Eight-segment brocade virtual training method and system based on 8K set top box

By combining an 8K set-top box and the MoveNet algorithm with an external USB camera, high-definition video data capture and motion evaluation were achieved, solving the problems of inaccurate motion capture and insufficient personalized evaluation in traditional technologies, and improving the accuracy and interactivity of Baduanjin training.

CN118968567AInactive Publication Date: 2024-11-15SICHUAN TIANYI COMHEART TELECOM

Patent Information

Application Number
CN202411452201.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-17
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In existing sports training and rehabilitation technologies, traditional video capture equipment has insufficient resolution, making it difficult to accurately capture subtle movements, and lacks personalized movement assessment and feedback mechanisms, resulting in poor training effects.

Method used

It uses an 8K set-top box combined with an external USB camera and MoveNet algorithm to capture high-definition video data, extract key points of the human body, and evaluate the accuracy of the action by comparing feature vectors, providing real-time feedback.

Benefits of technology

It improves motion capture accuracy and personalized assessment, enhancing the interactivity and effectiveness of training, and is especially suitable for diverse exercise programs such as Baduanjin.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968567A_ABST
    Figure CN118968567A_ABST
Patent Text Reader

Abstract

The invention discloses an eight-segment brocade virtual training method and system based on an 8K set top box, relates to the technical field of posture assistance, and aims to solve the technical problem of low accuracy when a training process is guided in the prior art. The eight-segment brocade virtual training method based on the 8K set top box is used for an 8K set top box training system. The method comprises the following steps: acquiring an eight-segment brocade standard action data set; extracting a human body target standard key point in each standard action in the eight-segment brocade standard action data set based on a MoveNet algorithm; acquiring human body target actual key points of the user; and performing feature vector comparison on the human body target actual key point of the user and the corresponding human body target standard key point to obtain an action accuracy value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of posture assistance technology, and in particular to an Eight-Section Brocade virtual training method and system based on an 8K set-top box. Background Art

[0002] In existing sports training and rehabilitation technologies, it is common to use video capture and computer vision technology to assist sports training. However, these technologies have some limitations and challenges, mainly including: traditional video capture devices are often limited by resolution and are not accurate enough to capture subtle movements. This is particularly problematic when highly accurate movement analysis and training feedback are required; and traditional methods often lack effective standardized means when dealing with individual differences in body shape, height, and movement habits, resulting in insufficient personalization and accuracy in movement evaluation, and a lack of effective movement comparison and evaluation mechanisms, making it difficult for trainers to obtain accurate guidance on movement improvement. Summary of the invention

[0003] The present application provides a virtual training method and system for Ba Duan Jin based on an 8K set-top box, which at least solves the technical problem of low accuracy in guiding the training process in the prior art.

[0004] In order to solve the above technical problems, the embodiment of the present application provides: a virtual training method of Ba Duan Jin based on an 8K set-top box, which is used in an 8K set-top box training system; the method comprises the following steps: Obtain the Ba Duan Jin standard action dataset; Extracting standard key points of human targets in each standard action in the Ba Duan Jin standard action dataset based on the MoveNet algorithm; Obtain the actual key points of the user's human body target; The actual key points of the human body target of the user are compared with the corresponding standard key points of the human body target by feature vector to obtain the action accuracy value.

[0005] As some optional embodiments of the present application, the standard action data set of Ba Duan Jin includes a preparatory standard action data set, a standard action data set of raising both hands straight up, a standard action data set of raising both hands horizontally, a standard action data set of wrist-crossing, a standard action data set of left-drawn bow, a standard action data set of right-drawn bow, a standard action data set of left-hand lifting, and a standard action data set of right-hand lifting.

[0006] As some optional implementation modes of the present application, the human body target standard key points include nose target standard key points, left eye target standard key points, right eye target standard key points, left ear target standard key points, right ear target standard key points, left shoulder target standard key points, right shoulder target standard key points, left elbow target standard key points, right elbow target standard key points, left wrist target standard key points, right wrist target standard key points, left hip target standard key points, right hip target standard key points, left knee target standard key points, right knee target standard key points, left ankle target standard key points and right ankle target standard key points; The human body target actual key points include the nose target actual key point, the left eye target actual key point, the right eye target actual key point, the left ear target actual key point, the right ear target actual key point, the left shoulder target actual key point, the right shoulder target actual key point, the left elbow target actual key point, the right elbow target actual key point, the left wrist target actual key point, the right wrist target actual key point, the left hip target actual key point, the right hip target actual key point, the left knee target actual key point, the right knee target actual key point, the left ankle target actual key point and the right ankle target actual key point.

