Badminton action data analysis method and system in combination with big data

Abstract

The invention relates to the technical field of motion analysis, and particularly discloses a badminton motion data analysis method and system combined with big data, which are used for capturing joint angle, speed and acceleration information of an athlete in real time, and calculating a consistency coefficient on a time sequence by using continuous wavelet transform and a phase synchronization coefficient to ensure the accuracy and reliability of the data. Based on accurate data, extracting short-term fine adjustment features and long-term trend features, constructing comprehensive time dimension features, extracting joint angle change features and body posture features, forming comprehensive space dimension features, and integrating the joint angle change features and the body posture features into comprehensive feature representation; the method comprises the following steps: inputting a gesture recognition score of an athlete into a clustering model, classifying the gesture of the athlete based on a scoring result, providing a real-time feedback instruction based on a space-time position for a serious abnormal gesture, guiding the athlete to adjust the action, and establishing a closed-loop feedback system to monitor a correction effect.

Description

Technical Field

[0001] The present invention relates to the technical field of motion analysis, and in particular to a badminton motion data analysis method and system combined with big data. Background Art

[0002] Badminton is a widely popular sport, and the mastery and training of movement techniques significantly impact an athlete's overall performance. The development of big data technology has opened up new possibilities for the collection, analysis, and application of badminton data. One study proposed a "wireless intelligent motion correction system" based on a "triple friction nanogenerator" to monitor an athlete's pedaling frequency and force in real time. This data analysis provides feedback to coaches to help improve their technique. The use of "big data stream" analysis technology to obtain train braking performance information for optimizing railway operations has also been explored. These studies demonstrate the broad application prospects of "big data" technology in sports training monitoring and optimization.

[0003] The existing technology has the following deficiencies:

[0004] Athletes' motion data contains information at multiple temporal and spatial scales, such as short-term subtle adjustments and long-term overall trends. Effectively fusing these multi-scale features so that they complement rather than interfere with each other during feature extraction is a challenging problem. Failure to properly fuse these multi-scale features can lead to overlooking important movement details or overfitting, resulting in poor performance when exposed to new data and an inability to provide effective personalized guidance. Summary of the Invention

[0005] The purpose of the present invention is to provide a badminton motion data analysis method and system combined with big data to solve the above-mentioned problems.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The badminton action data analysis method combined with big data includes the following steps:

[0008] S1: Capture the athlete's motion data in real time and analyze the accuracy of the motion data to ensure the accuracy of the data;

[0009] The motion data includes: athlete's joint angle, speed, and acceleration information;

[0010] S2: Based on accurate motion data, extract short-term fine-tuning features for the data in each small window, merge multiple consecutive small windows into a larger window, and extract long-term trend features. The short-term fine-tuning features and long-term trend features are constructed into a comprehensive time dimension feature.

[0011] S3: Based on accurate motion data, the joint angle change features and body posture features are extracted by performing geometric transformation and topological analysis on the motion data, and the joint angle change features and body posture features are constructed into comprehensive spatial dimension features;

[0012] S4: constructing the comprehensive feature representation of the integrated time dimension features and the comprehensive space dimension features as the input of the action model, and outputting the athlete's posture recognition score according to the action model;

[0013] S5: Based on the output of the model, the athlete’s posture recognition is divided into normal posture, slightly abnormal posture, and seriously abnormal posture;

[0014] S6: Provide early warning and correction based on serious abnormal posture.

[0015] As a further solution of the present invention: the real-time capture of the athlete's motion data and the accuracy analysis of the motion data are performed to ensure the accuracy of the data;

[0016] The athlete's motion data, including the athlete's joint angle, speed, and acceleration information, are collected in real time. The consistency of the motion data in the time series is analyzed. Based on the mutual relationship and error range of the motion data, the consistency coefficient of the motion data in the time series is calculated. It is determined whether the consistency coefficient of the motion data is greater than or equal to the preset threshold. If so, it is recorded as accurate motion data. If not, it is recorded as inaccurate motion data.

[0017] As a further solution of the present invention: the logic for obtaining the consistency coefficient is:

[0018] Collect athlete's motion data in real time, including joint angle, velocity and acceleration information, organize it into time series and record the value at each moment;

[0019] The time series is analyzed at multiple scales by continuous wavelet transform to generate wavelet transform coefficients reflecting signal characteristics, which contain signal amplitude and phase information.

[0020] Phase information is extracted from the wavelet transform results, the phase differences among joint angles, velocities, and accelerations are calculated, and the degree of synchronization between the signals is analyzed based on the phase differences to obtain the phase synchronization coefficient.

