Multi-dimensional data-driven volleyball teaching effect feedback method and system

Abstract

This application relates to the field of intelligent motion analysis technology, and discloses a multi-dimensional data-driven method and system for volleyball teaching effect feedback. The method includes: collecting real-time training motion and ball-hitting trajectory data to form a raw dataset; extracting features from the raw dataset and analyzing the training motion chain to obtain motion analysis results; if the motion deviation exceeds a threshold, classifying the data to determine a problem classification label; combining the problem classification label with subjective feelings and training load data to generate a personalized feedback report; monitoring subsequent training data according to the feedback report and updating the motion analysis results; comparing the motion analysis results before and after the update to calculate a continuity evaluation index; and generating a visualization chart based on the continuity evaluation index to determine a personalized training optimization path. This application solves the problems of delayed feedback, fragmented data, and lack of continuous evaluation in traditional volleyball teaching, improving the accuracy of volleyball motion diagnosis, the personalization level of training programs, and overall teaching efficiency.

Description

Technical Field

[0001] This application relates to the field of intelligent motion analysis technology, and in particular to a multi-dimensional data-driven method and system for providing feedback on volleyball teaching effectiveness. Background Technology

[0002] Volleyball, as a competitive sport demanding extremely high levels of precision, coordination, and explosive power, relies heavily on the accuracy of technical movements and the scientific nature of training load to ensure effective instruction. With the rapid development of sports technology, utilizing sensors, machine vision, and artificial intelligence algorithms to assist sports training has become a significant trend in the field of smart sports. Currently, volleyball instruction is transitioning from traditional experience-driven to data-driven approaches, urgently requiring technological means to comprehensively capture athletes' dynamic characteristics and provide quantitative feedback to support the construction of a scientific training system.

[0003] However, existing volleyball teaching methods have significant shortcomings in practical application, making it difficult to meet the refined requirements of high-level training. First, traditional teaching relies heavily on coaches' visual observation and experience-based judgment. This method is highly subjective and struggles to capture millisecond-level movement details. It often fails to accurately identify subtle deviations in complex dynamic processes such as swing sequence and lower limb power chains, leading to vague corrective suggestions and hindering athletes from developing proper muscle memory. Second, many existing auxiliary analysis methods are based on single data sources, such as video playback or single inertial sensor data, lacking multimodal data fusion analysis. Simple video analysis struggles to obtain internal mechanical parameters (such as joint torque and cushioning force), while single sensor data lacks intuitive verification of spatial trajectories. This results in a one-sided assessment of the coordination and continuity of each link in the movement chain, failing to comprehensively reflect the true quality of technical movements. Furthermore, existing technical solutions generally lack a closed-loop mechanism of "monitoring-feedback-re-evaluation," mostly performing static analysis only on single movements, failing to effectively correlate objective motion data with athletes' subjective feelings (such as fatigue and pain) and real-time training load. This disconnect makes feedback reports lack personalized guidance and prevents dynamic adjustments to training strategies based on athletes' real-time performance. Furthermore, existing technologies often neglect continuous tracking of long-term training data, lack quantitative assessment indicators for the extent of movement improvement, and struggle to generate visualized long-term training optimization paths. This leads to delayed adjustments to training plans and may even increase the risk of sports injuries due to undetected movement compensation, severely hindering the efficiency of volleyball instruction and the improvement of athletes' competitive levels.

[0004] To address the above shortcomings, this application combines wearable sensors, high-definition visual capture, and CNN and SVM artificial intelligence algorithms to construct a multi-dimensional data-driven closed-loop feedback system. This solves the problems of delayed feedback, fragmented data, and lack of continuous evaluation in traditional teaching, thereby improving the accuracy of volleyball movement diagnosis, the personalization level of training programs, and the overall teaching efficiency. Summary of the Invention

[0005] This application provides a multi-dimensional data-driven method and system for volleyball teaching effectiveness feedback, which solves the problems of delayed feedback, fragmented data, and lack of continuous evaluation in traditional teaching, and improves the accuracy of volleyball movement diagnosis, the personalization level of training programs, and the overall teaching efficiency.

[0006] Firstly, this application provides a multi-dimensional data-driven method for providing feedback on volleyball teaching effectiveness, the method comprising:

[0007] Step S1: Collect real-time motion data and ball trajectory data of volleyball players during training to form the original dataset;

[0008] Step S2: Extract features from the original dataset, analyze the training action chain, and obtain action analysis results;

[0009] Step S3: Based on the motion analysis results, determine whether the deviation of the training motion exceeds the preset deviation threshold. If so, classify the real-time motion data and ball trajectory data and determine the problem classification label.

[0010] Step S4: Based on the problem classification tags, integrate the volleyball player's subjective feeling records and training load data to generate a personalized feedback report;

[0011] Step S5: Based on the adjustment suggestions in the personalized feedback report, monitor the training data in real time during the subsequent training process, and update the action analysis results based on the training data obtained from the monitoring.

[0012] Step S6: Compare and analyze the action analysis results before and after the update, and calculate the continuity evaluation index;

[0013] Step S7: Generate a visualization chart based on the continuity evaluation index, and determine the personalized training optimization path for volleyball players based on the visualization chart.

[0014] Secondly, this application provides a multi-dimensional data-driven volleyball teaching effectiveness feedback system for implementing the aforementioned multi-dimensional data-driven volleyball teaching effectiveness feedback method. The system includes:

[0015] The data acquisition module is used to collect real-time motion data and ball trajectory data of volleyball players during training to form the raw dataset;

[0016] The feature extraction module is used to extract features from the original dataset, analyze the training action chain, and obtain action analysis results.

[0017] The action judgment module is used to determine whether the deviation of the training action exceeds a preset deviation threshold based on the action analysis results. If so, the real-time motion data and ball trajectory data are classified to determine the problem classification label.

[0018] The report generation module is used to generate personalized feedback reports by integrating the subjective feelings records and training load data of volleyball players based on the problem classification tags.

[0019] The real-time monitoring module is used to monitor the training data in the subsequent training process in real time according to the adjustment suggestions in the personalized feedback report, and update the action analysis results based on the training data obtained from the monitoring.

[0020] The comparison and evaluation module is used to compare and analyze the action analysis results before and after the update, and to calculate the continuity evaluation index.

[0021] The path planning module is used to generate a visualization chart based on the continuous evaluation index, and to determine the personalized training optimization path for volleyball players based on the visualization chart.

[0022] This application proposes a multi-dimensional data-driven method and system for volleyball teaching effectiveness feedback, solving the problems of delayed feedback, fragmented data, and lack of continuous evaluation in traditional teaching. It improves the accuracy of volleyball movement diagnosis, the personalization level of training programs, and overall teaching efficiency. Compared with existing technologies, the beneficial effects of this application's technical solution are at least as follows:

[0023] First, by collecting real-time motion data such as lower limb force, body coordination, and landing cushioning through wearable sensors, and combining it with ball trajectory data such as swing sequence and ball trajectory recorded by high-definition cameras, multimodal data fusion acquisition is achieved. This overcomes the limitation of a single data source being unable to simultaneously acquire internal mechanical parameters and spatial trajectory information, making the analysis of the coordination and coherence of each link in the action chain more comprehensive and accurate.

[0024] Secondly, convolutional neural networks are used to mine key features and quantify displacement, velocity, and angle parameters. Support vector machines are used to classify and identify the deviation values ​​of weak links, transforming vague motion perception into precise quantitative indicators. This allows coaches and athletes to clearly understand the specific links and degrees of motion deviations, providing actionable data for targeted correction.

[0025] Third, the motion analysis results are correlated with the athlete's subjective feelings and training load data to generate a personalized feedback report that includes a description of weaknesses and specific adjustment suggestions. This achieves an effective integration of objective motion data and subjective human feelings, making the feedback suggestions more in line with the athlete's actual condition and avoiding training programs that are not suitable due to ignoring individual differences.

