Physical exercise track data analysis system based on AI (Artificial Intelligence) identification

Abstract

The invention relates to the technical field of data analysis, in particular to a physical exercise trajectory data analysis system based on AI recognition. Comprising a data acquisition module for acquiring multi-dimensional data of a table tennis player in a movement process through a multi-modal sensor; according to the method, a multi-dimensional complete data set is formed by accurately capturing the biomechanical data of the acceleration, the angular velocity and the force output of the ping-pong athlete in the actions of running, jumping and swinging, a basis is provided for accurate analysis of the force exerting skill of the ping-pong athlete, and the force exerting skill of the ping-pong athlete is accurately analyzed through the AI model. The exercise efficiency, physical fitness condition and action precision of the ping-pong athlete are predicted, so that the power generation skill and promotion space of the ping-pong athlete are better analyzed, a personalized training plan is generated according to the defects of the athlete, the ping-pong athlete is assisted in identifying and improving the defects, and the training efficiency is improved. Powerful support is provided for improving the power generation efficiency and the ball hitting stability of the ping-pong athletes, and the training effect of the power generation skills of the ping-pong athletes can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly relates to a sports movement trajectory data analysis system based on AI recognition. Background Art

[0002] A sports movement trajectory refers to the movement route or path of an athlete's body or an object in space over time during a sports activity. This trajectory can be a straight line, a curved route, or a complex path, depending on the type of sport, the athlete's movements, and the sports environment. For example, in sports such as basketball, football, and tennis, the flight trajectory of the ball in the air can be recorded and analyzed.

[0003] The patent with the publication number CN118887739A records in its specification that "the present invention discloses a sports movement trajectory data analysis system based on AI recognition, which relates to the technical field of data analysis. The in-apparatus boundary analysis module constructs a risk coefficient Fxxs through analysis to preliminarily judge the possible net-touching risk during the ball transmission process, and provides a more accurate assessment for the later stage by optimizing the safety range. After the in-apparatus out-of-bounds analysis module analyzes the net-touch risk, it constructs a trajectory influence coefficient Gyxs based on the training data set and combines it with the training analysis model to accurately fit the training level evaluation index Sgzs, which helps to further evaluate the performance of the athlete during training to ensure the training effect and safety. The feedback module compares and analyzes the preset evaluation threshold U with the training level evaluation index Sgzs, comprehensively judges the training status, and provides targeted feedback according to the comparison result to help the athlete timely understand the training effect and adjust the training strategy and goal." Although the above technology captures the movement trajectory of the athlete during training in real time through image acquisition technology and analyzes the training data in combination with AI technology to achieve the purpose of helping the athlete and the coach timely understand the training effect and optimize the training strategy, it does not involve in-depth data analysis of the athlete's force application skills. Although factors such as the hitting action and the swing speed of the racket are crucial for the hitting performance, the core of the force application skill lies in how to efficiently utilize the strength of various parts of the body to increase the initial velocity and stability of the ball. The above technology does not accurately capture the biomechanical data of the athlete's force application, nor does it analyze the optimization space of the force application efficiency and skills in combination with key indicators such as the initial velocity of the ball. Therefore, this technology has certain limitations in helping the athlete improve the force application skill and fails to comprehensively reflect how the athlete improves the performance by optimizing the force application method in actual competitions.

[0004] In summary, developing a sports movement trajectory data analysis system based on AI recognition is still a key problem that urgently needs to be solved in the technical field of data analysis. Summary of the Invention

[0005] The object of the present invention is to solve the problems existing in the prior art. Although the above-mentioned technology captures the movement trajectory of athletes during training in real time through image acquisition technology and analyzes the training data in combination with AI technology to help athletes and coaches understand the training effect in a timely manner and optimize the training strategy, it does not involve in-depth data analysis of the athletes' force application skills. Although factors such as the hitting action and the swing speed of the racket are crucial for the hitting performance, the core of the force application skill lies in how to efficiently utilize the strength of various parts of the body to improve the initial velocity and stability of the ball. The above-mentioned technology does not accurately capture the biomechanical data of the athletes' force application, nor does it analyze the optimization space of the force application efficiency and skills in combination with key indicators such as the initial velocity of the ball. Therefore, this technology has certain limitations in helping athletes improve their force application skills and fails to comprehensively reflect the problem of how athletes improve their performance by optimizing the force application method in actual competitions.

