An interactive LED display system for smart sports
By integrating multimodal sensors and environmental perception technology with machine learning algorithms, the LED display system solves the problems of lack of personalized feedback and environmental adaptability in existing technologies. It enables real-time analysis of athletes' posture and environment and personalized training suggestions, thereby improving training effectiveness and safety.
Patent Information
- Application Number
- CN202411110490.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-14
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-14
AI Technical Summary
Existing LED display systems in the field of smart sports lack the ability to analyze and display athletes' dynamic postures and detailed physiological feedback, cannot provide personalized training suggestions based on complex data analysis, and do not effectively integrate multimodal data or consider the impact of environmental variables.
It employs a combination of a multimodal motion posture capture unit, an environmental perception unit, an adaptive motion standard database, a posture comparison and analysis unit, a feedback generation unit, and an LED display unit. Through wearable sensors, high-speed cameras, pressure sensors, environmental sensors, and machine learning algorithms, it captures and analyzes athletes' motion postures and environmental data in real time, generates personalized posture correction instructions, and displays them on the LED screen.
It enables comprehensive capture and analysis of athletes' movement postures and environmental data, providing personalized training feedback, improving training efficiency, reducing injury risks, and enhancing training interactivity and educational effectiveness.
Smart Images

Figure CN118628620B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of LED display screens, and in particular to an interactive LED display system for smart sports. Background Art
[0002] In the field of smart sports, existing technologies have begun to apply LED displays to improve the interactivity and feedback efficiency of athlete training. These display systems are usually used to display basic training data, such as time, speed, and heart rate, to help athletes and coaches monitor basic training effects. However, these systems mainly display static data or simple animations, lacking in-depth data integration and personalized dynamic feedback. They are usually used as a single-function device rather than as an intelligent system that integrates multiple data sources.
[0003] The main problem with existing technologies is that, although LED display technology has been adopted, its application is usually limited to displaying simple numerical values or charts. These systems lack the ability to analyze and display athletes' dynamic postures and detailed physiological feedback, and cannot provide personalized training suggestions based on complex data analysis. In addition, existing LED display systems do not consider the impact of environmental variables on sports performance, and fail to effectively integrate multimodal data from different sensors, limiting their potential in improving training quality and athlete performance.
[0004] Therefore, there is an urgent need to develop a new interactive LED display system for smart sports. Summary of the invention
[0005] The present application provides an interactive LED display system for smart sports to improve the effectiveness and efficiency of sports training.
[0006] The present application provides an interactive LED display system for smart sports, comprising:
[0007] A multimodal motion posture capture unit configured to capture the athlete's motion posture data and contact force data with the ground or equipment in real time through wearable sensors, high-speed cameras and pressure sensors;
[0008] An environmental perception unit, which uses environmental sensors to capture environmental data of the training environment, wherein the real-time environmental data includes temperature, humidity, wind speed, and light intensity;
[0009] The adaptive sports standard database is used to store standard posture data of various sports actions; and dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level and environmental data provided by the environmental perception unit to obtain the adjusted standard posture data;
[0010] a posture comparison and analysis unit configured to compare and analyze the data captured by the multimodal motion posture capture unit with the adjusted standard posture data provided by the adaptive motion standard database in real time to obtain a posture analysis result;
[0011] A feedback generating unit, used for generating a posture correction instruction according to the posture analysis result provided by the posture comparing and analyzing unit;
[0012] The LED display unit is used to display the image of the athlete's current posture, the image corresponding to the adjusted standard posture data, specific posture adjustment suggestions and the simulated animation of the athlete's posture adjustment according to the posture correction instructions provided by the feedback generation unit.
[0013] Furthermore, the multimodal motion posture capture unit also includes an electromyography sensor and an electroencephalography sensor; the electromyography sensor is configured to monitor the athlete's muscle activity intensity in real time to provide muscle activity data about the athlete's muscle usage efficiency and fatigue level; the electroencephalography sensor is configured to monitor the athlete's brain electrical activity to provide brain electrical activity data about the athlete's mental concentration and psychological state.
[0014] Furthermore, the adaptive sports standard database is also configured to dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data to obtain adjusted standard posture data.
[0015] Furthermore, the adaptive sports standard database uses a support vector machine model to obtain adjustment parameters for standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data; and dynamically adjusts the stored standard posture data according to the adjustment parameters to obtain adjusted standard posture data;
[0016] The kernel function used by the support vector machine model is shown in the following formula 1:
[0017] ;
[0018] in, and Represent the new input feature vector and the feature vector in the training data respectively; The first feature vector representing the new input elements; Represents the feature vector in the training data The elements; the first part of the kernel function in formula 1 uses the environmental data feature vector, where Represent temperature, humidity, wind speed and light intensity respectively; the kernel function in the second part of formula 1 uses the athlete's physical characteristics, skill level, muscle activity data and EEG activity data, among which, They represent physical characteristics, skill levels, muscle activity data, and EEG activity data respectively; and It is the parameter obtained through data training; is a regularization parameter used to adjust the contribution ratio of the two kernel function parts.
[0019] Furthermore, the posture comparison and analysis unit is specifically used for:
[0020] Fast Fourier transform and wavelet transform are used to extract key frequency and time features from the time series data captured by the multimodal motion posture capture unit to obtain a structured feature vector;
[0021] Applying dynamic time warping algorithm to synchronize structured feature vectors and adjusted standard posture data, obtaining feature vector pairs of real-time data and standard posture data after time synchronization;
[0022] According to the time-synchronized real-time data and standard posture data feature vector pairs, a pre-trained deep learning model combining a convolutional neural network and a long short-term memory network is used to obtain a posture deviation analysis result.
