Fitness live broadcast action annotation processing system and method based on big data

Abstract

The present invention discloses a fitness exercise live broadcast action labeling and processing system and method based on big data, which relates to the field of intelligent fitness technology. The technical points of the scheme are: accurately labeling fitness actions in live broadcasts, detecting the positions of key points of the human body, constructing feature vectors that describe the actions, calculating the similarity between the real-time action feature vectors and the standard action feature vectors, calculating fatigue through real-time heart rate variability and exercise duration, as well as a heart rate variability baseline value, taking action standardization, action intensity, heart rate variability and fatigue as risk features, calculating the weights of various risk features through a risk assessment matrix, combining the values ​​of the risk features to calculate action risk assessment values, comparing the action risk assessment values ​​with risk thresholds, taking corresponding measures based on the comparison results, and realizing real-time analysis of exercise data of different users under specific actions, thereby effectively avoiding the occurrence of sports injuries.

Description

Technical Field

[0001] The present invention relates to the field of intelligent fitness technology, and in particular to a system and method for labeling and processing fitness exercise live broadcast movements based on big data. Background Art

[0002] With the rapid development of information technology and the advent of the big data era, the field of fitness has also ushered in unprecedented changes. Traditional fitness methods can no longer meet modern people's pursuit of health, efficiency and personalization. The big data-based fitness live action annotation and processing system and method have emerged, providing fitness enthusiasts with more accurate and scientific exercise guidance and experience.

[0003] In the Chinese invention application with application publication number CN117274319A, a method and system for live broadcast of fitness exercises based on big data is disclosed, including: constructing human body contour models of online fitness users and live broadcast coaches; constructing a standard human body tracking frame that is compatible with the fitness users; establishing a three-dimensional coordinate system of the fitness user's human body contour model and the human body tracking frame, recording the initial coordinates of the fitness user's human body contour model and the human body tracking frame and the initial distance between the two; dynamically tracking the human body contour of the fitness user during exercise, monitoring physiological sign data during exercise; obtaining the deviation distance and physiological sign deviation data of the fitness user during exercise; obtaining the fitness user's unqualified movements and movements with inappropriate exercise intensity; issuing movement correction instructions or intensity adjustment instructions to the coach end based on statistical data; the coach end makes corresponding adjustments to the live broadcast exercise based on the received instructions.

[0004] In view of the above invention application, the prior art has the following deficiencies:

[0005] 1. In fitness exercises, different movements have different intensity levels due to their mechanical properties and physiological effects. The same movement can also have different intensities for different people. However, existing technologies often overlook this key difference and fail to conduct detailed risk assessments for specific movements. This can lead users to blindly follow live streams without understanding the actual intensity of the movements, increasing the risk of injury.

[0006] 2. Although existing fitness live streaming systems have made some progress in motion recognition and labeling, most of them still remain at the level of simple recognition of the movements themselves. They fail to fully integrate personalized factors such as the user's physical condition, physical fitness level, and exercise habits, and fail to conduct in-depth analysis of the potential risks that users may face in specific movements. Summary of the Invention

[0007] (1) Technical problems solved

[0008] In response to the deficiencies in the prior art, the present invention provides a fitness exercise live broadcast action labeling and processing system and method based on big data, which accurately labels fitness actions in live broadcasts, detects the positions of key points of the human body, constructs feature vectors describing the actions through the coordinates of the key points, obtains action standardization by calculating the similarity between the real-time action feature vector and the standard action feature vector, calculates fatigue through the user's heart rate variability and exercise duration during fitness exercise, and the heart rate variability baseline value, establishes a risk assessment matrix using action standardization, action intensity, heart rate variability and fatigue as risk features, calculates the weight of each risk feature through the risk assessment matrix, calculates the action risk assessment value by combining the value of each risk feature, compares the action risk assessment value with the risk threshold, and takes corresponding measures based on the comparison results, thereby solving the problems mentioned in the background technology.