[0006] To achieve the above object, the present invention provides a sports trajectory data analysis system based on AI recognition, including:

[0007] A data acquisition module that acquires multi-dimensional data of a table tennis athlete during movement through multi-modal sensors;

[0008] A data transmission module that transmits the multi-dimensional data acquired by the multi-modal sensors to the data processing module through wireless transmission;

[0009] A data processing module that filters, denoises, and synchronizes the received multi-dimensional data to ensure the accuracy of the multi-dimensional data and obtains optimized data;

[0010] An AI analysis module that models and analyzes the optimized data based on artificial intelligence algorithms to identify the movement indicators of table tennis athletes and obtain analysis results;

[0011] A data interaction module that actively conducts voice interaction with coaches and table tennis athletes according to the analysis results and provides training plans.

[0012] Further, the operation process of the data acquisition module includes:

[0013] Acceleration sensors are deployed on the wrists, shoulders, waists, thighs, calves, ankles, and soles of table tennis players. The acceleration formula is used to measure the acceleration changes of table tennis players during running, jumping, and racket swinging. Inertial measurement units are deployed on the rackets, wrists, and elbows of table tennis players to measure the rotation angles and speeds of the athletes. The swing actions of table tennis players are analyzed through a rigid body kinematics model. Pressure sensors are deployed on the soles and wrists of the athletes. The force output formula is used to measure the force output of the athletes during hitting, starting, and running. The force output formula: Where is the total applied force, is a time variable, is the force applied by the sole, is the force applied by the wrist, acceleration formula: where q z is the body coordinate system acceleration measurement value, is the rotation matrix from the body coordinate system to the world coordinate system, q r is the acceleration in the world coordinate system, T r is the gravitational acceleration vector, z q is the accelerometer zero bias, is the Gaussian white noise of the accelerometer, rigid body kinematic model: where is the body coordinate system angular velocity measurement value, is a time variable, r r is the angular velocity in the world coordinate system, z r is the zero bias of the gyroscope, is the Gaussian white noise of the gyroscope, y z is the magnetic field measurement value in the body coordinate system, is the rotation matrix from the body coordinate system to the world coordinate system, y r the magnetic field in the world coordinate system, is the Gaussian white noise of the magnetic field sensor.

[0014] Furthermore, the operation process of the data acquisition module includes:

[0015] Deploy high-speed cameras on the sports field, use the position change formula to record the position information of athletes in real time, analyze the movement trajectory and reaction ability of athletes on the field, and then synthesize the data of acceleration, swing angle, force output and position change through the data fusion formula to obtain multi-dimensional data, position change formula: where is the position vector of the athlete, are the horizontal and vertical position coordinates of the athlete on the field, is a time variable, data fusion formula: where is the fused multi-dimensional data, is the acceleration data, is the angular velocity data, is the total applied force, is the position data, is a time variable.

[0016] Furthermore, the operation process of the data transmission module includes:

[0017] Perform noise filtering, data collation, and timestamp marking on the multi-dimensional data obtained by the multi-modal sensor to ensure the quality and format uniformity of the multi-dimensional data. Integrate and package the multi-dimensional data. The process of integrating and packaging multi-dimensional data includes organizing the data from different sensors into a data packet according to certain rules. The data packet contains time information, sensor type, and the acquired data values. Transmit the data packet from the multi-modal sensor to the data transmission module through wireless communication technology. The data transmission module is responsible for summarizing the received data packets, compressing and encrypting the data packets, and then sending the compressed and encrypted data packets to the data processing module through the wireless network. Noise filtering formula: where is the data value after filtering, is the i-th raw data, h i is the weight coefficient, N represents the number of data points used for filtering, is the time variable, Timestamp marking formula: where is the data packet with timestamp and sensor type, T represents the acquisition time of this data point, J i represents the sensor from which the data originates, is the i-th raw data.

