Intelligent online correction method and system for pull-up

Abstract

The invention provides an intelligent online correction method and system for pull-up, and relates to the technical field of data processing, and the method comprises the steps: obtaining real-time to-be-corrected pull-up video data; extracting skeleton point coordinate data in the pull-up video data through a posture recognition algorithm; missing value completion processing is carried out on the skeleton point coordinate data; establishing a joint state space model according to the skeleton point coordinate data after completion processing; according to the joint state space model, a target angle is predicted in combination with a Kalman filtering algorithm and a curve fitting algorithm; and outputting a correction prompt through a preset correction rule under the condition that the predicted target angle which does not conform to the pull-up motion standard index exists. According to the method and the system, a relatively good recognition and correction effect can be achieved on nonstandard actions in pull-up sports, correction feedback can be given to a user in real time according to a correction rule, and the user is assisted to better independently complete pull-up sports training.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an intelligent online correction method and system for pull-ups. Background Art

[0002] Pull-ups is a common physical exercise method, mainly exercising the muscles of the back, shoulders and arms. Intelligent online correction is to analyze and correct a certain action or behavior online in real time through intelligent technologies (such as computer vision, artificial intelligence algorithms, etc.). The intelligent online correction method for pull-ups refers to a technical solution that uses intelligent technologies to be able to analyze the pull-up action in real time online and automatically provide correction suggestions.

[0003] Intelligent online correction based on the pull-up action can achieve personalized guidance, is suitable for various fitness enthusiasts, especially in the case where professional coach guidance cannot be obtained, can effectively improve the training quality and efficiency of athletes, and is of great significance for effectively reducing sports injuries caused by incorrect postures.

[0004] At present, much progress has been made in the research of combining artificial intelligence and sports science to assist sports, but most have only completed the tasks of classification recognition and counting. Although the existing intelligent online correction methods for pull-ups can provide real-time feedback through video analysis and pose recognition, they have poor adaptability when facing athletes with different body types and individual sports levels, resulting in inaccurate correction suggestions. In addition, devices relying on cameras or sensors are easily affected by factors such as environmental light and device angles, leading to a decrease in recognition accuracy. Moreover, existing methods mainly focus on technical implementation, ignoring the psychological feedback and personalized needs of athletes, and may not be able to comprehensively consider the physiological characteristics and training goals of athletes. Summary of the Invention

[0005] In order to solve the technical problems that the existing intelligent online correction methods for pull-ups have poor adaptability when facing athletes with different body types and individual sports levels, resulting in inaccurate correction suggestions, the recognition accuracy decreases due to relying on cameras or sensors to collect motion images, and the physiological characteristics and training goals of athletes cannot be comprehensively considered, the present invention provides an intelligent online correction method and system for pull-ups.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First aspect:

[0008] An intelligent online correction method for pull-ups provided by an embodiment of the present invention includes:

[0009] S1: Obtain real-time video data of pull-ups to be corrected;

[0010] S2: Extract the skeletal point coordinate data in the chin-up video data through the pose recognition algorithm;

[0011] S4: Perform missing value filling processing on the skeletal point coordinate data;

[0012] S7: Establish a joint state space model based on the filled skeletal point coordinate data;

[0013] S5: Predict the target angle according to the joint state space model, combining the Kalman filter algorithm and the curve fitting algorithm;

[0014] S6: In the case where there is a predicted target angle that does not meet the standard index of the chin-up movement, output a correction prompt through the preset correction rule.

[0015] Second aspect:

[0016] An intelligent online correction system for chin-ups provided by an embodiment of the present invention includes:

[0017] A processor;

[0018] A memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, the intelligent online correction method for chin-ups as in the first aspect is implemented.

[0019] Third aspect:

[0020] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored. When the program is executed by the processor, the intelligent online correction method for chin-ups as in the first aspect is implemented.

[0021] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0022] In the embodiment of the present invention, by obtaining real-time chin-up video data to be corrected, and then, through the pose recognition algorithm, extracting the skeletal point coordinate data in the chin-up video data, so as to ensure the accurate capture of the athlete's movements. Further, perform missing value filling processing on the skeletal point coordinate data, establish a joint state space model based on the skeletal point coordinate data, and combine the Kalman filter algorithm and the curve fitting algorithm to predict the target angle, providing a scientific basis for action correction. Finally, compare the predicted target angle with the standard action and give a correction prompt in a timely manner. Through precise pose analysis and intelligent algorithms, monitor and correct non-standard actions in real time, improve the training effect and reduce sports injuries, and give users correction feedback in real time according to the correction rules, assisting users to better independently complete the chin-up movement training, meeting the personalized requirements of different athletes. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A schematic flow chart of an intelligent online correction method for pull-ups provided by an embodiment of the present invention;

[0025] Figure 2 A schematic structural diagram of an intelligent online correction system for pull-ups provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0027] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0028] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0029] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0030] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0031] Reference Manual Attached Figure 1 , showing a flow chart of an intelligent online correction method for pull-ups provided by an embodiment of the present invention.