[0026] Fourth, a closed-loop mechanism of monitoring, feedback, re-monitoring, and evaluation is established. By comparing the changes in motion parameters before and after training, weights are assigned according to the importance of each step, and the overall improvement is calculated. This is standardized into a continuous evaluation indicator, enabling quantitative tracking of long-term training effects and allowing training plan adjustments to be based on actual improvement data rather than experience-based judgments.

[0027] Fifth, by mapping continuous assessment indicators to the timeline to generate multi-dimensional visualization charts, potential problems such as movement compensation and load discomfort can be identified, corrective methods and load strategies can be matched, and phased achievement goals can be formulated to form a personalized training optimization path. This can improve teaching efficiency while also identifying sports injury risks in advance and ensuring the safety of athletes during training. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the multi-dimensional data-driven volleyball teaching effectiveness feedback method in this application.

[0030] Figure 2 This is a comparison chart of the training results in this application;

[0031] Figure 3 This is a schematic diagram of the multi-dimensional data-driven volleyball teaching effectiveness feedback system in this application. Detailed Implementation

[0032] This application provides a multi-dimensional data-driven method and system for providing feedback on volleyball teaching effectiveness. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of a multi-dimensional data-driven volleyball teaching effectiveness feedback method in this application includes:

[0034] Step S1: Collect real-time motion data and ball trajectory data of volleyball players during their training process to form the original dataset.

[0035] In one specific embodiment, performing step S1 includes the following steps:

[0036] Wearable sensors are used to collect data on lower limb force exertion, body coordination, and landing cushioning during volleyball players' training to obtain real-time motion data.

[0037] High-definition cameras are used to record the swing sequence and trajectory data of volleyball players during training to obtain ball trajectory data;

[0038] The collected real-time motion data and recorded ball trajectory data are integrated and preprocessed to generate the original dataset.

[0039] Specifically, wearable sensors are deployed on key joints of the lower limbs, such as the hip, knee, and ankle, as well as the soles of the feet of volleyball players. Throughout the entire training process, including spiking, setting, serving, and blocking, data on lower limb force generation, body coordination, and landing cushioning are continuously collected. Lower limb force generation data includes torque changes in each joint, acceleration values ​​of muscle exertion, and time-series data of force generation. Body coordination data covers the relative motion angles of the hip, knee, and ankle joints, the phase difference of joint movements, and the coordination between the trunk and lower limbs. Landing cushioning data includes plantar pressure distribution data, peak impact force upon landing, force decay rate during the cushioning phase, and cushioning duration. This data, which directly reflects the athlete's internal biomechanical state, is integrated into real-time motion data, providing internal parameters for subsequent biomechanical characteristic analysis of the movement chain. Supported by a large number of high-definition cameras, which are arrayed in multiple fixed angles such as the side, back, and net of the training field, the entire training process of the athletes is recorded at a data acquisition rate of 30 frames per second. By analyzing and extracting features from each frame of the continuous image, swing sequence data and ball trajectory data are obtained. The swing sequence data includes the time sequence nodes of each limb movement from the athlete's preparation to the completion of the swing, the time difference between the sequential connection of each limb movement, and the speed change sequence of the arm swing. The ball trajectory data is tracked in real time by image processing algorithms to obtain the three-dimensional coordinates of the hitting point, the speed value of the volleyball flight, the flight angle, the trajectory curvature, and the coordinates of the landing point. This type of data intuitively reflects the external spatial characteristics of the athlete's movements, making up for the technical deficiency of single sensor data lacking intuitive verification of spatial trajectory.

[0040] After collecting real-time motion data and ball trajectory data, the two types of data are integrated using timestamps as the core correlation basis. Lower limb force data, body coordination data, and landing cushioning data collected at the same time point are matched one-to-one with the corresponding swing sequence data and ball trajectory data. This achieves precise alignment of internal mechanical parameters and external spatial trajectory parameters in the time dimension. For example, the knee joint force acceleration and the phase difference between the hip and knee joints collected at 0.5 seconds are bound to the arm swing speed and volleyball flight angle at that moment, allowing different types of data to form corresponding correlations. The integrated data forms a raw data sequence containing multi-dimensional parameters. Preprocessing is then performed on the integrated data, first using a median filtering algorithm to remove noise from the real-time motion data. For example, by setting the filter window size to 9, sliding filtering is applied to continuously collected data such as force acceleration, joint angle, and plantar pressure to eliminate abnormal peak data caused by environmental vibration and body swaying during sensor acquisition. Then, linear interpolation is used to supplement missing values ​​in the ball trajectory data caused by frame intervals during high-definition camera acquisition. Interpolation coordinates are calculated based on the volleyball spatial coordinates of two adjacent time points to ensure the temporal continuity of the ball trajectory data. At the same time, all data are standardized, mapping all values ​​of lower limb force data, body coordination data, landing cushioning data, swing sequence data, and ball trajectory data to the [0,1] interval to ensure data uniformity and avoid interference from parameter value range differences in subsequent feature extraction and analysis.

[0041] After filtering, interpolating, and standardizing the data, the processed real-time motion data and ball trajectory data are classified and stored according to the training action type. Multi-dimensional data of different actions such as spiking, passing, serving, and blocking are formed into independent data subsets. Within each data subset, standardized mechanical parameters and trajectory parameters are arranged in chronological order using timestamps as indexes. The dataset corresponding to each timestamp contains all standardized values ​​of lower limb force data, body coordination data, landing cushioning data, swing sequence data, and ball trajectory data at that moment. Data subsets of all action types together constitute a structured original dataset. The dataset is stored in the form of a two-dimensional table, with the row dimension being the timestamp and the column dimension being the standardized parameters of various types, so that each data record can be accurately matched with the specific training action and time node.

[0042] Step S2: Extract features from the original dataset, analyze the training action chain, and obtain action analysis results.

[0043] In one specific embodiment, performing step S2 includes the following steps:

[0044] Convolutional neural networks are used to mine key feature information in the training action chain in the original dataset, and the extracted key feature information is quantified to obtain the displacement parameters, velocity parameters and angle parameters corresponding to each action link in the action chain.

[0045] Based on the calculated displacement, velocity, and angle parameters, a dynamic model of the motion chain is constructed.

[0046] The dynamic model is analyzed to obtain the coordination and coherence analysis results between the various action links in the action chain;

[0047] The results of coordination analysis and coherence analysis are integrated to form the results of motion analysis.

[0048] Specifically, the input layer of the convolutional neural network receives multi-dimensional raw data, including data on lower limb force, body coordination, and landing cushioning collected by wearable sensors, and swing sequence and ball trajectory data captured by a high-definition camera. All data undergoes timestamp alignment and standardization. The data dimensions are constructed into a multi-channel matrix according to time series and parameter type. The convolutional neural network uses three convolutional layers and two pooling layers arranged alternately. The convolutional layers use 3×3 convolutional kernels with a stride of 1 and the same padding method. The pooling layers use 2×2 max pooling kernels with a stride of 2. Through convolutional operations, the temporal and spatial features in the raw dataset are locally perceived, extracting key feature information from the volleyball training action chain, such as hip, knee, and ankle joint movements, arm swing and ball hitting sequence, and changes in volleyball flight trajectory. The features are then reduced in dimensionality by the pooling layers. While preserving key information, the fully connected layer integrates the extracted features and transforms them into specific parameters for each link in the motion chain through quantization calculations. The displacement parameter is based on the three-dimensional coordinates of the volleyball training court, calculating the spatial position changes of each joint and the hitting point during the execution of the action. The velocity parameter is obtained by solving the time difference of the displacement parameter to obtain the rate of position change per unit time. The angle parameter calculates the relative angles between joints such as hip, knee, ankle, shoulder, elbow, and wrist, as well as the angle between the hitting racket face and the flight direction of the volleyball. Each action link, such as pushing off the ground, jumping, swinging, hitting the ball, and landing cushioning, corresponds to a set of independent displacement, velocity, and angle parameters. The multi-dimensional data of each time node in the original dataset and the quantized single-link parameters form a one-to-one mapping relationship, ensuring that the parameters can accurately reflect the motion state of the corresponding action link.

