Action normalization real-time evaluation method, system and device and storage medium

Abstract

The invention discloses an action normalization real-time evaluation method, system and device and a storage medium, and the method comprises the steps: obtaining human body action image data, and carrying out the preprocessing of the data, and obtaining a preprocessed human body action image; performing skeleton point detection on each frame of preprocessed human body action image by using a pre-trained human body posture estimation model to obtain two-dimensional coordinate information of main joints of a human body in each frame of image, and constructing a time sequence skeleton key point sequence; and a time sequence feature analysis algorithm is adopted to analyze the time sequence skeleton key point sequence, multi-joint angle normalization and dynamic fluency evaluation is carried out, and a joint correction suggestion is customized according to an evaluation result. According to the method, quantitative analysis and objective scoring of motion normalization and fluency can be realized without additional hardware equipment, so that the practicability of the evaluation system in scenes of physical education, rehabilitation training, fitness guidance and the like is remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of multi-joint action evaluation and intelligent processing, and relates to a motion specification real-time evaluation method, system, device and storage medium. BACKGROUND

[0002] At present, human motion specification evaluation is an important research direction in the fields of computer vision, pattern recognition and intelligent sports. The task aims to automatically judge whether the motion execution of the measured person conforms to the preset specification standard through image or video analysis, and is widely used in sports training, medical rehabilitation, fitness guidance, human-computer interaction and other scenes. Traditional human motion evaluation usually relies on expert manual scoring or sensor data based on wearable devices to analyze joint motion trajectory, motion amplitude and coordination, but these methods have many limitations such as strong subjectivity, poor portability or high cost of acquisition equipment, which is difficult to meet the current needs of large-scale, real-time and automatic evaluation. In recent years, with the rapid development of deep learning and computer vision technology, human pose estimation and motion recognition have made great progress. Multi-joint skeleton point detection based on camera video stream can stably obtain the spatial coordinate information of the main joints of the human body, greatly enriching the analysis means of motion timing features. However, how to effectively use these skeleton key point data for motion segmentation, specification analysis and real-time feedback still faces many challenges.

[0003] A large number of studies have shown that it is difficult to accurately evaluate the specification of complex actions with strong continuity and coordination based on single-frame pose features alone, especially when dealing with actions that require high coordination and timing continuity, the capture and modeling of timing information are particularly critical. On the one hand, the original joint coordinates have many interference factors such as pose changes, occlusions, detection errors, etc., which put higher requirements on the accurate extraction of skeleton points and subsequent feature normalization and smoothing processing. On the other hand, existing methods often only focus on the state of a joint or a moment, without fine-grained evaluation of whole-body multi-joint coordination and complete action cycles, which can easily lead to missing key links or inaccurate overall scores.

[0004] In addition, the motion specification real-time evaluation system must have the characteristics of low latency, high robustness and scalability to meet the changing scene requirements in actual applications. For example, in online sports teaching or rehabilitation assistance training, the evaluation system not only needs to accurately obtain the skeleton information, but also needs to provide fast and visual feedback on key action stages to help users correct errors in time. Although some current researches have introduced timing modeling (such as RNN, LSTM, timing convolution, etc.), they often lack knowledge-driven specification reference template modeling and lack open multi-scene evaluation systems and automated data export and visual analysis capabilities, making it difficult to fully improve the generality and practicality of the system. SUMMARY

[0005] The purpose of this invention is to provide a method, system, device and storage medium for real-time evaluation of action standardization, which solves the problem of low universality and practicality in the prior art.

[0006] To achieve the above objectives, the present invention employs the following technical solution: A real-time evaluation method for motion standardization includes: Human motion image data is acquired and preprocessed to obtain preprocessed human motion images; Using a pre-trained human pose estimation model, skeleton point detection is performed on each pre-processed human motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, and a temporal skeleton key point sequence is constructed. A temporal feature analysis algorithm is used to analyze the temporal skeletal key point sequence, evaluate the standardization of multi-joint angles and dynamic smoothness, and customize joint correction suggestions based on the evaluation results.

[0007] Furthermore, the preprocessing of human motion image data includes Gaussian filtering for noise reduction, contrast adjustment, and size normalization.

[0008] Furthermore, the pre-trained human pose estimation model is the yolov8-pose open-source model.

[0009] Furthermore, the temporal feature analysis algorithm's analysis process of the temporal skeletal keypoint sequence includes: Action phase recognition is performed on temporal skeleton key point sequences. By analyzing the spatial displacement, velocity and angle changes of key points between consecutive frames, action label sequences are formed.