[0007] As some optional implementations of the present application, the step of obtaining the actual key points of the user's human body target includes: Send a startup command to the external USB camera module; Receive video stream data returned by the external USB camera module; Based on the video stream data, determining whether a target user is detected; If the target user has been detected, a start command is sent to the gesture recognition module; Receive the actual key points of the user's human body target returned by the gesture recognition module.

[0008] As some optional implementation manners of the present application, the step of determining whether a target user is detected based on the video stream data further includes: If the target user is not detected, the latest video stream data returned by the external USB camera module is received again; Based on the latest video stream data, it is determined whether a target user is detected.

[0009] As some optional implementations of the present application, the step of sending a start instruction to the gesture recognition module if the target user has been detected includes: If the target user has been detected, a start command is sent to the gesture recognition module; Determining whether the gesture recognition module successfully recognizes the gesture of the target user; If the recognition is successful, receiving the actual key points of the user's human body target returned by the posture recognition module; If the recognition is unsuccessful, the latest video stream data returned by the external USB camera module is received again; Based on the latest video stream data, determining whether a target user is detected; If the target user has been detected, a start command is sent to the gesture recognition module; Receive the actual key points of the user's human body target returned by the gesture recognition module.

[0010] As some optional implementations of the present application, the step of performing feature vector comparison between the actual key points of the human body target of the user and the corresponding standard key points of the human body target to obtain the action accuracy value includes: After centering the actual key points of the user's human body target and the corresponding standard key points of the human body target, scaling them to the same scale, and then rotating them so that the actual key points of the user's human body target are aligned with the corresponding standard key points of the human body target; The aligned actual key points of the human body target of the user are compared with the corresponding standard key points of the human body target by feature vector to obtain the action accuracy value.

[0011] As some optional implementations of the present application, the step of comparing the aligned actual key points of the human body target of the user with the corresponding standard key points of the human body target by performing feature vector comparison to obtain the action accuracy value includes: Based on the cosine similarity algorithm, feature vectors of the aligned actual key points of the user's human body target are compared with the corresponding standard key points of the human body target to obtain the action accuracy value.

[0012] On the other hand, the embodiment of the present application provides: an 8K set-top box training system, comprising: Obtain standard data module, used to obtain the Ba Duan Jin standard action data set; A standard key point extraction module is used to extract the standard key points of human targets in each standard action in the Ba Duan Jin standard action data set based on the MoveNet algorithm; The actual key point extraction module is used to obtain the actual key points of the user's human body target; The key point comparison module is used to compare the actual key points of the human body target of the user with the corresponding standard key points of the human body target by performing feature vector comparison to obtain the action accuracy value.

[0013] As some optional implementations of the present application, the actual key point extraction module includes: The first control module is used to send a start instruction to the external USB camera module; receive the video stream data returned by the external USB camera module; A determination module, used to determine whether a target user is detected based on the video stream data; The second control module is used to send a start instruction to the gesture recognition module if the target user has been detected; and receive the actual key points of the user's human body target returned by the gesture recognition module.

