Human body training scheme generation method and system based on visual evaluation

Abstract

The invention relates to the technical field of human body exercise training, in particular to a human body training scheme generation method and system based on visual evaluation, and the method comprises the steps: continuously collecting human body images through a depth camera, and recognizing a plurality of mark points of a human body; obtaining mark point sequence data and joint angle sequence data; comprehensive scores of four dimensions of activity, stability, symmetry and sport injury risk of each group are evaluated through mark point sequence data and joint angle sequence data; according to the comprehensive score of each group, a periodic training scheme for a period of time in the future is automatically generated, and the periodic training scheme comprises basic training and improved exercise training for improving the current weak part corresponding to the group with the low comprehensive score. According to the invention, a personalized training scheme can be output through internal logic.

Description

Technical Field

[0001] The present invention relates to the technical field of human motion training, and in particular to a method and system for generating a human training program based on visual evaluation. Background Art

[0002] With the development of medical technology, we are gradually moving from the past era of seeing a doctor for treatment to the era of pre-disease prevention. Therefore, it is very important for people to conduct reasonable human function assessment tests and to conduct training and improvement based on the subjects' existing physical foundation.

[0003] The data used for human function assessment and training program generation are currently mostly based on subjective description and human evaluation. Human function assessment and training program formulation are generally carried out through questions and answers and the professional knowledge of trainers. This form and logic are highly dependent on the professionalism of the training program formulation personnel. Therefore, it is necessary to develop a system that can output personalized training programs through internal logic.

[0004] At present, the patent document with the publication number CN118016242A discloses a method and system for generating a corrective training program for human motor function. The method calculates the functional problem score according to the evaluation parameters of the test action after each person under test performs the test action, calculates the functional problem score again after the person under test performs the corrective training action, and calculates the action effectiveness score of each corrective training action according to the difference between the two times, and generates a general training library; generates a corrective training program according to the general training library of each person under test, optimizes the general training library of the person under test according to the execution of the corrective training program, and continuously optimizes the corrective training program of the person under test; the invention realizes the organic linkage between evaluation and training, avoids the one-sided emphasis on flexibility training and strength training in traditional training, and ignores the sports injuries caused by the special needs of the human body, realizes the personalized output of the corrective training program, and can continuously and adaptively adjust the corrective training program based on the individual's movement tendency and training effect. However, in this training program, there is a lack of continuous change of exercise load, and the generated training program lacks timeliness. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method and system for generating a human training program based on visual assessment, which solves the technical problem that the existing human function assessment and training programs have high professional requirements for the personnel formulating the training programs.

[0007] (II) Technical solution

[0008] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0009] In a first aspect, an embodiment of the present invention provides a method for generating a human training program based on visual assessment, comprising the following steps:

[0010] Prompt the subject to complete a specified standard action, continuously collect human body images through a depth camera, and identify multiple marker points of the human body from the human body images according to a visual recognition system;

[0011] Record the spatial coordinates corresponding to multiple marking points in the time series to obtain marking point sequence data; and calculate the corresponding joint angles to obtain joint angle sequence data;

[0012] The multiple markers are grouped according to the human body parts, and the marker sequence data and joint angle sequence data are used to evaluate the comprehensive scores of the four dimensions of mobility, stability, symmetry and sports injury risk of each group;

[0013] According to the comprehensive scores of each group, a periodic training plan for the future period is automatically generated. The periodic training plan includes basic training and improvement exercise training for improving the current weak parts corresponding to the group with low comprehensive scores.

[0014] The embodiment of the present invention proposes a method for generating a human training program based on visual evaluation. The method evaluates the sports fitness level of multiple parts of the body by the quality of the subject's completion of standard movements, and then automatically generates a set of solutions for targeted training plans based on the evaluation results. The method can output personalized training programs through internal logic.

[0015] Optionally, the multiple human body markers include: head, neck, mid-clavicle, tenth thoracic vertebra, sacral vertebra, left / right shoulder, left / right elbow, left / right wrist, left / right hand, left / right thumb, left / right middle finger, left / right hip, left / right knee, left / right ankle and left / right foot.

[0016] Optionally, the multiple marking points are respectively used to form multiple joint angles according to the starting point-middle point-end point, including: neck, upper spine, right upper spine, left upper spine, middle spine, right spine bottom, left spine bottom, right wrist, right elbow, right shoulder, right hip, right knee, right ankle, left wrist, left elbow, left shoulder, left hip, left knee and left ankle;

[0017] The joint angles are grouped according to the human body segments, including five groups: neck, upper limbs, trunk, pelvis and lower limbs, among which:

[0018] The neck includes the angles of joints: the neck;

[0019] Upper limbs include joint angles: left shoulder, right shoulder, left elbow, right elbow, left wrist and right wrist;

[0020] The trunk includes the joint angles: upper spine, middle spine, right spine base, and left spine base; and right spine upper and left spine upper;

[0021] The pelvis includes the joint angles: left and right hip;

[0022] Lower limbs including joint angles: left knee, right knee, left ankle, and right ankle;

[0023] The five groups' corresponding body parts trained in the periodized training program were: lower extremities, upper extremities, and core; the core included the neck, trunk, and pelvis.