[0023] Furthermore, the dynamic time warping algorithm calculates the minimum distance between two time series by the following formula 2:
[0024] ;
[0025] in, Indicates The structured feature vector and The Euclidean distance between the adjusted standard posture data; Indicates arrival The cumulative minimum distance of points; is a weighting function adjusted based on the athlete's fatigue and environmental factors, expressed as follows:
[0026] ;
[0027] in, Represents at a point in time Athlete fatigue; Represents at a point in time Ambient temperature; The ideal temperature for athletes to perform at their best; Represents at a point in time Relative humidity; ideal humidity for optimal performance of athletes; and It is a parameter that adjusts the influence of each factor and is obtained through experimental data.
[0028] Furthermore, the deep learning model includes a preprocessing layer, a convolutional neural network layer, a long short-term memory network layer, and a feature fusion and analysis layer;
[0029] The preprocessing layer is used to process the time-synchronized real-time data and standard posture data feature vector pairs to obtain preprocessed normalized feature vectors; the preprocessing layer uses an autoencoder to reduce noise and standardize input data to enhance the generalization ability of the model;
[0030] The convolutional neural network layer is used to process the preprocessed normalized feature vector to obtain a feature map; the convolutional neural network layer adopts a combination of multiple convolutional layers and pooling layers, and each convolution layer is followed by batch normalization and a ReLU activation function;
[0031] The long short-term memory network layer is used to process the feature map to obtain a high-level abstraction of time-dependent features;
[0032] The feature fusion and analysis layer is used to process the high-level abstraction of the time-dependent features to obtain a posture deviation classification result.
[0033] Furthermore, the feedback generating unit is specifically used for:
[0034] Obtaining key deviation parameters according to the posture analysis results provided by the posture comparison and analysis unit, wherein the key deviation parameters include angle, position and force;
[0035] Determining posture parameters that need to be adjusted based on the key deviation parameters, the posture parameters including changes in joint angles, body alignment, or balance points;
[0036] Generate a posture correction instruction according to the posture parameters that need to be adjusted.
[0037] Furthermore, the LED display unit is specifically used for:
[0038] receiving a posture correction instruction from a feedback generation unit;
[0039] According to the posture correction instruction, the image of the athlete's current posture is adjusted to obtain an image corresponding to the adjusted standard posture data, a specific posture adjustment suggestion, and a simulation animation of the athlete's posture adjustment.
[0040] Furthermore, the wearable sensor includes a wearable accelerometer and a gyroscope.
[0041] The beneficial effects of the present application mainly include: (1) By integrating wearable sensors, high-speed cameras and pressure sensors, the system can comprehensively capture the athlete's movement posture and contact force data with the equipment or the ground during training. This multimodal data capture capability enables coaches and athletes to obtain more comprehensive training feedback, helping them to more accurately analyze sports techniques, thereby optimizing training methods and improving sports performance. (2) The design of the environmental perception unit enables the system to consider factors such as temperature, humidity and light intensity in the actual training environment, and dynamically adjust the standard posture data based on these environmental data. This function ensures that training suggestions can adapt to different environmental conditions, thereby improving the safety and effectiveness of training, especially in outdoor or changing environments. (3) The adaptive motion standard database combines machine learning algorithms for posture comparison and analysis, and can provide personalized training feedback based on the athlete's personal physical characteristics and skill level. This personalized adaptation can not only improve training efficiency, but also reduce the risk of injury, making training more in line with the athlete's personal needs. (4) The LED display unit not only displays the comparison between the athlete's current posture and the standard posture, but also includes specific posture adjustment suggestions and simulation animations. This intuitive visual feedback greatly enhances the interactive and educational effects of training, allowing athletes to see in real time how their movements differ from the ideal standard and make immediate adjustments. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is a schematic diagram of an interactive LED display system for smart sports provided in the first embodiment of the present application. DETAILED DESCRIPTION
[0043] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.
[0044] The first embodiment of the present application provides an interactive LED display system for smart sports. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 The first embodiment of the present application provides an interactive LED display system for smart sports for detailed description.
[0045] The interactive LED display system includes a multimodal motion posture capturing unit 101, an environment sensing unit 102, an adaptive motion standard database 103, a posture comparison and analysis unit 104, a feedback generation unit 105 and an LED display unit 106.
[0046] The multimodal motion posture capturing unit 101 is configured to capture the motion posture data of the athlete and the contact force data with the ground or the equipment in real time through wearable sensors, high-speed cameras and pressure sensors.
[0047] The multimodal motion posture capture unit 101 is a core component of the interactive LED display system for smart sports, and is designed to provide athletes with real-time, comprehensive motion data analysis. The unit integrates wearable sensors, high-speed cameras, and pressure sensors, and through the joint work of these devices, it can capture in detail the athlete's body movements, motion postures, and interactions with equipment or the ground during training.
[0048] The role of wearable sensors in this unit is to monitor the athlete's motion parameters, such as body displacement and speed, in real time. These sensors are usually configured in key locations of the athlete, such as the wrist, chest strap or waist, to ensure the accuracy and practicality of the data.
[0049] Wearable sensors play a vital role in smart sports devices. They are key tools for real-time monitoring of athletes' movement parameters. These sensors are specially designed to capture data about athletes' body displacement and speed, which is extremely important for analyzing sports performance, optimizing training methods and preventing injuries.
[0050] In a specific implementation, the type and configuration of wearable sensors will depend on the specific sports parameters and types of sports to be monitored. For example, sensors used to capture speed and displacement usually include accelerometers and gyroscopes, which can accurately measure the acceleration and body rotation angle of athletes during exercise.
[0051] To ensure the data collected is highly accurate and reliable, sensors are typically placed at key locations on athletes:
[0052] Wrist: The wrist is an ideal location for motion sensors, especially during sports that involve frequent arm movements, such as tennis or basketball. In these locations, sensors can capture the speed and direction of the arm swing, which is particularly useful for analyzing throwing or hitting motions.