[0014] Compared with the prior art, the progress of the virtual training method of Ba Duan Jin based on 8K set-top box provided by this application is reflected in the following aspects: 1) By using the external USB camera of the 8K set-top box, this technical solution can capture higher-definition video data, greatly improving the accuracy of motion capture. In capturing subtle movements, this high-resolution camera can record more details, making the motion analysis more accurate, especially in sports training and rehabilitation scenarios, where the requirements for motion accuracy are extremely high. 2) As an efficient and lightweight posture estimation tool, the MoveNet algorithm can quickly and accurately identify the standard key points of human targets. The application of this algorithm not only improves the speed and accuracy of key point detection, but also, due to its lightweight characteristics, is very suitable for deployment on resource-constrained 8K set-top boxes, ensuring the real-time and response speed of the system. 3) By comparing the feature vectors of the user's actual key points with the standard key points, the system can calculate a motion accuracy value, which not only provides a quantitative evaluation of the training effect, but also provides instant feedback to the user to help the user understand which specific parts of the action need to be improved; this real-time feedback mechanism greatly enhances the interactivity and effect of training. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the specific implementation of the present application or the technical solution in the prior art, the following is a brief introduction to the drawings required for the specific implementation or the prior art description. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0016] Figure 1 A flowchart of the virtual training method of Ba Duan Jin based on an 8K set-top box involved in this application; Figure 2 A schematic diagram of the 17 core key points of the human body involved in this application; Figure 3 A schematic diagram of the process of obtaining the actual key points of the user's human body target involved in this application; Figure 4 This is a schematic diagram of the structure of the 8K set-top box training system involved in this application; In the figure, 0 represents nose, 1 represents left eye, 2 represents right eye, 3 represents left ear, 4 represents right ear, 5 represents left shoulder, 6 represents right shoulder, 7 represents left elbow, 8 represents right elbow, 9 represents left wrist, 10 represents right wrist, 11 represents left hip, 12 represents right hip, 13 represents left knee, 14 represents right knee, 15 represents left ankle, and 16 represents right ankle.

[0017] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In today's smart home ecosystem, high-definition set-top boxes and ultra-large-screen TVs are core terminal devices, and their main responsibilities focus on video entertainment services, with relatively simple functions. However, as mobile devices such as smartphones and tablets gradually occupy more of users' viewing time, the regular frequency of TV usage faces challenges. Despite this, TV still occupies a pivotal position in smart home scenarios with its unparalleled visual impact. Especially with the advent of 8K set-top boxes, equipped with high-performance CPUs and GPUs, unprecedented computing power has been injected into the system, which has opened up new paths for exploring the diversified functions of TV terminals and prompted manufacturers to think deeply about how to maximize the use of these hardware advantages.

[0020] In this context, the booming development of artificial intelligence (AI) technology, especially large-scale model technology, is leading a cross-industry innovation storm and profoundly changing the face of traditional fields. This application combines AI technology with the characteristics of 8K set-top boxes and large-screen TVs. Among them, human posture recognition technology, as a high-end application of deep integration of computer vision, image processing and deep learning, aims to accurately capture and track human movements and postures in videos, analyze individual behavior patterns, motor skills and even emotions and health status, and provide a new interface for the integration of technology and human life.

[0021] Integrating human posture recognition technology into the 8K set-top box, by connecting an external USB camera to capture dynamic images, can not only achieve high-precision recognition and analysis of human movements, with an accuracy of tracking up to 17 key body points, but also create a new paradigm for interactive family fitness. For example, as a virtual personal trainer, it can guide users in real time in the living room environment to perform dance, yoga, Ba Duan Jin and other diverse exercise programs, and provide instant feedback and correction based on precise movement comparison, thereby improving the efficiency and fun of family fitness, and also opening a new chapter of intelligent health management.

[0022] like Figure 1 As shown, a virtual training method of Ba Duan Jin based on an 8K set-top box is used in an 8K set-top box training system; the method comprises the following steps: Step S10, obtaining a data set of standard movements of Ba Duan Jin.

[0023] It should be noted that the Ba Duan Jin standard movement data set is a Ba Duan Jin movement data set composed of standardized practice videos of professionals.

[0024] Step S20, extracting the standard key points of the human body target in each standard action in the Ba Duan Jin standard action data set based on the MoveNet algorithm.

[0025] It should be noted that the human body target standard key points include the nose target standard key point, the left eye target standard key point, the right eye target standard key point, the left ear target standard key point, the right ear target standard key point, the left shoulder target standard key point, the right shoulder target standard key point, the left elbow target standard key point, the right elbow target standard key point, the left wrist target standard key point, the right wrist target standard key point, the left hip target standard key point, the right hip target standard key point, the left knee target standard key point, the right knee target standard key point, the left ankle target standard key point and the right ankle target standard key point.

[0026] Step S30: Acquire the actual key points of the user's human body target.