[0024] Optionally, the visual recognition system is trained on a data set through machine learning methods, and the data set includes a database of key points of the human body in motion in open source data and a motion training movement library obtained through self-training, and the motion training movement library contains motion record data of test movements and training movements.

[0025] Optionally, the method further comprises:

[0026] After identifying the spatial coordinates of multiple markers of the human body, the wavelet threshold method is first used to perform data noise reduction. Then, based on the spatial coordinates and time series of the markers, the acceleration of the markers is obtained. Based on the upper limit of acceleration that the human body can reach and the results of empirical evaluation, the data limit value is set again, and the spatial coordinates that exceed the limit value are deleted.

[0027] After calculating the corresponding joint angles, a threshold is set based on the empirical evaluation results, and angles exceeding the threshold are deleted.

[0028] Optionally, the method further comprises: scoring and ranking multiple indicators of a single joint for assessing flexibility of muscle tension or risk scoring for assessing stability of muscle insufficiency according to the score of sports injury risk; finding out the indicators with the highest ranking and the muscle groups whose functions are mutually antagonistic; and determining the main muscles according to the muscle groups;

[0029] The single-cycle training program includes six training plans per week with lower body strength, upper body strength and core strength as the training targets; each training plan includes activation movements, loading movements and stretching movements;

[0030] According to the different training goals of the training plan in the exercise cycle, with the training part as the anchor point, the muscles with insufficient strength are assigned to the corresponding loading movements, and the muscles with tight strength are assigned to the corresponding stretching movements.

[0031] Optionally, the automatic generation of a periodic training plan for the future period also includes: conducting a physical fitness assessment every two weeks after the subject follows the previous training plan (i.e., assessing the comprehensive scores of the four dimensions of mobility, stability, symmetry and sports injury risk), and making corresponding adjustments to the training plan based on the comprehensive scores of the latest status of each body part and the completion of the previous training plan. The training plan adjustments include adjustments to individual training items and training intensity.

[0032] Optionally, the training intensity is adjusted according to the rules for changing the intensity of exercise:

[0033] When the intensity is low, if you can complete 16 movements in a row in the last training session, the movement will be upgraded to a medium-intensity movement; if you cannot complete 16 movements in a row, the intensity of the movement will not be upgraded;

[0034] When the intensity is medium, if you can complete 12 consecutive reps in the last training session, the action will be upgraded to a high-intensity action; if you cannot complete 6 consecutive reps, the action will be downgraded to a low-intensity action.

[0035] When the intensity is high, if you fail to complete 4 consecutive reps in the last training session, the action will be downgraded to a medium-intensity action; if you can complete 4 consecutive reps, the action will not be downgraded;

[0036] When the quality of action completion in the last training program conflicts with the training intensity matched by the current sports test index score, the training intensity matched by the test index score shall prevail.

[0037] Optionally, the training intensity in the first generated periodic training plan is based on the current comprehensive score, and the matching rules are as follows: low intensity is matched for high risk, medium intensity is matched for medium risk, and high intensity is matched for low risk.

[0038] In a second aspect, an embodiment of the present invention provides a human training program generation system based on visual assessment, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the computer program.

[0039] (III) Beneficial effects

[0040] The beneficial effect of the present invention is that the human training program generation method and system based on visual recognition of the present invention adopts a set of solutions that evaluates the sports fitness level of multiple parts of the body by the quality of the subject's completion of standard movements, and then automatically generates a targeted training plan based on the evaluation results. Compared with the existing technology, it solves the technical problem that the existing human function evaluation and training programs have high professional requirements for the developers of the training programs, and achieves the ability to output personalized training programs through internal logic. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of a flow chart of a method for generating a human training program based on visual assessment according to Embodiment 1 of the present invention;

[0042] Figure 2 This is a schematic diagram of the positions of multiple marking points on a human body according to Example 1 of the present invention;

[0043] Figure 3 This is a schematic diagram of the process of physical fitness assessment according to Example 1 of the present invention;

[0044] Figure 4 This is a schematic diagram of a process for generating a periodic training program according to Embodiment 1 of the present invention;

[0045] Figure 5 A flow chart of training intensity adjustment of a periodic training program according to Embodiment 1 of the present invention;

[0046] Figure 6 This is a schematic diagram of the training intensity variation rule of Example 1 of the present invention;

[0047] Figure 7 This is a muscle group analysis flow chart of Example 1 of the present invention. DETAILED DESCRIPTION

[0048] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0049] The embodiment of the present invention proposes a method for generating a human training program based on visual evaluation. The method evaluates the motor fitness level of multiple parts of the body by the quality of the subjects' completion of standard movements, and then automatically generates a set of solutions for targeted training plans based on the evaluation results. The method solves the technical problem that the existing human function evaluation and training programs have high professional requirements for the developers of the training programs, and can output personalized training programs through internal logic.