[0032] An embodiment of the present invention provides an intelligent online correction method for pull-ups. This method can be implemented by an intelligent online correction device for pull-ups, which can be a terminal or a server. The processing flow of the intelligent online correction method for pull-ups can include the following steps:

[0033] S1: Obtain real-time video data of pull-ups to be corrected.

[0034] Among them, "to be corrected" refers to the pull-up actions of athletes that need to be analyzed and corrected by intelligent methods. Pull-up video data refers to the video records of athletes doing pull-ups captured by devices such as cameras, including the real-time images and motion information of athletes during the execution of actions.

[0035] It should be noted that by obtaining the video data of pull-ups, a comprehensive basis for motion analysis is provided. Video data is the key to the subsequent correction process. It can accurately record the action details of athletes, capture every moment of the motion posture. Compared with traditional manual supervision or device assistance, video data provides a large amount of information that is convenient for analysis and can be easily played back and processed multiple times.

[0036] S2: Extract the bone point coordinate data in the pull-up video data through a pose recognition algorithm.

[0037] Among them, the pose recognition algorithm is an algorithm that analyzes the human pose through computer vision technology, used to extract the positions of the key bones of the human body (such as joints, limb ends, etc.) from images or videos. Bone point coordinate data refers to the two-dimensional or three-dimensional coordinate data of the human key points (such as head, shoulders, wrists, etc.) extracted through the pose recognition algorithm.

[0038] It should be noted that the bone point coordinate data in the pull-up action is accurately extracted through the pose recognition algorithm. This process can achieve the precise capture of the positions of each joint of the athlete's body, providing detailed information for subsequent motion analysis. Different from traditional manual recording or using complex sensors, the pose recognition algorithm can monitor athletes in real time through ordinary camera devices, reducing equipment costs and operation complexity.

[0039] In a possible implementation manner, the pose recognition algorithm is specifically: the lightweight OpenPose algorithm.

[0040] The extraction of bone point coordinate data through the lightweight OpenPose algorithm specifically includes:

[0041] S201: Based on the lightweight OpenPose algorithm, extract the image frames in the pull-up video data.

[0042] Among them, the lightweight OpenPose algorithm is an optimized version of the original OpenPose algorithm, aiming to reduce the consumption of computing resources and improve real-time performance. The OpenPose algorithm extracts key skeletal points of the human body, such as the head, shoulders, elbows, wrists, knees, ankles, etc., from images or videos through a deep learning model for pose recognition. An image frame refers to each static image in video data, which is called a frame.

[0043] S202: Extract the image features of the image frame through a lightweight convolutional network:

[0044] F(t) = MobileNet(I(t))

[0045] Among them, F(t) represents the image features at time t, MobileNet represents the lightweight convolutional neural network, and I(t) represents the image frame at time t.

[0046] S203: According to the image features, output the heatmaps of each skeletal point through the target point regression algorithm:

[0047] H k (t) = Heatmap(F(t), k), k ∈ {1, 2,..., N}

[0048] Among them, H k (t) represents the heatmap of the k-th skeletal point at time t, Heatmap(F(t), k) represents the heatmap generation function, which is used to extract the heatmap of the k-th skeletal point from the input image features F(t), k = 1, 2,..., N, and N represents the total number of skeletal points.

[0049] S204: Extract the maximum position of each heatmap to generate skeletal point coordinate data:

[0050] (x k (t), y k (t)) = arg max(H k (t))

[0051] Among them, x k (t) represents the abscissa of the k-th skeletal point at time t, and y k (t) represents the ordinate of the k-th skeletal point at time t, and argmax represents the independent variable for taking the maximum value.

[0052] It should be noted that the lightweight OpenPose is optimized by reducing the complexity of the network model, reducing the amount of calculation, or adopting a more efficient network architecture (such as MobileNet, EfficientNet, etc.), so that the algorithm can run more efficiently on resource-constrained devices (such as mobile devices, embedded devices, etc.).

[0053] S3: Perform missing value filling processing on the bone point coordinate data.