[0049] Based on these quantified displacement, velocity, and angle parameters, a dynamic model is constructed according to the execution sequence of the volleyball training action chain. The parameters of each action link are constructed into a time-series vector sequence, with each vector containing the displacement, velocity, and angle components of the corresponding link. The potential missing values ​​in the parameter sequence are filled by linear interpolation to ensure the continuity of the vector sequence. Then, a node graph of the dynamic model is generated with action links as nodes and parameter transition relationships as edges. The parameter change trend between adjacent nodes in the node graph reflects the connection state between action links. For example, the slope of the change of velocity parameters in the push-off link and the jump link directly reflects the force transmission state from push-off to jump. The construction of the dynamic model transforms discrete parameters into a holistic model that reflects the continuous execution process of the action, solving the technical problem in traditional volleyball teaching that can only perform static analysis on a single action and cannot capture the continuous changes of the action chain.

[0050] The dynamic model is analyzed by first extracting the parameter transition sequences between each motion link in the model, and then calculating the Pearson correlation coefficient (denoted as γ) of this sequence to measure the coordination between the links in the motion chain. The correlation coefficient is calculated based on the parameter sequences of two adjacent motion links, obtained by dividing the covariance by the product of their respective standard deviations. Specific mapping rules and thresholds are established: when γ ≥ 0.85, the qualitative evaluation is "excellent coordination," and the quantitative label is "coordination coefficient ≥ 0.85," indicating a high degree of synchronization in the motion states of adjacent links; when γ ≥ 0.7... When 0 ≤ γ < 0.85, the qualitative evaluation is "good coordination," and the quantitative label is "0.70 ≤ coordination coefficient < 0.85," indicating that the parameter changes of adjacent links are basically synchronized, with a slight coordination deviation. When 0.55 ≤ γ < 0.70, the qualitative evaluation is "average coordination," and the quantitative label is "0.55 ≤ coordination coefficient < 0.70," indicating a significant coordination deviation between adjacent links. When γ < 0.55, the qualitative evaluation is "poor coordination," and the quantitative label is "coordination coefficient < 0.55," indicating a disconnect between the movement states of adjacent links. For example, in the spiking action, the Pearson correlation coefficient γ = 0.82 between the velocity parameter sequence of the push-off link and the velocity parameter sequence of the jump link corresponds to "good coordination," labeled "0.70 ≤ coordination coefficient < 0.85." Simultaneously, a smoothness index is calculated by taking the average of the second derivatives of the parameter curves (denoted as θ) to assess the continuity of the action chain. The second derivative reflects the smoothness of parameter changes. The mapping rules and threshold settings are as follows: When |θ|≤0.3, the qualitative evaluation is "excellent continuity", and the quantitative label is "|continuity index|≤0.3"; when 0.3<|θ|≤0.6, the qualitative evaluation is "good continuity", and the quantitative label is "0.3<|continuity index|≤0.6"; when 0.6<|θ|≤1.0, the qualitative evaluation is "average continuity", and the quantitative label is "0.6<|continuity index|≤1.0"; when |θ|>1.0, the qualitative evaluation is "poor continuity", and the quantitative label is "|continuity index|>1.0". For example, in the passing motion, the average value of the second derivative of the angle parameter curve of the arm swing segment is θ = 0.75° / s², corresponding to "general continuity," marked as "0.6 < |Continuity Index| ≤ 1.0." By incorporating ball trajectory data into the analysis, the deviation of the volleyball's flight trajectory is correlated with the parameter changes in the motion segment. If the ball trajectory deviation exceeds a preset range, the continuity score at the corresponding motion segment connection is simultaneously reduced. For instance, a sudden change in the angle parameter during the swing segment directly causes the volleyball's flight trajectory to deviate, thus lowering the continuity score from the swing to the hitting segment. This analytical method links the assessment results of coordination and continuity with the actual movement effects in volleyball training, solving the technical problem of traditional teaching methods that rely solely on visual observation and cannot quantitatively assess motion coordination and continuity.The above thresholds can be dynamically adjusted according to the athlete's level and the type of movement: the threshold for youth athletes can be relaxed by 15% to 20%, and the threshold for professional athletes can be tightened by 10% to 15%, to ensure the adaptability and practicality of the evaluation criteria.

[0051] The correlation coefficients, qualitative evaluations, and quantitative annotations of each link obtained from the coordination analysis are integrated with the smoothness index results, corresponding qualitative evaluations, and quantitative annotations obtained from the coherence analysis. The two types of results are then spliced ​​together according to the link sequence of the action chain to form an action analysis result that includes the coordination of each link and the overall coherence. During the integration process, the correspondence between the two types of results and the action links is maintained.

[0052] Step S3: Based on the motion analysis results, determine whether the deviation of the training motion exceeds the preset deviation threshold. If so, classify the real-time motion data and ball trajectory data and determine the problem classification label.

[0053] In one specific embodiment, performing step S3 includes the following steps:

[0054] Extract the deviation values ​​corresponding to the weak links in the training motion chain from the motion analysis results;

[0055] Determine whether the deviation value exceeds the preset deviation threshold. If so, use the support vector machine model to classify the real-time motion data and ball trajectory data, identify the dynamic coherence problems, and generate corresponding classification features.

[0056] Determine the problem classification label based on classification characteristics.

[0057] Specifically, deviation values ​​corresponding to weak links in the training movement chain are extracted from the motion analysis results. The motion analysis results already contain data on the coordination and continuity of each movement link. Weak links need to be screened in conjunction with the execution logic of the movement chain, focusing on links that significantly affect the overall quality of the movement, such as the push-off force link and the swing connection link in a spiking motion. Deviation value extraction is based on the quantitative parameters in the motion analysis results. The actual parameters of each weak link are compared and calculated with standard parameters. The standard parameters come from the best training data of athletes of the same level stored in the database or the technical movement standard template set by the coach. For example, the standard force acceleration of the push-off force link is 4.5 m / s². If the actual collected push-off force acceleration is 3.2 m / s², then the deviation value of this link is calculated as 1.3 m / s² by the absolute value of the difference between the actual parameter and the standard parameter. The standard phase difference of the swing connection link is 0.2 seconds, the actual phase difference is 0.5 seconds, and the corresponding deviation value is 0.3 seconds. The calculation of different types of parameter deviation values ​​all follow this logic, ensuring that the deviation of each weak link can be accurately quantified, thus solving the technical problem of difficulty in quantifying and evaluating movement deviations in traditional teaching.

[0058] The system determines whether the deviation value exceeds a preset deviation threshold. This threshold is dynamically adjusted based on the training scenario, athlete level, and movement type. It is automatically generated by the system using historical training data from the database and experience values ​​input by the coach, or set manually. For example, for a junior athlete's passing motion, the ground-pushing force deviation threshold is set to 1.5 m / s², and the swing connection phase difference threshold is set to 0.4 seconds. For a professional athlete's spiking motion, the corresponding thresholds are tightened to 0.8 m / s² and 0.25 seconds, respectively. The extracted deviation values ​​of each weak point are compared one by one with the corresponding preset threshold. If the deviation value of any weak point exceeds the threshold, the subsequent data classification process is triggered. This judgment process achieves objective screening of movement deviations, avoiding the limitations of traditional teaching methods that rely on subjective experience to determine whether deviations need correction.

[0059] Support Vector Machine (SVM) models were used to classify real-time motion data and ball trajectory data. First, the input feature vectors of the model were determined. The input data included lower limb force acceleration sequences, joint angle changes, and landing cushioning pressure distribution data from the real-time motion data, as well as multi-dimensional parameters such as the three-dimensional coordinates of the hitting point, flight speed, and trajectory curvature from the ball trajectory data. All input data were standardized and mapped to the [0,1] interval to ensure data uniformity. The SVM model used radial basis functions as kernel functions, with the kernel parameter σ determined to be 0.8 using a grid search method. The penalty factor C was set to 10. Historical data labeled with dynamic continuity issues such as "discontinuous push-off and hip rotation," "broken arm swing and hitting connection," and "unbalanced landing cushioning" were used as training samples, with a sample size of no less than 5000 sets. During training, 5-fold cross-validation was used to optimize model parameters, ensuring a classification accuracy of no less than 92%. For example, when real-time data of an athlete's spiking training is input, the model identifies the dynamic continuity problem of "insufficient hip rotation-arm swing connection" in the data by calculating the distance between the sample and the hyperplane, and generates corresponding classification features. The classification features include quantitative information such as the parameter change pattern corresponding to the problem, the time node of the deviation, and the affected movement links, which realizes the accurate identification of dynamic continuity problems and solves the technical defects of traditional teaching in that it is difficult to locate specific continuity problems in the movement chain.