[0053] Chest strap: Chest sensors are primarily used to monitor heart rate and respiratory rate, but can also be equipped with sensors to measure velocity and displacement. Sensors located on the chest strap can continuously monitor cardiopulmonary function and core body movement without interfering with the athlete's natural movements.
[0054] Waist: The waist is another important sensor installation point, especially for tracking the overall movement and speed of athletes. The waist is a relatively stable position, suitable for placing devices that can measure running speed and distance.
[0055] Each sensor needs to have wireless transmission capabilities so that the collected data can be sent to the central processing unit in real time or directly displayed on the user interface. In addition, in order to adapt to various training environments and ensure the stability of the equipment during intense exercise, these sensors should be designed to be lightweight, durable, and have certain waterproof and sweat-proof functions.
[0056] In short, the reasonable configuration and accurate data collection of wearable sensors are crucial to the functional realization of modern smart sports systems. By placing efficient sensors at key locations, high-quality data can be obtained, which can help athletes and coaches gain a deeper understanding of training effects, optimize training plans, improve sports performance and reduce the risk of injury.
[0057] High-speed cameras are used to capture the details of athletes' movements and provide visual data support. These cameras are able to record every detail of the movement at a high frame rate, thus providing high-definition video for sports technical analysis. This is particularly important because it allows coaches and analysts to observe fast and subtle movements that may be missed by ordinary speed cameras.
[0058] Pressure sensors are installed on the soles of athletes’ shoes or on the sports equipment they use to measure the contact force with the ground or equipment. This data is crucial for analyzing athletes’ balance, stability, and force application, helping to evaluate the efficiency of technology and possible room for improvement.
[0059] The multimodal motion posture capture unit 101 collects data synchronously through these sensors and transmits the data in real time to the system's posture comparison and analysis unit 104. These data are used to instantly generate the athlete's motion posture image, analyze the correctness and efficiency of the posture, and compare with the preset standard posture data to provide instant correction feedback during training.
[0060] Furthermore, the multimodal motion posture capture unit also includes an electromyography sensor and an electroencephalography sensor; the electromyography sensor is configured to monitor the athlete's muscle activity intensity in real time to provide muscle activity data about the athlete's muscle usage efficiency and fatigue level; the electroencephalography sensor is configured to monitor the athlete's brain electrical activity to provide brain electrical activity data about the athlete's mental concentration and psychological state.
[0061] By introducing electromyography and electroencephalography sensors, the system can further capture detailed data about athletes' muscle activity and brain electrical activity, providing coaches and sports scientists with more in-depth physiological and psychological monitoring data.
[0062] The function of the EMG sensor is to monitor the electrical activity of the athlete's muscles in real time. By analyzing the strength and pattern of the EMG signal, the sensor can provide data on the efficiency of the athlete's muscle use and information on the level of muscle fatigue. This data is crucial for evaluating the athlete's physical state and adjusting the training plan, which can help avoid overtraining and muscle injuries.
[0063] The introduction of EEG sensors is to monitor the brain electrical activity of athletes during training or competition, especially brain wave patterns related to concentration, emotional state and psychological stress. EEG activity data can be used to assess athletes' mental concentration and psychological preparation, which is very valuable for optimizing sports performance and formulating psychological training interventions.
[0064] The integration of these sensors not only adds a new dimension to system monitoring, but also supports more complex data analysis and decision-making processes by providing real-time and accurate physiological and psychological data.
[0065] Furthermore, the wearable sensor includes a wearable accelerometer and a gyroscope.
[0066] In the interactive LED display system of smart sports, wearable sensors are key components, mainly used to capture and analyze athletes' movement postures and movements in real time. These sensors include accelerometers and gyroscopes, which work together to provide dynamic and static data about various parts of the athlete's body, thereby achieving accurate evaluation and optimization guidance of their sports performance.
[0067] An accelerometer is a device that measures acceleration, or how quickly an object's speed changes. In sports, an accelerometer can capture acceleration information of any part of an athlete's body, including linear acceleration and acceleration due to gravity. This data is extremely helpful in analyzing an athlete's starting speed, jump height, instantaneous explosive power, etc. For example, in sports such as running or long jump, by analyzing the acceleration data of an athlete's legs, one can assess their explosive power and the efficiency of their sports technique.
[0068] Gyroscopes are used to measure and maintain orientation. They determine the rotational motion of a device by measuring angular velocity. In sports technology applications, gyroscopes help monitor the body rotation of athletes, such as turns and tumbles. This is especially important for sports that require a high degree of body control and balance, such as gymnastics and diving. Gyroscope data can help coaches analyze the coordination and balance of various parts of an athlete's body when performing these movements, thereby providing specific guidance for improving training.
[0069] Comprehensive analysis of the data from the accelerometer and gyroscope can provide comprehensive dynamic information about the athlete during exercise. After processing, these data can be displayed as the specific movement status of each part of the athlete's body, such as speed, direction, angle change, etc., which is very suitable for in-depth analysis of movement skills and optimization of training effects.
[0070] In the system implementation, wearable sensors are usually fixed on key parts of athletes, such as wrists, waist or ankles. These sensors are small and light, and will not interfere with the natural movement of athletes. Through wireless technology, the sensors send the captured data to the central processing system in real time. The processing system processes this data according to the preset analysis model, such as comparing the real-time data with the standard posture model through the dynamic time warping algorithm, identifying possible movement deviations, and feeding back to athletes or coaches through the system to help them adjust their training plans or game strategies in real time.
[0071] In summary, wearable sensors play a vital role in smart sports systems. They not only enhance the scientificity and systematicness of training, but also help athletes achieve the best training results and competitive state through precise data analysis.