[0021] Combining the signal amplitude and phase synchronization coefficients of different data dimensions, the consistency of the time series at each scale and time point is weighted and calculated to obtain the time series consistency coefficient of the motion data;

[0022] The calculation expression of the synchronization coefficient is:

[0023]

[0024] Where R(a,b) represents the phase synchronization coefficient, N represents the total length of the time series, t represents the time acquisition point, j represents the imaginary unit, and Δφ xy (a, b) represents the phase difference between the joint angle signal and the velocity signal;

[0025] The calculation expression of the consistency coefficient is:

[0026]

[0027] Where C represents the consistency coefficient, A represents the range of wavelet scale, a represents the scale parameter, b represents the translation parameter, and W x (a, b) represents the wavelet coefficients of the joint angle, W y (a, b) represents the wavelet coefficients of velocity, W z (a, b) represents the wavelet coefficients of acceleration, |W x (a, b)| represents the amplitude of the wavelet coefficient of the joint angle, |W y (a,b)| represents the amplitude of the wavelet coefficient of velocity, |W z (a,b)| represents the amplitude of the wavelet coefficient of acceleration.

[0028] As a further solution of the present invention: the process of obtaining the comprehensive time dimension feature is as follows:

[0029] The real-time motion data is divided into multiple small windows according to fixed time intervals. Each small window contains the time series data of joint angle, velocity and acceleration.

[0030] For each small window, calculate the short-term fine-tuning features of the action data, including the statistical characteristics of the change range, first-order difference average, and volatility;

[0031] Combine multiple consecutive small windows into a larger window, and use recursive aggregation within the larger window to extract the most significant short-term features based on the time series change trend and construct long-term trend features;

[0032] The short-term fine-tuning features and long-term trend features are directly spliced ​​in chronological order to form comprehensive time dimension features.

[0033] As a further solution of the present invention: the process of acquiring the comprehensive spatial dimension features is as follows:

[0034] The joint angle change data is embedded into the three-dimensional Euclidean space. The local curvature, first-order derivative curvature, second-order derivative curvature and third-order derivative curvature of the trajectory are calculated by constructing a local coordinate system to extract high-order geometric change features.

[0035] A topological graph model is constructed based on the joint trajectory point cloud, and topological invariants are extracted through persistent homology analysis to characterize the global topological characteristics of the body posture.

[0036] Combining the results of high-order curvature analysis and topological invariant features, joint angle change features and body posture features are extracted, and comprehensive spatial dimension features are constructed by directly splicing time series.

[0037] As a further solution of the present invention, the method of constructing a comprehensive feature representation from the integrated time dimension features and the integrated space dimension features as the action model input and outputting the athlete's posture recognition score based on the action model specifically includes:

[0038] Based on the comprehensive time dimension features and the comprehensive space dimension features, a motion model is constructed. The motion model is a clustering model. The model determines the local density distribution and global connectivity of the feature points by analyzing the distribution characteristics of each feature point in the multidimensional feature space and the relative relationship between the points, thereby determining the cluster center point and the corresponding feature point set;

[0039] For each cluster center point, the dynamic impact factor is calculated based on the local distribution of the surrounding feature points. The dynamic impact factor reflects the influence of the feature point on the cluster center.

[0040] Combining the dynamic impact factor of each feature point with the local density information of the cluster center, the overall distribution characteristics of the cluster are comprehensively evaluated, and the characteristic relationship between different clusters is further analyzed;

[0041] The comprehensive evaluation results are converted into the athlete's posture recognition score.

[0042] As a further solution of the present invention: the specific calculation process of the posture recognition score is:

[0043] Calculate the distance between feature points. The calculation expression is:

[0044]

[0045] Where w k represents the local distribution weight of the k-th dimension feature, k represents the number of dimensions of the feature, D cd represents the distance between the cth sample point and the dth sample point, Indicates the value of feature k at the cth sample point, The value of feature k at the dth sample point, α represents the nonlinear expansion coefficient, and c and d represent the sample points;

[0046] Based on the distance between feature points, combined with the dynamic changes of the features in the comprehensive time dimension and the comprehensive space dimension, the dynamic impact factor of each cluster center point is calculated. The calculation expression is:

[0047]

[0048] Among them, γ c Represents the local density of the c-th center point, and the calculation expression is: γ c = σ represents the kernel function parameter;

[0049] Based on the dynamic impact factors, the model adaptive optimization strategy is used to determine the label of each cluster and complete the classification, forming the basic structure of the action model;

[0050] According to the clustering results, the posture recognition score is calculated, and the calculation expression is:

[0051]

[0052] Where N c represents the number of cluster centers, |C c | represents the number of samples in the cth cluster, δ d Denotes the dynamic impact factor of the d-th sample, D df Represents the distance between feature points, S p represents the posture recognition score, d and f represent sample points, and e represents the logarithm of the natural number base.