[0060] Based on classification features, problem classification labels are determined, and a mapping relationship library between classification features and problem labels is established. This library contains feature vector templates corresponding to various dynamic and coherent problems. The classification features generated by support vector machines are matched with these templates using a cosine similarity algorithm. If the similarity exceeds 0.85, the corresponding problem label is directly matched. If multiple feature templates have similar matching degrees, a secondary determination is made based on coordination and coherence data from the motion analysis results. For example, if classification features show parameter changes concentrated in the transition between hip rotation and arm swing, with abrupt changes in joint angle rate and interrupted speed transmission, and the similarity to the feature template "insufficient hip rotation-arm swing connection" reaches 0.91, then the problem classification label is determined. If classification features simultaneously contain information about abnormal peak landing cushioning pressure and shortened cushioning time, combined with a low smoothness index between the landing and previous stages in the coherence analysis results, a composite label of "overly stiff landing cushioning leading to discontinuous motion" can be determined. The problem classification tags are presented in a structured format, containing key information such as problem type, influencing factors, and degree of deviation. This provides a precise basis for identifying problems in order to generate personalized feedback reports, and solves the technical problems of vague and untargeted corrective suggestions in traditional teaching.

[0061] Step S4: Based on the problem classification tags, integrate the volleyball players' subjective feelings records and training load data to generate a personalized feedback report.

[0062] In one specific embodiment, performing step S4 includes the following steps:

[0063] Obtain information on weaknesses in volleyball training techniques corresponding to the problem category tags;

[0064] Extract subjective feelings and training load data of volleyball players from the database;

[0065] The correlation analysis of subjective feelings records, training load data and problem classification labels was carried out to clarify the correspondence among the three.

[0066] Based on the corresponding relationships, a personalized feedback report is generated that includes a detailed description of the weaknesses and adjustment suggestions.

[0067] Specifically, the system retrieves information on weak points in volleyball training movements corresponding to problem classification tags. These tags clearly define the specific type and scope of impact of dynamic and continuous problems. By accessing a pre-defined tag-weak point mapping database, the corresponding information can be directly obtained. The mapping database stores a one-to-one correspondence between various tags and weak points. Each tag corresponds to a specific movement segment and the parameters that need attention in that segment. For example, the tag "disjointed push-off and hip rotation" corresponds to the push-off force generation segment and the hip rotation segment, with associated parameters being push-off acceleration, hip rotation angle, and the timing difference between the two. The tag "disconnected arm swing and hitting" corresponds to the arm swing segment and the hitting segment, with associated parameters being arm swing speed, the three-dimensional coordinates of the hitting point, and the phase difference between the two. This mapping relationship ensures rapid location of specific weak points from the tags, solving the technical problem of difficulty in identifying the root cause of movement problems in traditional teaching.

[0068] Subjective feeling records and training load data of volleyball players were extracted from a database, which categorized and stored data according to the athlete's unique identifier and training period. Subjective feeling records were entered by athletes via terminals after each training session and included information such as fatigue level, areas of discomfort (e.g., knees, shoulders), and difficulty assessment of movements. Each record was timestamped and labeled with the training movement type. Training load data was automatically recorded by the system, including quantitative data such as training duration, number of repetitions, peak force per movement, and training interval, also linked to timestamps and movement types. The extraction process used structured queries to filter data by athlete identifier, training period, and movement type, ensuring that the extracted data accurately matched the training scenarios corresponding to the problem classification tags. For example, for the tag "incoherent push-off and hip rotation during spiking," only the subjective feeling and load data of that athlete during the same spiking training period were extracted, providing targeted data support for subsequent correlation analysis and avoiding analytical bias caused by data fragmentation.

[0069] The association analysis of subjective feeling records, training load data, and problem classification labels was first performed. The three types of data were standardized to unify the data format and time dimension. Qualitative data such as fatigue level and movement difficulty evaluation in the subjective feeling records were converted into quantitative values ​​in the [0,1] interval, and the body discomfort areas were coded according to sports anatomy. Parameters such as duration and repetitions in the training load data were normalized to eliminate dimensional differences. Problem classification labels were converted into feature vectors containing weaknesses and related parameters. The association analysis used the Pearson correlation coefficient algorithm and time-series alignment method to calculate the correlation between the subjective feeling quantification values, training load parameters, and label feature vectors. A correlation coefficient threshold of 0.6 was set; when the correlation coefficient between a certain load parameter and the label feature vector exceeded the threshold, a strong correlation was determined. For example, if the correlation coefficient between the label "disjointed push-off and hip rotation" and the training load data "more than 30 spikes in a single training session" is 0.75, and the correlation coefficient between the label and the subjective feeling "hip joint fatigue score of 8" is 0.82, then the relationship among the three is clear: high spike count leads to hip joint fatigue, which in turn causes disjointed push-off and hip rotation. This correlation analysis integrates objective data and subjective feelings, solving the technical problems of separation between the two and lack of personalized feedback in traditional teaching.

[0070] Personalized feedback reports are generated based on the correlation relationships. The report structure includes two parts: a detailed description of weaknesses and adjustment suggestions. The description of weaknesses is based on the correlation analysis results, combined with problem classification tags and data characteristics, to quantitatively present the problem manifestations and influencing factors. For example, "In spiking training, the number of spiking attempts in a single training session reached 35 (20% exceeding the appropriate load), resulting in a hip joint fatigue score of 8, a push-off acceleration 1.2 m / s² lower than the standard value, a hip rotation angle deviation of 15°, and a problem of discontinuity in the push-off and hip rotation movements." The adjustment suggestions are formulated based on the correlation relationships, including load adjustment, movement correction, and training frequency optimization. For example, "Adjust the number of spiking attempts in a single training session to 25, add hip joint relaxation training (5 minutes each time), and adopt a specific training method that decomposes the push-off and hip rotation movements and uses resistance bands to assist in force generation, conduct it 3 times a week, and extend the training interval to 90 seconds." During report generation, the system calls upon a pre-set suggestion template library. Based on the problem type, workload level, and subjective feelings obtained from correlation analysis, it matches the corresponding adjustment scheme template and then refines the content with specific data parameters to ensure that the suggestions are operable and targeted. This solves the technical problems of vague correction suggestions and lack of data support in traditional teaching. For example, for the tag "break in the connection between arm swing and hitting in blocking action", if the correlation analysis shows that it is related to "insufficient arm swing speed" and "shoulder muscle soreness", the report suggestion will clearly state "increase shoulder explosive power training (such as resistance arm swing with elastic band, 15 times per set, 3 sets / time), adjust the blocking training load, and the number of blocks in a single training session should not exceed 20 times", and indicate the expected improvement goals, such as "increase arm swing speed by 0.8m / s and reduce the connection deviation to within 5°".

[0071] Step S5: Based on the adjustment suggestions in the personalized feedback report, monitor the training data in real time during the subsequent training process, and update the motion analysis results based on the training data obtained from the monitoring.

[0072] In one specific embodiment, performing step S5 includes the following steps:

[0073] Based on the adjustment suggestions in the personalized feedback report, set the corresponding sports data monitoring parameters for subsequent volleyball training.

[0074] The system continuously collects subsequent training data according to the set motion data monitoring parameters using sensors and cameras to obtain monitoring training data.

[0075] Compare the monitoring training data with the adjustment suggestions in the personalized feedback report to determine whether the monitoring training data meets the parameter requirements corresponding to the adjustment suggestions;

[0076] If not, a convolutional neural network is used to extract features from the monitoring training data to obtain the displacement parameters, velocity parameters, and angle parameters corresponding to the monitoring training data.