[0072] The environment sensing unit 102 uses an environment sensor to capture real-time environment data of the training environment, wherein the real-time environment data includes temperature, humidity, wind speed and light intensity.
[0073] The environmental sensing unit 102 is a key component designed for the interactive LED display system of smart sports to monitor and analyze real-time environmental data in the training environment. The unit is equipped with a variety of sensors, including temperature sensors, humidity sensors, wind speed sensors, and light intensity sensors, to ensure that the system can capture and respond to changes in the training environment, thereby providing athletes with optimized training feedback and adjustments.
[0074] Specifically, temperature sensors are used to monitor the temperature conditions of the training environment, as temperature directly affects athletes' body temperature regulation and performance. Humidity sensors monitor the water vapor content in the air, which is critical to understanding athletes' heat stress and water loss, especially when training for long endurance training or in high humidity environments. Light intensity sensors detect the light level in the environment, which affects athletes' visual clarity and safety during training. Wind speed measurements can help trainers understand the potential impact of the environment on athletes' performance. For example, training in windy conditions may affect athletes' speed and endurance because wind direction and wind force can increase or decrease the resistance to the athlete's forward movement. For throwing events, wind speed and direction are also important for predicting and adjusting throwing techniques. In addition, wind speed data is also critical to ensuring athlete safety, especially when performing high-speed and high-risk outdoor sports.
[0075] These sensors continuously collect environmental data and transmit the data to the adaptive motion standards database in real time via wireless or wired means.
[0076] The sensors should have high stability and long-term reliability to cope with changing environmental conditions. In addition, the system should be flexible enough to configure sensors according to different sports types and training venues. When installing sensors, the characteristics of their location and environment should be considered to ensure that the accuracy of the data is not disturbed.
[0077] The adaptive sports standard database 103 is used to store standard posture data of various sports movements; and dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level and environmental data provided by the environmental perception unit to obtain adjusted standard posture data.
[0078] The adaptive sports standard database 103 is an advanced data management system designed to store and dynamically adjust standard posture data of sports movements. The core function of this database is that its ability is not limited to static data storage, but also includes real-time adjustment of data according to individual differences of athletes and environmental factors to provide training feedback that best suits current conditions.
[0079] The database contains a wide range of standard sports posture data, covering a variety of sports types from basic sports skills to advanced competitive movements. The data of each sports action records the ideal posture parameters in detail, including but not limited to the angles of various parts of the body, force distribution and action sequence.
[0080] In order to adapt these standard data to the specific needs of individual athletes, the adaptive sports standard database can adjust the data according to the environmental data (such as temperature, humidity, wind speed and light intensity) collected from the environmental sensing unit 102, as well as the physical characteristics and skill level of the athletes. This dynamic adjustment is achieved through advanced algorithms that take into account the principles of sports physiology and sports training, ensuring that each adjustment can maximize the training effect and reduce the risk of injury.
[0081] An athlete's physical characteristics generally refer to the athlete's body structure and physiological conditions, including but not limited to height, weight, muscle mass, bone density, and other physiological indicators that may affect athletic performance. For example, a taller basketball player may require different training standards to optimize his speed and agility, while a lighter long-distance runner may focus more on endurance and speed training. In addition, physical characteristics also cover possible physical limitations or existing health conditions of athletes, such as joint flexibility, cardiopulmonary function, and chronic injury history.
[0082] Skill level refers to an athlete's technical proficiency and experience in a specific sport. This includes the athlete's mastery of sports techniques, tactical understanding, and psychological qualities related to competitive level. For example, an experienced soccer player may have a high level of skill in ball control and shooting techniques, while a beginner needs to start with basic skill training. The assessment of skill level can help coaches and training systems design more personalized training plans to improve athletes' weaknesses and further enhance their strengths.
[0083] The Adaptive Sports Standard Database integrates information about these physical characteristics and skill levels to dynamically adjust standard posture data during training. This adjustment is achieved through algorithms that are based on the principles of sports physiology and sports training to ensure that the training suggestions provided are both in line with the individual conditions of the athlete and can effectively improve their sports performance while minimizing the risk of injury.
[0084] For example, if the data shows that an athlete is unable to achieve the predetermined standard posture in the execution of a certain action, the system may automatically adjust the standard of this posture according to the athlete's specific physical characteristics and skill level, or recommend an alternative action that is more suitable for his current ability level. This flexibility and adaptability is the core advantage of modern smart sports training systems, which can ensure that every athlete can train under the most suitable conditions to achieve the best training effect.
[0085] The implementation of the database involves building a scalable data architecture, where the data input module is responsible for receiving real-time data from the environmental perception unit and the motion posture capture unit. A central processing unit (CPU) is responsible for processing these inputs and using a preset machine learning model to predict and implement the best adjustments to the standard posture data. All data updates and queries are performed through high-speed data interfaces to ensure the system's responsiveness and processing efficiency.
[0086] The database system needs to be able to handle large amounts of data input and have efficient data processing and query capabilities. The database design should take into account ease of use and security to ensure data confidentiality and integrity. In addition, the system should have a user-friendly interface that allows coaches and sports science experts to easily adjust the parameters in the database as needed to adapt to the training needs of different athletes.
[0087] Through such a configuration, the adaptive motion standard database 103 not only improves the personalization and adaptability of training, but also improves sports performance and safety by precisely adjusting the training plan, making sports training more scientific and effective.
[0088] Furthermore, the adaptive sports standard database is also configured to dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data to obtain adjusted standard posture data.
[0089] The adaptive motion standard database first stores detailed standard motion posture data that provide ideal execution standards for various sports actions. Then, the system collects comprehensive data about athletes and their training environment in real time through an integrated sensor network, including environmental sensors, electromyography sensors, and electroencephalography sensors.