[0053] As a further solution of the present invention: the athlete's posture recognition is divided into normal posture, slightly abnormal posture, and seriously abnormal posture according to the output of the model, specifically including:

[0054] The athlete's posture recognition score is compared with the first threshold. If the athlete's posture recognition score is greater than or equal to the first threshold, it is recorded as a normal posture. If the athlete's posture recognition score is less than the first threshold, the athlete's posture recognition score is compared with the second threshold. If the athlete's posture recognition score is greater than or equal to the second threshold, it is recorded as a slightly abnormal posture. If the athlete's posture recognition score is less than the second threshold, it is recorded as a seriously abnormal posture.

[0055] As a further solution of the present invention: the aforementioned early warning and correction based on serious abnormal posture specifically includes:

[0056] For athletes with severe abnormal posture, real-time feedback instructions based on time and space position are provided to athletes. The feedback instructions are displayed in real time through wearable devices and virtual reality training environments to guide athletes to make adjustments;

[0057] By monitoring the effect of the corrective action in real time, the athlete's posture recognition score is re-output through the action model and compared with the first threshold. If it is greater than or equal to the first threshold, it means that the correction is successful. The correction intensity is adjusted according to the athlete's adaptability to form a self-optimizing correction process.

[0058] The badminton action data analysis system combined with big data includes:

[0059] A data acquisition module, wherein the data acquisition module is used to capture the athlete's movement data in real time;

[0060] A data accuracy analysis module, which is used to perform accuracy analysis on the action data to ensure the accuracy of the data;

[0061] A data feature extraction module, based on accurate motion data, extracts short-term fine-tuning features for the data in each small window, merges multiple consecutive small windows into a larger window, and extracts long-term trend features; constructs the short-term fine-tuning features and long-term trend features into comprehensive time dimension features, extracts joint angle change features and body posture features by performing geometric transformation and topological analysis on the motion data, and constructs the joint angle change features and body posture features into comprehensive spatial dimension features;

[0062] An action model construction and recognition module, which constructs a comprehensive feature representation from the comprehensive temporal dimension features and the comprehensive spatial dimension features as an action model input and outputs an athlete's posture recognition score based on the action model;

[0063] Athlete posture classification module, which classifies the athlete's posture recognition according to the output of the model, including normal posture, slightly abnormal posture, and seriously abnormal posture;

[0064] The correction and warning module performs warning and correction based on serious abnormal posture.

[0065] Beneficial effects of the present invention:

[0066] (1) The present invention captures the joint angle, velocity and acceleration information of athletes during exercise in real time and accurately. By combining continuous wavelet transform with the calculation method of phase synchronization coefficient, the consistency analysis of time series data is performed to ensure the high accuracy and reliability of the collected data. Furthermore, by performing geometric transformation and topological analysis on these raw data, the subtle joint angle change characteristics and body posture characteristics are extracted to construct comprehensive spatial dimension characteristics. This multi-level data processing method can not only accurately describe the athlete's movement pattern, but also identify potential technical defects or irregular movements. Based on these detailed and accurate analysis results, coaches and athletes can scientifically evaluate movement posture, timely discover and correct subtle movement deviations, thereby optimizing training effects and significantly improving sports performance. More importantly, this method helps prevent chronic injuries caused by long-term maintenance of improper postures, and ensures the health of athletes and the sustainable development of their careers. By organically combining modern sensing technology and complex data analysis methods, the present invention provides unprecedented technical support and scientific basis for badminton training, and opens a new era of personalized and refined training.

[0067] (2) The present invention not only covers advanced data collection and analysis technologies, but also innovatively introduces a closed-loop feedback system to achieve real-time monitoring and dynamic adjustment of athletes' movements. When a serious abnormal posture is detected, the system can instantly generate precise feedback instructions based on time and space position, and intuitively convey these adjustments to the athletes through wearable devices or virtual reality training environments. In addition, the system has an intelligent adjustment function that can automatically optimize the correction intensity and frequency according to the specific adaptation and training progress of each athlete, forming a process of self-iteration and continuous improvement. This highly personalized training guidance method ensures that each athlete can obtain a training plan that best suits their needs, improves training efficiency and effectiveness, and promotes continuous progress and development in technology. At the same time, coaches can formulate more accurate and effective training plans through the detailed data analysis and scientific basis provided by the system, thereby achieving highly customized and scientific management of the training process. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The present invention will be further described below with reference to the accompanying drawings.