[0054] Among them, missing value filling processing means that during the data processing process, there may be situations where some bone points cannot be correctly recognized or data cannot be obtained. Missing value filling processing is to fill these missing data points through algorithms to ensure the integrity and consistency of the data. Through missing value filling processing, the integrity of the bone point coordinate data is ensured. When the pose recognition algorithm cannot completely extract the coordinates of some bone points, missing value filling processing can fill these missing data through interpolation, prediction or other technical methods, thereby avoiding analysis errors caused by incomplete data.

[0055] S4: Establish a joint state space model based on the bone point coordinate data after filling processing.

[0056] Among them, the joint state space model is a mathematical model used to describe and predict the motion state of joints. It considers physical quantities such as joint angles, angular velocities and angular accelerations, and describes their changes over time through mathematical equations. This model is commonly used in dynamic system analysis and can simulate the motion trajectories and behaviors of joints. By establishing a joint state space model, the motion behavior of joints can be accurately described and predicted. The joint state space model can not only predict the dynamic changes of actions, but also help to judge whether the actions are standard, thereby providing timely correction suggestions.

[0057] In a possible implementation manner, S4 specifically includes:

[0058] S401: Calculate the distances between the coordinates of each bone point:

[0059]

[0060] Among them, a represents the distance between bone point B and bone point C, b represents the distance between bone point A and bone point C, c represents the distance between bone point A and bone point B, (x a , y a ) represents the coordinates of bone point A, (x b , y b ) represents the coordinates of bone point B, (x c , y c ) represents the coordinates of bone point C.

[0061] S402: Calculate the target angle according to the distances using the cosine theorem:

[0062]

[0063] Among them, cosα represents the cosine value of the target angle α, and arccos represents the inverse cosine operation.

[0064] S403: Based on the target angle and combined with physics knowledge, establish the motion equation of the target angle:

[0065]

[0066] ω(k) = ω(k - 1)+β(k - 1)·Δt

[0067] (x k (t), y k (t)) = arg max(H k (t))

[0068] Among them, θ(k) represents the angle value of the mover at the k-th moment, ω(k - 1) represents the angular velocity of the mover at the (k - 1)-th moment, Δ represents the increment operator, β(k - 1) represents the angular acceleration of the mover at the (k - 1)-th moment, and ω(k) represents the angular velocity of the mover at the k-th moment.

[0069] S404: Based on the motion equation, establish the joint state space model.

[0070] It should be noted that by calculating the distance between bone points and using the cosine theorem to calculate the target angle, the system can accurately measure the angle change of the joint, thereby providing a reliable data basis for subsequent motion analysis. This precise dynamic analysis helps to improve the effect of motion correction, reduce sports injuries, and optimize the training process.

[0071] In a possible implementation manner, when the target angle is a periodically changing angle, the state space model is specifically:

[0072]

[0073] Among them, Θ i (k) represents the state vector of the i-th target angle at the k-th moment, θ i (k) represents the position of the i-th target angle at the k-th moment, ω i (k) represents the angular velocity of the i-th target angle at the k-th moment, Ai represents the state transition matrix of the i-th target angle, Bi represents the control input matrix of the i-th target angle, U i (k - 1) represents the control input of the i-th target angle at the (k - 1)-th moment, ni represents the noise of the i-th target angle, T represents the time step, and β i (k - 1) represents the angular acceleration of the i-th target angle at the (k - 1)-th moment.

[0074] The output equation is specifically:

[0075]

[0076] Among them, Y i (k) and y i (k) both represent the predicted value of the i-th target angle, and H i represents the observation matrix of the i-th target angle.

[0077] When the target angle is the stable angle, the state space model is specifically:

[0078]

[0079] The output equation is specifically:

[0080] Y i (k) = y i (k) = θ i (k).

[0081] It should be noted that in the action of pulling oneself up, some angles (such as the angles of the elbows and shoulders) will change periodically with the movement, while other angles (such as the relative position of the chin and the horizontal bar) may tend to be stable at certain stages. Separating these two types of angles for modeling can more accurately capture the characteristics and variation laws of each angle, thereby improving the adaptability of the model to different types of actions, and can effectively improve the accuracy, computational efficiency and adaptability of the model, and further enhance the effect of motion correction.

[0082] S5: According to the joint state space model, combined with the Kalman filtering algorithm and the curve fitting algorithm, predict the target angle.

[0083] Among them, the Kalman filtering algorithm is an algorithm commonly used for the estimation and prediction of dynamic systems, especially suitable for real-time data with noise. In motion analysis, the Kalman filter can accurately predict the joint state based on the current and historical data, and suppress the influence of noise, improving the stability and accuracy of the prediction. The curve fitting algorithm refers to fitting the data points into a curve through mathematical methods, and is commonly used to predict the future state of the system. The target angle refers to the angle formed between joints during the movement. For example, the angle between the elbows, shoulders and wrists is the target angle.