[0009] (2) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for labeling and processing live fitness exercise movements based on big data, comprising the following steps:

[0011] Collect the user's motion data during live fitness broadcasts in real time, perform real-time recognition on the motion data, identify the user's fitness movements, and accurately label the fitness movements in the live broadcast based on the recognition results;

[0012] Extract key frames from real-time video streams. The key frames correspond to the marked fitness movements. Detect the positions of key points on the human body. Construct feature vectors describing the movements based on the coordinates of the key points. Calculate the similarity between the real-time movement feature vector and the standard movement feature vector to obtain the movement standardization.

[0013] Obtain ECG signals or heartbeat interval data, extract the heartbeat interval time series, calculate heart rate variability, and calculate fatigue based on the user's heart rate variability, exercise duration, and heart rate variability baseline value during fitness exercise. The calculation formula is as follows:

[0014]

[0015] Among them, RCH represents fatigue, HRV represents heart rate variability, represents the baseline value of heart rate variability, and T represents the duration of exercise;

[0016] Taking movement standardization, movement intensity, heart rate variability and fatigue as risk characteristics, a risk assessment matrix is ​​established. The weight of each risk characteristic is calculated through the risk assessment matrix. Combined with the value of each risk characteristic, the movement risk assessment value is calculated and compared with the risk threshold, and corresponding measures are taken according to the comparison results.

[0017] Furthermore, key frames are extracted from the real-time video stream. The key frames correspond to the annotated actions. Background subtraction is used to remove the background, retaining only the human body part. The human body image is adjusted to a uniform size and proportion. The position of the key points of the human body is detected using the human posture estimation model, and the feature vector describing the action is constructed through the coordinates of the key points.

[0018] Furthermore, the action standardization is obtained by calculating the similarity between the real-time action feature vector and the standard action feature vector. The calculation formula is as follows:

[0019]

[0020] Among them, S represents the action standard, x i and y i They represent the i-th element of the real-time action and standard action feature vectors respectively, i = 1, 2, ..., N, and N represents the total number of elements.

[0021] Furthermore, the ECG signal or heartbeat interval data is acquired in real time, the heartbeat interval time series is extracted, and the heart rate variability is calculated based on the heartbeat interval data. The calculation formula is as follows:

[0022]

[0023] Among them, HRV represents heart rate variability, RR j represents the jth heartbeat interval, j = 1, 2,, M, and M represents the total number of heartbeat intervals.

[0024] Furthermore, taking movement standard, movement intensity, heart rate variability, and fatigue as risk characteristics, a risk assessment matrix A is established to assess the risk of the current fitness movement. The risk assessment matrix A is as follows:

[0025]

[0026] Among them, a mn It is expressed as the importance of the m-th risk feature relative to the n-th risk feature, where m is the number of rows in matrix A and n is the number of columns in matrix A.

[0027] Furthermore, the weight ω of each risk feature is calculated through the risk assessment matrix A m , the calculation formula is as follows:

[0028]

[0029] Where S is the number of risk characteristics, k is a constant, and 0 <k<S。

[0030] Furthermore, based on the value of each risk feature and its corresponding weight, each risk feature is weighted and summed to calculate the action risk assessment value. The calculation formula is as follows:

[0031]

[0032] Among them, ARV represents the action risk assessment value, V m represents the value of the mth risk feature, ω m represents the weight of the mth risk feature, m = 1, 2, …, S, and S represents the number of risk features.

[0033] Furthermore, a risk threshold is pre-set, the action risk assessment value is compared with the risk threshold, and corresponding measures are taken according to the comparison results, including:

[0034] When the action risk assessment value is less than or equal to the risk threshold, the user action execution status is continuously monitored;

[0035] When the action risk assessment value is greater than the risk threshold, timely feedback will be given to the user, suggesting adjusting exercise intensity, taking a rest, or seeking professional guidance.