[0077] Based on the displacement, velocity, and angle parameters corresponding to the monitoring training data, the dynamic model of the motion chain is reconstructed, and the coordination and continuity of each motion link are analyzed to update the motion analysis results.

[0078] Specifically, when setting monitoring parameters for subsequent volleyball training data based on adjustment suggestions in personalized feedback reports, the suggestions clearly define the improvement directions and target parameters for weak areas. Monitoring parameters must be precisely set based on this information, covering data collection dimensions, collection frequency, parameter thresholds, and anomaly detection rules. For example, regarding the adjustment suggestion of "insufficient arm swing speed during spiking," if the suggested target is to increase arm swing speed to 8 m / s, the monitoring parameters would be set as follows: arm swing speed collection frequency 50 Hz, threshold range 7.5 m / s-8.5 m / s, and anomaly detected when the collected data is below 7.5 m / s for three consecutive times. Regarding the adjustment suggestion of "landing cushioning angle deviation," if the suggested target angle is 45°, the landing cushioning angle collection accuracy would be set to 0.1°, with a threshold range of 43°~47°. Simultaneously, the collection of plantar pressure distribution data would be correlated, with a pressure peak threshold of 500 N, forming a multi-parameter collaborative monitoring system. The setting of monitoring parameters quantifies the adjustment suggestions, solving the technical problem of lacking clear goals and standards in traditional teaching training monitoring, and ensuring that subsequent data collection is targeted.

[0079] When sensors and cameras continuously collect subsequent training data according to set motion data monitoring parameters, the sensors and cameras adjust their working modes based on these parameters. Wearable sensors collect data at set frequencies for monitoring dimensions such as lower limb force, body coordination, and landing cushioning. For example, the accelerometer for arm swing speed monitoring collects data at 50Hz, and the pressure sensor for landing cushioning monitoring collects foot pressure distribution at 100Hz. High-definition cameras record training footage at set resolutions and frame rates. For example, for monitoring ball trajectory, the camera frame rate is set to 30 frames per second, and the resolution to 1920×1080. Image processing algorithms track the volleyball position and athlete's limb movements in real time. The collected data is categorized and stored by timestamp and training action type, forming monitoring training data. The data format is consistent with the original dataset to ensure compatibility for subsequent comparison and analysis. This parameter-based precise data collection method solves the problems of messy and untargeted traditional data collection, providing a high-quality data source for subsequent data processing.

[0080] When comparing the monitoring training data with the adjustment suggestions in the personalized feedback report, the monitoring training data is first preprocessed. A median filtering algorithm (e.g., filter window size 7) is used to remove sensor noise, and linear interpolation is used to supplement missing data to ensure data integrity. The comparison process is performed step-by-step according to the action stage. The parameters in the monitoring training data are compared with the target parameters in the adjustment suggestions, and a difference threshold is set to determine whether the requirements are met. For example, if the target value for arm swing speed in the adjustment suggestions is 8 m / s, and the difference threshold is set to 0.5 m / s, if the arm swing speed in the monitoring data is 7.3 m / s, the calculated difference is 0.7 m / s, exceeding the threshold, and is therefore considered non-compliant. Similarly, if the target value for the landing buffer angle in the adjustment suggestions is 45°, and the difference threshold is set to 2°, if the angle in the monitoring data is 42.5°, the difference is 2.5°, and is therefore considered non-compliant. The comparison process also needs to consider the correlation of multiple parameters. For example, if the arm swing speed fails to meet the target, and the lower limb force acceleration is also lower than the target value, it is marked as a multi-stage coordination anomaly, providing more comprehensive information for subsequent feature extraction. This precise comparison method solves the technical problem of difficulty in quantifying whether movements have improved in traditional teaching, and enables real-time evaluation of training effects.

[0081] If the monitoring training data does not meet the parameter requirements corresponding to the adjustment suggestions, when using a convolutional neural network (CNN) to extract features from the monitoring training data, the structure and training process of the CNN must be adapted to the characteristics of the monitoring data. The network consists of 4 convolutional layers, 2 pooling layers, and 2 fully connected layers. The convolutional layers use 3×3 kernels with a stride of 1 and the same padding method. The pooling layers use 2×2 max pooling kernels with a stride of 2, and the activation function is ReLU. The training process uses historical monitoring data and corresponding standard parameter data as samples, with a sample size of no less than 3000 sets, divided into training and validation sets in a 7:3 ratio. The optimizer used is Adam, with a learning rate of 0.001 and 100 iterations. Training stops when the accuracy of the validation set stabilizes above 90%. During feature extraction, preprocessed monitoring training data is input into the network. Convolutional layers capture temporal and spatial features in the data, such as the temporal variation of arm swing velocity and the spatial correlation of joint angles. After dimensionality reduction by pooling layers and integration by fully connected layers, the output is the displacement, velocity, and angle parameters corresponding to the monitoring training data. Each parameter corresponds one-to-one with a specific link in the motion chain, such as the push-off displacement, arm swing velocity, and hitting angle in a spiking motion. This feature extraction process transforms complex monitoring data into quantified motion parameters, solving the technical problem of difficulty in extracting key motion features from raw data in traditional analysis.

[0082] When reconstructing the dynamic model of the motion chain based on the displacement, velocity, and angle parameters corresponding to the monitoring training data, each parameter is constructed as a multi-dimensional vector sequence according to the execution sequence of the motion. Each vector contains the displacement, velocity, and angle components of the corresponding stage. Missing values ​​in the parameter sequence are filled using cubic spline interpolation to ensure model continuity. The dynamic model uses motion stages as nodes and parameter transition relationships as edges to construct a node graph. For example, the velocity parameters of the push-off stage and the take-off stage are related through a transition function to reflect the efficiency of force transmission. When performing coordination and coherence analysis on the dynamic model, the Pearson correlation coefficient of parameters of adjacent stages is calculated to assess coordination. The threshold for the correlation coefficient is set at 0.75; values ​​below the threshold are marked as insufficient coordination. Coherence is assessed by calculating the average of the first derivatives of the parameter curves; a larger average value indicates more drastic parameter changes and poorer coherence. For example, in the reconstructed dynamic model of the spiking motion, if the correlation coefficient of the velocity parameters between the arm swing and the hitting phase is 0.68, which is below the threshold, it is judged as insufficient coordination; if the average value of the first derivative of the angle parameter of the landing and cushioning phase is 15° / s, it indicates poor continuity. Integrating the results of coordination and continuity analysis forms an updated motion analysis result. This result not only includes the quantitative parameters of each phase but also clarifies the existing problems, solving the technical problems of one-sided motion analysis results and lack of dynamic continuity assessment in traditional teaching, and providing a precise basis for subsequent comparative evaluation.

[0083] Step S6: Compare and analyze the action analysis results before and after the update, and calculate the continuity evaluation index.

[0084] In one specific embodiment, performing step S6 includes the following steps:

[0085] The updated motion analysis results are matched one by one with the previous motion analysis results according to each link of the motion chain, and the changes in displacement parameters, velocity parameters and angle parameters of each corresponding link are calculated.

[0086] Weights are assigned to each action link according to their importance in the action chain, and the parameter changes of each link are weighted and calculated in combination with the weights to evaluate the overall improvement of the action chain.

[0087] The overall improvement of the action chain is standardized and quantified to obtain continuous evaluation indicators.