[0084] It should be noted that the Kalman filtering algorithm and the curve fitting algorithm are combined to accurately predict the target angle. The Kalman filtering algorithm can remove noise from the real-time data and provide stable angle predictions, while the curve fitting algorithm can predict the change of the angle according to the motion trend. The combination of the two enables the model to provide accurate and reliable feedback in a dynamic environment, especially in the case of fast movement, which can reduce errors and improve the prediction accuracy.

[0085] In the present invention, first, it is determined whether there is data input. If there is no input data, the process directly enters the end. When the input data is θ5, prediction is directly performed through the Kalman filtering algorithm. When the input data is θ3, θ4, and θ6, if the angular change Δθ is greater than 5°, the linear fitting algorithm is used for prediction; otherwise, the Kalman filtering algorithm is used for prediction, and Δθ is checked again to see if it is greater than 5°. If so, the linear fitting algorithm is continuously used for prediction; otherwise, the Kalman filtering algorithm is used for prediction. When the input data is θ1 and θ2, if the angular change Δθ is less than 5°, the sine curve fitting algorithm is used for prediction; otherwise, the Kalman filtering algorithm is used for prediction, and Δθ is checked again to see if it is less than 5°. If so, the sine curve fitting algorithm is continuously used for prediction; otherwise, the Kalman filtering algorithm is used for prediction.

[0086] Combined with algorithms such as linear fitting, sine curve fitting, and Kalman filtering, it can select the best prediction method according to different situations, avoid the limitations brought by a single algorithm, improve the modeling ability of motion data. Then, by judging conditions, the prediction methods adopted in different situations are clarified, making the entire correction and prediction process have good logic, facilitating understanding and implementation, and thus providing good technical support for intelligent pull-up correction.

[0087] In a possible implementation manner, the target angles specifically include: the angle θ1 with the left elbow as the vertex and the left wrist and left shoulder as the two endpoints respectively; the angle θ2 with the right elbow as the vertex and the right wrist and right shoulder as the two endpoints respectively; the angle θ3 with the left knee as the vertex and the left ankle and left hip as the two endpoints respectively; the angle θ4 with the right knee as the vertex and the right ankle and right hip as the two endpoints respectively; the angle θ5 with the chin as the vertex and the left and right wrists as the two endpoints respectively; the angle θ6 with the midpoint of the left and right hips as the vertex and the left and right ankles as the two endpoints respectively.

[0088] In a possible implementation manner, the curve fitting algorithm includes a linear fitting algorithm and a sine curve fitting algorithm.

[0089] Among them, the linear fitting algorithm is a method of fitting a straight line model of a data point set through methods such as the least squares method. Specifically, linear fitting finds a straight line such that the sum of the squares of the perpendicular distances from the data points to the straight line is minimized. This algorithm is usually used when there is an approximate linear relationship between data to predict or describe the trend of the data through this straight line. The sine curve fitting algorithm refers to using the sine function to fit a data set, which is particularly suitable for situations where the data shows periodic fluctuations. The sine function has a specific period, amplitude, and phase, and can accurately describe the periodically changing data. The sine curve fitting algorithm fits the data points through the least squares method so that the fitted sine function can best represent the fluctuation trend of the data.

[0090] For the prediction of θ1 and θ2, the Kalman filtering algorithm and the sine curve fitting algorithm are combined for prediction.

[0091] Among them, during the standard chin-up movement, the angular values of angles θ1 and θ2 show periodic changes similar to a sine pattern, and there will be some errors in the Kalman filter at the inflection points of the curve. Based on this sine period characteristic of the curve, an algorithm combining the Kalman filter and the sine curve fitting method is adopted. When the curve slope is small, that is, near the inflection point, the sine curve is used to fit the angular value at the next moment.

[0092] For the prediction of θ3, θ4, and θ6, the Kalman filtering algorithm and the linear fitting algorithm are combined for prediction.

[0093] It should be noted that during the standard chin-up movement, the angular values of angles θ3, θ4, and θ6 are basically unchanged. When the slope of the angular value curve suddenly becomes large, it is very likely due to misdetection, and the value predicted by directly using the Kalman filtering algorithm may deviate greatly from the actual value. Therefore, the method of combining the Kalman filter and the linear fitting is used. When the curve slope changes greatly, the linear curve is used to fit the angular value at the next moment, and the data at the subsequent moment can be predicted more accurately.

[0094] For the prediction of θ5, the Kalman filtering algorithm is used for prediction.