[0069] Figure 1 This is a flowchart of the specific steps of the badminton action data analysis method combined with big data of the present invention;

[0070] Figure 2 This is a flowchart of the badminton action data analysis system combined with big data in the present invention. DETAILED DESCRIPTION

[0071] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0072] See also Figure 1 As shown, the present invention is a badminton action data analysis method combined with big data, comprising the following steps:

[0073] S1: Capture the athlete's motion data in real time and analyze the accuracy of the motion data to ensure the accuracy of the data;

[0074] The motion data includes: athlete's joint angle, speed, and acceleration information;

[0075] S2: Based on accurate motion data, extract short-term fine-tuning features for the data in each small window, merge multiple consecutive small windows into a larger window, and extract long-term trend features. The short-term fine-tuning features and long-term trend features are constructed into a comprehensive time dimension feature.

[0076] S3: Based on accurate motion data, the joint angle change features and body posture features are extracted by performing geometric transformation and topological analysis on the motion data, and the joint angle change features and body posture features are constructed into comprehensive spatial dimension features;

[0077] S4: constructing the comprehensive feature representation of the integrated time dimension features and the comprehensive space dimension features as the input of the action model, and outputting the athlete's posture recognition score according to the action model;

[0078] S5: Based on the output of the model, the athlete’s posture recognition is divided into normal posture, slightly abnormal posture, and seriously abnormal posture;

[0079] S6: Provide early warning and correction based on serious abnormal posture.

[0080] In S1, the athlete's motion data is captured in real time and analyzed for accuracy to ensure the accuracy of the data, including:

[0081] This invention uses a self-powered sensor called a "triple triboelectric nanogenerator" and integrates it into sports equipment to collect real-time data on athletes' movements, including joint angles, speeds, and accelerations. The sensor requires no external power supply, making it easy for athletes to wear.

[0082] Collecting the athlete's motion data in real time, including the athlete's joint angle, velocity, and acceleration information, analyzing the consistency of the motion data in time series, calculating the consistency coefficient of the motion data in time series based on the mutual relationship and error range of the motion data, and determining whether the consistency coefficient of the motion data is greater than or equal to a preset threshold. If so, it is recorded as accurate motion data; if not, it is recorded as inaccurate motion data;

[0083] The logic for obtaining the consistency coefficient is:

[0084] Collect the athlete's joint angle, velocity and acceleration information in real time and construct a time series: S = {x t ,y t ,z t}, where t represents the time acquisition point, x t Indicates the value of the joint angle at time t, y t represents the value of velocity at time t, z t represents the value of acceleration at time t;

[0085] For the time series S={x t ,y t ,z t} Perform continuous wavelet transform to obtain wavelet coefficients W x (a,b),W y (a,b),W z (a,b), the calculation expression is:

[0086]

[0087] Where a represents the scale parameter, b represents the translation parameter, ψ represents the mother wavelet function, and W x (a, b) represents the wavelet coefficients of the joint angle, W y (a, b) represents the wavelet coefficients of velocity, W z (a, b) represent the wavelet coefficients of acceleration;

[0088] The phase information of each dimension is extracted by wavelet transform, and the calculation expression is:

[0089] φ x (a,b)=arg(W x (a,b));

[0090] φ y (a,b)=arg(W y (a,b));

[0091] φ z (a,b)=arg(W z (a,b));

[0092] Where, φ x (a, b) represents the phase information of the joint angle, φ y (a, b) represents the phase information of velocity, φ z (a, b) represents the bit information of acceleration, arg represents the phase angle operation of complex number;

[0093] Calculate the phase difference, the calculation expression is:

[0094] Δφ xy (a,b)=φ x (a,b)-φ y (a,b);

[0095] Δφ yz (a,b)=φ y (a,b)-φ z (a,b);

[0096] Where Δφ xy (a, b) represents the phase difference between the joint angle signal and the velocity signal, Δφ yz (a, b) represents the phase difference between the velocity signal and the acceleration signal;

[0097] Construct the phase synchronization coefficient, and the calculation expression is:

[0098]

[0099] Where R(a,b) represents the phase synchronization coefficient, N represents the total length of the time series, t represents the time acquisition point, j represents the imaginary unit, and Δφ xy (a, b) represents the phase difference between the joint angle signal and the velocity signal;

[0100] Combining the wavelet coefficient and the phase synchronization coefficient, the consistency coefficient of the time series is calculated. The calculation expression is:

[0101]

[0102] Where C represents the consistency coefficient, A represents the range of wavelet scale, a represents the scale parameter, b represents the translation parameter, and W x (a, b) represents the wavelet coefficients of the joint angle, W y (a, b) represents the wavelet coefficients of velocity, W z (a, b) represents the wavelet coefficients of acceleration, |W x (a, b)| represents the amplitude of the wavelet coefficient of the joint angle, |W y (a,b)| represents the amplitude of the wavelet coefficient of velocity, |W z (a,b)| represents the amplitude of the wavelet coefficient of acceleration.