[0090] Environmental sensors provide data such as temperature, humidity, wind speed, and light intensity, which help the system understand the external conditions the athlete is in. Electromyography sensors monitor the intensity of muscle activity during training, providing data on muscle efficiency and fatigue levels. Electroencephalography sensors assess the athlete's mental concentration and psychological state, obtaining this information by analyzing brain wave activity.
[0091] The Adaptive Sports Standard Database uses this multi-dimensional data to dynamically adjust the stored standard posture data through a built-in algorithm. This process takes into account the individual differences of athletes, such as physical characteristics and skill levels, as well as their physiological and psychological conditions. For example, if an athlete exhibits high muscle fatigue, the database may adjust their training posture to reduce stress on specific muscle groups and prevent overtraining.
[0092] These adjustments are made using advanced data analysis and machine learning technology, ensuring that each adjustment is precisely matched to the athlete's current state and needs. In this way, each athlete can receive fully customized training guidance to help them reach their peak performance at their best.
[0093] Furthermore, the adaptive sports standard database uses a support vector machine model to obtain adjustment parameters for standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data; and dynamically adjusts the stored standard posture data according to the adjustment parameters to obtain adjusted standard posture data;
[0094] The kernel function used by the support vector machine model is shown in the following formula 1:
[0095] ;
[0096] in, and Represent the new input feature vector and the feature vector in the training data respectively; The first feature vector representing the new input elements; Represents the feature vector in the training data The elements; the first part of the kernel function in formula 1 uses the environmental data feature vector, where Represent temperature, humidity, wind speed and light intensity respectively; the kernel function in the second part of formula 1 uses the athlete's physical characteristics, skill level, muscle activity data and EEG activity data, among which, They represent physical characteristics, skill levels, muscle activity data, and EEG activity data respectively; and It is the parameter obtained through data training; is a regularization parameter used to adjust the contribution ratio of the two kernel function parts.
[0097] Formula 1 provides a support vector machine (SVM) kernel function, which is specially designed to dynamically adjust the motion posture data based on the individual characteristics of the athlete and environmental factors. This kernel function combines the two parts of environmental data and athlete physiological data to find the optimal hyperplane in high-dimensional space for effective classification or regression analysis. The following is a detailed explanation of each part of the formula:
[0098] Components of the kernel function:
[0099] 1. Eigenvectors x and :
[0100] A feature vector representing new input, which is a real-time dataset at the current point of analysis, including the athlete’s environmental and physiological information.
[0101] Represents the feature vectors in the training data, which are previously collected and labeled data points used to train the SVM model.
[0102] 2. Characteristic elements and :
[0103] is the first elements such as temperature, humidity, etc.
[0104] is the training data feature vector The elements, and Corresponding.
[0105] 3. Two parts of the kernel function:
[0106] Part 1: Using the environmental data feature vector (temperature ,humidity , wind speed , light intensity ), this part is processed by Gaussian kernel calculation, and the expression is as follows:
[0107] ;
[0108] This reflects the nonlinear relationship between environmental factors and athletic performance.
[0109] Part II: Physical Characteristics of Athletes , Skill Level , muscle activity data and EEG activity data , processed by the hyperbolic tangent kernel, the form is as follows:
[0110] ;
[0111] This part of the kernel function processes more individual physiological and skill characteristics.
[0112] Function and acquisition method of parameters:
[0113] and : These are parameters obtained through data training and are used to adjust the influence and sensitivity of each part of the kernel function.
[0114] and Mainly adjust the impact of environmental factors.
[0115] Adjust the kernel function part for individual physiological and skill characteristics.
[0116] is the regularization parameter, which helps balance the contribution of the two parts in the kernel function to prevent overfitting and ensure the generalization ability of the model when processing new data.
[0117] The adaptive sports standard database uses a support vector machine (SVM) model to dynamically adjust the standard posture data of sports movements. This process uses a variety of athletes' physiological and environmental data to ensure that the posture adjustment not only meets the individual needs of athletes, but also adapts to their training environment. The implementation of this function is of great significance to improving athletes' training effects and reducing sports injuries.
[0118] In this system, the support vector machine model is the core technology used to analyze and process input feature vectors, which include athletes' physiological data and environmental data. SVM is a supervised learning method widely used in classification and regression analysis. It distinguishes different categories or makes predictions by finding the optimal hyperplane in the data.
[0119] Specifically, the SVM model uses a specially designed kernel function, and formula 1 describes the calculation method of this kernel function. The role of the kernel function is to map the input data into a high-dimensional space, making it easier to linearly segment the data in the new space. The kernel function used here is a combined kernel, which consists of two parts: one is a Gaussian kernel that uses environmental data (temperature, humidity, wind speed, and light intensity), and the other is a hyperbolic tangent kernel that uses the athlete's physical characteristics, skill level, muscle activity data, and EEG activity data.
[0120] In the kernel function, each part has its own parameter settings. These parameters It is obtained through data training to adjust the model's sensitivity to different data types. The regularization parameter σ is used to balance the contribution of the two kernel function parts to ensure that the model can capture both environmental effects and individual physiological differences.
[0121] When implementing this system, the following steps need to be performed:
[0122] 1. Data collection: Collect necessary environmental and physiological data through an integrated sensor system.
[0123] 2. Data preprocessing: Clean and format the collected data to ensure that it is suitable for input into the SVM model.
[0124] 3. Model training: Use historical data to train the SVM model and optimize the parameters in the kernel function to ensure the accuracy and robustness of the model.
[0125] 4. Dynamic adjustment: Use the SVM model to process newly collected data in real time, and dynamically adjust the stored standard posture data based on the model output.
[0126] 5. Output updated posture data: The adjusted posture data is used by coaches and athletes in training and competition to help them optimize performance and reduce the risk of injury.
[0127] In the interactive LED display system of smart sports, using the support vector machine (SVM) model to dynamically adjust the standard posture data is a key technology. The following is a reference implementation of the model, including the definition and parameter setting of the kernel function, and how to use these parameters for calculation in the model.