[0095] For angle θ5, there are many inflection points, and the effect is poor after two-step prediction. Therefore, the Kalman filtering algorithm is adopted to only make a single-step prediction scheme, so as to improve the stability and accuracy of the prediction.

[0096] In a possible implementation manner, the Kalman filtering algorithm and the sine curve fitting algorithm are combined for prediction, specifically including:

[0097] When the angular change of θ1 or the angular change of θ2 is greater than the preset threshold, the sine curve fitting algorithm is used for prediction:

[0098] θ k = Asin(ωt k + φ)+ D

[0099] Among them, θ k represents the target angular value at time k, A represents the amplitude, ω represents the frequency, t k represents time k, φ represents the phase shift, and D represents the overall angular shift.

[0100] Otherwise, the Kalman filtering algorithm is used for prediction.

[0101] The Kalman filtering algorithm and the linear fitting algorithm are combined for prediction, specifically including:

[0102] When the angular changes of θ3, θ4 or θ6 are less than a preset threshold, prediction is performed through a linear fitting algorithm:

[0103] θ k = at k + b

[0104] where θ k represents the target angular value at time k, a represents the slope of the angular change, t k represents time k, and b represents the initial angle.

[0105] Otherwise, prediction is performed using the Kalman filter algorithm.

[0106] Performing prediction through the Kalman filter algorithm specifically includes:

[0107] Based on the Kalman filter state prediction equation, calculate the target angular value:

[0108]

[0109] where represents the state prediction value at the k-th moment, that is, the target angular value, represents the state estimate value at the (k - 1)-th moment, A represents the state transition matrix, B represents the control input matrix, u k-1 represents the control vector at the (k - 1)-th moment, represents the state prediction error covariance matrix at the k-th moment, P k-1 represents the state estimate error covariance matrix at the (k - 1)-th moment, and Q represents the process noise covariance matrix.

[0110] Update the target angular value and the state prediction error covariance matrix:

[0111]

[0112] where K k represents the Kalman gain at the k-th moment, H represents the measurement matrix, R represents the measurement noise covariance matrix, represents the state estimate value at the k-th moment, z k represents the measurement vector at the k-th moment, and P k represents the state estimate error covariance matrix at the k-th moment after measurement update.

[0113] S6: In the case where there is a predicted target angle that does not meet the standard index of the chin-up movement, output a correction prompt through a preset correction rule.

[0114] Among them, the standard indicators for the chin-up exercise are standards set for aspects such as the joint angles and movement amplitudes in the chin-up action, aiming to ensure that the action meets the correct biomechanical requirements, avoid sports injuries, and maximize the exercise effect. The standard indicators may include elbow angles, knee angles, whether the chin crosses the horizontal bar, etc. The preset correction rules are a series of rules preset by the system according to the exercise standards, used to judge whether the athlete's actions are standardized and give improvement suggestions when the actions do not meet the standards. The correction prompt is the adjustment suggestion given by the system according to the athlete's real-time action data and based on the preset correction rules when the action does not meet the standards.

[0115] It should be noted that according to the comparison between the target angle and the exercise standard, the correction prompt is automatically output to help the athlete immediately adjust the non-standard actions. Through the preset correction rules, the system can real-time evaluate the changes in key angles during the exercise process and provide feedback according to the athlete's specific performance. This intelligent correction method eliminates the deficiencies of relying on manual observation and judgment in traditional training and improves the training efficiency.

[0116] In a possible implementation manner, the exercise standard indicators specifically include: Indicator One to Indicator Five.

[0117] Indicator One is specifically: During the exercise, when reaching the lowest point, the included angles θ1 and θ2 of the two arms with the elbow as the vertex should both be close to 180 degrees.

[0118] Indicator Two is specifically: During the exercise, the exerciser cannot borrow strength by kicking the legs forward and backward, and the included angles θ3 and θ4 of the two legs with the knee as the vertex should both be close to 180 degrees.

[0119] Indicator Three is specifically: During the exercise, one cannot borrow strength by swinging the legs left and right, that is, the included angle θ5 between the two legs and the midpoint of the hip is less than the preset included angle value.

[0120] Indicator Four is specifically: During the exercise, when reaching the highest point, the exerciser's mandible must be higher than the horizontal bar, that is, θ6 is less than 180 degrees, and the ordinate of the mandible should be higher than the coordinates of the two wrists.

[0121] Indicator Five is specifically: During the exercise, the process from grasping the bar to rising and then lowering to the lowest point by the exerciser cannot exceed the preset time.

[0122] It should be noted that when judging whether a chin-up exercise is standardized, it is usually judged according to some standards, such as whether there is borrowing strength by body swinging during the chin-up process, and the chin not exceeding the horizontal bar when reaching the highest point, etc. The setting of the exercise standard indicators can not only reduce sports injuries caused by non-standard actions of athletes, but also provide personalized improvement suggestions for athletes to help them achieve the ideal training effect faster.