[0010] Furthermore, the displacement velocity of key points between consecutive frames is:

[0011] in, Indicates the first frame, Indicates the first A key joint Indicates the first The first frame Displacement velocity at key points and They represent the first The first frame The x and y coordinates of the key points in the image coordinate system and These represent the first and second frames in the previous frame, respectively. A key point.

[0012] The joint angles between keypoints in consecutive frames are:

[0013] in, Indicates the first frame, Indicates the first A key joint Indicates the actual angle. , and These represent the points that make up a certain joint angle.

[0014] Furthermore, the method for evaluating the standardization and dynamic smoothness of multi-joint angles is as follows: Calculate the canonical score of a key joint in a given frame; Calculate the weighted normalization total score of all joints in the whole body in this frame; Calculate the score for the smoothness of the full-body multi-joint movements in this frame; The final overall score for the frame is calculated using a weighted algorithm.

[0015] Furthermore, the canonical score for a specific key joint in a given frame is:

[0016] in, Indicates the first frame, Indicates the first A key joint Indicates the first Frame number The standardization score of each key joint, Indicates the actual angle. Indicates the standard angle. Indicates the allowed normalization threshold; The weighted normalization score is:

[0017] in, Indicates the first frame, Indicates the first A key joint This indicates the total number of key joints contained in the current frame. This represents the weighted normative total score. This represents the weighting coefficient for each joint; The score for motion smoothness is:

[0018] in, Indicates the first frame, This represents the standard deviation of the canonical score sequence of the most recent frames. The normalized parameter representing the fluency score, Indicates fluency score; The final overall score is:

[0019] in, Indicates the first frame, The weighting coefficients representing the balance between normality and fluency. This represents the weighted normative total score. Indicates fluency score, This indicates the final overall score.

[0020] A real-time motion standardization evaluation system, comprising: The preprocessing module is used to acquire human motion image data and perform preprocessing to obtain preprocessed human motion images. The detection module is used to perform skeleton point detection on each frame of pre-processed human motion image using a pre-trained human pose estimation model, to obtain the two-dimensional coordinate information of the main joints of the human body in each frame image, and to construct a temporal skeleton key point sequence. The evaluation module uses a temporal feature analysis algorithm to analyze the temporal skeletal key point sequence, evaluate the standardization and dynamic smoothness of multiple joint angles, and customize joint correction suggestions based on the evaluation results.

[0021] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method.

[0022] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method.

[0023] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a real-time evaluation method for motion standardization. By preprocessing static images or multi-frame human motion image data, and utilizing a pre-trained human pose estimation model, skeleton point detection is performed on each frame of the motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, constructing a temporal sequence of skeletal key points. A temporal feature analysis algorithm is then used to analyze the temporal sequence of skeletal key points, evaluating the standardization and dynamic smoothness of multi-joint angles, and customizing joint correction suggestions based on the scoring results. This invention combines a multi-joint pose recognition model with a temporal deep learning algorithm, employing an optimized skeleton feature extraction and motion evaluation process. The real-time acquired skeletal key point sequence is input into a finely tuned motion classification network. Through multi-joint angle transformation, continuity analysis, and multi-task loss function optimization, the temporal features of the skeleton data are fully utilized, supporting accurate, efficient, and real-time motion standardization calculation, greatly improving the practical value and applicability of the evaluation. Compared with existing methods, this invention requires no additional hardware equipment; it can efficiently and accurately extract the dynamic temporal features of multiple human joints directly from ordinary image or video input, achieving quantitative analysis and objective scoring of motion standardization and smoothness. This invention utilizes an end-to-end visual feature modeling and evaluation mechanism to effectively improve the intelligence level of motion segmentation detection, anomaly identification, real-time feedback, and correction suggestions, thereby significantly enhancing the practicality of the evaluation system in scenarios such as sports teaching, rehabilitation training, and fitness guidance. While ensuring real-time processing speed, this invention possesses high accuracy, strong robustness, and good scalability, enabling it to adapt to the needs of multi-person, multi-type motion detection and evaluation in complex environments. Furthermore, the automated archiving and visual analysis of key evaluation data facilitates subsequent historical tracking of sports training and big data mining, effectively improving the automation and scientific basis of motion standardization evaluation, and has significant application value in promoting the development of intelligent sports and human-computer interaction.