[0050] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0051] Example 1

[0052] Figure 1 1 is a flow chart of a method for generating a human training program based on visual assessment in Example 1 of the present invention. The method for generating a human training program based on visual assessment in Example 1 of the present invention comprises the following steps:

[0053] S1: Prompt the subject to complete the specified standard action, continuously collect human body images through the depth camera, and identify multiple landmarks of the human body from the human body images according to the visual recognition system.

[0054] Figure 2 It is a schematic diagram of multiple markers of the human body. During implementation, the steps of joint marker recognition are as follows:

[0055] Human body images are collected by a camera, which is a depth camera with 3D function. The spatial coordinates of human body markers are extracted according to a visual recognition system. The 25 markers include: head, neck, midpoint of clavicle, tenth thoracic vertebra, sacral vertebra, left / right shoulder, left / right elbow, left / right wrist, left / right hand, left / right thumb, left / right middle finger, left / right hip, left / right knee, left / right ankle and left / right foot.

[0056] ι i =(x i ,y i , z i ) represents the coordinates of the i-th marker point, and the subject marker set is represented by R(x, y, z). During the test, 25 human body markers in each time series were collected. The 25 markers include: head, neck, mid-clavicular point, 10th thoracic vertebra, sacral vertebra, left / right acromion joint, left / right elbow, left / right wrist, left / right hand, left / right thumb, left / right middle finger, left / right anterior superior iliac part, left / right knee, left / right ankle, left / right foot.

[0057] The marking points and numbers are shown in Table 1 and Figure 2 As shown:

[0058] Table 1 Human body landmarks and numbers

[0059] serial number Marking Points serial number Marking Points 0 head 13 Right middle finger 1 neck 14 Right thumb 2 Mid-clavicular point 15 Tenth thoracic vertebra 3 Left shoulder 16 Sacrum 4 left elbow 17 Left hip 5 left wrist 18 Left knee 6 Left Hand 19 Left ankle 7 Left middle finger 20 Left foot 8 Left thumb 21 Right hip 9 Right shoulder 22 right knee 10 right elbow 23 Right ankle 11 right wrist 24 Right foot 12 Right hand

[0060] Figure 2In the system, the human body markers defined by the system usually correspond to joints with a certain degree of freedom on the human body or bone markers on the surface. The current state of the human body is estimated by calculating the relative positions of the key points of the human body in three-dimensional space. Due to the particularity of the human body structure, the lines between the markers are only connected by the human body structure.

[0061] Then collect the marker points:

[0062] Due to the different fixed points of human muscles when working, they are divided into near fixation, far fixation, upper fixation, lower fixation and no fixation. The specific definition of muscle origin and insertion is: the attachment point of the muscle close to the trunk or close to the median plane of the human body is called the origin, and the attachment point far away from the trunk or far away from the median plane is called the insertion. The muscle origin and insertion are constant. When the muscle contracts, if the starting point is relatively fixed, it is called near fixation, and if the insertion point is relatively fixed, it is called far fixation; the trunk muscles (such as the rectus abdominis) are parallel to the long axis of the human body. When these muscles contract, if the attachment point at the upper end is relatively fixed, it is called upper fixation, if the attachment point at the lower end is relatively fixed, it is called lower fixation, and if both the upper and lower ends are not fixed, it is called no fixation contraction.

[0063] In this embodiment, multiple marking points are respectively used to form multiple joint angles according to the starting point-middle point-end point, including: neck, upper spine, right upper spine, left upper spine, middle spine, right spine bottom, left spine bottom, right wrist, right elbow, right shoulder, right hip, right knee, right ankle, left wrist, left elbow, left shoulder, left hip, left knee and left ankle. When static, posture and action recognition only adopts the lower fixation method, so the joint angles and their starting and ending points (starting point and end point) and middle points are shown in Table 2:

[0064] Table 2 Definition of joint angles

[0065] Serial number Joint Angle Name starting point Midpoint end 1 neck Mid-clavicular point neck head 2 Upper spine Tenth thoracic vertebra Mid-clavicular point neck 3 Right upper spine neck Mid-clavicular point Right shoulder 4 Left upper spine neck Mid-clavicular point Left shoulder 5 Mid-spine Mid-clavicular point Tenth thoracic vertebra Sacrum 6 Right spine base Tenth thoracic vertebra Sacrum Right hip 7 Base of left spine Tenth thoracic vertebra Sacrum Left hip 8 right wrist right elbow right wrist Right thumb 9 right elbow Right shoulder right elbow right wrist 10 Right shoulder Mid-clavicular point Right shoulder right elbow 11 Right hip Sacrum Right hip right knee 12 right knee Right hip right knee Right ankle 13 Right ankle right knee Right ankle Right foot 14 left wrist left elbow left wrist Left thumb 15 left elbow Left shoulder left elbow left wrist 16 Left shoulder Mid-clavicular point Left shoulder left elbow 17 Left hip Sacrum Left hip Left knee 18 Left knee Left hip Left knee Left ankle 19 Left ankle Left knee Left ankle Left foot

[0066] Use marker points ik0 , ik and ik1 Represents the spatial angle θ of joint point k k , ι ik0 and l ik1 The joint angles of two adjacent markers are calculated as follows:

[0067] ι ik0 =(x ik0 ,y ik0 , z ik0 ) represents the coordinates of the ik0th marker point;

[0068] ι ik =(x ik ,y ik , z ik ) represents the coordinates of the ikth marking point;

[0069] ι ik1 =(x ik1 ,y ik1 , z ik1 ) represents the coordinates of the ik1th marker point;

[0070] Vector m = (x1, y1, z1), n ​​= (x2, y2, z2)

[0071] x1,y1,z1=(x ik0 -x ik ), (y ik0 -y ik ), (z ik0 -z ik )

[0072] x2,y2,z2=(x ik1 -x ik ), (y ik1 -y ik ), (z ik1 -z ik )

[0073]

[0074] The spatial coordinates of the subject's marker points are extracted, and the quality of the subject's test movements is restored through modeling. The motion trajectory of the subject's body marker points and the angles of 19 joints are calculated, and the image information of the subject during the test is saved.

[0075] The camera and system extract the spatial coordinates R corresponding to the time series of the calculated marker points and the angle θ corresponding to the joint angle.

[0076] S2: Record the spatial coordinates corresponding to the multiple marking points in the time series to obtain the marking point sequence data; and calculate the corresponding joint angles to obtain the joint angle sequence data.

[0077] Mark point spatial coordinate position data processing:

[0078] When training the system to automatically identify markers, in addition to using the source data set, the system also trains the recognition system based on its own exercise action library. Different from the open source action library, this action library includes a large number of test actions and exercise actions, which can improve the recognition accuracy of the system for markers.

[0079] After the marker points are identified, data denoising is first performed. After denoising using the wavelet threshold method, the acceleration of the marker points is obtained based on the marker point coordinates and time series. The data limit value is set again based on the upper limit of acceleration that the human body can reach and the results of expert discussion. If the limit value is exceeded, the data is deleted. After these two processes, a marker point sequence that is more consistent with the human body is obtained.

[0080] During the test, the subjects make changes in movements and postures according to the system prompts, and the system performs subsequent calculations on the collected and processed data.

[0081] Joint angle data processing method:

[0082] During the test, the subjects performed changes in movements and postures according to the system prompts, obtained spatial marker point sequence data, and obtained joint angle sequence data through point-to-angle calculation. After the joint angles were calculated, thresholds were set based on discussions between experts and project personnel, and unreasonable joint angles were deleted, thereby reducing the frequency of unreasonable angles and obtaining a sequence that is more in line with the human joint activity pattern.

[0083] S3: Multiple marker points are grouped according to human body parts, and the comprehensive scores of the four dimensions of mobility, stability, symmetry and sports injury risk of each group are evaluated through the marker point sequence data and joint angle sequence data.

[0084] After the test, the system will score the subjects in four dimensions: range of motion, stability, symmetry and sports injury risk. The test report is scored according to the human body segments. In this embodiment, the human body segments are divided into five parts based on the joint angles, including the neck (neck), upper limbs (shoulder joints, elbow joints, wrist joints), trunk (spine: upper spine, middle spine, right spine bottom, left spine bottom; upper limb belt: upper right spine, upper left spine), pelvis (hip joint), lower limbs (knee joint, ankle joint). The test report shows the scores of the five parts in four dimensions and graded them. The stability and symmetry calculated by the test indicators are calculated using the PCMHS algorithm to calculate the indicator scores, the joint range of motion percentile method is used to calculate the scores, and the sports injury risk uses the AHP-CRITIC mixed weighted method to determine the weight coefficients of each evaluation indicator, and the comprehensive score is obtained as the final evaluation indicator.

[0085] According to the patent document "Physical health assessment method based on joint point identification (publication number: CN115497626A)", a comprehensive risk score is used to evaluate the exercise risks of five major parts of the body. The higher the score, the higher the exercise risk of the subject and the lower the load intensity of the exercise training plan.

[0086] Based on the sports risk score, multiple indicators of individual joints are further scored for flexibility (muscle tightness) or stability (muscle weakness).

