Video running posture analysis method based on movable video equipment and running device

Abstract

The invention relates to the technical field of running data processing, and particularly discloses a video running posture analysis method based on mobile video equipment, which comprises the following steps: S1, collecting video frame data of a runner; s2, extracting joint point features, and identifying, filtering, forming and storing key frame data; s3, performing statistical analysis based on the key frame data to obtain running posture data; and S4, outputting the running posture data and guidance suggestions. Through comparison with a standard value, the running posture can be fed back to the runner in real time, the runner can dynamically adjust the posture or avoid safety risks, the risk of sports injuries is reduced, the runner can enjoy healthier and cheerful running experience, other functions such as sports prescriptions can be combined, the user experience is improved, and the user experience is improved. And the requirements of scientific sports or interesting sports of the user are met.

Description

Technical Field

[0001] The present invention relates to the technical field of running data processing, and in particular to a video running posture analysis method based on a mobile video device. Background Art

[0002] With the improvement of living standards, people pay more and more attention to body management, so fitness has become a way for most people to improve their physical fitness. However, how to exercise scientifically and effectively is often not easy for most people. Low-intensity fitness may not achieve the desired results, and excessive fitness may lead to counterproductive effects. Therefore, running, as a widely popular form of exercise, the correctness of its posture is not only related to the exercise effect, but also directly affects the physical health and sports safety of the participants.

[0003] China's public patent document (CN113076002A) discloses an interconnected fitness competition system and method based on multi-part motion recognition, including a four-in-one motion recognition unit, a smart phone client, a display device and an Internet server, which is realized through data collection, motion analysis, virtual competition, screen projection display, fitness data analysis and Internet communication, wherein the data collection is collected by the four-in-one motion recognition unit to collect the three-axis acceleration component data, three-axis angular velocity data and three-axis magnetic field component of the limb movement of the wearing part. The interconnected fitness competition system and method based on multi-part motion recognition can accurately identify the whole body motion posture of the human body and assist the fitness person to correctly complete the fitness movement. However, wearing smart wearable sports equipment is bound to increase the extra burden on the user during the running exercise, resulting in many inconveniences in the entire running exercise process and affecting the user's running experience. Summary of the invention

[0004] The purpose of the present invention is to provide a video running posture analysis method based on a mobile video device, so that runners can gradually adjust and optimize their running posture, improve running efficiency, and reduce the risk of sports injuries.

[0005] To achieve the above object, the technical solution adopted by the present invention is: The video running posture analysis method based on a mobile video device comprises the following steps: S1, collecting video frame data of the runner; S2, extract joint point features and identify, filter, form and store key frame data; S3, obtaining running posture data based on statistical analysis of key frame data; S4. Output running posture data and guidance.

[0006] Preferably, the step S1 is captured by a movable video device with a camera.

[0007] Preferably, the joint point feature extraction in step S2 adopts an open source neural network model.

[0008] Preferably, the filtering in step S2 uses statistical maximum values ​​to eliminate dirty data.

[0009] Preferably, the key frame data of step S2 includes a key frame sequence composed of real-time key frames within a specified time period, and step S3 instantly obtains the current running posture joint angle data based on the real-time key frames and obtains the running posture dynamic data within the specified time period based on the key frame sequence.

[0010] Preferably, the current running posture joint angle data uses the three joint points A, B, and C to calculate the angle θ data of the second point B, and the calculation formula is: .

[0011] Preferably, the running posture dynamic data includes left / right shoulder vertical amplitude, left / right hip vertical amplitude, ground contact method, ground contact time, air time, step frequency, step length, and hand and foot time.

[0012] Preferably, step S3 further includes scoring based on running posture data, specifically: setting a standard value S based on different postures, setting a score D, the default score is 100 points, the score deviation is ±V, and the score decreases by 10 points for every deviation value greater than 10, that is, the score decreases by one level, and the final score is P, and the formula is as follows: .

[0013] Preferably, step S4 includes displaying running posture data and instant suggestions in real time through a UI display screen during running, and generating and sending a running posture analysis report after the running is completed.