[0027] It should be noted that the human body target actual key points include the nose target actual key point, the left eye target actual key point, the right eye target actual key point, the left ear target actual key point, the right ear target actual key point, the left shoulder target actual key point, the right shoulder target actual key point, the left elbow target actual key point, the right elbow target actual key point, the left wrist target actual key point, the right wrist target actual key point, the left hip target actual key point, the right hip target actual key point, the left knee target actual key point, the right knee target actual key point, the left ankle target actual key point and the right ankle target actual key point.

[0028] The human body target standard key points and the user's human body target actual key points are extracted based on the MoveNet algorithm. The MoveNet algorithm can extract 17 core key points of the human body, such as Figure 2 As shown, Figure 2 Middle: 0 represents nose, 1 represents left eye, 2 represents right eye, 3 represents left ear, 4 represents right ear, 5 represents left shoulder, 6 represents right shoulder, 7 represents left elbow, 8 represents right elbow, 9 represents left wrist, 10 represents right wrist, 11 represents left hip, 12 represents right hip, 13 represents left knee, 14 represents right knee, 15 represents left ankle, and 16 represents right ankle.

[0029] The acquired video data is processed using the MoveNet algorithm to output the coordinates of the key points of the human body in each frame. The following steps are detailed below to describe the application of the MoveNet algorithm and the process of extracting key point data of the human body: Before processing the video, first load the MoveNet model. Initialize the model parameters, set the appropriate computing resources (such as CPU or GPU). Make sure the model has been warmed up, that is, run with a small amount of data to achieve the performance of real-time video processing.

[0030] Perform necessary preprocessing on the input video frames. This includes resizing the frames, normalizing pixel values, and possible color space conversion (such as from RGB to grayscale). These preprocessing steps ensure that the video frames are consistent with the data format used during training, thereby optimizing the output accuracy of the model.

[0031] Apply the MoveNet model to each frame of the video one by one. This involves passing the pre-processed frames as input to the model, which runs its internal neural network architecture to detect key points. Ensure real-time processing frame rate to ensure the continuity of video streaming and real-time user interaction.

[0032] For each frame, MoveNet outputs the coordinates of 17 key points. These coordinates are usually (x, y) pairs in pixels and may include confidence scores, indicating how confident the model is in the detection results. The confidence scores can be used to filter out key points with low detection quality, retaining only key point data with high confidence for subsequent analysis.

[0033] Post-processing of keypoint coordinates, such as smoothing (through time series analysis methods such as sliding average or Kalman filtering) to reduce jitter and noise. If necessary, convert pixel coordinates to actual physical size (such as based on known size references in the scene), which helps achieve keypoint data consistency across different devices and environments.

[0034] The processed key point data is recorded for long-term analysis or immediate feedback. Depending on the application requirements, the data may be stored on the local device or uploaded to a cloud server to support more complex data analysis and remote training support.

[0035] Through these detailed steps, the user's movements can be captured and analyzed more accurately, providing a scientific basis for training and rehabilitation. This method not only improves the accuracy of training, but also helps coaches and physical therapists better understand and guide the user's movement execution.

[0036] It can be seen that this application uses the advanced algorithm of the MoveNet model. This application can accurately locate the 17 core key points of the human body, laying the technical foundation for personalized feedback and correction of Ba Duan Jin exercise. This process is not just about capturing data, but about integrating high technology into traditional health-preserving methods, transforming it into an exclusive virtual coach for each practitioner, and guiding them towards a path of exercise with precise movements and standard postures. Let everyone enjoy high-quality guidance at home as if they were having a private coach by their side, making the learning journey of Ba Duan Jin both efficient and fun, and truly realizing the harmonious coexistence of technology and health.

[0037] Specifically, the step of obtaining the actual key points of the user's human body target in step S30 includes: Step S31, sending a start command to the external USB camera module; Step S32, receiving the video stream data returned by the external USB camera module; Step S33: judging whether a target user is detected based on the video stream data; Step S34: if the target user has been detected, a start instruction is sent to the gesture recognition module; it is determined whether the gesture recognition module successfully recognizes the gesture of the target user; if the recognition is successful, the actual key points of the human body target of the user returned by the gesture recognition module are received; if the recognition is unsuccessful, the latest video stream data returned by the external USB camera module is received again; based on the latest video stream data, it is determined whether the target user is detected; if the target user has been detected, a start instruction is sent to the gesture recognition module; the actual key points of the human body target of the user returned by the gesture recognition module are received; if the target user is not detected, the latest video stream data returned by the external USB camera module is received again; based on the latest video stream data, it is determined whether the target user is detected; Step S35: receiving the actual key points of the user's human body target returned by the posture recognition module.