[0087] See also Figure 3, through visual recognition, big data, edge computing, cloud computing and database technologies, based on sports measurement, sports training, functional anatomy, sports rehabilitation, etc. as professional basis, the quality of the subjects' completion of standard movements is automatically evaluated to evaluate their body's "sports fitness" level (the range of motion of the main joints, movement stability, symmetry and the risk of sports injuries of the main body parts), so as to automatically generate a set of solutions for targeted training plans based on the evaluation results.

[0088] S4: According to the comprehensive scores of each group, a periodic training plan for a period of time in the future is automatically generated. The periodic training plan includes basic training and improvement exercise training for improving the current weak parts corresponding to the group with a low comprehensive score.

[0089] When implementing, see Figure 4 , and an evaluation report can also be generated. According to the test evaluation result report, based on the training action library, training action matching rules, training program generation rules and exercise intensity change rules, a training plan (periodic training plan) of six times a week and lasting for two weeks will be given. On the basis of normal training, this plan will improve the current weak parts of the subjects. At the same time, the subjects can follow the training plan on the online platform to learn and practice the movements. The exercise intensity change rules apply to subsequent exercise intensity changes in training plans other than the first one.

[0090] The periodic training program in this embodiment can generate a two-week, six-times-per-week training program. It is recommended that the subject conduct a physical fitness assessment every two weeks after following the training program (i.e., re-evaluate: the comprehensive score of the four dimensions of mobility, stability, symmetry and sports injury risk) so that the system can make corresponding adjustments to the training program based on the latest status of each part of the subject's body and the completion of the previous training cycle.

[0091] Training plan generation method:

[0092] When formulating a training plan, this embodiment modifies the plan for all parts of the body for some people who only train for a few parts of the body during fitness. This avoids the problem of muscle imbalance caused by long-term training of a single part. The training plan formulated in this embodiment covers all major muscle groups and combines aerobic exercise and strength training to train the body in all aspects.

[0093] In the training program of this embodiment, a default training program is output first, and then individual training items and training intensity are adjusted based on the training program according to the test results of the subjects.

[0094] 1) Basic structure of training program output:

[0095] This embodiment will output a training plan for six times a week, and each training includes warm-up exercises, formal exercises, and stretching exercises. The functional parts in the training process are divided into five categories, and the five categories can be summarized into three types, which are three training objectives, namely lower limb strength (lower limbs), upper limb strength (shoulder and upper limbs), and core strength (neck, trunk, pelvis). With the three training objectives as the main actions, two cycles are performed per week, and the training actions are matched according to the subject's weak links and training objectives.

[0096] The training objectives of the first and fourth training sessions are to stretch and strengthen the muscles of the lower limbs; the training objectives of the second and fifth training sessions are to stretch and strengthen the muscles of the upper limbs; and the training objectives of the third and sixth training sessions are to stretch and strengthen the muscles of the core trunk. The specific training objectives are shown in Table 3:

[0097] Table 3 Main objectives of a single training cycle per day

[0098] Number of training sessions Training Purpose First training Lower body strength Second training Upper body strength The third training Core Strength Fourth training Lower body strength Fifth training Upper body strength Sixth training Core Strength

[0099] Each training plan includes two activation movements, four loading movements and two stretching movements. According to the different training objectives of the training plan in the exercise cycle, the training parts (links) are used as anchor points to assign muscles with insufficient strength to the corresponding loading movements, and muscles with tight strength to the corresponding stretching movements.

[0100] Each time a training program is output, a text description is given. For example, the first training description is: The main goal of this training day is to strengthen the lower limb strength. And, based on your test results, the **muscle (deficient) strength is strengthened in the plan, while the **muscle (tense) is relaxed.

[0101] The training program attribute code tags include training items, training equipment, training parts, training forms, movement difficulty, left and right sides, part functions, and involved muscles. Some tables are shown in Table 4:

[0102] Table 4 Attribute code tags

[0103]

[0104] 2) Rules for changing training intensity:

[0105] The training intensity in the training plan is time-sensitive. The intensity of the training plan output for the first time is based on the test index score of this time, matching low intensity (75-100 points for high risk), medium intensity (36-74 points for medium risk), and high intensity (0-35 points for low risk). The intensity of each action is matched with the risk score to generate the first training plan.

[0106] Taking the left elbow flexion score as an example, if the subject's elbow flexion risk score is 80 during the activity, the training program output will proceed as follows:

[0107] Analysis result output:

[0108] ① In the "Indicator Value Configuration" table, search for the positioning category of elbow flexion, and find the "flexible" positioning, and the function number is left elbow flexion F41;

[0109] ② Based on step ①, classify all muscles into “flexible” and “stable” and sort the risk scores. If elbow flexion is in the top three movements in terms of flexibility risk score, proceed to the next step;

[0110] ③ In the "Part Function Configuration" table, find the antagonist function number and the corresponding active muscle group corresponding to F41, and get the antagonist function number as F43, and call the active muscle group M12 of F43;

[0111] ④ Find the corresponding muscle in the "Muscle Configuration" table, which is M12 triceps brachii (left);

[0112] ⑤ Output corresponding text content according to positioning classification and risk score sorting, and interpret the results.