[0018] Furthermore, the operation process of the data processing module includes:

[0019] Receive the data packet from the data transmission module for further unpacking and parsing, extract the multi-dimensional data from the data packet, perform filtering processing on the multi-dimensional data through the low-pass filter formula to remove high-frequency noise and interference, perform denoising processing on the multi-dimensional data using the median filter formula, and detect and remove abnormal data by taking the median of adjacent data points. Low-pass filter formula: where represents the signal after filtering at time moment, K(α) is the data value collected at time α, α is the time constant, is the current time point, is the weight function of the filter, mα is the infinitesimal increment of the integral, Median filter formula: where is the data after denoising, is the current data point and its adjacent data points before and after.

[0020] Furthermore, the operation process of the data processing module includes:

[0021] Since the multi-modal sensors work independently and have different sampling frequencies, it is necessary to compare the timestamps of multi-dimensional data through the time synchronization formula and align them according to a unified time axis. After completing filtering, denoising, and synchronization processing, the processed multi-dimensional data is further analyzed and refined through the principal component analysis formula, including data smoothing, feature extraction, and dimensionality reduction, to reduce redundant information in the multi-dimensional data and output optimized data. The time synchronization formula: where is the synchronized data, is the data from multiple sensors at time , Interp{·} is used for interpolation between data at different timestamps. The principal component analysis formula: K PCA = L·Z, where K PCA is the data after principal component analysis, L is the eigenvector of the data matrix, and Z is the eigenvalue of the data matrix.

[0022] Furthermore, the operation process of the AI analysis module includes:

[0023] Taking the optimized data as the AI model training data and scaling it to between [0, 1] through the normalization formula. Based on the artificial intelligence algorithm, by analyzing the acceleration data, step data, and total applied force data of the table tennis player in the AI model training data, calculate the racket swing speed, step frequency, and hitting force. The AI model uses regression models and classification models to predict continuous numerical values and discrete categories, and performs training and learning according to the gradient descent formula based on the AI model training data to adjust the AI model parameters. The normalization formula: where C norm is the normalized data, C is the original data, C min is the minimum value in the original data,

[0024] C max is the maximum value in the original data. The gradient descent formula: where V is the weight parameter of the model, β is the learning rate, is the gradient of the loss function B with respect to the weight Q.

[0025] Furthermore, the operation process of the AI analysis module includes:

[0026] After the AI model training is completed, predictions and analyses are performed based on the real-time optimized data, and the analysis results are output. The analysis results include the motion efficiency data, physical fitness data, and action accuracy data of the table tennis player. Through the accumulation of AI model training data, the AI model will continue to be adjusted and optimized. The prediction formula: where H′ pred is the predicted value, W′ 1 ,W′2 ,...,W′ n′ are weight coefficients, E′ 1 ,E′ 2 ,…,E′ n′ are input features, R′ is a constant term, Y′(U′=O′|E′) is the probability that the output class U′ equals O′ given the input feature E′, is the score of class O′, W′ O′ is the weight vector of class O′, E′ is the input feature vector, R′ O′ is the bias term of class O′, L′ is the total number of classes, is the normalization term, adjusting and optimizing the formula: where J′ is the loss function, N′ is the number of training samples, is the true label of the i-th sample, is the predicted value of the i-th sample.

[0027] Furthermore, the operation process of the data interaction module includes:

[0028] Receiving the analysis result, the analysis result includes the indexes of the athlete's sports performance, physical fitness status, and action precision. After the AI model recognizes that the table tennis player is in the rest time through the high-speed camera, it uses the speech synthesis technology to communicate with the table tennis player and / or coach by voice, and actively reminds the table tennis player of the deficiencies and progress in training. The speech synthesis formula: where is the text content to be converted into speech, and V′ is the synthesized speech signal.