[0123] In a possible implementation, the preset correction rules specifically include: Rule 1 to Rule 5.

[0124] Rule 1 is specifically: By evaluating the degree of arm bending and the body lowering situation, different degrees of reminders are output.

[0125] Rule 2 is specifically: Classify and feedback the degree of leg bending to help avoid relying on strength.

[0126] Rule 3 is specifically: Check the angle between the midpoint of the two legs and the hip to ensure the standard of the movement.

[0127] Rule 4 is specifically: Ensure that the position of the lower jaw is higher than the horizontal bar and adjust the rising height.

[0128] Rule 5 is specifically: Based on the movement speed, prevent the movement from being too slow or non-standard.

[0129] It should be noted that the preset correction rules can detect non-standard movements in real time when the athlete is doing pull-ups and give timely feedback, avoiding the delay of relying on manual observation, improving the efficiency of correction. By systematically correcting non-standard movements, the preset correction rules can effectively reduce sports injuries caused by incorrect movements and ensure the safety of athletes during training.

[0130] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0131] In the embodiment of the present invention, by obtaining the real-time video data of the pull-up to be corrected, then, through the pose recognition algorithm, the bone point coordinate data in the pull-up video data is extracted, so as to ensure the accurate capture of the athlete's movements. Further, the missing value filling process is performed on the bone point coordinate data, and a joint state space model is established according to the bone point coordinate data, and the target angle is predicted by combining the Kalman filter algorithm and the curve fitting algorithm, providing a scientific basis for action correction. Finally, by comparing the predicted target angle with the standard action, a correction prompt is given in time. Through precise pose analysis and intelligent algorithms, non-standard movements are monitored and corrected in real time, improving the training effect and reducing sports injuries. According to the correction rules, correction feedback is given to the user in real time to assist the user to better independently complete the pull-up exercise training, meeting the personalized requirements of different exercisers.

[0132] Refer to the attached Figure 2 illustrates the structural schematic diagram of an intelligent online correction system for pull-ups provided by the present invention.

[0133] The present invention also provides an intelligent online correction system 20 for pull-ups, which is applied to the above-mentioned intelligent online correction method for pull-ups, including:

[0134] A processor 201.

[0135] A memory 202 stores computer-readable instructions thereon. When the computer-readable instructions are executed by a processor 201, an intelligent online correction method for pull-ups as in the method embodiment is implemented.

[0136] The intelligent online correction system 20 for pull-ups provided by the present invention can execute the above-mentioned intelligent online correction method for pull-ups and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0137] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0138] In the embodiments of the present invention, by acquiring real-time video data of pull-ups to be corrected, and then, through a pose recognition algorithm, extracting the bone point coordinate data in the pull-up video data, so as to ensure accurate capture of the athlete's movements. Further, missing value filling processing is performed on the bone point coordinate data, and a joint state space model is established based on the bone point coordinate data, and the target angle is predicted by combining the Kalman filtering algorithm and the curve fitting algorithm, providing a scientific basis for action correction. Finally, by comparing the predicted target angle with the standard action, a correction prompt is given in a timely manner. Through precise pose analysis and intelligent algorithms, non-standard actions are monitored and corrected in real time, the training effect is improved and sports injuries are reduced, and correction feedback is given to the user in real time according to the correction rules, assisting the user to better independently complete the pull-up exercise training, meeting the personalized requirements of different exercisers.

[0139] It should be understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0140] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0141] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, or magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0142] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be understood specifically by referring to the context before and after.

[0143] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or plural.

[0144] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0145] Those of ordinary skill in the art will realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0146] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0147] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0148] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0150] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0151] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the intelligent online correction method for pull-ups as in the method embodiment.

[0152] The computer-readable storage medium provided by the present invention can implement the steps and effects of the intelligent online correction method for pull-ups in the above method embodiment. To avoid repetition, the present invention will not elaborate further.

[0153] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0154] In the embodiment of the present invention, by acquiring real-time video data of pull-ups to be corrected, and then, through a pose recognition algorithm, extracting the bone point coordinate data in the pull-up video data, so as to ensure accurate capture of the athlete's movements. Further, missing value filling processing is performed on the bone point coordinate data, and a joint state space model is established based on the bone point coordinate data, and the target angle is predicted by combining the Kalman filter algorithm and the curve fitting algorithm, providing a scientific basis for action correction. Finally, by comparing the predicted target angle with the standard action, a correction prompt is given in a timely manner. Through accurate pose analysis and intelligent algorithms, non-standard actions are monitored and corrected in real time, improving the training effect and reducing sports injuries. According to the correction rules, correction feedback is given to the user in real time to assist the user in better independently completing the pull-up exercise training, meeting the personalized requirements of different exercisers.