[0113] Training movements and intensity output:

[0114] ① The risk score of elbow flexion is 80. According to the "Attribute Code Label" table, the corresponding "Training Part" is 03 Elbow, "Part Function" is 41 Elbow Flexion (Left), "Action Difficulty" is 01 Level 1, and "Left and Right Side" is 01 Left.

[0115] ② In the "Action Screening" table, the following actions are obtained after screening in step ①: No. 1115 Left Dumbbell Isolated Curl, No. 1120 Bent-over Left Dumbbell Extension;

[0116] ③ According to the main purpose of each day in a single training cycle, allocate the movements: allocate the elbow flexion movements that need to be strengthened to the second and fifth training sessions (see Figure 5 ).

[0117] Principles of training intensity changes:

[0118] This system recommends that subjects perform an exercise test every two weeks to monitor their physical condition in real time. In the subsequent training program generation, follow the principle of training intensity change, see Figure 6 The specific principles are as follows:

[0119] ① When the intensity is low, if you can complete 16 movements in a row in the last training, the movement can be upgraded to a medium-intensity movement; if you fail to complete 16 movements in a row, the intensity of the movement will not be upgraded;

[0120] ② When the intensity is medium, if you can complete 12 consecutive movements in the last training, the movement can be upgraded to a high-intensity movement; if you cannot complete 6 consecutive movements, the movement can be downgraded to a low-intensity movement;

[0121] ③ When the intensity is high, if you fail to complete 4 consecutive reps in the last training session, downgrade the action to a medium-intensity action; if you can complete 4 consecutive reps, the action will not be downgraded;

[0122] ④ When the quality of action completion in the last training program conflicts with the training intensity matched by the current sports test index score, the training intensity matched by the test index score shall prevail.

[0123] Training Action Matching Method

[0124] The training action first obtains the scores of various indicators through visual recognition detection, and then classifies the indicator definitions, and finally sorts the sub-item scores and selects the top three scores for subsequent training plan formulation.

[0125] When the human body moves, the joints involved in the movement have a certain degree of flexibility and stability. Flexibility refers to the ability of a joint to move freely within the range of motion, and stability refers to the ability of a joint to resist movement and control joint movement. Some joints are mainly flexible, while some joints are mainly stable. For example, in human movement, the shoulder, ankle, and hip joints require higher flexibility, while the knee and lumbar joints require higher stability. Therefore, in this embodiment, the indicators are defined as "flexible" or "stable" to describe the flexibility or stability of each joint, that is, muscle tension or lack of muscle strength. See Figure 7 After visual recognition detection, training action matching is first performed. The training action matching method ranks the risk scores by the scores of each indicator and the corresponding functional positioning of the indicator (flexible or stable). The top three flexibility indicators and stability indicators with the highest risk scores are selected, and the muscle groups that are antagonistic to the indicator functions are searched in the database and then output. The muscle groups are output as text analysis functions in the evaluation results, and the six indicators with higher risk scores are output as the focus of training in the subsequent training plan in the evaluation results and training plans.

[0126] Training action calling logic:

[0127] 1) “Flexibility” and “stability” were ranked in order of scores from low to high, and the three functions with the lowest flexibility and the lowest stability were obtained.

[0128] 2) Find the corresponding “antagonistic functions” for the three functions with the lowest flexibility and record them as “Tension 1”, “Tension 2” and “Tension 3” respectively.

[0129] 3) Find the corresponding “antagonistic functions” for the three functions with the lowest stability and record them as “less than 1”, “less than 2”, and “less than 3” respectively.

[0130] 4) "Tense 1", "Tense 2", "Tense 3" and "Insufficient 1", "Insufficient 2", "Insufficient 3" are the training action call parts. In this embodiment, the output muscle indexes are all the main antagonistic muscle groups in the scoring index.

[0131] At the same time, the functions of different parts of the body are different, and the "flexibility" and "stability" of each indicator are different during the test. The configuration table of the indicator value for the hip is used as an example (Table 5) for explanation:

[0132] Table 5 Hip index value configuration table

[0133] Indicator name Calculation formula description position Function No. Right hip range of motion The absolute value of angle 11 (starting position - peak position) in the sagittal plane projection is X flexible F58 Left hip range of motion The absolute value of angle 17 (starting position - peak position) in the sagittal plane projection is X flexible F57 Left hip extension The absolute value of the projection of angle 17 (starting position - peak position) on the sagittal plane is X flexible F59 Right hip extension The absolute value of angle 11 (starting position - peak position) in the sagittal plane projection is X flexible F60

[0134] According to the different functions of the parts corresponding to each indicator, different indicators correspond to different main muscles in the output results.