[0103] In S2, based on accurate action data, short-term fine-tuning features are extracted for the data in each small window. Multiple consecutive small windows are merged into a larger window to extract long-term trend features. The short-term fine-tuning features and long-term trend features are constructed into a comprehensive time dimension feature, specifically including:

[0104] The real-time motion data is divided into multiple small windows according to fixed time intervals. Each small window contains the time series data of joint angle, velocity and acceleration.

[0105] For each small window, the short-term variation characteristics of joint angle, velocity, and acceleration are calculated, including the variation range (maximum value minus minimum value), first-order difference average value, and volatility (ratio of standard deviation to mean), and a short-term feature vector is constructed.

[0106] Through recursive aggregation, multiple consecutive small windows are combined into a larger window. Within the larger window, the most significant short-term features (features with the largest change amplitude) are selected according to the changing trend of the time series data and extracted as long-term trend features.

[0107] In a larger window, short-term fine-tuning features and long-term trend features are directly spliced ​​into a comprehensive time dimension feature vector in chronological order to fully retain short-term and long-term feature information and ensure the structural integrity of the data;

[0108] Use the sliding window method to gradually move the boundaries of the larger window, introduce a new small window each time, recursively recalculate the short-term features and long-term trend features, and update the comprehensive time dimension features.

[0109] In S3, based on accurate motion data, joint angle change features and body posture features are extracted through geometric transformation and topological analysis of the motion data. These features are then transformed into comprehensive spatial dimensional features, specifically including:

[0110] Collect athletes' joint angles, speeds, and accelerations in real time, and perform interpolation, completion, and noise reduction on the collected data to ensure data accuracy and continuity;

[0111] Normalize the data to the same dimension for subsequent geometric analysis and topological modeling;

[0112] The joint angle change data is embedded into the three-dimensional Euclidean space to generate the joint trajectory point cloud corresponding to each time step, which is recorded as

[0113] In the formula, i represents the number of data points, and M represents the total number of data points;

[0114] By constructing a local coordinate system (based on the normal vector and tangent vector of the neighborhood point), the local curvature, second-order derivative curvature, and third-order derivative curvature of the joint trajectory are calculated to extract high-order geometric change features;

[0115] Construct a topological graph model G = (V, E) of the motion data based on the trajectory point cloud, where V represents the point set in the trajectory point cloud and E represents the connection relationship between the points;

[0116] Perform persistent homology analysis on the topological graph to extract the topological invariants of the trajectory (0-dimensional and 1-dimensional persistence barcodes) to describe the global topological characteristics of the body posture;

[0117] For each time step, the curvature change trajectory of the feature point is calculated based on the local curvature to generate a high-order curvature vector Among them, κ (i) represents the curvature vector, represents the first-order curvature, represents the second-order curvature, represents the third-order curvature;

[0118] In the topological graph model, the curvature differences and connection relationships between nodes are used to extract key turning points and stable segments in the trajectory and construct body posture features;

[0119] The joint angle change features and body posture features are directly spliced ​​into comprehensive spatial dimension features based on time series. A dynamic recursive method is used to gradually analyze the newly added trajectory data points. The comprehensive spatial dimension features are updated in real time by recalculating high-order curvature and topological invariants.

[0120] In S4, the comprehensive temporal dimension features and the comprehensive spatial dimension features are constructed into a comprehensive feature representation as the action model input. The athlete's posture recognition score is output based on the action model, specifically including:

[0121] Based on the comprehensive time dimension features and the comprehensive space dimension features, a motion model is constructed. The motion model is a clustering model. The model determines the local density distribution and global connectivity of the feature points by analyzing the distribution characteristics of each feature point in the multidimensional feature space and the relative relationship between the points, thereby determining the cluster center point and the corresponding feature point set;

[0122] For each cluster center point, the dynamic impact factor is calculated based on the local distribution of the surrounding feature points. The dynamic impact factor reflects the influence of the feature point on the cluster center.

[0123] Combining the dynamic impact factor of each feature point with the local density information of the cluster center, the overall distribution characteristics of the cluster are comprehensively evaluated, and the characteristic relationship between different clusters is further analyzed;

[0124] Convert the comprehensive evaluation results into the athlete's posture recognition score;

[0125] The specific calculation process of the posture recognition score is as follows:

[0126] Standardize the comprehensive time dimension features and comprehensive space dimension features separately to ensure that all feature dimensions have the same dimension;

[0127] A multimodal feature clustering model is constructed. The model is based on the local density distribution and global connectivity of the feature subspace. The improved density peak clustering algorithm is used to define the distance between feature points. The calculation expression is:

[0128]

[0129] Where w k represents the local distribution weight of the k-th dimension feature, k represents the number of dimensions of the feature, D cd represents the distance between the cth sample point and the dth sample point, Indicates the value of feature k at the cth sample point, The value of feature k at the dth sample point, α represents the nonlinear expansion coefficient, and c and d represent the sample points;