[0024] Furthermore, this invention employs an advanced human pose estimation algorithm to obtain multi-joint skeleton data in real time. By introducing the YOLOv8-pose lightweight skeleton detection model, complete human keypoint sequences are extracted directly from regular video streams, greatly reducing hardware dependence and system deployment complexity, and providing a fundamental guarantee for large-scale, multi-scenario, and low-cost real-time motion evaluation.

[0025] Furthermore, the multi-joint angle standardization and dynamic smoothness evaluation method of the present invention can automatically segment and label different action stages based on continuously collected human multi-joint information, perform temporal recognition and segmentation of skeletal action data, and perform weighted calculation and scoring of key joint angles, amplitudes, rhythms, and other indicators for each segment of action to quantitatively evaluate the standardization of action at each stage. Abnormal and low-scoring actions are highlighted on the interface in real time, thereby achieving continuous monitoring and intelligent feedback on the standardization of user action execution. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the real-time evaluation method for motion standardization of the present invention.

[0028] Figure 2 This is the original image of Embodiment 1 of the present invention.

[0029] Figure 3 This is the original image of Embodiment 2 of the present invention.

[0030] Figure 4 This is a result image of the upright state in Embodiment 1 of the present invention.

[0031] Figure 5 This is a result image of the knee being slightly bent in Embodiment 1 of the present invention.

[0032] Figure 6 This is a result image of the standard squatting state in Embodiment 1 of the present invention.

[0033] Figure 7 This is a result image of a non-standard squatting state in Embodiment 2 of the present invention.

[0034] Figure 8 This is a result image of the standard squatting state in Embodiment 2 of the present invention.

[0035] Figure 9 This is a schematic diagram of the structure of a preferred embodiment of the real-time evaluation system for action standardization of the present invention.

[0036] Figure 10 This is a schematic diagram of the electronic device structure according to a preferred embodiment of the present invention. Detailed Implementation

[0037] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0038] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0039] It should be noted that the terminals involved in the embodiments of this application may include, but are not limited to, mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers, personal computers (PCs), MP3 players, MP4 players, wearable devices (e.g., smart glasses, smartwatches, smart bracelets), smart home devices, and other smart devices.

[0040] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0041] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1 This invention provides a real-time evaluation method for action standardization, specifically including the following steps: S1. Acquire static images or multi-frame human motion image data captured by a camera and perform preprocessing.

[0042] The experimental image set comes from open-source static images of motion poses or dynamic video streams captured in real time by built-in or external cameras on the computer. To obtain preprocessed images, Gaussian filtering and contrast adjustment are performed on the input static images or multi-frame motion images of the dynamic video stream. At the same time, the image size is adjusted using the size normalization function built into YOLOv8 to improve the accuracy and processing efficiency of subsequent skeleton keypoint detection and enhance the reliability of the results.

[0043] Preferably, Gaussian filtering is performed for noise reduction according to the following formula (1): (1) in, Indicates the neighborhood. This represents the horizontal coordinate of the current pixel in the image. The vertical coordinate of the current pixel in the image. Indicates the current pixel in the neighborhood The relative horizontal coordinate offset within. Indicates the current pixel in the neighborhood The relative vertical coordinate offset within. This indicates that the image after filtering is in The pixel value at a given location represents the filter kernel weight. Indicates the original image at position Pixel values; Calculate the two-dimensional Gaussian kernel according to the following formula (2). : (2) in, Indicates in Two-dimensional Gaussian kernel at the location, The standard deviation of the Gaussian kernel is represented. Depend on Calculated; The contrast is adjusted by performing an affine transformation according to the following formula (3): (3) in, This indicates adjusting the contrast. This indicates adjusting the brightness. This indicates the pixel values ​​after affine transformation (after adjusting contrast and brightness). Pixel value at that location, This indicates the original image at the pixel level. Pixel value at; Image size normalization is performed according to the following formula (4): (4) in, Indicates the width of the input image. Indicates the height of the input image. Indicates the target size width. Indicates the target dimension height. Represents the horizontal coordinates of the normalized target image. Represents the vertical coordinates of the normalized target image. This indicates the target dimension coordinates after dimension normalization. Image pixel values ​​at that location.

[0044] S2. Using a pre-trained human pose estimation model, skeleton point detection is performed on each pre-processed human motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, and a temporal skeleton key point sequence is constructed.

[0045] The pre-trained human pose estimation model uses the yolov8-pose open-source model. The yolov8n-pose.pt file is downloaded using pip, and the pose detection model built into yolov8 is invoked. Skeleton point detection is performed on each frame of the human pose image, extracting the two-dimensional coordinate information of major human joints, including the hip, knee, and ankle joints.