[0036] The big data-based fitness live broadcast action annotation processing system includes: action recognition and annotation module, action standard evaluation module, physiological fatigue monitoring module and risk assessment and response module;

[0037] The action recognition and annotation module collects the user's action data during live fitness broadcasts in real time, identifies the user's fitness movements in real time, and accurately annotates the fitness movements in the live broadcast based on the recognition results;

[0038] The action standardization evaluation module extracts key frames from the real-time video stream, detects the positions of key points on the human body, constructs a feature vector describing the action based on the coordinates of the key points, and obtains the action standardization by calculating the similarity between the real-time action feature vector and the standard action feature vector.

[0039] The physiological fatigue monitoring module obtains ECG signals or heartbeat interval data, extracts the heartbeat interval time series, calculates heart rate variability, and calculates fatigue based on the user's heart rate variability and exercise duration during fitness exercise, as well as the heart rate variability baseline value;

[0040] The risk assessment and response measures module takes movement standardization, movement intensity, heart rate variability, and fatigue as risk characteristics, establishes a risk assessment matrix, calculates the weight of each risk characteristic through the risk assessment matrix, combines the value of each risk characteristic, calculates the movement risk assessment value, compares the movement risk assessment value with the risk threshold, and takes corresponding measures based on the comparison results.

[0041] (3) Beneficial effects

[0042] The present invention provides a system and method for labeling and processing fitness exercise live broadcast movements based on big data, which has the following beneficial effects:

[0043] (1) By collecting the user's motion data in real time, the timeliness and continuity of the data are ensured, which helps to more accurately reflect the user's actual status during the live fitness process. By using artificial intelligence technology to perform real-time recognition of the pre-processed motion data, the user's ongoing fitness movements can be quickly and accurately identified, which not only improves the efficiency of annotation, but also ensures the timeliness and accuracy of annotation.

[0044] (2) By using high-precision human key point detection technology, the key parts of the human body can be accurately identified, thereby constructing an accurate motion feature vector. By calculating the similarity between the real-time motion feature vector and the standard motion feature vector, a specific motion standardization score can be given, allowing users to intuitively understand whether their movements are standard.

[0045] (3) By monitoring the user's heart rate variability in real time, the user's physical reaction and fatigue status can be assessed more accurately, and possible sports risks such as abnormal heart rate fluctuations can be discovered in a timely manner, thereby preventing the occurrence of sports injuries. According to the real-time changes in heart rate variability, the exercise intensity and movement difficulty can be dynamically adjusted to avoid unnecessary fatigue or injury caused by excessive exercise.

[0046] (4) Identify user actions through action annotation, conduct real-time analysis of the motion data of different users under specific actions, and evaluate the risk of sports injuries incurred by users under these actions. Once the risk exceeds the preset safety threshold, feedback will be immediately issued to remind users to adjust their actions or reduce the intensity of exercise, thereby effectively avoiding the occurrence of sports injuries. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a schematic diagram of the steps of the method for labeling and processing live fitness exercise movements based on big data of the present invention;

[0048] Figure 2 This is a structural diagram of the big data-based fitness exercise live broadcast action annotation processing system of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] See also Figure 1 The present invention provides a method for labeling and processing live fitness exercise movements based on big data, comprising the following steps:

[0051] Step 1: Collect the user's motion data during live fitness broadcasts in real time, perform real-time recognition on the motion data, identify the user's fitness movements, and accurately label the fitness movements in the live broadcast based on the recognition results;

[0052] The step 1 includes the following contents:

[0053] Step 101: Using sensors such as smart wearable devices and cameras to collect real-time motion data of the user during live fitness broadcast, including body posture, movement trajectory, heart rate, breathing rate, etc.

[0054] Step 102: Preprocessing the collected raw data, including data cleaning, denoising, and standardization. The cleaning step is used to remove abnormal data, the denoising step is used to eliminate interference in the data, and the standardization step is used to convert the data into a unified format.