[0130] Based on the distance between feature points, combined with the dynamic changes of the features in the comprehensive time dimension and the comprehensive space dimension, the dynamic impact factor of each cluster center point is calculated. The calculation expression is:

[0131]

[0132] Among them, γ c Represents the local density of the c-th center point, and the calculation expression is: σ represents the kernel function parameter;

[0133] Based on the dynamic impact factors, the model adaptive optimization strategy is used to determine the label of each cluster and complete the classification, forming the basic structure of the action model;

[0134] According to the clustering results, the posture recognition score is calculated, and the calculation expression is:

[0135]

[0136] Where N c represents the number of cluster centers, |C c | represents the number of samples in the cth cluster, δ d Denotes the dynamic impact factor of the d-th sample, D df Represents the distance between feature points, S prepresents the posture recognition score, d and f represent sample points, and e represents the logarithm of the natural number base;

[0137] The comprehensive feature representation is input into the clustering algorithm model to obtain the athlete's posture recognition score.

[0138] In S5, the athlete's posture recognition is divided into normal posture, slightly abnormal posture, and seriously abnormal posture according to the output of the model, specifically including:

[0139] The athlete's posture recognition score is compared with the first threshold. If the athlete's posture recognition score is greater than or equal to the first threshold, it means that the athlete's posture is normal and is recorded as a normal posture. If the athlete's posture recognition score is less than the first threshold, the athlete's posture recognition score is compared with the second threshold. If the athlete's posture recognition score is greater than or equal to the second threshold, it means that the athlete's posture is slightly abnormal and is recorded as a slightly abnormal posture. If the athlete's posture recognition score is less than the second threshold, it means that the athlete's posture is seriously abnormal and needs to be adjusted in time, and is recorded as a seriously abnormal posture.

[0140] In S6, based on serious abnormal posture, early warning and correction are carried out, including:

[0141] For athletes with severe abnormal posture, real-time feedback instructions based on time and space position are provided to athletes. The feedback instructions are displayed in real time through wearable devices or virtual reality training environments to guide athletes to make adjustments. This feedback is provided not only through visual or auditory means, but also in combination with a tactile feedback system to remind athletes to make timely adjustments;

[0142] A closed-loop feedback system is established to monitor the effect of corrective actions in real time, re-output the athlete's posture recognition score through the action model, and compare it with the first threshold. If it is greater than or equal to the first threshold, it means that the correction is successful. The correction intensity is adjusted according to the athlete's adaptability to form a self-optimizing correction process.

[0143] See also Figure 2 As shown, the badminton action data analysis system combined with big data includes:

[0144] A data acquisition module, wherein the data acquisition module is used to capture the athlete's movement data in real time;

[0145] A data accuracy analysis module, which is used to perform accuracy analysis on the action data to ensure the accuracy of the data;

[0146] A data feature extraction module, based on accurate motion data, extracts short-term fine-tuning features for the data in each small window, merges multiple consecutive small windows into a larger window, and extracts long-term trend features; constructs the short-term fine-tuning features and long-term trend features into comprehensive time dimension features, extracts joint angle change features and body posture features by performing geometric transformation and topological analysis on the motion data, and constructs the joint angle change features and body posture features into comprehensive spatial dimension features;

[0147] An action model construction and recognition module, which constructs a comprehensive feature representation from the comprehensive temporal dimension features and the comprehensive spatial dimension features as an action model input and outputs an athlete's posture recognition score based on the action model;

[0148] Athlete posture classification module, which classifies the athlete's posture recognition according to the output of the model, including normal posture, slightly abnormal posture, and seriously abnormal posture;

[0149] Correction warning module, which provides warning and correction based on serious abnormal posture

[0150] The proposed method works by capturing an athlete's joint angle, velocity, and acceleration information in real time, and calculating the consistency coefficient over the time series using continuous wavelet transform and phase synchronization coefficients to ensure data accuracy. Next, based on accurate data, a combination of short and long window methods is used to extract short-term fine-tuning features and long-term trend features, constructing a comprehensive temporal dimension feature. Simultaneously, geometric transformations and topological analysis are performed on the motion data to extract joint angle variation features and body posture features, forming a comprehensive spatial dimension feature. These two features are then integrated into a comprehensive feature representation and input into a clustering model to generate a posture recognition score for the athlete. Based on the scoring results, the athlete's posture is classified into normal, slightly abnormal, and severely abnormal. For severely abnormal postures, real-time feedback instructions based on temporal and spatial location are provided, guiding the athlete to adjust their movements through wearable devices or a virtual reality training environment. A closed-loop feedback system is then established to monitor the effectiveness of the corrections, enabling a self-optimizing correction process. This technical solution not only improves the accuracy and efficiency of motion data analysis but also provides scientific evidence and technical support for athletes, helping to improve athletic performance and reduce injury risk. Throughout the entire process, through the fusion of multi-source data and the application of intelligent algorithms, we achieved an in-depth understanding and precise optimization of the details of badminton players' movements, reflecting the important role of modern technology in sports training.