[0088] Specifically, the updated motion analysis results are mapped one-to-one with the unupdated motion analysis results according to each link of the motion chain. The motion chain covers continuous links such as push-off, take-off, arm swing, ball hitting, and landing cushioning. Each link contains three types of core data: displacement parameters, velocity parameters, and angle parameters. The mapping process is based on the execution sequence of the motion, establishing a two-way mapping between link names and parameter types to ensure accurate alignment of the same type of parameters in the same link before and after the update. For example, the displacement parameters of the push-off link before the update correspond one-to-one with the displacement parameters of the push-off link after the update, and the velocity parameters of the arm swing link correspond one-to-one with the velocity parameters of the arm swing link after the update, avoiding comparison errors caused by link misalignment or parameter type confusion. When calculating the parameter changes for each corresponding stage, a difference calculation logic is used. The change in displacement parameter is the updated displacement value minus the original displacement value. The changes in velocity and angle parameters are calculated using the same logic. For example, the displacement parameter for the push-off stage was 0.8 meters before the update and 1.0 meter after the update, corresponding to a displacement change of 0.2 meters; the arm swing velocity parameter was 7.5 m / s before the update and 8.3 m / s after the update, resulting in a velocity change of 0.8 m / s; the hitting angle parameter was 38° before the update and 42° after the update, resulting in an angle change of 4°. This step-by-step, parameter-by-parameter calculation of changes enables precise quantification of movement improvement, solving the technical problem of accurately measuring the magnitude of movement improvement in traditional teaching.

[0089] Weights are assigned to each action segment based on its importance in the motion chain. This weighting is based on the technical characteristics and mechanical principles of volleyball, and the weight value for each segment is determined using the analytic hierarchy process (AHP). The total weight is 1. Core power-generating segments such as the push-off and the hit are weighted at 0.3; key connecting segments such as the jump and the arm swing are weighted at 0.2; and auxiliary cushioning segments such as landing cushioning are weighted at 0.1. For example, in a spiking motion, the push-off, as the foundation of power generation, has a weight of 0.3; the hit, as the core technique, has a weight of 0.3; the jump, connecting the push-off and the arm swing, has a weight of 0.2; the arm swing, transmitting power, has a weight of 0.2; and the landing cushioning, ensuring the safety of the motion, has a weight of 0.1. When calculating the weighted changes in parameters for each segment, the average value of the displacement, velocity, and angle changes for each segment is first taken to obtain the comprehensive change value for that segment. Then, this comprehensive change value is multiplied by the corresponding segment weight. Finally, the weighted change values ​​of all segments are summed to obtain the overall improvement of the motion chain. For example, the overall change value for the push-off phase is 0.3, with a weight of 0.3, resulting in a weighted change value of 0.09; the overall change value for the take-off phase is 0.2, with a weight of 0.2, resulting in a weighted change value of 0.04; the overall change value for the arm swing phase is 0.4, with a weight of 0.2, resulting in a weighted change value of 0.08; the overall change value for the ball striking phase is 0.5, with a weight of 0.3, resulting in a weighted change value of 0.15; and the overall change value for the landing cushioning phase is 0.1, with a weight of 0.1, resulting in a weighted change value of 0.01. The overall improvement is 0.09 + 0.04 + 0.08 + 0.15 + 0.01 = 0.37. This weighted calculation method fully considers the different impacts of each phase on the overall movement quality, ensuring that the assessment of the overall improvement is more in line with the actual training effect and solving the problem of distorted results caused by treating each phase equally in traditional assessments.

[0090] The overall improvement magnitude of the motion chain is standardized and quantified to obtain a continuous evaluation index. First, the range of values ​​for the overall improvement magnitude is determined. By analyzing a large amount of historical training data, the minimum improvement magnitude is determined to be -0.5 (indicating significant deterioration of the motion), and the maximum is 0.5 (indicating that the motion has reached the optimal standard). Standardization is performed using the Min-Max normalization algorithm, with the formula: Continuous Evaluation Index = (Overall Improvement Magnitude - Minimum) / (Maximum - Minimum), or Continuous Evaluation Index = (Overall Improvement Magnitude + 0.5) / 1.0, mapping the overall improvement magnitude to the [0,1] interval. For example, when the overall improvement magnitude is 0.37, the continuous evaluation index = (0.37 + 0.5) / 1.0 = 0.87. After quantization, the value of the continuous evaluation index directly reflects the sustained effect and stability of the motion improvement. The closer the value is to 1.0, the more significant and stable the motion improvement; a value below 0.5 indicates a risk of deterioration or insufficient improvement. This process transforms abstract improvement metrics into intuitive quantitative indicators, solving the technical problem of lacking continuous evaluation standards for long-term training effects in traditional teaching, and providing core data support for the subsequent generation of visual charts and personalized training optimization paths.

[0091] Step S7: Generate visualization charts based on the continuity assessment indicators, and determine the personalized training optimization path for volleyball players based on the visualization charts.

[0092] In one specific embodiment, performing step S7 includes the following steps:

[0093] Obtain the parameter change values ​​and overall improvement data of each link of the motion chain corresponding to the continuous evaluation index, and perform time-axis correlation mapping with the subjective feeling records and training load data of volleyball players;

[0094] Based on the correlation mapping results, generate multi-dimensional visualization charts that include the trend curve of action improvement, training load distribution and changes in link deviation;

[0095] By analyzing visualization charts, we can identify potential problems in volleyball players' training caused by load adaptation or movement compensation, as well as persistent weak links and areas for significant improvement in the movement chain.

[0096] By combining potential problems with persistent weaknesses and significant areas for improvement in the motion chain, corresponding motion correction methods, training load adjustment strategies, and training frequency planning are matched.

[0097] Based on the matching results and the adjustment suggestions in the personalized feedback report, phased training parameter achievement targets are formulated.

[0098] Based on the phased training parameter achievement targets, a personalized training optimization path for volleyball players is determined.

[0099] Specifically, the changes in parameters and the overall improvement rate of each link in the motion chain corresponding to the continuous evaluation indicators are obtained. These data have been quantified and generated through comparative analysis in step S6, covering the changes in displacement, velocity, and angle parameters of each link, including push-off, take-off, arm swing, ball hitting, and landing cushioning, as well as the overall improvement rate after weighted calculation and standardization. When mapping these data with the volleyball players' subjective feeling records and training load data along a time axis, the training sessions or training timestamps are used as a unified benchmark to establish a correspondence between multi-dimensional data. Quantitative data such as fatigue level, discomfort area, and movement difficulty evaluation in the subjective feeling records, and parameters such as training duration, number of repetitions, peak force, and rest intervals in the training load data are all matched one-to-one with the changes in parameters and the overall improvement rate of each link in the motion chain according to the timestamp. For example, in a training session, the change in speed during the push-off phase was 0.8 m / s, the overall improvement was 0.37, the corresponding subjective fatigue score was 7, and the number of repetitions of the movement in the training load was 30. Through time-axis correlation mapping, a three-dimensional data set of "parameter change - subjective feeling - training load" is formed. This correlation mapping solves the technical problem of the separation between objective motion data and subjective feelings and training load in traditional teaching, and provides an integrated data foundation for subsequent multi-dimensional analysis.

[0100] Based on the correlation mapping results, multi-dimensional visualization charts are generated. These charts employ a composite presentation format, with the main axis representing time on the horizontal axis and the vertical axis divided into three dimensions: the movement improvement trend curve, training load distribution, and stage deviation changes. The movement improvement trend curve, based on the overall improvement magnitude, is presented as a smooth line chart, intuitively reflecting the overall change in training effectiveness over different time periods. The training load distribution is presented as a bar chart, with different colored bars corresponding to load parameters such as training duration and number of repetitions. Stage deviation changes are presented as a stacked area chart, with the deviation change value for each stage corresponding to a different stacked layer, clearly showing the improvement or deterioration trend of each stage. For example, in the visualization chart, if the movement improvement trend curve shows an upward trend over a certain period, the height of the bar for the number of repetitions in the training load distribution decreases, and the stacked area of ​​the deviation change in the landing buffer stage shrinks, indicating a significant improvement in the landing buffer stage after reducing the training load. During chart generation, a data visualization algorithm is used to analyze the three-dimensional data set after correlation mapping, and the chart update frequency is set to be synchronized with the training sessions to ensure that the charts can reflect training dynamics in real time, solving the technical problem of traditional teaching methods lacking intuitive and quantitative presentation of training effects.