[0029] Furthermore, the operation process of the data interaction module includes:

[0030] Generating a targeted training plan according to the items where the table tennis player performs poorly, converting the voice into text through the speech recognition technology, understanding the needs and / or questions of the table tennis player and / or coach, and generating a reply through the speech synthesis technology. After each training session, the AI model records the performance of the athlete, generates a complete training report, and summarizes it with the athlete and coach by voice. The speech recognition formula: where V″ is the input speech signal, is the output text information.

[0031] Beneficial effects

[0032] Adopting the technical solution provided by the present invention, compared with the known public technology, it has the following

[0033] Beneficial effects:

[0034] When in use, the present invention accurately captures the biomechanical data of acceleration, angular velocity, and force output of table tennis players during running, jumping, and racket-swinging actions, forming a complete multi-dimensional data set, which provides a basis for the precise analysis of the table tennis players' force application skills. Through the AI model, it predicts the movement efficiency, physical fitness, and movement accuracy of table tennis players, thereby better analyzing the force application skills and improvement space of table tennis players, generating personalized training plans according to the deficiencies of the players, assisting table tennis players in identifying and improving deficiencies, providing strong support for improving the force application efficiency and hitting stability of table tennis players, and being beneficial to improving the training effect of table tennis players' force application skills. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a system diagram of a sports trajectory data analysis system based on AI recognition of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0038] The present invention will be further described in detail below in conjunction with the accompanying drawings:

[0039] Embodiment:

[0040] As Figure 1 shown, the present invention provides a sports trajectory data analysis system based on AI recognition, including: a data acquisition module that acquires multi-dimensional data of table tennis players during the movement process through multi-modal sensors;

[0041] Furthermore, the operation process of the data acquisition module includes:

[0042] Acceleration sensors are deployed on the wrists, shoulders, waists, thighs, calves, ankles and soles of table tennis players. The acceleration formula is used to measure the acceleration changes of table tennis players during running, jumping and swinging the racket. Inertial measurement units are deployed on the rackets, wrists and elbows of table tennis players to measure the rotation angles and speeds of the players. The swing actions of table tennis players are analyzed through the rigid body kinematics model. Pressure sensors are deployed on the soles and wrists of the players. The force output formula is used to measure the force output of the players during hitting, starting and running. The force output formula: Where is the total applied force, is the time variable, is the force applied by the sole, is the force applied by the wrist. The acceleration formula: Where q z is the body coordinate system acceleration measurement value, is the rotation matrix from the body coordinate system to the world coordinate system, q r is the acceleration in the world coordinate system, T r is the gravitational acceleration vector, z q is the accelerometer zero bias, is the Gaussian white noise of the accelerometer. The rigid body kinematics model: Where is the body coordinate system angular velocity measurement value, is the time variable, r r is the angular velocity in the world coordinate system, z r is the zero bias of the gyroscope, is the Gaussian white noise of the gyroscope, y z is the magnetic field measurement value in the body coordinate system, is the rotation matrix from the body coordinate system to the world coordinate system, y r is the magnetic field in the world coordinate system, is the Gaussian white noise of the magnetic field sensor.

[0043] Furthermore, the operation process of the data acquisition module includes:

[0044] High-speed cameras are deployed on the sports field. The position change formula is used to record the position information of the athletes in real time, analyze the movement trajectories and reaction capabilities of the athletes in the field, and then synthesize the data of acceleration, swing angle, force output and position change through the data fusion formula to obtain multi-dimensional data. The position change formula: Where is the position vector of the athlete, is the horizontal and vertical position coordinates of the athlete in the field, is a time variable, and the data fusion formula is: where is the multi-dimensional data after fusion, is the acceleration data, is the angular velocity data, is the total applied force, is the position data, is the time variable;

[0045] Specifically, the data acquisition module uses technical means such as the acceleration formula, rigid body kinematic model, and force output formula to facilitate the accurate capture of key data on the acceleration, angular velocity, and force output of athletes. These data are integrated through the position change formula and data fusion formula to obtain a complete multi-dimensional data set, which is beneficial to accurately capture the biomechanical indicators of athletes in various motion states, providing accurate data support for subsequent training optimization, motion analysis, and sports performance evaluation. At the same time, the diversified layout of sensors and the application of data fusion technology enable the comprehensive recording and accurate analysis of the multi-dimensional performance of athletes.