[0155] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0156] The following points need to be explained:

[0157] (1) The accompanying drawings of the embodiments of the present invention only relate to the structures involved in the embodiments of the present invention, and other structures can refer to the general design.

[0158] (2) For clarity, in the accompanying drawings used to describe the embodiments of the present invention, the thickness of layers or regions is enlarged or reduced, that is, these drawings are not drawn to actual scale. It can be understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element can be "directly" on or under the other element or there can be intermediate elements.

[0159] (3) Without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0160] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An intelligent online correction method for pull-ups, characterized in that: include: S1: Acquire real-time pull-up video data to be corrected; S2: extracting the coordinate data of the skeleton points in the pull-up video data through a posture recognition algorithm; S3: performing missing value filling processing on the coordinate data of the skeleton points; S4: establishing a joint state space model according to the bone point coordinate data after the padding process; S5: predicting the target angle according to the joint state space model by combining a Kalman filter algorithm and a curve fitting algorithm; S6: When there is a predicted target angle that does not meet the standard index of the pull-up movement, a correction prompt is outputted through a preset correction rule.

2. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The posture recognition algorithm is specifically: lightweight OpenPose algorithm; Extracting the skeleton point coordinate data by the lightweight OpenPose algorithm specifically includes: S201: extracting image frames from the pull-up video data based on the OpenPose algorithm; S202: Extracting image features of the image frame through a lightweight convolutional network: F(t)=MobileNet(I(t)) Among them, F(t) represents the image features at time t, MobileNet represents a lightweight convolutional neural network, and I(t) represents the image frame at time t; S203: Outputting the heat map of each bone point through the target point regression algorithm according to the image features: H k (t)=Heatmap(F(t),k),k∈{1,2,...,N} Among them, H k (t) represents the heat map of the kth bone point at time t, Heatmap(F(t),k) represents the heat map generation function, which is used to extract the heat map of the kth bone point from the input image feature F(t), k = 1, 2, ..., N, N represents the total number of bone points; S204: Extract the maximum value position of each heat map to generate the bone point coordinate data: (x k (t),y k (t))=argmax(H k (t)) Among them, x k (t) represents the horizontal coordinate of the kth bone point at time t, y k (t) represents the ordinate of the kth bone point at time t, and argmax represents the independent variable with the maximum value.

3. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The S4 specifically includes: S401: Calculate the distance between the coordinates of each bone point: Among them, a represents the distance between bone point B and bone point C, b represents the distance between bone point A and bone point C, and c represents the distance between bone point A and bone point B. (x a ,y a ) represents the coordinates of the bone point A, (x b ,y b ) represents the coordinates of the bone point B, (x c ,y c ) represents the coordinates of the skeleton point C; S402: Calculate the target angle according to the distance using the law of cosines: Wherein, cosα represents the cosine value of the target angle α, and arccos represents the inverse cosine operation; S403: According to the target angle, combined with physics knowledge, establish the motion equation of the target angle: ω(k)=ω(k-1)+β(k-1)·Δt (x k (t),y k (t))=argmax(H k (t)) Among them, θ(k) represents the angle value of the athlete at the kth moment, ω(k-1) represents the angular velocity of the athlete at the k-1th moment, Δ represents the increment operator, β(k-1) represents the angular acceleration of the athlete at the k-1th moment, and ω(k) represents the angular velocity of the athlete at the k-1th moment; S404: Based on the motion equation, establish the joint state space model.

4. The intelligent online correction method for pull-ups according to claim 1, characterized in that: When the target angle is a periodically changing angle, the state space model is specifically: Among them, Θ i (k) represents the state vector of the i-th target angle at the k-th moment, θ i (k) represents the position of the i-th target angle at the k-th moment, ω i (k) represents the angular velocity of the i-th target angle at the k-th moment, A i represents the state transfer matrix of the i-th target angle, B i represents the control input matrix of the i-th target angle, U i (k-1) represents the control input of the i-th target angle at the k-1th time, ni represents the noise of the i-th target angle, T represents the time step, β i (k-1) represents the angular acceleration of the i-th target angle at the k-1th moment; The output equation is: Among them, Y i (k) and y i (k) represents the predicted value of the i-th target angle, H i Represents the observation matrix of the i-th target angle; When the target angle is a stable angle, the state space model is specifically: The output equation is: Y i (k)=y i (k)=θ i (k)。 5. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The target angles specifically include: angle θ1 with the left elbow as the vertex, the left wrist and the left shoulder as the two endpoints; angle θ2 with the right elbow as the vertex, the right wrist and the right shoulder as the two endpoints; angle θ3 with the left knee as the vertex, the left ankle and the left hip as the two endpoints; angle θ4 with the right knee as the vertex, the right ankle and the right hip as the two endpoints; angle θ5 with the jaw as the vertex, the left and right wrists as the two endpoints; angle θ6 with the midpoint of the left and right hips as the vertex, the left and right ankles as the two endpoints.

6. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The curve fitting algorithm includes a linear fitting algorithm and a sinusoidal curve fitting algorithm; For the prediction of θ1 and θ2, the Kalman filter algorithm and the sinusoidal curve fitting algorithm are combined to perform the prediction; For the prediction of θ3, θ4 and θ6, the Kalman filter algorithm and the linear fitting algorithm are combined to make predictions; The prediction of θ5 is performed using the Kalman filter algorithm.

7. The intelligent online correction method for pull-ups according to claim 6, characterized in that: Combining the Kalman filter algorithm and the sinusoidal curve fitting algorithm to perform prediction specifically includes: When the angle change of θ1 or the angle change of θ2 is greater than the preset threshold, the sinusoidal curve fitting algorithm is used to predict: i k =Asin(ωt k +φ)+D Among them, θ k represents the target angle value at time k, A represents the amplitude, ω represents the frequency, and t k represents the k moment, φ represents the phase shift, and D represents the overall angle shift; Otherwise, the Kalman filter algorithm is used for prediction; Combining the Kalman filter algorithm and the line fitting algorithm to perform prediction specifically includes: When the angle change of θ3, θ4 or θ6 is less than the preset threshold, the linear fitting algorithm is used to make a prediction: θ k =at k +b Among them, θ k represents the target angle value at time k, a represents the slope of the angle change, and t k represents the k moment, and b represents the initial angle; Otherwise, the Kalman filter algorithm is used for prediction; Prediction by the Kalman filter algorithm specifically includes: Based on the Kalman filter state prediction equation, the target angle value is calculated: in, represents the state prediction value at the kth moment, that is, the target angle value, represents the state estimate at the k-1th moment, A represents the state transfer matrix, B represents the control input matrix, u k-1 represents the control vector at the k-1th moment, represents the state prediction error covariance matrix at the kth moment, P k-1 represents the state estimation error covariance matrix at the k-1th moment, Q represents the process noise covariance matrix; The target angle value and the state prediction error covariance matrix are updated: Among them, K k represents the Kalman gain at the kth moment, H represents the measurement matrix, R represents the measurement noise covariance matrix, represents the estimated state value at the kth moment, z k represents the measurement vector at the kth moment, P k Represents the state estimation error covariance matrix at the kth moment after measurement update.

8. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The sports standard indicators specifically include: indicators one to five; The first indicator is specifically: during the exercise, when reaching the lowest point, the angles θ1 and θ2 of the two arms with the elbows as the apex should be close to 180 degrees; The second indicator is specifically: during the exercise, the athlete cannot use the force of the legs to push forward and backward, and the angles θ3 and θ4 of the two legs with the knee as the vertex must be close to 180 degrees; The third indicator is specifically: during the exercise, the force cannot be gained by swinging the legs left and right, that is, the angle θ5 between the two legs and the midpoint of the hips is less than the preset angle value; The fourth indicator is specifically: during the exercise, when reaching the highest point, the athlete's lower jaw must be higher than the horizontal bar, that is, θ6 is less than 180 degrees, and the longitudinal coordinate of the lower jaw must be higher than the coordinates of both wrists; The fifth indicator is specifically: during the exercise, the time from the athlete starting to rise from grabbing the pole to the athlete descending to the lowest point cannot exceed the preset time.

9. The intelligent online correction method for pull-ups according to claim 1, characterized in that: The preset correction rules specifically include: Rule 1 to Rule 5; Rule 1 is as follows: Output different levels of reminders by evaluating the degree of arm bending and body lowering; Rule 2 is as follows: provide classified feedback on the degree of leg bending to help avoid leveraging; Rule three is as follows: Check the angle between the legs and the midpoint of the hips to ensure the movement is standard; Rule 4 is as follows: Make sure your chin is above the bar and adjust the height of the rise; Rule five is specifically: Based on the speed of movement, prevent movements that are too slow or irregular.

10. An intelligent online correction system for pull-ups, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the intelligent online correction method for pull-ups as described in any one of claims 1 to 9 is implemented.

Images