[0038] That is Figure 3 As shown, the steps to obtain the actual key points of the user's human body target are as follows: After the model is deployed, the inference engine is initialized and the external USB camera is started; Get the camera video stream and determine whether a human body is detected; If NO, the camera video stream is re-acquired and the subsequent steps are performed; if YES, the inference engine performs gesture recognition; Determine whether the gesture recognition is successful; If NO, reacquire the camera video stream and proceed to the subsequent steps; if YES, the reasoning ends and 17 key points of the human body are obtained.

[0039] Specifically, the steps include: Deploy MoveNet or other posture recognition models on the device, ensuring that the model files are complete and best suited for the current hardware configuration; configure the inference engine to support fast data processing, including setting the appropriate number of threads, memory allocation, and GPU acceleration (if available); connect an external camera through the USB interface, initialize the camera settings, and ensure that the video stream can be smoothly transmitted to the processing unit.

[0040] Receive the video stream from the USB camera in real time and decode the video frames; run the human detection algorithm in each frame to determine whether the current frame contains a human body; if no human body is detected, the system will continue to obtain new video streams from the camera; if a human body is detected, proceed to the next step for posture recognition. More specifically, the USB camera continuously transmits real-time video data to the system to ensure that the data transmission between the camera and the computing unit is delay-free and high-bandwidth to support high-resolution and high-frame-rate video streams; decode the received video stream in real time and convert it into single-frame image data for subsequent processing; use efficient decoding technology (such as H.264 or VP9) to ensure that the decoding process does not become a performance bottleneck. Run a human detection algorithm, such as YOLO or SSD, in each frame of the image to identify whether there is a human body in the image.

[0041] Use gesture recognition models such as MoveNet to analyze the detected human body and extract key point information; check the results of gesture recognition to confirm whether all key points are correctly identified and located; if gesture recognition is unsuccessful or key point information is incomplete, the system will re-acquire the video stream from the camera and try recognition again; if recognition is successful, extract key point data. More specifically, use models such as MoveNet to perform gesture recognition on detected human images, and the model should be pre-trained and optimized to accurately identify human key points. Extract the coordinates and possible confidence of 17 key points from the gesture recognition results; including scaling and translation correction of key point coordinates to ensure that the coordinates are consistent on different devices and resolutions. Verify the gesture recognition results to confirm whether all key points are correctly identified and accurately located. The judgment logic is: if gesture recognition is unsuccessful or key point information is incomplete: re-acquire the camera video stream and perform gesture recognition again. If recognition is successful: go to the next step to process and output key point data. Format the successfully identified key point data (such as JSON or XML) for subsequent application use, and ensure that the key point data can be called by other applications or modules through API or direct memory access. The system adjusts parameters or algorithm configuration according to the key point recognition results to optimize recognition accuracy and processing speed; based on the analysis of past recognition data, the system can self-adjust to adapt to different environmental variables, such as lighting, movement speed, etc.

[0042] The coordinates of 17 main key points are extracted from the recognition results; the key point data are processed, such as denoising, smoothing and standardization, to ensure the quality and consistency of the data.

[0043] The processed key point data is presented to the user or passed to the application for further analysis; the key point data can be optionally saved to a local database or cloud server for historical data analysis or remote monitoring.

[0044] Through these detailed steps, your system can efficiently and accurately acquire and analyze human posture from video streams, providing reliable data support for subsequent applications. This process is particularly critical to ensure the stability and reliability of the action recognition system in various practical environments.

[0045] Step S40: perform feature vector comparison on the actual key points of the human body target of the user and the corresponding standard key points of the human body target to obtain a motion accuracy value.

[0046] This step mainly involves comparing the user's human body key points with predefined standard human body key points to evaluate the accuracy of the action. In some embodiments, the following steps may be included: Obtained from the posture recognition module in the previous step, including the coordinates and confidence of the key points; the key point data is formatted and normalized as necessary to adapt to the requirements of the comparison algorithm.

[0047] Get standard key point data stored in a database or file, which can be called according to different actions or action categories.