[0055] Step 103: Using artificial intelligence technology, the pre-processed motion data is recognized in real time to identify the user's ongoing fitness movements. Based on the recognition results, the fitness movements in the live broadcast are accurately labeled, including the movement name, start time, end time, etc.;

[0056] Using artificial intelligence technology, the pre-processed motion data is recognized in real time, including:

[0057] Step 1031: Use a high-definition camera to capture video data of the fitness person in real time, and perform pre-processing operations such as frame extraction, image enhancement, noise removal, human body detection and segmentation on the video data to improve the accuracy of subsequent recognition;

[0058] Step 1032: Using YOLOv5s or other deep learning models, collect and annotate a large amount of video frame data containing different fitness movements, including push-ups, squats, sit-ups, etc., use the annotated data set to train the deep learning model, and adjust the model parameters through an optimization algorithm (such as a genetic algorithm) to improve recognition accuracy;

[0059] Step 1033: Input the pre-processed video frames into the trained deep learning model to detect the user's ongoing fitness movements in real time. Based on the model recognition results, the action name, start time, and end time are automatically marked in the live broadcast screen.

[0060] When using, combine the contents of steps 101 to 103:

[0061] By collecting users' motion data in real time, the timeliness and continuity of the data are ensured, which helps to more accurately reflect the users' actual status during live fitness. By using artificial intelligence technology to perform real-time recognition of pre-processed motion data, it can quickly and accurately identify the users' ongoing fitness movements, which not only improves the efficiency of labeling, but also ensures the timeliness and accuracy of labeling.

[0062] Step 2: Extract key frames from the real-time video stream. The key frames correspond to the labeled fitness movements. Detect the positions of key points on the human body. Construct a feature vector describing the movement based on the coordinates of the key points. Calculate the similarity between the real-time movement feature vector and the standard movement feature vector to obtain the movement standardization.

[0063] The second step includes the following contents:

[0064] Step 201: extract key frames from the real-time video stream, the key frames corresponding to the marked fitness movements, use background subtraction technology to remove the background, retain only the human body part, and adjust the human body image to a uniform size and ratio for subsequent processing;

[0065] It should be noted that background subtraction technology is a commonly used technology in image processing, video analysis and computer vision, such as Gaussian mixture model and K-nearest neighbor algorithm, which is mainly used to separate foreground objects (such as moving objects) and background from video sequences;

[0066] Step 202: using a human posture estimation model to detect the positions of key points (such as joints) of the human body, and constructing a feature vector describing the action based on the coordinates of the detected key points;

[0067] It should be noted that human pose estimation models, such as OpenPose and HRNet, use images or videos to identify and locate the positions of various human joints (such as the head, shoulders, elbows, wrists, hips, knees, ankles, etc.).

[0068] Step 203: The action standardization is obtained by calculating the similarity between the real-time action feature vector and the standard action feature vector. The calculation formula is as follows:

[0069]

[0070] Among them, S represents the action standard, x i and y i They represent the i-th element of the real-time action and standard action feature vectors, i = 1, 2, ..., N, where N represents the total number of elements;

[0071] When using, combine the contents of step 201 to step 203:

[0072] By using high-precision human key point detection technology, we can accurately identify the key parts of the human body, thereby constructing an accurate motion feature vector. By calculating the similarity between the real-time motion feature vector and the standard motion feature vector, we can give a specific motion standardization score, allowing users to intuitively understand whether their movements are standard.