[0135] The specific parts and their corresponding index numbers and muscle numbers are shown in Table 6:

[0136] Table 6 Human body parts and their function matching

[0137]

[0138]

[0139] This embodiment evaluates the sports fitness level of multiple parts of the body by the quality of the subjects' completion of standard movements, and then automatically generates a set of solutions for targeted training plans based on the evaluation results, and can output personalized training plans through internal logic.

[0140] Example 2

[0141] In this embodiment, the five major parts of the subject and the comprehensive physical movement risk score are calculated by the patent "physical health assessment method based on joint point recognition (publication number: CN115497626A)" as follows:

[0142] Table 7 Risk score and evaluation

[0143] Location Risk Score evaluate neck 36.0 Medium risk Shoulder and upper limbs 21.0 Low risk trunk 25.7 Low risk pelvis 30.3 Low risk Lower limbs 23.3 Low risk

[0144] According to the risk score evaluation, when training the neck position, the system will select from the medium-intensity movement library, and when training the shoulders, upper limbs, trunk, pelvis and lower limbs, the system will select from the high-intensity movement library.

[0145] At the same time, in this test, the top three index functions with poor flexibility are: neck flexion, neck lateral flexion (left), hip extension (right); the top three index functions with poor stability are: neck extension, spinal rotation (left), neck rotation (right). That is, the top three poor muscles in terms of flexibility (muscle tension) are: upper trapezius (right), upper trapezius (left), iliopsoas (right); the top three poor muscles in terms of stability (lack of muscle strength) are: rectus capitis anterior (left), internal oblique muscle (right), sternocleidomastoid (left).

[0146] The default training plan will be modified in terms of exercise intensity and exercise movements based on the test results of the subjects, and six muscle groups will be output as analysis results.

[0147] That is, after modifying the training intensity in the default training plan, in the third and sixth times (core muscle training), the training movements were modified to stretching exercises including: neck flexion, neck lateral flexion (left), hip extension (right); the formal exercises included: neck extension, spinal rotation (left), neck rotation (right).

[0148] The following are the test results of the subject indicators and the output results of the training report:

[0149] You have a good foundation of physical activity quality; you have a good level of left lower limb mobility and pelvic (lumbar) symmetry, but the trunk symmetry quality needs to be strengthened.

[0150] Your body flexibility and mobility are relatively good; however, your body muscle strength is relatively weak.

[0151] Among them, the flexibility of the left neck is relatively weak, the strength of the left neck is relatively weak, and the symmetry of the torso is weak.

[0152] Among physical activities, exercise risk levels are lower.

[0153] There is a certain risk of cervical pain, which needs to be paid attention to during training

[0154] Table 8 The third training program

[0155] Serial number action Within group Number of groups Between groups (seconds) 1 Neck forward and backward movement 10 times 1 30 2 Nod and look up 10 times 1 30 3 Neck back against 10 times 3 30 4 Lunge Back Twist 10 times 3 30 5 Hip rotation with bridge 10 times 3 30 6 Chin reduction 10 times 3 30 7 Standing back leg raise 30 seconds 1 10 8 Prone Leg Raise 30 seconds 1 10

[0156] Embodiment 3:

[0157] This embodiment provides a human training program generation system based on visual assessment, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the above embodiments are implemented.

[0158] In summary, the method and system for generating a human training program based on visual recognition of the present invention adopts a set of solutions that evaluates the sports fitness level of multiple parts of the body by the quality of the subject's completion of standard movements, and then automatically generates a targeted training plan based on the evaluation results. Compared with the prior art, it solves the technical problem that the existing human function evaluation and training programs have high professional requirements for the personnel formulating the training programs, and achieves the ability to output personalized training programs through internal logic. In the preferred embodiment, the training intensity is adjusted in stages during the training process, so that the training program is more in line with the subject's physical quality and fitness.

[0159] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0160] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0161] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0162] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0163] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for generating a human training program based on visual assessment, characterized in that: The following steps are involved: Prompting the subject to complete a specified standard action, continuously collecting human body images through a depth camera, and identifying multiple marking points of the human body from the human body images according to a visual recognition system; Record the spatial coordinates corresponding to multiple marking points in the time series to obtain marking point sequence data; and calculate the corresponding joint angles to obtain joint angle sequence data; The multiple markers are grouped according to the human body parts, and the marker sequence data and joint angle sequence data are used to evaluate the comprehensive scores of the four dimensions of mobility, stability, symmetry and sports injury risk of each group; According to the comprehensive scores of the groups, a periodic training program for a period of time in the future is automatically generated, and the periodic training program includes basic training and improvement exercise training for improving the current weak parts corresponding to the groups with low comprehensive scores.