[0014] The present invention also provides a running device, including a treadmill host equipped with a UI display screen, the treadmill host communicates with a mobile video device, and the treadmill host stores instructions for executing some or all of the above steps S2-S4.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses a mobile video device to provide real-time video images, extracts key frame data through model data trained by a neural network, and forms joint point data without the need for wearable devices; then, through the joint point data, according to business needs, detailed runner posture data is calculated, and synchronously displayed on relevant devices, so that runners can examine their running posture from a third-party perspective, or get professional guidance, etc. In addition, by connecting with the background and incorporating it into the data system, after data analysis, more detailed posture problem points or exercise prescriptions can be given to runners, to assist runners in correcting their postures, improving their exercise experience, and improving their exercise health level.

[0016] By comparing with standard values, the present invention can provide instant feedback of running posture to the runner. The runner can dynamically adjust the posture or avoid safety risks, reduce the risk of sports injuries, and enjoy a healthier and more enjoyable running experience. Other functions such as exercise prescription can also be combined to improve user experience and meet the user's requirements for scientific or fun exercise. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of the video running posture analysis method.

[0018] Figure 2 This is a workflow diagram of video running posture analysis in Example 1 of the present invention.

[0019] Figure 3 This is a workflow diagram for video running posture analysis in Example 2 of the present invention.

[0020] Figure 4 This is a workflow diagram of video running posture analysis in Example 3 of the present invention.

[0021] Figure 5 This is a workflow diagram for video running posture analysis in Example 4 of the present invention.

[0022] Figure 6 This is a real picture taken during running of the present invention.

[0023] Figure 7 It is a numbered diagram of human joints of the present invention.

[0024] Figure 8 It is the joint point information diagram in the key frame of the present invention.

[0025] Fig. 9 for Figure 7 Selected angle diagram of lower body joints. DETAILED DESCRIPTION

[0026] In order to make the above features and advantages of the present invention more obvious and easy to understand, embodiments are given below with reference to the accompanying drawings for detailed description as follows.

[0027] Example 1 like Figure 1 As shown, the present invention provides a video running posture analysis method based on a mobile video device, comprising the following steps: S1, collecting video frame data of a runner; S2, extracting joint point features and identifying, filtering, forming and storing key frame data; S3, obtaining running posture data based on statistical analysis of the key frame data; S4, outputting the running posture data and guidance opinions.

[0028] Specifically in this embodiment, the above analysis method is performed by a mobile video device and a running device. The running device of this embodiment includes a treadmill host equipped with a UI display screen, and the mobile video device communicates with the treadmill host. The mobile video device and the treadmill host can communicate through any appropriate standard or proprietary hardware and connection protocol, such as: Universal Serial Bus, High Definition Multimedia Interface, Bluetooth, wireless WiFi, Near Field Communication, Global System for Mobile Communications, Enhanced Data GSM Environment and other wireless connection protocols.

[0029] In this embodiment, step S1 is completed by a mobile video device, and the collected video frame data is transmitted to the treadmill host, and then steps S2-S4 are completed by the treadmill host. Among them, step S4 includes real-time display of running posture data and instant suggestions through the UI display screen during running, and generating and sending a running posture analysis report after the running is completed. The entire workflow is as follows Figure 2 shown.

[0030] In step S1, a mobile video device with a camera dynamically collects and analyzes the video frame data of the runner at a specified position and direction. The mobile video device is a mobile phone or other mobile terminal device with an operating system, storage medium, processor and camera, or a camera device of the treadmill itself, which is set at a position of about 1.5 meters on both sides of the runner, and the full body image of the runner is captured as much as possible through the mobile video device, such as Figure 6 shown.

[0031] In step S2, the human joint information contained in the video frame data is extracted through the open source neural network model to form key frame data, which includes a key frame sequence composed of real-time key frames within a specified time period. The joint feature extraction uses the pose_landmark module source code in the open source neural network model mediapipe (based on neural network and deep learning) to compile and generate according to the requirements, so that it can quickly capture the human skeleton features and form key frame data, in which the human joints are numbered, such as Figure 7-8As shown. While forming the key frame data, various data are calculated, counted, and cached according to the coordinates of the joint point information on the key frame. The statistical maximum value is used to eliminate dirty data, such as the maximum and minimum values ​​of the left and right ankle X coordinates and the maximum distance between the two ankles. The purpose of the statistical maximum value is to use the filtering algorithm to eliminate unreasonable data and avoid the wrong frame data from being identified as correct data; at the same time, a key frame queue is formed for detection operations such as air touchdown in the running posture method.