[0048] Convert the actual key point coordinates into a vector, with the coordinates (x, y) of each key point as an element of the vector. In the same way, convert the standard key points into vector form.

[0049] Use vector similarity measurement methods, such as Euclidean distance, cosine similarity, or other suitable distance measurement methods. Among them, Euclidean distance refers to calculating the straight-line distance between two feature vectors, which is suitable for quantifying the precise deviation of key point positions. Cosine similarity refers to evaluating the similarity of two vectors in direction, which is suitable for comparing the directionality of actions.

[0050] Based on the result of the feature vector comparison, a scalar value is calculated to represent the accuracy of the action; the scalar value can be a direct distance value, or the accuracy represented by converting the distance into a percentage. The camera's capture strategy or algorithm parameters are adjusted based on the accuracy value to optimize the accuracy of key point capture. The accuracy result is displayed to the user, which can be a numerical display or a visual graphic, such as a progress bar or a heat map.

[0051] Through this series of detailed steps, the system can not only effectively compare and evaluate the accuracy of actions, but also optimize and adjust itself based on the results, ultimately improving user experience and system performance.

[0052] In some other embodiments, the step of performing feature vector comparison between the actual key points of the human body target of the user and the corresponding standard key points of the human body target to obtain the action accuracy value includes: Step S41, centering the actual key points of the human body target of the user and the corresponding standard key points of the human body target, scaling them to the same scale, and rotating them so that the actual key points of the human body target of the user are aligned with the corresponding standard key points of the human body target; Step S42, compare the actual key points of the human body target of the user after alignment with the corresponding standard key points of the human body target to obtain a motion accuracy value. Specifically, this step may include: based on a cosine similarity algorithm, compare the actual key points of the human body target of the user after alignment with the corresponding standard key points of the human body target to obtain a motion accuracy value.

[0053] Specifically, step S41 may include the following steps: Calculate the mean of all key points and then subtract this mean from each key point coordinate; thereby moving the actual key point and the standard key point coordinates to a common center position, usually the center of the body or the center of gravity.

[0054] All keypoint coordinates are scaled equally using a scaling factor, usually based on a key distance (such as shoulder width or height); this ensures that the keypoints are consistently proportioned to eliminate size differences between individuals.

[0055] The optimal rotation angle between the actual key point set and the standard key point set is calculated, usually using principal component analysis (PCA) or least squares method; the actual key points are rotated to make their directions consistent with those of the standard key points.

[0056] Specifically, step S42 may include the following steps: For the aligned key point set, the coordinates of each point are converted into a long vector to form the actual feature vector and the standard feature vector. The cosine similarity is used to compare the directional similarity of the two vectors.

[0057] Convert the cosine similarity to an action accuracy value. Generally, the closer the cosine similarity is to 1, the closer the two vectors are, and the higher the accuracy of the action. You can use the cosine similarity directly as the accuracy value, or further adjust this value through some function (such as linear or nonlinear mapping) to adapt to a specific scoring system.

[0058] It can be seen that the purpose of step S41 is to ensure the spatial alignment of the actual key points with the standard key points through centering, scaling and rotation processing, while step S42 uses the cosine similarity algorithm to accurately compare the feature vectors of the two sets of key points to evaluate the accuracy of the action. This method makes the action analysis more consistent and comparable across different individuals and environments.

[0059] It can be seen that the progress of the virtual training method of Ba Duan Jin based on 8K set-top box provided by this application is reflected in the following aspects: 1) By using the external USB camera of the 8K set-top box, this technical solution can capture higher-definition video data, greatly improving the accuracy of motion capture. In capturing subtle movements, this high-resolution camera can record more details, making the motion analysis more accurate, especially in sports training and rehabilitation scenarios, where the requirements for motion accuracy are extremely high. 2) As an efficient and lightweight posture estimation tool, the MoveNet algorithm can quickly and accurately identify the standard key points of human targets. The application of this algorithm not only improves the speed and accuracy of key point detection, but also, due to its lightweight characteristics, is very suitable for deployment on resource-constrained 8K set-top boxes, ensuring the real-time and response speed of the system. 3) By comparing the feature vectors of the user's actual key points with the standard key points, the system can calculate a motion accuracy value, which not only provides a quantitative evaluation of the training effect, but also provides instant feedback to the user to help the user understand which specific parts of the action need to be improved; this real-time feedback mechanism greatly enhances the interactivity and effect of training.