[0128] import numpy as np
[0129] from sklearn.svm import SVC
[0130] from scipy.spatial.distance import cdist
[0131] # Custom kernel function
[0132] def custom_kernel(X, Y):
[0133] """
[0134] Customized SVM kernel function, combining environmental factors and individual physiological characteristics.
[0135] X, Y: Input feature vector matrix, each row represents a sample, the first four columns are environmental data, and the last four columns are physiological data.
[0136] """
[0137] # Extract environmental data and physiological data
[0138] X_env = X[:, :4] # The first four columns are environment data
[0139] Y_env = Y[:, :4]
[0140] X_bio = X[:, 4:] # The last four columns are physiological data
[0141] Y_bio = Y[:, 4:]
[0142] # Environmental data core calculation part
[0143] gamma_env = 0.1 # gamma value of environmental data
[0144] alpha = np.array([1, 1, 1, 1]) # Environmental factor weight
[0145] env_kernel = np.exp(-gamma_env np.sum(alpha (cdist(X_env,Y_env,'sqeuclidean')), axis=1))
[0146] # Physiological data core calculation part
[0147] gamma_bio = 0.1 # gamma value of physiological data
[0148] beta = np.array([1, 1, 1, 1]) # Physiological factor weight
[0149] omega = np.array([1, 1, 1, 1]) # Hyperbolic tangent parameter
[0150] b = np.array([0, 0, 0, 0]) # Hyperbolic tangent bias
[0151] bio_kernel = np.sum(beta np.tanh(omega np.abs(cdist(X_bio,Y_bio,'euclidean')) + b), axis=1)
[0152] # Combine the two kernel functions
[0153] sigma = 0.5 # Regularization parameter
[0154] combined_kernel = sigma (env_kernel + bio_kernel)
[0155] return combined_kernel
[0156] # SVM model using custom kernel function
[0157] model = SVC(kernel=custom_kernel)
[0158] # Assume there is training data X_train and label y_train
[0159] X_train = np.random.rand(100, 8) # 100 samples, the first 4 columns are environmental data, the last 4 columns are physiological data
[0160] y_train = np.random.randint(0, 2, 100) # Binary classification problem
[0161] # Train the model
[0162] model.fit(X_train, y_train)
[0163] # Use the model to make predictions
[0164] X_test = np.random.rand(10, 8) # test data
[0165] predictions = model.predict(X_test)
[0166] print("Prediction results:", predictions)
[0167] This code defines a custom SVM kernel function, which includes two parts: one part processes environmental data and uses a Gaussian kernel (RBF kernel) for calculation; the other part processes individual physiological data and uses a hyperbolic tangent function. This combined kernel design allows the model to show better flexibility and adaptability when dealing with environmental factors and individual differences.
[0168] Through the above codes and instructions, technicians can better understand how to implement and adjust this support vector machine model to meet the needs of dynamic posture adjustment in smart sports systems.
[0169] The posture comparison and analysis unit 104 is configured to use a machine learning algorithm to perform real-time comparison and analysis on the data captured by the multimodal motion posture capture unit and the adjusted standard posture data provided by the adaptive motion standard database to obtain a posture analysis result.
[0170] The posture comparison and analysis unit 104 is a key component of the smart sports interactive LED display system, responsible for real-time analysis of the athlete's posture and comparison with the preset standard posture. The unit uses machine learning algorithms to process and analyze complex data collected from the multimodal motion posture capture unit 101, which includes dynamic motion data obtained through wearable sensors, high-speed cameras and pressure sensors.
[0171] To achieve this function, the posture comparison and analysis unit contains high-performance computing hardware and software, which are specially designed to process high-frequency and high-complexity data. The application of machine learning algorithms here is mainly to identify the deviation between the athlete's current movements and the standard movements stored in the adaptive sports standard database 103. These standard movement data have been adjusted in advance according to the athlete's personal characteristics and environmental factors to ensure the accuracy and practicality of the comparison.
[0172] The algorithm first receives input data, including the athlete's motion posture data. Then, it performs data preprocessing, including normalization, denoising, and feature extraction, to optimize the data analysis effect. Next, the algorithm uses the trained model to analyze the input data in real time, comparing the athlete's current posture with the standard posture in the database.
[0173] To implement this unit, first, it is necessary to select appropriate machine learning models and algorithms (such as neural networks, decision trees, or support vector machines) that can handle time series data and can be trained and predicted on multi-dimensional data. Secondly, the model must be properly trained, using historical data sets to train the algorithm to ensure its accuracy and robustness in practical applications.
[0174] In addition, the design of the entire system should take into account ease of use and maintainability. The selection of software and hardware should be able to support fast data processing and real-time feedback, while also considering the scalability of the system so that more types of sensor data or more complex analysis models can be introduced in the future.
[0175] With this configuration, the posture comparison and analysis unit can not only provide athletes with precise training feedback, but also continuously optimize its performance through continuous learning and adjustment as more data is accumulated, thereby achieving long-term improvement in athlete training and performance.
[0176] Furthermore, the posture comparison and analysis unit is specifically used for:
[0177] Fast Fourier transform and wavelet transform are used to extract key frequency and time features from the time series data captured by the multimodal motion posture capture unit to obtain a structured feature vector;
[0178] Applying dynamic time warping algorithm to synchronize structured feature vectors and adjusted standard posture data, obtaining feature vector pairs of real-time data and standard posture data after time synchronization;
[0179] According to the time-synchronized real-time data and standard posture data feature vector pairs, a pre-trained deep learning model combining a convolutional neural network and a long short-term memory network is used to obtain a posture deviation analysis result.