[0032] In step S3, the current running posture joint angle data is instantly obtained based on the real-time key frame using an instant algorithm, such as the vertical inclination angle of the head, the elbow angle, the vertical inclination angle of the upper body, etc. The current running posture joint angle data uses three joint point lines to calculate the second point angle data. Assuming there are three joint points A, B, and C, the calculation formula for the angle θ about point B is: .

[0033] Regarding the real-time algorithm, the logic is to obtain intuitive data that is easy to obtain from the current key frame data, mainly most of the angle information. The user's running posture can also be inferred through the angle information of each joint, which is mainly used to give users real-time data feedback to obtain a good experience in sports.

[0034] In step S3, a buffer value algorithm is used based on the key frame sequence to obtain running posture dynamic data within a specified time period, including left / right shoulder vertical amplitude, left / right hip vertical amplitude, ground contact method, ground contact duration, flight duration, step frequency, step length, and hand and foot duration.

[0035] Regarding the buffer value algorithm, the logic is to cache a series of key frame data to calculate the data status under the precise user posture, or the data status that changes over time, such as: left and right shoulder vertical amplitude data, ground contact method data, etc.

[0036] Regarding the calculation of amplitude data, it is necessary to obtain the difference between the previous and next frames on the y-coordinate axis in the cache to obtain the amplitude data. It can also be compared based on the body data nodes to form comparative data for calculation. Fig. 9 For example, the difference between the last frame in the buffer (i.e., the current frame) and the previous frame based on the y coordinate of point 24 (right hip) can be calculated to obtain the vertical amplitude data of the right hip.

[0037] The method for determining the touchdown and takeoff mode is relatively complicated, which is explained here: according to step S2, the joint data of the key frame has been obtained through the neural network model, and the coordinate system established with the center position of the runner (the center position of the waist) is obtained. After the characteristics of the runner's picture are collected through the video, the runner's coordinate system relative to itself is obtained. After restoring the coordinate system with the runner's own center point as the origin, various actions can be analyzed through eigenvalues ​​(no longer affected by the moving state). There are extreme coordinate values ​​for the running actions of the left and right feet. For example, when the camera captures the picture of the runner facing the left, the runner runs forward and the left and right legs need to be stretched forward. At this time, the X coordinate values ​​of the joints of the left and right legs gradually decrease. The moment of touching the ground, the other foot needs to move forward with the foot touching the ground as the fulcrum; therefore, in the process of capturing the left running posture, the center of the runner's body is taken as the origin, and the moment of touching the ground is the minimum X value of each joint point of the foot. After touching the ground, the foot is used as a fulcrum to change the foot, and the X value will gradually increase. When it reaches the maximum value, it will be in the air and then move forward to complete the running action cycle; based on the above conclusions, the last frame in the buffer sequence is taken as the real-time frame, and it is determined whether the penultimate frame in the buffer sequence is the maximum value of the X axis of the key point of the foot, so as to determine the touching and flying state of the user's foot, and at the same time, the angle between the connecting line of the toe joint point and the heel joint point and the horizontal direction can be used to determine whether the heel or the toe touches the ground, such as Figure 8 , the foot that touches the ground is the right foot, and the angle between the line connecting the 30-point position and the 32-point position and the horizontal direction indicates that the heel touches the ground (this method is verified by comparing running videos, and restoring videos of data collected while running to specific action frames for manual comparison, and it is basically determined to be the logic for judging the touchdown and take-off).

[0038] It should be noted that the buffer value algorithm also needs to introduce the concept of distance between the two feet, and use the data cached in step S2 to determine whether the data is legal, so as to avoid dirty data pollution, which will lead to reduced data accuracy and poor user experience. For example, if the deviation between the horizontal coordinate X value in the real-time data collected and the dynamic statistical X value is too large, the data can be discarded. The step S3 also includes scoring based on the running posture data, specifically: setting a standard value S based on different postures, setting a score D, the default score is 100 points, the score deviation is ±V, and the score decreases by 10 points when the deviation value is greater than 10, that is, the score decreases by one level, and the final score is P, and the formula is as follows: .