[0060] On the other hand, Figure 4 As shown, the embodiment of the present application provides an 8K set-top box training system, including: Obtain standard data module, used to obtain the Ba Duan Jin standard action data set; A standard key point extraction module is used to extract the standard key points of human targets in each standard action in the Ba Duan Jin standard action data set based on the MoveNet algorithm; The actual key point extraction module is used to obtain the actual key points of the user's human body target; The key point comparison module is used to compare the actual key points of the human body target of the user with the corresponding standard key points of the human body target by performing feature vector comparison to obtain the action accuracy value.

[0061] Wherein, the actual key point extraction module includes: The first control module is used to send a start instruction to the external USB camera module; receive the video stream data returned by the external USB camera module; A determination module, used to determine whether a target user is detected based on the video stream data; The second control module is used to send a start instruction to the gesture recognition module if the target user has been detected; and receive the actual key points of the user's human body target returned by the gesture recognition module.

[0062] It should be noted that all directional indications in the embodiments of the present application (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0063] In this application, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" 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, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0064] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "and / or" appearing in the full text includes three parallel schemes. Taking "A and / or B" as an example, it includes scheme A, or scheme B, or a scheme that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0065] The embodiment of the present application also provides a computer device, which may include: a processor, such as a central processing unit (CPU), a communication bus, a user interface, a network interface, and a memory. Among them, the communication bus is used to realize the connection and communication between these components. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface may also include a standard wired interface and a wireless interface. The network interface may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory may be a high-speed random access memory (Random Access Memory, RAM) memory, or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory may also be a storage device independent of the aforementioned processor.

[0066] In the computer device, the network interface is mainly used for data communication with the network server; the user interface is mainly used for data interaction with the user; the processor and memory in the computer device of this embodiment can be set in the computer device, and the computer device calls the 8K set-top box training system stored in the memory through the processor, and executes the 8K set-top box training method provided in this embodiment.

[0067] It should be noted that the above-mentioned computer device can be an external hardware device that can run independently, or it can be a hardware device that comes with the 8K set-top box training system itself.

[0068] Based on the same inventive concept as the aforementioned embodiment, this embodiment provides a computer-readable storage medium, on which a computer program is stored, and a processor executes the computer program to implement the aforementioned method.

[0069] In some embodiments, the computer readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. The computer may be various computing devices including intelligent terminals and servers.

[0070] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A virtual training method of Ba Duan Jin based on 8K set-top box, characterized in that: Used for 8K set-top box training system; the method comprises the following steps: Obtain the Ba Duan Jin standard action dataset; Extracting standard key points of human targets in each standard action in the Ba Duan Jin standard action dataset based on the MoveNet algorithm; Obtain the actual key points of the user's human body target; The actual key points of the human body target of the user are compared with the corresponding standard key points of the human body target by feature vector to obtain the action accuracy value.

2. The virtual training method of Ba Duan Jin based on 8K set-top box according to claim 1 is characterized in that: The Ba Duan Jin standard action data set includes a preparatory standard action data set, a standard action data set for raising both hands straight up, a standard action data set for raising both hands horizontally, a standard action data set for putting wrists together, a standard action data set for drawing a bow with the left hand, a standard action data set for drawing a bow with the right hand, a standard action data set for lifting the left hand, and a standard action data set for lifting the right hand.

3. According to the Ba Duan Jin virtual training method based on 8K set-top box according to claim 1, it is characterized in that: The human body target standard key points include nose target standard key points, left eye target standard key points, right eye target standard key points, left ear target standard key points, right ear target standard key points, left shoulder target standard key points, right shoulder target standard key points, left elbow target standard key points, right elbow target standard key points, left wrist target standard key points, right wrist target standard key points, left hip target standard key points, right hip target standard key points, left knee target standard key points, right knee target standard key points, left ankle target standard key points and right ankle target standard key points; The human body target actual key points include the nose target actual key point, the left eye target actual key point, the right eye target actual key point, the left ear target actual key point, the right ear target actual key point, the left shoulder target actual key point, the right shoulder target actual key point, the left elbow target actual key point, the right elbow target actual key point, the left wrist target actual key point, the right wrist target actual key point, the left hip target actual key point, the right hip target actual key point, the left knee target actual key point, the right knee target actual key point, the left ankle target actual key point and the right ankle target actual key point.

