera, obtains the coordinates and scores of each key point of the user's body, and after removing the head point, the average score of other key points is used as the overall confidence. If the overall confidence is lower than the set value, it prompts that no person is captured, that is, no one is in the camera; S62. In the evaluation once every t seconds, if the movement deviation degree of the user within t seconds is greater than the set value, and the movement deviation degrees of the user within t + 2 seconds are all less than the set value, it means that the user stands still. If the user stands still continuously for multiple times, voice prompts will be given and the video playback will be paused; the set value in this embodiment is 2; S63. When doing one of the small movements in the exercise video, if the number of times the movement is not standard is greater than or equal to the parameter of "if the movement is not up to standard for multiple evaluations, the whole movement is unqualified" configured for this movement, prompt the user that the current whole movement is not standard; S64. When in the evaluation once every t seconds, when the overall deviation degree of the user then give an unqualified prompt for the body part with the deviation degree display the movement essentials of the current body part on the page, and conduct voice broadcast; S65. Obtain the average movement deviation degree in the previous t seconds. When it corresponding prompts and broadcasts will be made, such as prompting "Perfect, please continue", etc., that is, the exercise evaluation.

[0046] The above content is Figure 1 the content of the fifth point on the right side of

[0047] See Figure 5 and further, the step S7 specifically includes the following steps: S71. After the original action video follow - up is completed, for the actions with a deviation degree greater than or equal to 40, replace them with the action videos of similar actions in the exercise plan. The similar actions are actions with the same action type and main exercise part but with a decreased intensity; For example, if the actions before and after replacement are both resistance exercises, the exercise types are the same; if the exercise parts before and after replacement are both the upper arms, the main exercise parts are the same; regarding the decrease in intensity, in this application, actions with a 10% decrease in intensity (METs) are used, and the replaced actions will be combined with the actions corresponding to the deviation degrees of ≥20 and <40 to form a new exercise plan; S72. Extract the actions corresponding to the deviation degrees of ≥20 and <40, and add warm - up exercises and cool - down exercises before and after these actions to form a complete compensation video list, so that the user can further follow - up according to the compensation video list to ensure the exercise effect of the user on the same day.

[0048] The above content is Figure 1 the content of the sixth point on the right side of

[0049] See Figure 6 and further, the step S8 specifically includes the following steps: S81. Dynamic adjustment based on various body parts: When the deviation degree of a certain part is < 20 for consecutive days, the training intensity of this part will be increased by 5% in the next generated plan until the maximum intensity; when the deviation degree ≥ 40, a new plan will be generated immediately, and the intensity of this part will be decreased by 10% until the low intensity; when the deviation degree ≥ 20 for consecutive days, the intensity of this part will be decreased by 5% in the next generated plan; in other cases, the current intensity will be maintained. S82. Dynamic adjustment based on action types: When the deviation degree of a certain action type is < 20 for consecutive days, the intensity of this part will be increased by 5% in the next generated plan until the maximum intensity; when the deviation degree ≥ 40, a new plan will be generated immediately, and the intensity of this action type will be decreased by 10% until the low intensity; when the deviation degree ≥ 20 for consecutive days, the intensity of this action type will be decreased by 5% in the next generated plan; in other cases, the current intensity will be maintained. S83. Dynamic adjustment based on the whole: When the deviation degree of the whole is < 20 for consecutive days, the overall intensity will be increased by 5% in the next generated plan until the maximum intensity; when the deviation degree ≥ 40, a new plan will be generated immediately, and the overall intensity will be decreased by 10% until the low intensity; when the deviation degree ≥ 20 for consecutive days, the overall intensity will be decreased by 5% in the next generated plan; in other cases, the current intensity will be maintained.

[0050] When there are action intensity data for each body part and each action type, the above method shall prevail; if there is no data, the overall intensity shall prevail. The "consecutive days" in the above content can specifically be 7 days.

[0051] The above content is Figure 1 the content of the seventh point on the right side of

[0052] Further, the step S3 specifically includes the following steps: S31. Obtain the comprehensive physical fitness index of the user - Physical function test indicators - Physical form test indicators - ; S32. Calculate the above indicators through the comprehensive rating score calculation formula: ; α + β + γ = 1; where , , represents the sum of the standardized values of each indicator, α, β, and γ are weight coefficients, is the sum of the standardized values of the comprehensive physical fitness index, The sum after standardization of the physical function test indicators The sum after standardization of the physical form test indicators.

[0053] If the conditions do not permit, the exercise test can be skipped, and the exercise plan can be generated based on the results in the health screening and the exercise intensity. The exercise test data can be collected through methods such as questionnaire surveys, importing physical examination reports, synchronizing with the national national physical fitness testing platform and third-party platforms.