[0101] When analyzing visualization charts, potential problems and persistent weaknesses are identified and significant areas for improvement by setting data thresholds and trend analysis rules. Potential problem identification focuses on load adaptation and motion compensation. Thresholds are set for training load parameters (e.g., a threshold of 35 repetitions and a peak force threshold of 600N) and a correlation threshold for parameter changes (correlation coefficient threshold of 0.6). When the load parameter exceeds the threshold and the corresponding deviation value of the link shows an inverse trend, it is identified as a load adaptation problem. When the deviation value of one link decreases but the deviation values ​​of other links increase abnormally, it is identified as a motion compensation problem. For example, if the visualization chart shows that the training load repetition count reaches 40 (exceeding the threshold of 35), the corresponding swing link deviation change value is -0.5m / s (increased deviation), and the correlation coefficient is 0.72 (exceeding the threshold of 0.6), then it is identified as a swing link adaptation problem caused by excessive load. If the ball-hitting link deviation value decreases but the knee joint angle deviation value increases, then it is identified as a ball-hitting motion compensation problem. The identification of persistent weaknesses is achieved by calculating the average value of the deviation changes in each stage. Stages with an average value higher than a preset deviation threshold (e.g., 0.2) are identified as persistent weaknesses. Stages of significant improvement are identified by calculating the decrease in the deviation changes in each stage. Stages with a decrease greater than 30% are identified as stages of significant improvement. This analytical method solves the technical problem of accurately identifying potential training problems and key action stages in traditional teaching, enabling targeted positioning of problems and stages.

[0102] By combining potential problems and related aspects with matching movement correction methods, training load adjustment strategies, and training frequency planning, a pre-defined problem-element-strategy matching database is established. In the database, each potential problem and each movement element corresponds to a specific correction method, adjustment strategy, and frequency plan. For example, the load adaptation problem corresponds to a load reduction strategy (reducing the number of repetitions by 20%-30% and extending the rest interval to 90 seconds); a weak push-off element corresponds to an elastic band resistance push-off correction method, with a training frequency plan of 3 specific training sessions per week; a movement compensation problem corresponds to a decomposition movement correction method, with a load adjustment strategy of reducing the intensity of each training session and a training frequency plan of 4 low-intensity training sessions per week. The matching process uses a fuzzy matching algorithm. The potential problem type and element name are input, and the algorithm searches the database for the strategy combination with the highest similarity. The similarity threshold is set to 0.85 to ensure the relevance of the matching results. For example, if the potential problem is "excessive load leading to increased deviation in the arm swing element" and the persistently weak element is the arm swing element, the matching result would be "arm swing decomposition movement correction method + reducing the number of repetitions by 25% + 3 specific training sessions per week". This matching method solves the technical problems of traditional teaching methods and training strategies lacking data support and personalization, ensuring that intervention measures are accurately adapted to the actual situation.

[0103] When setting phased training parameter achievement goals based on the matching results and adjustment suggestions in the personalized feedback report, the training cycle is divided into three phases: the basic improvement phase, the reinforcement and consolidation phase, and the stabilization and improvement phase. Each phase sets specific parameter achievement values ​​and time periods. The basic improvement phase aims to address potential problems and improve persistent weaknesses, with the parameter achievement value set at approximately 70% of the standard parameters, and a time period of 2 weeks. The reinforcement and consolidation phase aims to consolidate the improvement effects and enhance movement stability, with the parameter achievement value set at approximately 85% of the standard parameters, and a time period of 3 weeks. The stabilization and improvement phase aims to optimize the movement and reach the optimal standard, with the parameter achievement value set at over 95% of the standard parameters, and a time period of 4 weeks. For example, if the standard arm swing speed is 8 m / s, the achievement target for the basic improvement phase is set at 5.6 m / s, the reinforcement and consolidation phase at 6.8 m / s, and the stabilization and improvement phase at 7.6 m / s. During the target setting process, the parameter target values ​​are fine-tuned based on the adjustment suggestions in the personalized feedback reports. For example, if the feedback report suggests "focusing on improving push-off acceleration," then the push-off acceleration target values ​​for each stage are increased by 10%. This phased target setting solves the technical problems of vague training objectives and lack of gradual planning in traditional teaching, and provides clear guidance for optimizing personalized training paths.

[0104] When determining personalized training optimization paths based on phased training parameter achievement targets, the paths are organized in phases, integrating movement correction methods, training load adjustment strategies, training frequency planning, and achievement targets to form a complete training implementation framework. The basic improvement phase focuses on addressing potential problems and training weak points in the basics, such as "using arm swing decomposition movement correction methods + adjusting the number of repetitions to 25 per session + 3 specialized training sessions per week + achieving an arm swing speed of 5.6 m / s within 2 weeks"; the reinforcement and consolidation phase focuses on consolidating the improvement effect and training the continuity of movement, such as "using arm swing-hitting connection reinforcement training + maintaining the number of repetitions at 25 per session + 4 training sessions per week (including 1 continuous specialized training session) + achieving an arm swing speed of 6.8 m / s within 3 weeks"; the stable improvement phase focuses on movement optimization and load adaptation training, such as "using variable load arm swing training + adjusting the number of repetitions to 30 per session (gradually increasing) + 3 training sessions per week + achieving an arm swing speed of 7.6 m / s within 4 weeks". During the path determination process, a dynamic adjustment mechanism is established. After each stage, the parameter changes are updated based on the visualization charts, and the actual values ​​are compared with the target values. If the actual values ​​do not meet the target, the time period of that stage is extended or the correction methods and load strategies are adjusted; if the actual values ​​meet the target ahead of schedule, the next stage is started early. This personalized training optimization path solves the technical problems of rigid training plans, lack of dynamic adjustment and personalized adaptation in traditional teaching, and achieves precise and scientific guidance for the training process.

[0105] Please seeFigure 2 , Figure 2 The training effect comparison chart shows a quantitative comparison between the proposed method and traditional methods in terms of movement improvement indicators. It clearly demonstrates that the movement improvement indicators corresponding to the proposed method are not only significantly higher than those of the traditional method, but also exhibit extremely small fluctuations in the indicator values. This indicates that the proposed method, through the synergistic effect of multi-dimensional data fusion collection, closed-loop feedback mechanisms, and personalized optimization paths, improves core dimensions such as the reduction in movement deviation, the degree of coordination improvement, and the adaptability of training load, while achieving superior training stability. This stability is reflected in the smooth and gradual changes in parameters during the athlete's movement improvement process, avoiding the problem of inconsistent movement quality caused by feedback lag and data fragmentation in traditional teaching. This proves that the proposed method can continuously and stably promote athletes' technical movements towards standardization, providing strong support for the predictability of long-term training effects.

[0106] Please see Figure 3 The following describes the multi-dimensional data-driven volleyball teaching effect feedback system in the embodiments of this application. The multi-dimensional data-driven volleyball teaching effect feedback system includes:

[0107] The data acquisition module is used to collect real-time motion data and ball trajectory data of volleyball players during training to form the raw dataset;

[0108] The feature extraction module is used to extract features from the original dataset, analyze the action chain of the training, and obtain action analysis results.

[0109] The motion judgment module is used to determine whether the deviation of the training motion exceeds the preset deviation threshold based on the motion analysis results. If so, it classifies the real-time motion data and ball trajectory data and determines the problem classification label.

[0110] The report generation module is used to generate personalized feedback reports by integrating volleyball players' subjective feelings records and training load data based on problem classification tags.

[0111] The real-time monitoring module is used to monitor the training data in the subsequent training process in real time based on the adjustment suggestions in the personalized feedback report, and update the action analysis results based on the training data obtained from the monitoring.

[0112] The comparison and evaluation module is used to compare and analyze the action analysis results before and after the update, and to calculate the continuity evaluation index.

[0113] The path planning module is used to generate visual charts based on continuous evaluation indicators, and to determine personalized training optimization paths for volleyball players based on these charts.