[0073] Step 3: Obtain the ECG signal or heartbeat interval data, extract the heartbeat interval time series, calculate the heart rate variability, and calculate the fatigue level based on the user's heart rate variability and exercise duration during fitness exercise, as well as the heart rate variability baseline value;

[0074] The step three includes the following contents:

[0075] Step 301: Collect historical data of the user's ECG signal or heartbeat interval over a period of time (e.g., 24 hours), obtain the ECG signal or heartbeat interval data in real time, and extract the heartbeat interval time series, i.e., the time interval between two consecutive heartbeats;

[0076] Step 302: Calculate the heart rate variability using the heart rate interval data. The calculation formula is as follows:

[0077]

[0078] Among them, HRV represents heart rate variability, RR j represents the jth heartbeat interval, j = 1, 2, ..., M, and M represents the total number of heartbeat intervals;

[0079] Step 303: Calculate the fatigue level based on the user's heart rate variability and exercise duration during exercise, as well as the heart rate variability baseline value. The calculation formula is as follows:

[0080]

[0081] Among them, RCH represents fatigue, HRV represents heart rate variability, It represents the baseline value of heart rate variability, which is the average HRV value in a certain period of time before (such as the previous day or the previous week). T represents the duration of exercise.

[0082] It should be noted that, generally speaking, as exercise intensity increases, HRV values ​​may first increase and then decrease. During low- to moderate-intensity exercise, HRV values ​​may remain relatively stable or slightly increase; during high-intensity exercise, HRV values ​​may decrease. Therefore, when the HRV value is lower, the RCH value is higher, indicating a higher level of fatigue. Conversely, when the HRV value is higher, the RCH value is lower, indicating a lower level of fatigue.

[0083] When using, combine the contents of step 301 to step 303:

[0084] By monitoring the user's heart rate variability in real time, it is possible to more accurately assess the user's physical reaction and fatigue status, and promptly detect possible exercise risks of the user, such as abnormal heart rate fluctuations, thereby preventing the occurrence of sports injuries. According to the real-time changes in heart rate variability, the exercise intensity and movement difficulty are dynamically adjusted to avoid unnecessary fatigue or injury due to excessive exercise.

[0085] Step 4: Take movement standardization, movement intensity, heart rate variability, and fatigue as risk characteristics to establish a risk assessment matrix. Calculate the weight of each risk characteristic through the risk assessment matrix. Combine the values ​​of each risk characteristic to calculate the movement risk assessment value. Compare the movement risk assessment value with the risk threshold, and take corresponding measures based on the comparison results.

[0086] The fourth step includes the following contents:

[0087] Step 401: Using movement standardization, movement intensity, heart rate variability, and fatigue as risk characteristics, a risk assessment matrix A is established to assess the risk of the current fitness movement. The risk assessment matrix A is as follows:

[0088]

[0089] Among them, a mn It is expressed as the importance of the m-th risk feature relative to the n-th risk feature, where m is the number of rows in matrix A and n is the number of columns in matrix A;

[0090] The assessment process of movement intensity specifically includes:

[0091] S1: Calculate the user's maximum heart rate (HRmax), usually using the formula "HRmax = 220 - age" or the formula "HRmax = 206.9 - (0.67 × age)";

[0092] S2: Measure the user's heart rate (HR) during exercise and calculate its percentage of the maximum heart rate (%HRmax), i.e., "%HRmax = (HR / HRmax) × 100%";

[0093] S3: Use %HRmax as the user's action intensity;

[0094] Step 402: Calculate the weight ω of each risk feature using the risk assessment matrix A m , the calculation formula is as follows:

[0095]

[0096] Where S is the number of risk characteristics, k is a constant, and 0 <k<S;

[0097] Step 403: Based on the value of each risk feature and the corresponding weight, each risk feature is weighted and summed to calculate the action risk assessment value. The calculation formula is as follows:

[0098]

[0099] Among them, ARV represents the action risk assessment value, V m represents the value of the mth risk feature, ω m represents the weight of the mth risk feature, m = 1, 2, ..., S, S represents the number of risk features;

[0100] Step 404: Preset a risk threshold, compare the action risk assessment value with the risk threshold, and take corresponding measures based on the comparison results, including:

[0101] When the action risk assessment value is less than or equal to the risk threshold, it means that the action is probably safe under the current conditions, and the user action execution is continuously monitored;

[0102] When the action risk assessment value is greater than the risk threshold, it means that the action may be risky under the current conditions. The user will be given timely feedback and advised to adjust the exercise intensity, take a rest or seek professional guidance.