[0046] The data transmission module transmits the multi-dimensional data obtained by the multi-modal sensors to the data processing module through wireless transmission;

[0047] Furthermore, the operation process of the data transmission module includes:

[0048] Filter the noise, organize the data, and mark the time stamps for the multi-dimensional data obtained by the multi-modal sensors to ensure the quality and format uniformity of the multi-dimensional data. Integrate and package the multi-dimensional data. The process of integrating and packaging the multi-dimensional data includes organizing the data of different sensors into a data packet according to certain rules. The data packet contains time information, sensor type, and the obtained data values. Transmit the data packet from the multi-modal sensors to the data transmission module through wireless communication technology. The data transmission module is responsible for summarizing the received data packets, compressing and encrypting the data packets, and then sending the compressed and encrypted data packets to the data processing module through the wireless network. The noise filtering formula: where is the filtered data value, is the i-th raw data, h i is the weight coefficient, and N represents the number of data points used for filtering, is the time variable, and the time stamp marking formula: where is the data packet with time stamp and sensor type, T represents the acquisition time of this data point, J i represents the sensor from which the data comes, is the i-th raw data;

[0049] Specifically, the data transmission module filters the noise of the sensor data to remove unnecessary interference; then adds accurate time information to each data point through timestamp marking to ensure the timing of the data; then, integrates the data from different sensors into a data packet according to certain rules, compresses and encrypts the data packet to ensure the security and efficiency of the data during transmission. Finally, the processed data packet is sent to the data processing module through wireless communication, which helps to ensure the data quality and format uniformity during the data transmission process, avoids analysis deviation caused by inconsistent or lost data. At the same time, the security and efficiency of data transmission are improved through compression and encryption processing, enabling multi-modal data to be quickly and reliably transmitted to the data processing module.

[0050] The data processing module performs filtering, denoising, and synchronization processing on the received multi-dimensional data to ensure the accuracy of the multi-dimensional data and obtain optimized data;

[0051] Furthermore, the operation process of the data processing module includes:

[0052] The data packet received from the data transmission module is further unpacked and parsed to extract multi-dimensional data from the data packet, and the multi-dimensional data is filtered through the low-pass filter formula to remove high-frequency noise and interference. The median filter formula is used to denoise the multi-dimensional data, and abnormal data is detected and removed by taking the median of adjacent data points. The low-pass filter formula: where represents the signal after filtering at time K(α) is the data value collected at time α, α is the time constant, is the current time point, is the weight function of the filter, mα is the tiny increment of the integral. The median filter formula: where is the data after denoising, is the current data point and its adjacent data points before and after;

[0053] Furthermore, the operation process of the data processing module includes:

[0054] Since the multi-modal sensors work independently and have different sampling frequencies, it is necessary to compare the timestamps of the multi-dimensional data through the time synchronization formula and align them according to a unified time axis. After completing the filtering, denoising, and synchronization processing, the processed multi-dimensional data is further analyzed and refined through the principal component analysis formula, including data smoothing, feature extraction, and dimensionality reduction, to reduce the redundant information in the multi-dimensional data and output optimized data. The time synchronization formula: where is the data after synchronization, is data from multiple sensors at time , Interp{·} is used for interpolation between data at different timestamps, and the principal component analysis formula: K PCA = L·Z, where K PCA is the data after principal component analysis, L is the eigenvector of the data matrix, and Z is the eigenvalue of the data matrix;

[0055] Specifically, the data processing module removes high-frequency noise and interference through low-pass filtering to ensure the stability of the data; then, it denoises the data using the median filtering method to remove outliers and ensure the reliability of the data. At the same time, it uses time synchronization technology to align data at different timestamps to ensure the consistency and accuracy of the data. Principal component analysis reduces the data dimension, which helps to reduce the computational amount and improve the interpretability and application value of the optimized data.