[0180] The main task of the posture comparison and analysis unit is to extract key frequency and time features from the time series data obtained by the multimodal motion posture capture unit, and then use these features to synchronize and compare the real-time data with the adjusted standard posture data, and finally use deep learning technology to analyze the posture deviation. This process involves a series of complex data processing, analysis and machine learning steps, which are described in detail below.
[0181] First, the posture comparison and analysis unit uses fast Fourier transform (FFT) and wavelet transform to process the time series data collected from the motion posture capture unit. FFT is an effective method for converting time domain signals to frequency domain, which helps to identify periodic changes and main frequency components in the signal. Wavelet transform provides local information of time and frequency, allowing the system to detect non-periodic and instantaneous changes in the data. Through these two transforms, the system is able to extract key frequency and time features that characterize motion characteristics from the original motion capture data, which are further organized into structured feature vectors.
[0182] Next, a dynamic time warping (DTW) algorithm is used to synchronize these structured feature vectors with the adjusted standard posture data. DTW is a powerful algorithm used to measure the similarity between two time series and synchronize them, even if they differ in time or speed. In this system, DTW helps ensure that the real-time monitored athlete posture data can be effectively compared with the pre-defined standard posture data, thereby accurately identifying deviations in time and movement.
[0183] Finally, a deep learning model that combines convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) processes the synchronized data pairs to analyze and identify posture deviations. CNNs are very effective at processing spatial features in image and video data, while LSTMs excel at processing temporal dependencies in time series data, making them well suited for processing and analyzing complex action sequences. This deep learning model is pre-trained to learn to identify deviation patterns for various postures based on historical data.
[0184] The output of this deep learning model is a detailed analysis of the deviation between the athlete's current posture and the standard posture, including the type and degree of deviation and specific adjustment suggestions. These analysis results can then be provided to athletes and coaches in an intuitive way through the LED display system, helping them to adjust and improve training effects in real time.
[0185] Furthermore, the dynamic time warping algorithm calculates the minimum distance between two time series by the following formula 2:
[0186] ;
[0187] in, Indicates The structured feature vector and The Euclidean distance between the adjusted standard posture data; Indicates arrival The cumulative minimum distance of points; is a weighting function adjusted based on the athlete's fatigue and environmental factors, expressed as follows:
[0188] ;
[0189] in, Represents at a point in time Athlete fatigue; Represents at a point in time Ambient temperature; The ideal temperature for athletes to perform at their best; Represents at a point in time Relative humidity; ideal humidity for optimal performance of athletes; and It is a parameter that adjusts the influence of each factor and is obtained through experimental data.
[0190] The dynamic time warping (DTW) algorithm is used in the interactive LED display system for smart sports to analyze the athlete's posture data. The DTW algorithm is used here to quantify and analyze the similarity between two time series, that is, the difference between the real-time captured motion posture and the pre-set standard posture. This method is particularly suitable for processing sequence data that may differ in speed and time, making it possible to synchronize and compare data at different time points, thereby accurately identifying deviations and inconsistencies.
[0191] When implementing this algorithm, we first need to calculate the distance between the corresponding points in the two sequences. Here we use the Euclidean distance , which measures the The structured feature vector of the real-time data point is The Euclidean distance is the most direct way to quantify the similarity or difference between two data points in the feature space.
[0192] Next, the DTW algorithm finds the best correspondence between two time series by recursively calculating the path with the smallest cumulative distance among all possible paths. This is achieved through the following formula:
[0193] ;
[0194] here Represents the minimum cumulative distance from the previous data point to the current data point. This approach ensures that the algorithm considers all possible paths to find the matching method with the lowest cost.
[0195] In addition, a weight function is introduced here , which is based on the athlete's fatigue and environmental factors (temperature and humidity Adjusted. The weight function is in the form of:
[0196] ;
[0197] in and The weight adjustment mechanism allows the algorithm to take into account the possible impact of athletes' performance in different physical and psychological states, and then make appropriate adjustments to the data analysis.
[0198] The implementation of this dynamic time warping algorithm includes: first, it is necessary to collect and preprocess time series data suitable for DTW analysis. Then, according to the actual application scenario and requirements, adjust and optimize the parameters in the DTW algorithm, such as the weight function and ,as well as and Finally, this algorithm will be integrated into the smart sports system to ensure real-time analysis and feedback to athletes or coaches to help them optimize their training and competition performance.
[0199] The following is a reference implementation of the Dynamic Time Warping (DTW) algorithm. This code will simulate how to use DTW to compare two time series and calculate the minimum distance between them.