[0114] Through the collaborative efforts of the aforementioned components, the system constructs an intelligent training system with multi-dimensional data closed-loop feedback, achieving precise teaching management throughout the entire process from motion perception to path optimization. Specifically:

[0115] The data acquisition module, serving as the foundation of the system's perception layer, overcomes the limitations of traditional single data sources through the collaborative work of wearable sensors and high-definition cameras. It achieves simultaneous acquisition of athletes' internal mechanical parameters and external spatial trajectories, providing comprehensive and accurate data support for subsequent analysis. The feature extraction module utilizes the deep feature mining capabilities of convolutional neural networks to transform raw multidimensional data into quantifiable displacement, velocity, and angle parameters, and constructs a dynamic model of the motion chain, solving the technical challenge of quantifying and evaluating motion quality in traditional teaching. The motion judgment module intelligently classifies and identifies quantified parameters based on a support vector machine model. Through preset deviation thresholds and dynamic continuity problem feature extraction, it accurately locates the weak points and problem types in athletes' technical movements, achieving a leap from vague perception to precise diagnosis. The report generation module deeply integrates objective motion data with athletes' subjective feelings and training load data through correlation analysis, generating highly targeted reports. The highly operable personalized feedback reports break down the limitations of the separation between objective data and subjective experience in traditional teaching. The real-time monitoring module sets monitoring parameters based on the adjustment suggestions in the feedback reports, continuously tracks subsequent training, and dynamically updates the movement analysis results, forming a closed-loop mechanism of "monitoring-feedback-re-monitoring" to ensure that training adjustments can respond in real time to changes in the athlete's condition. The comparative evaluation module compares and weights the movement analysis results before and after the update step by step, transforming the abstract movement improvement effect into standardized continuous evaluation indicators, providing a scientific basis for the quantitative tracking of long-term training effects. The path planning module visualizes the continuous evaluation indicators and multi-dimensional data, identifies potential problems and key links by analyzing charts, matches targeted correction methods and load strategies, and formulates phased achievement goals, ultimately forming a dynamically adjustable personalized training optimization path that effectively prevents the risk of sports injuries while improving teaching efficiency.

[0116] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the methods and systems described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0117] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-dimensional data-driven method for providing feedback on volleyball teaching effectiveness, characterized in that, Includes the following steps: Step S1: Collect real-time motion data and ball trajectory data of volleyball players during training to form the original dataset; Step S2: Extract features from the original dataset, analyze the training action chain, and obtain action analysis results; Step S3: Based on the motion analysis results, determine whether the deviation of the training motion exceeds the preset deviation threshold. If so, classify the real-time motion data and ball trajectory data and determine the problem classification label. Step S4: Based on the problem classification tags, integrate the volleyball player's subjective feeling records and training load data to generate a personalized feedback report; Step S5: Based on the adjustment suggestions in the personalized feedback report, monitor the training data in real time during the subsequent training process, and update the action analysis results based on the training data obtained from the monitoring. Step S6: Compare and analyze the action analysis results before and after the update, and calculate the continuity evaluation index; Step S7: Generate a visualization chart based on the continuity evaluation index, and determine the personalized training optimization path for volleyball players based on the visualization chart.

2. The method according to claim 1, characterized in that, Step S1 includes: Wearable sensors are used to collect data on lower limb force exertion, body coordination, and landing cushioning during volleyball players' training to obtain real-time motion data. High-definition cameras are used to record the swing sequence and trajectory data of volleyball players during training to obtain ball trajectory data; The collected real-time motion data and recorded ball trajectory data are integrated and preprocessed to generate the original dataset.

3. The method according to claim 2, characterized in that, Step S2 includes: A convolutional neural network is used to mine key feature information in the training action chain in the original dataset, and the extracted key feature information is quantified to obtain the displacement parameters, velocity parameters and angle parameters corresponding to each action link in the action chain. Based on the calculated displacement, velocity, and angle parameters, a dynamic model of the motion chain is constructed. The dynamic model is analyzed to obtain the coordination and coherence analysis results between each action link in the action chain; The coordination analysis results are integrated with the coherence analysis results to form the motion analysis results.

4. The method according to claim 1, characterized in that, Step S3 includes: Extract the deviation values ​​corresponding to the weak links in the training motion chain from the motion analysis results; Determine whether the deviation value exceeds a preset deviation threshold. If so, use a support vector machine model to classify the real-time motion data and the ball trajectory data, identify the dynamic coherence problems, and generate corresponding classification features. The problem classification label is determined based on the classification features.

5. The method according to claim 1, characterized in that, Step S4 includes: Obtain information on weak points in volleyball training movements corresponding to the problem classification tags; Extract subjective feelings and training load data of volleyball players from the database; The subjective feelings recorded, the training load data, and the problem classification labels are correlated and analyzed to clarify the correspondence among the three. Based on the corresponding relationships, a personalized feedback report is generated that includes a detailed description of the weaknesses and adjustment suggestions.

6. The method according to claim 3, characterized in that, Step S5 includes: Based on the adjustment suggestions in the personalized feedback report, set the corresponding sports data monitoring parameters for subsequent volleyball training. The system continuously collects subsequent training data according to the set motion data monitoring parameters using sensors and cameras to obtain monitoring training data. The monitoring training data is compared with the adjustment suggestions in the personalized feedback report to determine whether the monitoring training data meets the parameter requirements corresponding to the adjustment suggestions; If not, a convolutional neural network is used to extract features from the monitoring training data to obtain the displacement parameters, velocity parameters, and angle parameters corresponding to the monitoring training data. Based on the displacement, velocity, and angle parameters corresponding to the monitoring training data, the dynamic model of the motion chain is reconstructed, and the coordination and continuity of each motion link are analyzed to update the motion analysis results.

7. The method according to claim 6, characterized in that, Step S6 includes: The updated motion analysis results are matched one by one with the previous motion analysis results according to each link of the motion chain, and the changes in displacement parameters, velocity parameters and angle parameters of each corresponding link are calculated. Weights are assigned to each action link according to their importance in the action chain, and the parameter changes of each link are weighted and calculated in combination with the weights to evaluate the overall improvement of the action chain. The overall improvement of the action chain is standardized and quantified to obtain continuous evaluation indicators.

8. The method according to claim 1, characterized in that, Step S7 includes: Obtain the parameter change values ​​and overall improvement data of each link of the motion chain corresponding to the continuity evaluation index, and perform time-axis correlation mapping with the subjective feeling records and training load data of volleyball players; Based on the correlation mapping results, a multi-dimensional visualization chart is generated that includes the action improvement trend curve, training load distribution, and changes in link deviation. By analyzing the visualization charts, potential problems caused by load adaptation or movement compensation in volleyball player training can be identified, as well as persistent weak links and areas for significant improvement in the movement chain. Based on the aforementioned potential problems and the persistent weak links and significant improvement links in the motion chain, corresponding motion correction methods, training load adjustment strategies, and training frequency planning are matched. Based on the matching results and the adjustment suggestions in the personalized feedback report, phased training parameter achievement targets are formulated. Based on the phased training parameter achievement targets, a personalized training optimization path for volleyball players is determined.

9. A multi-dimensional data-driven volleyball teaching effectiveness feedback system, used to implement the multi-dimensional data-driven volleyball teaching effectiveness feedback method as described in any one of claims 1 to 8, characterized in that, The aforementioned multi-dimensional data-driven volleyball teaching effectiveness feedback system includes: The data acquisition module is used to collect real-time motion data and ball trajectory data of volleyball players during training to form the raw dataset; The feature extraction module is used to extract features from the original dataset, analyze the training action chain, and obtain action analysis results. The action judgment module is used to determine whether the deviation of the training action exceeds a preset deviation threshold based on the action analysis results. If so, the real-time motion data and ball trajectory data are classified to determine the problem classification label. The report generation module is used to generate personalized feedback reports by integrating the subjective feelings records and training load data of volleyball players based on the problem classification tags. The real-time monitoring module is used to monitor the training data in the subsequent training process in real time according to the adjustment suggestions in the personalized feedback report, and update the action analysis results based on the training data obtained from the monitoring. The comparison and evaluation module is used to compare and analyze the action analysis results before and after the update, and to calculate the continuity evaluation index. The path planning module is used to generate a visualization chart based on the continuous evaluation index, and to determine the personalized training optimization path for volleyball players based on the visualization chart.

Images