[0046] The two-dimensional coordinate information of the major joints is time-sequentially superimposed to form a continuous frame sequence of human temporal skeletal key points. .

[0047] S3. The extracted multi-joint skeletal point sequence is analyzed using a time-series feature analysis algorithm.

[0048] In S3, a temporal feature analysis algorithm is used to analyze the temporal skeleton keypoint sequence. Action phase recognition is performed by analyzing the spatial displacement, velocity, and angle changes of key points between consecutive frames to form an action label sequence. }

[0049] Furthermore, the speed is calculated according to the following formula (5). : (5) in, Indicates the first frame, Indicates the first A key joint Indicates the first The first frame Displacement velocity at key points and They represent the first The first frame The x and y coordinates of the key points in the image coordinate system and These represent the previous frame (the first frame) In the first frame) One key point; Calculate the joint angle using the following formula (6). : (6) in, Indicates the first frame, Indicates the first A key joint Indicates the actual angle. , and These represent the points that make up a certain joint angle.

[0050] S4. Conduct an evaluation of the standardization and dynamic smoothness of multi-joint angles.

[0051] By designing standardized and dynamic smoothness evaluation criteria, and using the joint angles obtained in step 3... and speed Calculate the first Frame number Standardized scores for key joints Weighted normative total score of multiple joints throughout the body Scoring of the fluidity of movements across multiple joints throughout the body And the final overall score .

[0052] S41. Using actual angles and standard angles, calculate the normative score of the j-th key joint in frame t. S42. Using normative scores to identify joints, calculate the weighted normative total score of all joints in the whole body in frame t. S43. Using the standard deviation of the normalized score sequence, calculate the score of the smoothness of the whole-body multi-joint movement in frame t. S44. Using the weighted coefficients of normality and fluency, the weighted normality score, and the motion fluency score, calculate the final comprehensive score for frame t according to the weighted algorithm.

[0053] In S41, the normativity score is calculated according to the following formula (7): (7) in, Indicates the first frame, Indicates the first A key joint Indicates the first Frame number The standardization score of each key joint, Indicates the actual angle. Indicates the standard angle. This represents the allowable normalization threshold; the smaller the deviation from the standard deviation, the higher the score, with a maximum score of 1.

[0054] In S42, the weighted normalization total score is calculated according to the following formula (8): (8) in, Indicates the first frame, Indicates the first A key joint This indicates the total number of key joints contained in the current frame. This represents the weighted normative total score. This represents the weighting coefficient for each joint.

[0055] In S43, the score for motion fluency is calculated according to the following formula (9): (9) in, Indicates the first frame, This represents the standard deviation of the canonical score sequence of the most recent frames. The normalized parameter representing the fluency score, The standard deviation indicates fluency; the smaller the standard deviation, the smoother (more fluid) the performance.

[0056] In S44, the final overall score is calculated according to the following formula (10): (10) in, Indicates the first frame, The weighting coefficients representing the balance between normality and fluency. This represents the weighted normative total score. Indicates fluency score, This indicates the final overall score.

[0057] S5. Based on the scoring results, detect key joints in the current frame or action phase whose scores are lower than the preset threshold, highlight them on the visualization interface, overlay the action score progress, and customize joint correction suggestions.

[0058] According to the The scoring results for a frame or action phase, if a key joint scores... < If a point is identified as an abnormal key point, the percentage of quantitative abnormalities will be displayed in real time.

[0059] The abnormality rate is quantified according to the following formula (11): (11) in, This indicates the total number of key joints contained in the current frame. Indicates the first Frame number The standardization score of each key joint, The threshold for anomaly detection represents the normative score. Indicates the first The proportion of abnormal joints in a frame (abnormal percentage). This indicates an indicator function; a value of 1 indicates that the condition is true. Abnormal areas are highlighted in the visualization interface, and the interface displays them synchronously. Real-time progress and joint correction suggestions.

[0060] The present invention will be further described in detail below through specific embodiments: Example 1: This embodiment verifies the description generated by the above method through simulation to ensure its accuracy and consistency. Based on the verification results and actual needs, appropriate model parameters are selected and adjusted to improve and guarantee the description quality. The method of this invention has good flexibility.

[0061] like Figure 2 As shown, this embodiment selects open-source knee flexion images as the test object and verifies the effectiveness of the real-time evaluation method for motion standardization based on multi-joint temporal features through simulation experiments.