[0103] When using, combine the contents of step 401 to step 403:

[0104] It identifies user actions through action annotation, conducts real-time analysis of the motion data of different users under specific actions, and assesses the risk of sports injuries to users under these actions. Once it is found that the risk exceeds the preset safety threshold, it will immediately issue feedback to remind users to adjust their actions or reduce the intensity of exercise, thereby effectively avoiding the occurrence of sports injuries.

[0105] See also Figure 2 The present invention also provides a big data-based fitness exercise live broadcast action annotation processing system, including: an action recognition and annotation module, an action standard evaluation module, a physiological fatigue monitoring module, and a risk assessment and response module; wherein,

[0106] The action recognition and annotation module collects the user's action data during live fitness broadcasts in real time, identifies the user's fitness movements in real time, and accurately annotates the fitness movements in the live broadcast based on the recognition results;

[0107] The action standardization evaluation module extracts key frames from the real-time video stream, detects the positions of key points on the human body, constructs a feature vector describing the action based on the coordinates of the key points, and obtains the action standardization by calculating the similarity between the real-time action feature vector and the standard action feature vector.

[0108] The physiological fatigue monitoring module obtains ECG signals or heartbeat interval data, extracts the heartbeat interval time series, calculates heart rate variability, and calculates fatigue based on the user's heart rate variability and exercise duration during fitness exercise, as well as the heart rate variability baseline value;

[0109] The risk assessment and response measures module takes movement standardization, movement intensity, heart rate variability, and fatigue as risk characteristics, establishes a risk assessment matrix, calculates the weight of each risk characteristic through the risk assessment matrix, combines the value of each risk characteristic, calculates the movement risk assessment value, compares the movement risk assessment value with the risk threshold, and takes corresponding measures based on the comparison results.

[0110] In the application, the several formulas involved are all calculated by taking their numerical values ​​after removing the dimensions, and the formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the latest real situation. The coefficients in the formula are set by technical personnel in this field according to actual conditions.

[0111] The above embodiments can be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented using a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0112] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0113] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for labeling and processing fitness exercise live broadcast movements based on big data, characterized by: The following steps are involved: Collect the user's motion data during live fitness broadcasts in real time, perform real-time recognition on the motion data, identify the user's fitness movements, and accurately label the fitness movements in the live broadcast based on the recognition results; Extract key frames from real-time video streams. The key frames correspond to the marked fitness movements. Detect the positions of key points on the human body. Construct feature vectors describing the movements based on the coordinates of the key points. Calculate the similarity between the real-time movement feature vector and the standard movement feature vector to obtain the movement standardization. Obtain ECG signals or heartbeat interval data, extract the heartbeat interval time series, calculate heart rate variability, and calculate fatigue based on the user's heart rate variability, exercise duration, and heart rate variability baseline value during fitness exercise. The calculation formula is as follows: Among them, RCH represents fatigue, HRV represents heart rate variability, represents the baseline value of heart rate variability, and T represents the duration of exercise; Taking movement standardization, movement intensity, heart rate variability and fatigue as risk characteristics, a risk assessment matrix is ​​established. The weight of each risk characteristic is calculated through the risk assessment matrix. Combined with the value of each risk characteristic, the movement risk assessment value is calculated and compared with the risk threshold, and corresponding measures are taken according to the comparison results.

2. The method for labeling and processing fitness exercise live broadcasts based on big data according to claim 1, characterized in that: Key frames are extracted from the real-time video stream. The key frames correspond to the annotated actions. Background subtraction is used to remove the background, retaining only the human body part. The human body image is adjusted to a uniform size and proportion. The position of the key points of the human body is detected using the human posture estimation model. The feature vector describing the action is constructed based on the coordinates of the key points.