[0056] The AI analysis module models and analyzes the optimized data based on artificial intelligence algorithms to identify the motion indicators of table tennis players and obtain analysis results;

[0057] Furthermore, the operation process of the AI analysis module includes:

[0058] Taking the optimized data as AI model training data and scaling it to between [0, 1] through the normalization formula. Based on artificial intelligence algorithms, by analyzing the acceleration data, pace data, and total applied force data of table tennis players in the AI model training data, it calculates the swing speed, pace frequency, and hitting force. The AI model uses regression models and classification models to predict continuous numerical values and discrete categories, and trains and learns according to the gradient descent formula using the AI model training data to adjust the AI model parameters. The normalization formula: where C norm is the normalized data, C is the original data, C min is the minimum value in the original data,

[0059] C max is the maximum value in the original data, and the gradient descent formula: where V is the weight parameter of the model, β is the learning rate, is the gradient of the loss function B with respect to the weight Q;

[0060] Furthermore, the operation process of the AI analysis module includes:

[0061] After the AI model training is completed, it makes predictions and analyzes based on real-time optimized data and outputs the analysis results. The analysis results include the motion efficiency data, physical fitness status data, and action accuracy data of table tennis players. Through the accumulation of AI model training data, the AI model will continue to be adjusted and optimized. The prediction formula: where H′ pred is the predicted value, W′ 1 , W′ 2 , …, W′ n′ are the weight coefficients, E′ 1 , E′ 2 , …, E′ n′ are the input features, R′ is a constant term, and Y′(U′ = O′|E′) is the probability that the output class U′ equals O′ given the input feature E′. is the score for class O′, W′ O′ is the weight vector for class O′, E′ is the input feature vector, and R′ O′ is the bias term for class O′, and L′ is the total number of classes. is the normalization term, and the adjustment and optimization formula is: where J′ is the loss function and N′ is the number of training samples. is the true label of the i-th sample, is the predicted value of the i-th sample;

[0062] Specifically, the AI analysis module models and analyzes the optimized data through artificial intelligence algorithms, and outputs the analysis results of the movement efficiency, physical fitness, and movement accuracy of table tennis players. As the training data of the AI model accumulates, the AI model will continuously adjust and optimize to improve the prediction accuracy and precision, which is beneficial to improving the accuracy of analyzing various movement indicators of table tennis players.

[0063] The data interaction module actively conducts voice interaction with the coach and table tennis players based on the analysis results and provides training plans;

[0064] Furthermore, the operation process of the data interaction module includes:

[0065] Receiving the analysis results, which include the indicators of the athlete's movement performance, physical fitness, and movement accuracy. After the AI model recognizes through a high-speed camera that the table tennis player is in the rest time, it uses speech synthesis technology to communicate with the table tennis player and / or coach through voice, actively reminding the table tennis player of the deficiencies and progress in training. The speech synthesis formula is: where is the text content to be converted into voice, and V′ is the synthesized voice signal;

[0066] Furthermore, the operation process of the data interaction module includes:

[0067] Generate a targeted training plan based on the items where the table tennis players perform inadequately. Convert speech into text through speech recognition technology, understand the needs and / or questions of table tennis players and / or coaches, and generate responses through speech synthesis technology. After each training session, the AI model records the players' performances, generates a complete training report, and summarizes it with the players and coaches through speech. Speech recognition formula: where V″ is the input speech signal, is the output text information;

[0068] Specifically, the data interaction module actively provides feedback to the players and coaches through speech synthesis technology by receiving and analyzing data on the players' sports performances, physical fitness status, and action accuracy, reminding the players of the deficiencies and progress in training. Secondly, for the items where the players perform inadequately, a personalized training plan is generated, and the speech inputs of the players and coaches are converted into text through speech recognition technology, so as to understand their needs and questions, and then corresponding responses are generated through speech synthesis technology, which is conducive to analyzing the players' performances in real time and accurately, and helping the players identify deficiencies and improve them faster.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sports trajectory data analysis system based on AI recognition, characterized in that: include: The data acquisition module acquires multi-dimensional data of table tennis players during their movements through multi-modal sensors; A data transmission module transmits the multi-dimensional data acquired by the multi-modal sensor to the data processing module through wireless transmission; The data processing module filters, removes noise and performs synchronization processing on the received multi-dimensional data to ensure the accuracy of the multi-dimensional data and obtain optimized data; AI analysis module, which models and analyzes optimization data based on artificial intelligence algorithms, identifies the sports indicators of table tennis players and obtains analysis results; The data interaction module actively interacts with coaches and table tennis players through voice according to the analysis results and provides training plans.