[0062] I. Simulation Conditions The experiment was conducted on a high-performance workstation equipped with a 12-core Intel i7-12700H CPU (2.30GHz), 16GB of RAM, and an NVIDIA GeForce RTX 3070 Ti GPU. A virtual Python 3.9 environment was created using Anaconda3, and the yolov8n-pose model was used for related experiments.

[0063] First, Gaussian filtering is applied to the image frames of the static image, using a kernel size of 5. 5. Standard deviation Gaussian filtering is used to denoise each frame of the image; for some acquired images with uneven lighting, a linear stretching method is used to uniformly adjust the contrast range. The pixel grayscale dynamic range of the images before and after processing was statistically analyzed, with an average improvement rate of 21%. After filtering and contrast enhancement, the images were directly input into the YOLOv8 pose model, which utilizes its internal automatic resizing and normalization processes to standardize the input size. Pixel value range normalized to The average spatial interpolation error (PSNR) introduced during the model resizing process is 38.4 dB, far lower than the structural information interference introduced by filtering and enhancement. Gaussian filtering of the original image significantly reduces noise intensity, increasing the structural similarity index (SSIM) from 0.88 to 0.93. After contrast enhancement, the average pixel variance increases from 2034 to 2950, ​​which is beneficial for downstream feature extraction. The preprocessing in the simulation experiments effectively improves the robustness and accuracy of the human pose detection model and significantly reduces the interference of data noise on subsequent tasks.

[0064] Simulation experiments were conducted using a collected set of static pose images to generate skeletal keypoint plots and pose scientific accuracy scores. For example... Figure 4 and Figure 5As shown, the object is judged as upright and does not receive a score when standing or slightly bent at the knees (not squatting). Figure 6 The object shown was identified as squatting under standard squatting conditions and given a reasonable score. The results demonstrate that this invention can detect and score human posture in static posture images containing multiple targets and interference information, and store the test results.

[0065] Example 2: The difference from Example 1 is that the simulation experiment uses a real-time acquired video stream, such as... Figure 3 As shown, a frame from a real-time video stream is captured, and Gaussian filtering is applied to the image frames of the dynamic video stream. The skeleton of a moving human body is traced in real-time, and scoring, posture judgment, and scientific correction are performed. The experiment attempts to detect the posture of multiple moving individuals simultaneously, thus testing the system's versatility.

[0066] Table 1 shows the experimental data on posture accuracy, correction accuracy, and scoring reasonableness under different simulation objects and test objects.

[0067] Table 1. Experimental data on posture accuracy, correction accuracy, and scoring reasonableness.

[0068] like Figure 7 As shown, in cases of non-standard squatting, the result was classified as a squat, but a low score was not obtained, and corrective suggestions were given, namely that the knee flexion needs to be improved, such as... Figure 8 As shown, under standard squatting conditions, the result is determined to be a squat, and a reasonable score is given. Experimental results show that the present invention has significant advantages in posture detection accuracy, correction quality, and scoring reasonableness, and achieves better performance in real-time evaluation of motion standardization among various methods.

[0069] Example 3: This invention also provides a real-time evaluation system for motion standardization, such as... Figure 9 As shown, the system includes a preprocessing module, a detection module, and an evaluation module.

[0070] The preprocessing module is used to acquire human motion image data and perform preprocessing to obtain preprocessed human motion images. The detection module is used to perform skeleton point detection on each frame of pre-processed human motion image using a pre-trained human pose estimation model, to obtain the two-dimensional coordinate information of the main joints of the human body in each frame image, and to construct a temporal skeleton key point sequence. The evaluation module uses a temporal feature analysis algorithm to analyze the temporal skeletal key point sequence, evaluate the standardization and dynamic smoothness of multiple joint angles, and customize joint correction suggestions based on the evaluation results.

[0071] It is understood that the real-time motion standardization evaluation system provided by the present invention corresponds to the real-time motion standardization evaluation method provided in the foregoing embodiments. The relevant technical features of the real-time motion standardization evaluation system can be referred to the relevant technical features of the real-time motion standardization evaluation method, and will not be repeated here.

[0072] Another object of the present invention is to provide an electronic device, such as... Figure 10 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor performs the steps of the real-time evaluation method for behavioral conformity.