3. The method for labeling and processing live fitness exercise movements based on big data according to claim 2, characterized in that: The action standardization is obtained by calculating the similarity between the real-time action feature vector and the standard action feature vector. The calculation formula is as follows: Among them, S represents the action standard, x i and y i They represent the i-th element of the real-time action and standard action feature vectors respectively, i = 1, 2, ..., N, and N represents the total number of elements.

4. The method for labeling and processing live fitness exercise movements based on big data according to claim 1, characterized in that: Acquire ECG signals or heartbeat interval data in real time, extract the heartbeat interval time series, and calculate the heart rate variability based on the heartbeat interval data. The calculation formula is as follows: Among them, HRV represents heart rate variability, RR j represents the jth heartbeat interval, j = 1, 2, ..., M, and M represents the total number of heartbeat intervals.

5. The method for labeling and processing fitness exercise live broadcasts based on big data according to claim 1, characterized in that: Taking movement standard, movement intensity, heart rate variability and fatigue as risk characteristics, a risk assessment matrix A is established to evaluate the risk of current fitness movements. The risk assessment matrix A is as follows: Among them, a mn It is expressed as the importance of the m-th risk feature relative to the n-th risk feature, where m is the number of rows in matrix A and n is the number of columns in matrix A.

6. The method for labeling and processing live fitness exercise movements based on big data according to claim 5, characterized in that: Calculate the weight ω of each risk feature through the risk assessment matrix A m , the calculation formula is as follows: Where S is the number of risk characteristics, k is a constant, and 0 <k<S。 7. The method for labeling and processing live fitness exercise movements based on big data according to claim 6, characterized in that: According to the value of each risk feature and the corresponding weight, each risk feature is weighted and summed to calculate the action risk assessment value. The calculation formula is as follows: Among them, ARV represents the action risk assessment value, V m represents the value of the mth risk feature, ω m represents the weight of the mth risk feature, m = 1, 2, …, S, and S represents the number of risk features.

8. The method for labeling and processing live fitness exercise movements based on big data according to claim 7, characterized in that: Pre-set the risk threshold, compare the action risk assessment value with the risk threshold, and take corresponding measures based on the comparison results, including: When the action risk assessment value is less than or equal to the risk threshold, the user action execution status is continuously monitored; When the action risk assessment value is greater than the risk threshold, timely feedback will be given to the user, suggesting adjusting exercise intensity, resting or seeking professional guidance.

9. A big data-based fitness exercise live broadcast action annotation processing system, used to implement the method described in any one of claims 1 to 8, characterized in that: include: The action recognition and annotation module collects the user's action data during live fitness broadcasts in real time, identifies the user's fitness movements in real time, and accurately annotates the fitness movements in the live broadcast based on the recognition results; The action standardization evaluation module extracts key frames from the real-time video stream. The key frames correspond to the labeled fitness movements, detects the positions of key points on the human body, constructs a feature vector describing the movement based on the coordinates of the key points, and calculates the similarity between the real-time movement feature vector and the standard movement feature vector to obtain the movement standardization. The physiological fatigue monitoring module obtains ECG signals or heartbeat interval data, extracts the heartbeat interval time series, and calculates heart rate variability. The fatigue level is calculated based on the user's heart rate variability, exercise duration, and heart rate variability baseline value during fitness exercise. The calculation formula is as follows: Among them, RCH represents fatigue, HRV represents heart rate variability, represents the baseline value of heart rate variability, and T represents the duration of exercise; The risk assessment and response measures module takes movement standardization, movement intensity, heart rate variability, and fatigue as risk characteristics, establishes a risk assessment matrix, calculates the weight of each risk characteristic through the risk assessment matrix, combines the value of each risk characteristic, calculates the movement risk assessment value, compares the movement risk assessment value with the risk threshold, and takes corresponding measures based on the comparison results.

Images