2. The sports trajectory data analysis system based on AI recognition according to claim 1, characterized in that: The operation process of the data acquisition module includes: Acceleration sensors are deployed on the wrists, shoulders, waists, thighs, calves, ankles and soles of table tennis players. The acceleration formula is used to measure the acceleration changes of table tennis players during running, jumping and swinging. Inertial measurement units are deployed on table tennis players' rackets, wrists and elbows to measure the rotation angle and speed of the players. The swinging action of table tennis players is analyzed through the rigid body kinematic model. Pressure sensors are deployed on the soles and wrists of the players. The force output formula is used to measure the force output of the players during hitting, starting and running. The force output formula is: in is the total applied force, is the time variable, It is the force exerted by the sole of the shoe. is the force applied by the wrist, and the acceleration formula is: where q z is the measured value of the acceleration in the body coordinate system, is the rotation matrix from body coordinate system to world coordinate system, q r is the acceleration in the world coordinate system, T r is the gravitational acceleration vector, z q is the accelerometer bias, is the Gaussian white noise of the accelerometer, and the rigid body kinematic model is: in is the measured value of the angular velocity in the body coordinate system, is the time variable, r r is the angular velocity in the world coordinate system, z r is the zero bias of the angular velocity meter, is the Gaussian white noise of the angular velocity meter, y z is the magnetic field measurement value in the body coordinate system, is the rotation matrix from body coordinate system to world coordinate system, y r The magnetic field in the world coordinate system, is the Gaussian white noise of the magnetic field sensor.

3. A sports trajectory data analysis system based on AI recognition according to claim 2, characterized in that: The operation process of the data acquisition module includes: Deploy high-speed cameras on the sports field, and use the position change formula to record the position information of athletes in real time, analyze the movement trajectory and reaction ability of athletes in the field, and then use the data fusion formula to integrate the acceleration, swing angle, force output and position change data to obtain multi-dimensional data. The position change formula is: in is the player's position vector, are the horizontal and vertical coordinates of the athlete in the field, is a time variable, and the data fusion formula is: in It is the fused multi-dimensional data. is the acceleration data, is the angular velocity data, is the total applied force, is location data, is a time variable.

4. The sports trajectory data analysis system based on AI recognition according to claim 3, characterized in that: The operation process of the data transmission module includes: The multi-dimensional data acquired by the multi-modal sensor is subjected to noise filtering, data sorting and time stamping to ensure the quality and format uniformity of the multi-dimensional data, and the multi-dimensional data is integrated and packaged. The process of integrating and packaging multi-dimensional data includes organizing the data of different sensors into a data packet according to certain rules. The data packet contains time information, sensor type, and acquired data value. The data packet is transmitted from the multi-modal sensor to the data transmission module through wireless communication technology. The data transmission module is responsible for aggregating the received data packets, compressing and encrypting the data packets, and then sending the compressed and encrypted data packets to the data processing module through the wireless network. The noise filtering formula is: in is the filtered data value, is the i-th original data, h i is the weight coefficient, N is the number of data points used for filtering, is a time variable, and the timestamp marking formula is: in is a data packet with a timestamp and sensor type, T represents the acquisition time of the data point, and J i The sensor from which the data originates, is the i-th original data.

5. The sports trajectory data analysis system based on AI recognition according to claim 4, characterized in that: The operation process of the data processing module includes: The data packets received from the data transmission module are further unpacked and parsed, and multi-dimensional data are extracted from the data packets. The multi-dimensional data are filtered through the low-pass filter formula to remove high-frequency noise and interference. The multi-dimensional data are denoised using the median filter formula, and abnormal data are detected and removed by taking the median of adjacent data points. The low-pass filter formula is: in Indicates at time The signal after filtering at time α, K(α) is the data value collected at time α, α is the time constant, is the current time point, is the weight function of the filter, mα is a small increment of the integral, and the median filter formula is: in is the denoised data, is the current data point and its adjacent data points before and after it.