4. According to claim 1, the virtual training method of Ba Duan Jin based on 8K set-top box is characterized in that: The step of obtaining the actual key points of the user's human body target includes: Send a startup command to the external USB camera module; Receive video stream data returned by the external USB camera module; Based on the video stream data, determining whether a target user is detected; If the target user has been detected, a start command is sent to the gesture recognition module; Receive the actual key points of the user's human body target returned by the gesture recognition module.

5. According to claim 4, the virtual training method of Ba Duan Jin based on 8K set-top box is characterized in that: The step of determining whether a target user is detected based on the video stream data further includes: If the target user is not detected, the latest video stream data returned by the external USB camera module is received again; Based on the latest video stream data, it is determined whether a target user is detected.

6. The virtual training method of Ba Duan Jin based on 8K set-top box according to claim 4 is characterized in that: The step of sending a start instruction to the gesture recognition module if the target user has been detected includes: If the target user has been detected, a start command is sent to the gesture recognition module; Determining whether the gesture recognition module successfully recognizes the gesture of the target user; If the recognition is successful, receiving the actual key points of the user's human body target returned by the posture recognition module; If the recognition is unsuccessful, the latest video stream data returned by the external USB camera module is received again; Based on the latest video stream data, determining whether a target user is detected; If the target user has been detected, a start command is sent to the gesture recognition module; Receive the actual key points of the user's human body target returned by the gesture recognition module.

7. The virtual training method of Ba Duan Jin based on 8K set-top box according to claim 1, characterized in that: The step of performing feature vector comparison between the actual key points of the human body target of the user and the corresponding standard key points of the human body target to obtain the action accuracy value comprises: After centering the actual key points of the user's human body target and the corresponding standard key points of the human body target, scaling them to the same scale, and then rotating them so that the actual key points of the user's human body target are aligned with the corresponding standard key points of the human body target; The aligned actual key points of the human body target of the user are compared with the corresponding standard key points of the human body target by feature vector to obtain the action accuracy value.

8. The virtual training method of Ba Duan Jin based on 8K set-top box according to claim 7 is characterized in that: The step of comparing the aligned actual key points of the human body target of the user with the corresponding standard key points of the human body target by performing feature vector comparison to obtain the action accuracy value comprises: Based on the cosine similarity algorithm, feature vectors of the aligned actual key points of the user's human body target are compared with the corresponding standard key points of the human body target to obtain the action accuracy value.

9. An 8K set-top box training system, characterized in that: include: Obtain standard data module, used to obtain the Ba Duan Jin standard action data set; A standard key point extraction module is used to extract the standard key points of human targets in each standard action in the Ba Duan Jin standard action data set based on the MoveNet algorithm; The actual key point extraction module is used to obtain the actual key points of the user's human body target; The key point comparison module is used to compare the actual key points of the human body target of the user with the corresponding standard key points of the human body target by performing feature vector comparison to obtain the action accuracy value.

10. The 8K set-top box training system according to claim 9, characterized in that: The actual key point extraction module comprises: The first control module is used to send a start instruction to the external USB camera module; receive the video stream data returned by the external USB camera module; A determination module, used to determine whether a target user is detected based on the video stream data; The second control module is used to send a start instruction to the gesture recognition module if the target user has been detected; and receive the actual key points of the user's human body target returned by the gesture recognition module.

Citation Information

Patent Citations

  • Rehabilitation training method based on set top box, set top box and readable storage medium

    CN109011418A

  • Human body posture recognition method and device based on skeleton key points, storage medium and terminal

    CN111680562A

  • Intelligent monitoring method and system for closed-loop feedback of functional rehabilitation training

    CN115346640A

  • Motion posture correction method based on motion recognition

    CN116343335A

  • Operation control method and device of IPTV set top box, set top box and medium

    CN118132344A

Cited By

  • Multifunctional programmable visual terminal system supporting custom scene configuration

    CN120356264A

  • Intelligent evaluation system and method for rehabilitation training exercises

    CN120727287B