[0039] Step S4 includes displaying running posture data and instant suggestions in real time through the UI display screen during running, and generating and sending a running posture analysis report after the run is completed, giving the runner corresponding feedback or suggestions.

[0040] Example 2 The difference between this embodiment and the above-mentioned embodiment 1 is that the executors of steps S2-S4 are different. Specifically, in this embodiment, steps S1-S2 are completed by a mobile video device. The collected video frame data is processed by the mobile video device equipped with a running posture mode to obtain key frame data, which is then transmitted to the treadmill host, and then steps S3-S4 are completed by the treadmill host. The entire workflow is as follows: Figure 3 shown.

[0041] Example 3 The difference between this embodiment and the above-mentioned embodiment 1 is that the executors of steps S2-S4 are different. Specifically, in this embodiment, steps S1-S3 are completed by a mobile video device. The collected video frame data is processed by the mobile video device equipped with a running posture mode to obtain key frame data and statistically analyze the running posture data, which is then transmitted to the treadmill host, and then step S4 is completed by the treadmill host. The entire workflow is as follows: Figure 4 shown.

[0042] Example 4 The difference between this embodiment and the above-mentioned embodiment 1 is that the executors of steps S2-S4 are different. Specifically, in this embodiment, steps S1-S3 are completed by a mobile video device. After the mobile video device obtains the running posture data based on the statistical analysis of the key frame data, the real-time running posture data and instant suggestion communication are transmitted to the treadmill host during the running and displayed to the runner through the UI display screen. After the running, the running posture analysis report is displayed to the runner through the mobile video device and the treadmill host. The whole workflow is as follows. Figure 5 shown.

[0043] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for illustrating the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which shall fall within the scope of the present invention to be protected. The scope of protection of the present invention shall be defined by the attached claims and their equivalents.

Claims

1. A video running posture analysis method based on a mobile video device, characterized in that: The following steps are involved: S1, collecting video frame data of the runner; S2, extract joint point features and identify, filter, form and store key frame data; S3, obtaining running posture data based on statistical analysis of key frame data; S4. Output running posture data and guidance.

2. The video running posture analysis method based on a mobile video device according to claim 1 is characterized in that: The step S1 is to collect data through a mobile video device with a camera.

3. The video running posture analysis method based on a mobile video device according to claim 1, characterized in that: The joint point feature extraction in step S2 adopts an open source neural network model.

4. The video running posture analysis method based on a mobile video device according to claim 1 is characterized in that: The filtering in step S2 uses statistical maximum values ​​to eliminate dirty data.

5. The video running posture analysis method based on a mobile video device according to claim 1 is characterized in that: The key frame data of step S2 includes a key frame sequence composed of real-time key frames within a specified time period, and step S3 instantly obtains current running posture joint angle data based on the real-time key frames and obtains running posture dynamic data within the specified time period based on the key frame sequence.

6. The video running posture analysis method based on a mobile video device according to claim 5 is characterized in that: The current running posture joint angle data uses the three joint points A, B, and C to calculate the angle θ data of the second point B, and the calculation formula is: 。 7. The video running posture analysis method based on a mobile video device according to claim 1 is characterized in that: The running posture dynamic data includes left / right shoulder vertical amplitude, left / right hip vertical amplitude, ground contact method, ground contact time, air time, step frequency, step length, and hand and foot step time.

8. The video running posture analysis method based on a mobile video device according to claim 1 is characterized in that: The step S3 also includes scoring based on the running posture data, specifically: setting a standard value S based on different postures, setting a score D, the default score is 100 points, the score deviation is ±V, and the score decreases by 10 points when the deviation value is greater than 10, that is, the score decreases by one level, and the final score is P, and the formula is as follows: 。 9. The video running posture analysis method based on a mobile video device according to claim 1, characterized in that: The step S4 includes displaying running posture data and instant suggestions in real time through the UI display screen during running, and generating and sending a running posture analysis report after the running is completed.

10. A running device, characterized in that: It includes a treadmill host equipped with a UI display screen, the treadmill host communicates with a mobile video device, and the treadmill host stores instructions for executing some or all of steps S2-S4 in the video running posture analysis method based on a mobile video device as described in claim 1.

Images