6. The sports trajectory data analysis system based on AI recognition according to claim 5, characterized in that: The operation process of the data processing module includes: Since multimodal sensors work independently and have different sampling frequencies, it is necessary to compare the timestamps of multidimensional data through the time synchronization formula and align them according to a unified time axis. After filtering, denoising and synchronization, the processed multidimensional data is further analyzed and refined through the principal component analysis formula, including data smoothing, feature extraction and dimensionality reduction, to reduce redundant information in the multidimensional data and output optimized data. Time synchronization formula: in It is the synchronized data. is from multiple sensors at the moment data, Interp{·} is used to interpolate between data at different timestamps, and the principal component analysis formula is: K PCA =L·Z, where K PCA is the data after principal component analysis, L is the eigenvector of the data matrix, and Z is the eigenvalue of the data matrix.

7. The sports trajectory data analysis system based on AI recognition according to claim 6, characterized in that: The operation process of the AI ​​analysis module includes: The optimized data is used as AI model training data and scaled to [0, 1] through the normalization formula. Based on the artificial intelligence algorithm, the acceleration data, pace data and total applied force data of the table tennis players in the AI ​​model training data are analyzed to calculate the swing speed, pace frequency and hitting force. The AI ​​model uses regression model and classification model to predict continuous values ​​and discrete categories. According to the AI ​​model training data, the gradient descent formula is used for training and learning, and the AI ​​model parameters are adjusted. The normalization formula is: Among them C norm is the normalized data, C is the original data, and C min is the minimum value in the original data, C max is the maximum value in the original data, and the gradient descent formula is: Where V is the weight parameter of the model, β is the learning rate, is the gradient of the loss function B with respect to the weight Q.

8. The sports trajectory data analysis system based on AI recognition according to claim 7, characterized in that: The operation process of the AI ​​analysis module includes: After the AI ​​model training is completed, prediction and analysis are performed based on real-time optimization data, and the analysis results are output. The analysis results include the table tennis players' sports efficiency data, physical condition data, and movement accuracy data. Through the accumulation of AI model training data, the AI ​​model will continue to adjust and optimize. The prediction formula is: where H′ pred are the predicted values, W1′,W2′,...,W n ″ is the weight coefficient, E′1, E′2, …, E′ n′ is the input feature, R′ is a constant term, Y′(U′=O′|E′) is the probability that the output category U′ is equal to O′ given the input feature E′, is the score of category O′, W′ O′ is the weight vector of category O′, E′ is the input feature vector, R′ O′ is the bias term for category O′, L′ is the total number of categories, is the normalization term, adjustment and optimization formula: Where J′ is the loss function, N′ is the number of training samples, is the true label of the i-th sample, is the predicted value of the ith sample.

9. The sports trajectory data analysis system based on AI recognition according to claim 8, characterized in that: The operation process of the data interaction module includes: Receive analysis results, which include indicators of the athlete's sports performance, physical condition, and movement accuracy. After the AI ​​model recognizes that the table tennis player is in rest time through a high-speed camera, it uses speech synthesis technology to communicate with the table tennis player and / or coach through voice, and actively reminds the table tennis player of his or her performance deficiencies and progress in training. The speech synthesis formula is: in is the text content to be converted into speech, and V′ is the synthesized speech signal.

10. The sports trajectory data analysis system based on AI recognition according to claim 9, characterized in that: The operation process of the data interaction module includes: Generate targeted training plans based on the areas where table tennis players are lacking in performance. Use speech recognition technology to convert speech into text, understand the needs and / or questions of table tennis players and / or coaches, and generate responses through speech synthesis technology. After each training session, the AI ​​model records the athletes’ performance, generates a complete training report, and summarizes it with athletes and coaches through speech. Speech recognition formula: Where V″ is the input speech signal, It is the output text information.

Images