[0073] The real-time evaluation method for action standardization includes the following steps: Human motion image data is acquired and preprocessed to obtain preprocessed human motion images; Using a pre-trained human pose estimation model, skeleton point detection is performed on each pre-processed human motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, and a temporal skeleton key point sequence is constructed. A temporal feature analysis algorithm is used to analyze the temporal skeletal key point sequence, evaluate the standardization of multi-joint angles and dynamic smoothness, and customize joint correction suggestions based on the evaluation results.

[0074] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the real-time evaluation method for action conformity.

[0075] The real-time evaluation method for action standardization includes the following steps: Human motion image data is acquired and preprocessed to obtain preprocessed human motion images; Using a pre-trained human pose estimation model, skeleton point detection is performed on each pre-processed human motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, and a temporal skeleton key point sequence is constructed. A temporal feature analysis algorithm is used to analyze the temporal skeletal key point sequence, evaluate the standardization of multi-joint angles and dynamic smoothness, and customize joint correction suggestions based on the evaluation results.

[0076] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0077] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0079] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for real-time evaluation of motion standardization, characterized in that, include: Human motion image data is acquired and preprocessed to obtain preprocessed human motion images; Using a pre-trained human pose estimation model, skeleton point detection is performed on each pre-processed human motion image to obtain the two-dimensional coordinate information of the main joints of the human body in each frame, and a temporal skeleton key point sequence is constructed. A temporal feature analysis algorithm is used to analyze the temporal skeletal key point sequence, evaluate the standardization of multi-joint angles and dynamic smoothness, and customize joint correction suggestions based on the evaluation results.

2. The real-time evaluation method for motion standardization according to claim 1, characterized in that, Preprocessing of human motion image data includes Gaussian filtering for noise reduction, contrast adjustment, and size normalization.

3. The real-time evaluation method for motion standardization according to claim 1, characterized in that, The pre-trained human pose estimation model is the yolov8-pose open-source model.

4. The real-time evaluation method for motion standardization according to claim 1, characterized in that, The temporal feature analysis algorithm analyzes the temporal skeleton keypoint sequence, including: Action phase recognition is performed on temporal skeleton key point sequences. By analyzing the spatial displacement, velocity and angle changes of key points between consecutive frames, action label sequences are formed.

5. The real-time evaluation method for motion standardization according to claim 4, characterized in that, The displacement velocity of key points between consecutive frames is: in, Indicates the first frame, Indicates the first A key joint Indicates the first The first frame Displacement velocity at key points and They represent the first The first frame The x and y coordinates of the key points in the image coordinate system and These represent the first and second frames in the previous frame, respectively. A key point. The joint angles between keypoints in consecutive frames are: in, Indicates the first frame, Indicates the first A key joint Indicates the actual angle. , and These represent the points that make up a certain joint angle.

6. The real-time evaluation method for motion standardization according to claim 1, characterized in that, The method for evaluating the standardization and dynamic smoothness of multi-joint angles is as follows: Calculate the canonical score of a key joint in a given frame; Calculate the weighted normalization total score of all joints in the whole body in this frame; Calculate the score for the smoothness of the full-body multi-joint movements in this frame; The final overall score for the frame is calculated using a weighted algorithm.

7. The real-time evaluation method for motion standardization according to claim 6, characterized in that, The canonical score for a specific key joint in a given frame is: in, Indicates the first frame, Indicates the first A key joint Indicates the first Frame number The standardization score of each key joint, Indicates the actual angle. Indicates the standard angle. Indicates the allowed normalization threshold; The weighted normalization score is: in, Indicates the first frame, Indicates the first A key joint This indicates the total number of key joints contained in the current frame. This represents the weighted normative total score. This represents the weighting coefficient for each joint; The score for motion smoothness is: in, Indicates the first frame, This represents the standard deviation of the canonical score sequence of the most recent frames. The normalized parameter representing the fluency score, Indicates fluency score; The final overall score is: in, Indicates the first frame, The weighting coefficients representing the balance between normality and fluency. This represents the weighted normative total score. Indicates fluency score, This indicates the final overall score.

8. A real-time evaluation system for motion standardization, characterized in that, include: The preprocessing module is used to acquire human motion image data and perform preprocessing to obtain preprocessed human motion images. The detection module is used to perform skeleton point detection on each frame of pre-processed human motion image using a pre-trained human pose estimation model, to obtain the two-dimensional coordinate information of the main joints of the human body in each frame image, and to construct a temporal skeleton key point sequence. The evaluation module uses a temporal feature analysis algorithm to analyze the temporal skeletal key point sequence, evaluate the standardization and dynamic smoothness of multiple joint angles, and customize joint correction suggestions based on the evaluation results.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images