[0054] The above content is Figure 1 The part about the exercise test in the second point of the right half in

[0055] Preferably, the indicators in this embodiment are as follows: Physical fitness test indicators: Grip strength ( ), Standing long jump ( ), Push-ups ( , male) / Kneeling push-ups ( , female), Sit-ups ( ), Sit-and-reach ( ), Standing on one leg with eyes closed ( ), Choice reaction time ( ).

[0056] Physical function test indicators: Vital capacity ( ), Second-stage load test on a cycle ergometer ( ).

[0057] Physical form test indicators: Body mass index ( ), Body fat percentage ( ).

[0058] Among them, , and are both secondary weights, is the weight of the body mass index, is the weight of the body fat percentage, and γ is the weight of the physical form, that is, the primary weight, , The above calculation method can also be adopted, which will not be elaborated here. The values of α, β, and γ can be modified according to your own needs.

[0059] See Figure 7 , in the figure are the evaluation indicators and their weights for adults and the elderly aged 20-79. The specific weight values can refer to the content in the figure.

[0060] See Figure 8 , in the figure are the comprehensive physical fitness rating scores for adults and the elderly aged 20-79. You can refer to the content in the figure to clarify the level corresponding to the score.

[0061] See Figure 9 , which is a comparison chart of physical fitness assessment and exercise goals, used to clarify the test items, assessment levels, exercise purposes, and recommended exercise types.

[0062] Among them, Figure 7 、 Figure 8 、 Figure 9 The content in comes from the "National Physical Fitness Measurement Standard (Revised in 2023)" released by the National Physical Fitness Monitoring Center.

[0063] See Figure 10 , which is a comparison chart of exercise test item ratings and exercise intensities. You can refer to the content in the figure to clarify the exercise intensity and its corresponding exercise test item ratings.

[0064] Figure 10 The content in comes from the content about metabolic equivalents in Baidu Encyclopedia. The exercise test item ratings in the figure are divided into three levels: excellent, good / qualified, and unqualified, corresponding to high-intensity physical activities, moderate-intensity physical activities, and low-intensity physical activities in Baidu Encyclopedia respectively, and the corresponding metabolic equivalents are also the same.

[0065] The above is only an exemplary embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structural transformation made under the technical concept of the present invention by using the content of the specification and drawings of the present invention, or directly / indirectly applied in other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. A method for generating and adjusting a personalized exercise plan based on movement deviation, characterized in that: The following steps are involved: S1. Collect and classify sports videos, and mark relevant information of the sports videos and timestamps corresponding to key frames of the sports videos; S2. Conduct health screening on users to determine their physical condition and sports risks; S3, performing a sports test on the user, collecting data obtained from the sports test and calculating a comprehensive sports rating score; S4. Generate a personalized exercise plan based on the obtained data. The user follows the exercise video according to the exercise plan and records it by shooting. Different exercise intensities and durations are set according to different action types and different body parts of the user. S5, calculating the deviation between the user action and the standard action of the sports video in real time; S6. According to the magnitude of the user's movement deviation, a corresponding movement evaluation is given, and the user is guided in movement; S7, add motion compensation mechanism, when the user's motion deviation exceeds the preset threshold, dynamically adjust the motion plan; S8. Dynamically adjust the exercise plan based on the user's deviation data.

2. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 1, characterized in that: The step S5 comprises the following steps: S51, calculating the deviation between the user's action and each frame of the standard action in the action video; S52, calculating the deviation of each action; S53, calculating the average value of all movement deviations of the user during the entire exercise period; S54: Calculate the displacement of each part of the user's body.

3. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 2, characterized in that: The step S51 comprises the following steps: S511, performing skeleton point analysis on each frame of action in the video, and calculating the angle of each joint; S512: denote one of the joints as k, and for each joint k in the video, calculate the angle difference △ between the jth frame and the j-1th frame of the video ; S513, according to the angle difference △ between the j-th frame and the j-1-th frame , calculate the weight of joint k , the adaptive weight formula is constructed as follows: = ； in is the smoothing factor. The larger the angle difference is, the more important the joint is in the movement change. S514, calculating the angle difference of each joint between the i-th frame of the user video and the j-th frame of the action video △ , the calculation formula is as follows: △ =| - |; in, is the angle of the kth joint in the i-th frame of the user, is the angle of the kth joint in the jth frame of the user; S515: Based on the obtained adaptive weight , the angle difference of each joint △ The specific formula for weighted average is as follows: = ; S516, will Converted to the action offset from 0 to x, the specific formula is as follows: =x * ; Among them, x represents the set value, Indicates the maximum angle difference.

4. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 2, characterized in that: The step S52 comprises the following steps: S521: There is a difference between the time of the user and the standard action video. The best matching path between the user frame and the video frame is calculated in the following way: D(i,j)= ; Where w is the window size, D(i,j) represents the cumulative distance between the matching path of user 0-i frame and action video 0-j frame, and the path determined by the min in the search is the matching relationship between the user frame and the video frame; S522, calculation of action fluency: ； in, Represents the absolute value of the rate of change of joint angle from the i-th frame to the i+1-th frame, is the fluency factor; S523, calculate the angle change rate of a single joint from the i-th frame to the i+1-th frame: - | / △t; in, represents the angle of the jth joint in the i-th frame, and △t represents the time interval between two frames; Then calculate the weighted average angle change rate of all joints between the two frames ,in, represents the weight of the jth joint; S524, assuming there are N frames, the key frame is , the key frame timestamp is , then the weight of each frame i is The calculation formula is as follows: = ; Among them, λ is the attenuation coefficient, which is used to control the influence of time difference on weight; S525. Establish the action comprehensive deviation formula: =( )* ； in, Indicates the offset of each frame.

5. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 2, characterized in that: In step S54, the deviation of each body part is calculated by the following formula: = ； Where k is the body part number, j is the jth frame, i is the i-th frame, represents the overall deviation of the kth body part, and m is the total number of video frames; Afterwards, Converted to the action offset from 0 to x, the specific formula is as follows: =x* ； in, represents the maximum value, and x represents the set value.

6. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 1, characterized in that: The step S6 specifically comprises the following steps: S61, performing skeleton point analysis on the captured image, obtaining the coordinates and scores of each key point of the user's body, removing the head point, and taking the average score of other key points as the overall confidence. If the overall confidence is lower than the set value, it is prompted that no person is captured; S62, in the evaluation every t seconds, if the user's movement deviation within t seconds is greater than the set value, and the user's movement deviation within t+2 seconds is less than the set value, it means that the user is standing still. If the user stands still for multiple consecutive times, a voice prompt is given and the video playback is paused; S63, when performing one of the small movements in the sports video, if the number of times the movement is not standard is greater than or equal to the "multiple evaluations fail to meet the standard, the movement is unqualified as a whole" parameter configured for this movement, the user is prompted that the current movement is not standard as a whole; S64, when the user's overall deviation degree is When The page will show the exercise tips for the current body part and give voice announcements. ; S65, obtain the average action deviation before t seconds, when When the time comes, the corresponding prompt will be given and announced.

7. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 1, characterized in that: The step S7 specifically includes the following steps: S71, after the original action video training is completed, the action with a deviation greater than or equal to the set value is replaced with an action video of a similar action in the exercise plan, wherein the similar action is an action with the same action type and main exercised parts but with reduced intensity; S72, extracting the movements corresponding to the deviations ≥ b and < a, and adding warm-up movements and finishing movements, so that the user can further practice.

8. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 1, characterized in that: The step S8 specifically includes the following steps: S81. Dynamic adjustment based on various parts of the body: When the deviation of a certain part is less than b for many consecutive days, the next time a plan is generated, the strength of the part is increased until it reaches the maximum strength; when the deviation is ≥ a, a new plan is generated immediately, and the strength of the part decreases until it reaches a low strength; when the deviation is ≥ b for many consecutive days, the strength of the part decreases when a plan is generated next time; in other cases, the current strength is maintained; S82, dynamic adjustment based on action type: when the deviation of a certain action type is less than b for many consecutive days, the intensity of the part will be increased to the maximum intensity when the plan is generated next time; when the deviation is ≥ a, the plan will be regenerated immediately, and the intensity of the action type will be reduced to a low intensity; when the deviation is ≥ b for many consecutive days, the intensity of the action type will be reduced when the plan is generated next time; in other cases, the current intensity will be maintained; S83. Dynamic adjustment based on the whole: When the overall deviation is less than b for many consecutive days, the next time a plan is generated, the overall intensity will be increased to the maximum intensity; when the deviation is ≥a, the plan will be regenerated immediately, and the overall intensity will be reduced to a low intensity; when the deviation is ≥b for many consecutive days, the overall intensity will be reduced the next time a plan is generated; in other cases, the current intensity will be maintained.

9. The method for generating and adjusting a personalized exercise plan based on movement deviation according to claim 1, characterized in that: The step S3 specifically comprises the following steps: S31. Obtaining comprehensive indicators of the user's physical fitness - , Physical function test indicators - , Body shape test indicators - ; S32. Calculate the above indicators using the comprehensive rating score calculation formula: ； α+β+γ=1; in, , , It is expressed as the standardized sum of various indicators, and α, β, and γ are weight coefficients.

Images