2. The method for generating a human training program based on visual assessment according to claim 1, characterized in that: The multiple human body landmarks include: head, neck, mid-clavicle, tenth thoracic vertebra, sacral vertebra, left / right shoulder, left / right elbow, left / right wrist, left / right hand, left / right thumb, left / right middle finger, left / right hip, left / right knee, left / right ankle and left / right foot.

3. The method for generating a human training program based on visual assessment as claimed in claim 2, characterized in that: The multiple marking points respectively form multiple joint angles according to the starting point-middle point-end point, including: neck, upper spine, right upper spine, left upper spine, middle spine, right spine bottom, left spine bottom, right wrist, right elbow, right shoulder, right hip, right knee, right ankle, left wrist, left elbow, left shoulder, left hip, left knee and left ankle; The joint angles are grouped according to the human body segments, including five groups: neck, upper limbs, trunk, pelvis and lower limbs, among which: The neck includes the angles of joints: the neck; Upper limbs include joint angles: left shoulder, right shoulder, left elbow, right elbow, left wrist and right wrist; The trunk includes the joint angles: upper spine, middle spine, right spine base, and left spine base; and right spine upper and left spine upper; The pelvis includes the joint angles: left and right hip; Lower limbs including joint angles: left knee, right knee, left ankle, and right ankle; The five groups correspond to the following body parts in the periodic training program: lower limbs, upper limbs and core; the core includes the neck, trunk and pelvis.

4. The method for generating a human training program based on visual assessment according to claim 1, characterized in that: The visual recognition system is trained on a data set using a machine learning method. The data set includes a database of key points of the human body in motion in open source data and a motion exercise action library obtained through self-training. The motion exercise action library contains motion record data of test actions and exercise actions.

5. The method for generating a human training program based on visual assessment according to claim 1, characterized in that: The method further comprises: After identifying the spatial coordinates of multiple markers of the human body, the wavelet threshold method is first used to perform data noise reduction. Then, based on the spatial coordinates and time series of the markers, the acceleration of the markers is obtained. Based on the upper limit of acceleration that the human body can reach and the results of empirical evaluation, the data limit value is set again, and the spatial coordinates that exceed the limit value are deleted. After calculating the corresponding joint angles, a threshold is set based on the empirical evaluation results, and angles exceeding the threshold are deleted.

6. The method for generating a human training program based on visual assessment according to claim 3, characterized in that: The method further comprises: scoring and ranking multiple indicators of a single joint for assessing flexibility of muscle tension or risk scores for assessing stability of muscle insufficiency according to the score of sports injury risk; finding out the indicators with the highest ranking and the muscle groups whose functions are mutually antagonistic; and determining the main muscles according to the muscle groups; A single cycle training program includes six training plans per week with lower limb strength, upper limb strength and core strength as training targets respectively; each training plan includes activation movements, loading movements and stretching movements; According to the different training goals of the training plan in the exercise cycle, with the training part as the anchor point, the muscles with insufficient strength are assigned to the corresponding loading movements, and the muscles with tight strength are assigned to the corresponding stretching movements.

7. The method for generating a human training program based on visual assessment according to claim 1 or 6, characterized in that: The automatic generation of a periodic training plan for a period of time in the future also includes: conducting a physical fitness assessment every two weeks after the subject follows the previous cycle training plan, and adjusting the training plan accordingly based on the comprehensive score of the latest status of each body part and the completion of the previous cycle training plan, and the training plan adjustment includes the adjustment of individual training items and training intensity.

8. The method for generating a human training program based on visual assessment according to claim 7, characterized in that: The training intensity is adjusted according to the rules for changing the intensity of exercise: When the intensity is low, if you can complete 16 movements in a row in the last training session, the movement will be upgraded to a medium-intensity movement; if you cannot complete 16 movements in a row, the intensity of the movement will not be upgraded; When the intensity is medium, if you can complete 12 consecutive reps in the last training session, the action will be upgraded to a high-intensity action; if you cannot complete 6 consecutive reps, the action will be downgraded to a low-intensity action. When the intensity is high, if you fail to complete 4 consecutive reps in the last training session, the action will be downgraded to a medium-intensity action; if you can complete 4 consecutive reps, the action will not be downgraded. When there is a conflict between the quality of action completion of the last training program and the training intensity matched by the sports test index score of that time, the training intensity matched by the test index score shall prevail.

9. The method for generating a human training program based on visual assessment according to claim 8, characterized in that: The training intensity in the first generated periodic training plan is based on the comprehensive score of this time, and the matching rules are as follows: low intensity matches high risk, medium intensity matches medium risk, and high intensity matches low risk.

10. A system for generating a human training program based on visual assessment, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 9 are implemented.

Images