[0151] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0152] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. 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 program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., 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 available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0153] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0154] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0155] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A badminton action data analysis method based on big data, characterized in that: The following steps are involved: S1: Capture the athlete's motion data in real time and analyze the accuracy of the motion data to ensure the accuracy of the data; The motion data includes: athlete's joint angle, speed, and acceleration information; S2: Based on accurate motion data, extract short-term fine-tuning features for the data in each small window, merge multiple consecutive small windows into a larger window, and extract long-term trend features. The short-term fine-tuning features and long-term trend features are constructed into a comprehensive time dimension feature. S3: Based on accurate motion data, the joint angle change features and body posture features are extracted by performing geometric transformation and topological analysis on the motion data, and the joint angle change features and body posture features are constructed into comprehensive spatial dimension features; S4: constructing the comprehensive feature representation of the integrated time dimension features and the comprehensive space dimension features as the input of the action model, and outputting the athlete's posture recognition score according to the action model; S5: Based on the output of the model, the athlete’s posture recognition is divided into normal posture, slightly abnormal posture, and seriously abnormal posture; S6: Provide early warning and correction based on serious abnormal posture.

2. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The real-time capture of the athlete's motion data and the accuracy analysis of the motion data are performed to ensure the accuracy of the data; The athlete's motion data, including the athlete's joint angle, speed, and acceleration information, are collected in real time. The consistency of the motion data in the time series is analyzed. Based on the mutual relationship and error range of the motion data, the consistency coefficient of the motion data in the time series is calculated. It is determined whether the consistency coefficient of the motion data is greater than or equal to the preset threshold. If so, it is recorded as accurate motion data. If not, it is recorded as inaccurate motion data.

3. The badminton action data analysis method combined with big data according to claim 2, characterized in that: The logic for obtaining the consistency coefficient is: Collect athlete's motion data in real time, including joint angle, velocity and acceleration information, organize it into time series and record the value at each moment; The time series is analyzed at multiple scales by continuous wavelet transform to generate wavelet transform coefficients reflecting signal characteristics, which contain signal amplitude and phase information. Phase information is extracted from the wavelet transform results, the phase differences among joint angles, velocities, and accelerations are calculated, and the degree of synchronization between the signals is analyzed based on the phase differences to obtain the phase synchronization coefficient. Combining the signal amplitude and phase synchronization coefficients of different data dimensions, the consistency of the time series at each scale and time point is weighted and calculated to obtain the time series consistency coefficient of the motion data; The calculation expression of the synchronization coefficient is: Where R(a,b) represents the phase synchronization coefficient, N represents the total length of the time series, t represents the time acquisition point, j represents the imaginary unit, and Δφ xy (a, b) represents the phase difference between the joint angle signal and the velocity signal; The calculation expression of the consistency coefficient is: Where C represents the consistency coefficient, A represents the range of wavelet scale, a represents the scale parameter, b represents the translation parameter, and W x (a, b) represents the wavelet coefficients of the joint angle, W y (a, b) represents the wavelet coefficients of velocity, W z (a, b) represents the wavelet coefficients of acceleration, |W x (a, b)| represents the amplitude of the wavelet coefficient of the joint angle, |W y (a,b)| represents the amplitude of the wavelet coefficient of velocity, |W z (a,b)| represents the amplitude of the wavelet coefficient of acceleration.

4. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The process of obtaining the comprehensive time dimension feature is as follows: The real-time motion data is divided into multiple small windows according to fixed time intervals. Each small window contains the time series data of joint angle, velocity and acceleration. For each small window, calculate the short-term fine-tuning features of the action data, including the statistical characteristics of the change range, first-order difference average, and volatility; Combine multiple consecutive small windows into a larger window, and use recursive aggregation within the larger window to extract the most significant short-term features based on the time series change trend and construct long-term trend features; The short-term fine-tuning features and long-term trend features are directly spliced ​​in chronological order to form comprehensive time dimension features.

5. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The process of obtaining the comprehensive spatial dimension features is as follows: The joint angle change data is embedded into the three-dimensional Euclidean space. The local curvature, first-order derivative curvature, second-order derivative curvature and third-order derivative curvature of the trajectory are calculated by constructing a local coordinate system to extract high-order geometric change features. A topological graph model is constructed based on the joint trajectory point cloud, and topological invariants are extracted through persistent homology analysis to characterize the global topological characteristics of the body posture. Combining the results of high-order curvature analysis and topological invariant features, joint angle change features and body posture features are extracted, and comprehensive spatial dimension features are constructed by directly splicing time series.

6. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The method constructs the comprehensive time dimension features and the comprehensive space dimension features into a comprehensive feature representation as the action model input, and outputs the athlete's posture recognition score according to the action model, specifically including: Based on the comprehensive time dimension features and the comprehensive space dimension features, a motion model is constructed. The motion model is a clustering model. The model determines the local density distribution and global connectivity of the feature points by analyzing the distribution characteristics of each feature point in the multidimensional feature space and the relative relationship between the points, thereby determining the cluster center point and the corresponding feature point set; For each cluster center point, the dynamic impact factor is calculated based on the local distribution of the surrounding feature points. The dynamic impact factor reflects the influence of the feature point on the cluster center. Combining the dynamic impact factor of each feature point with the local density information of the cluster center, the overall distribution characteristics of the cluster are comprehensively evaluated, and the characteristic relationship between different clusters is further analyzed; The comprehensive evaluation results are converted into the athlete's posture recognition score.

7. The badminton action data analysis method combined with big data according to claim 6, characterized in that: The specific calculation process of the posture recognition score is as follows: Calculate the distance between feature points. The calculation expression is: Where w k represents the local distribution weight of the k-th dimension feature, k represents the number of dimensions of the feature, D cd represents the distance between the cth sample point and the dth sample point, Indicates the value of feature k at the cth sample point, The value of feature k at the dth sample point, α represents the nonlinear expansion coefficient, and c and d represent the sample points; Based on the distance between feature points, combined with the dynamic changes of the features in the comprehensive time dimension and the comprehensive space dimension, the dynamic impact factor of each cluster center point is calculated. The calculation expression is: Among them, γ c Represents the local density of the c-th center point, and the calculation expression is: σ represents the kernel function parameter; Based on the dynamic impact factors, the model adaptive optimization strategy is used to determine the label of each cluster and complete the classification, forming the basic structure of the action model; According to the clustering results, the posture recognition score is calculated, and the calculation expression is: Where N c represents the number of cluster centers, |C c | represents the number of samples in the cth cluster, δ d Denotes the dynamic impact factor of the d-th sample, D df Represents the distance between feature points, S p represents the posture recognition score, d and f represent sample points, and e represents the logarithm of the natural number base.

8. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The output of the model is used to classify the athlete's posture recognition into normal posture, slightly abnormal posture, and seriously abnormal posture, specifically including: The athlete's posture recognition score is compared with the first threshold. If the athlete's posture recognition score is greater than or equal to the first threshold, it is recorded as a normal posture. If the athlete's posture recognition score is less than the first threshold, the athlete's posture recognition score is compared with the second threshold. If the athlete's posture recognition score is greater than or equal to the second threshold, it is recorded as a slightly abnormal posture. If the athlete's posture recognition score is less than the second threshold, it is recorded as a seriously abnormal posture.

9. The badminton action data analysis method combined with big data according to claim 1, characterized in that: The aforementioned early warning and correction based on serious abnormal postures specifically include: For athletes with severe abnormal posture, real-time feedback instructions based on time and space position are provided to athletes. The feedback instructions are displayed in real time through wearable devices and virtual reality training environments to guide athletes to make adjustments; By monitoring the effect of the corrective action in real time, the athlete's posture recognition score is re-output through the action model and compared with the first threshold. If it is greater than or equal to the first threshold, it means that the correction is successful. The correction intensity is adjusted according to the athlete's adaptability to form a self-optimizing correction process.

10. A badminton action data analysis system based on big data is characterized by: The badminton action data analysis method combined with big data as described in any one of claims 1 to 9 comprises: A data acquisition module, wherein the data acquisition module is used to capture the athlete's movement data in real time; A data accuracy analysis module, which is used to perform accuracy analysis on the action data to ensure the accuracy of the data; A data feature extraction module, based on accurate motion data, extracts short-term fine-tuning features for the data in each small window, merges multiple consecutive small windows into a larger window, and extracts long-term trend features; constructs the short-term fine-tuning features and long-term trend features into comprehensive time dimension features, extracts joint angle change features and body posture features by performing geometric transformation and topological analysis on the motion data, and constructs the joint angle change features and body posture features into comprehensive spatial dimension features; An action model construction and recognition module, which constructs a comprehensive feature representation from the comprehensive temporal dimension features and the comprehensive spatial dimension features as an action model input and outputs an athlete's posture recognition score based on the action model; Athlete posture classification module, which classifies the athlete's posture recognition according to the output of the model, including normal posture, slightly abnormal posture, and seriously abnormal posture; The correction and warning module performs warning and correction based on serious abnormal posture.

Images