[0200] import numpy as np
[0201] def euclidean_distance(vec1, vec2):
[0202] """Calculate the Euclidean distance between two vectors"""
[0203] return np.sqrt(np.sum((vec1 - vec2) 2))
[0204] def calculate_weight(fatigue, temp, temp_opt, humidity, humidity_opt,alpha, beta, gamma):
[0205] """
[0206] Calculate the weight function w(i), taking into account the athlete's fatigue and environmental factors
[0207] fatigue: athlete fatigue
[0208] temp: current temperature
[0209] temp_opt: optimal temperature
[0210] humidity: current humidity
[0211] humidity_opt: optimal humidity
[0212] alpha, beta, gamma: weight adjustment parameters
[0213] """
[0214] return 1 / (1 + np.exp(alpha fatigue + beta (temp - temp_opt) 2 + gamma * (humidity - humidity_opt) 2))
[0215] def dynamic_time_warping(seq1, seq2, alpha, beta, gamma, temp_opt,humidity_opt):
[0216] """
[0217] Implement the Dynamic Time Warping (DTW) algorithm to calculate the similarity between two time series
[0218] seq1, seq2: two time series of input, assuming they have the same length
[0219] alpha, beta, gamma: weight adjustment parameters
[0220] temp_opt, humidity_opt: optimal temperature and humidity of the environment
[0221] """
[0222] n, m = len(seq1), len(seq2)
[0223] dtw_matrix = np.zeros((n + 1, m + 1))
[0224] dtw_matrix[0, :] = np.inf
[0225] dtw_matrix[:, 0] = np.inf
[0226] dtw_matrix[0, 0] = 0
[0227] for i in range(1, n + 1):
[0228] for j in range(1, m + 1):
[0229] fatigue = seq1[i-1][0] # Assume fatigue is at the first position in the sequence
[0230] temp = seq1[i-1][1] # Assume that the temperature is at the second position in the sequence
[0231] humidity = seq1[i-1][2] # Assume humidity is in the third position of the sequence
[0232] weight = calculate_weight(fatigue, temp, temp_opt,humidity, humidity_opt, alpha, beta, gamma)
[0233] cost = euclidean_distance(seq1[i-1], seq2[j-1])
[0234] dtw_matrix[i, j] = weight (cost + min(dtw_matrix[i-1,j], # Insertion operation
[0235] dtw_matrix[i, j-1],# delete operation
[0236] dtw_matrix[i-1, j-1])) # matching operation
[0237] return dtw_matrix[n, m]
[0238] # Example data
[0239] seq1 = np.array([[0.3, 22, 50], [0.4, 23, 51], [0.5, 24, 52]]) # Example sequence 1, including fatigue, temperature, humidity
[0240] seq2 = np.array([[0.3, 21, 50], [0.4, 22, 52], [0.6, 25, 53]]) # Example sequence 2, including fatigue, temperature, and humidity
[0241] # Parameters
[0242] alpha = 0.5
[0243] beta = 0.3
[0244] gamma = 0.2
[0245] temp_opt = 22
[0246] humidity_opt = 50
[0247] # Call DTW function
[0248] dtw_distance = dynamic_time_warping(seq1, seq2, alpha, beta, gamma,temp_opt, humidity_opt)
[0249] print("DTW distance:", dtw_distance)
[0250] The above code first defines the function for calculating the Euclidean distance and the weight function, which are used in the DTW algorithm to calculate the similarity between two time series. The initialization of the DTW matrix includes setting the boundary conditions, while the main loop calculates the minimum cumulative distance between each point, taking into account the weight factors that reflect the athlete's fatigue and the current environmental conditions. This approach ensures the flexibility and adaptability of the algorithm, enabling it to effectively analyze the athlete's posture data in changing environments and different physical states.
[0251] Furthermore, the deep learning model includes a preprocessing layer, a convolutional neural network layer, a long short-term memory network layer, and a feature fusion and analysis layer;
[0252] The preprocessing layer is used to process the time-synchronized real-time data and standard posture data feature vector pairs to obtain preprocessed normalized feature vectors; the preprocessing layer uses an autoencoder to reduce noise and standardize input data to enhance the generalization ability of the model;
[0253] The convolutional neural network layer is used to process the preprocessed normalized feature vector to obtain a feature map; the convolutional neural network layer adopts a combination of multiple convolutional layers and pooling layers, and each convolution layer is followed by batch normalization and a ReLU activation function;
[0254] The long short-term memory network layer is used to process the feature map to obtain a high-level abstraction of time-dependent features;
[0255] The feature fusion and analysis layer is used to process the high-level abstraction of the time-dependent features to obtain a posture deviation classification result.
[0256] The deep learning model in the interactive LED display system for smart sports integrates a preprocessing layer, a convolutional neural network (CNN) layer, a long short-term memory (LSTM) network layer, and a feature fusion and analysis layer. Each layer is designed for a specific data processing task to ensure that the most useful information is extracted from the motion data for effective posture analysis.
[0257] First, the main task of the preprocessing layer is to process the input feature vector pairs, which come from the dynamic time warping algorithm, including real-time data and standard posture data after time synchronization. In this layer, the autoencoder is used to reduce the noise in the data and perform normalization. The autoencoder is a neural network that removes unimportant variations and noise by learning a compressed representation of the input data, thereby improving the quality of the data and the generalization ability of the subsequent model. The processed data forms a normalized feature vector, ready for subsequent deep learning analysis.
[0258] Next, the convolutional neural network layer takes over to process these normalized feature vectors. CNN is a powerful image and video processing network suitable for extracting spatial features from normalized feature vectors. This layer contains multiple convolutional layers and pooling layers. The convolutional layers are responsible for extracting local features in the data, while the pooling layers are used to reduce the spatial dimension of the features, enhance the model's abstraction ability and reduce the amount of computation. After each convolution operation, batch normalization and ReLU activation are performed to speed up the network's convergence and increase nonlinear processing capabilities. These steps work together to generate feature maps, which are deeper spatial feature representations of the input data.
[0259] The Long Short-Term Memory network layer is key to processing time series data. LSTM networks are specifically designed to process and learn temporal dependencies in data, which is particularly important for posture data that contains temporal information. In this layer, feature maps passed from the CNN layer are used to capture and model the temporal dynamics in the action sequence, thereby generating a high-level abstract representation of temporal dependent features.
[0260] Finally, the feature fusion and analysis layer integrates and analyzes the results of the CNN and LSTM layers. The purpose of this layer is to combine spatial and temporal features to form a comprehensive posture analysis result. Through specific data fusion techniques, this layer can optimize the information obtained from the first two layers of the network and ultimately produce a decision for posture deviation classification.
[0261] The following is the reference implementation code of the deep learning model and the specific training steps:
[0262] import numpy as np
[0263] import scipy.fftpack
[0264] import pywt
[0265] from keras.layers import Input, Conv1D, BatchNormalization,Activation, LSTM, Dense, Flatten
[0266] from keras.models import Model
[0267] from keras.layers import Layer
[0268] import keras.backend as K
[0269] class NoiseReductionAutoencoder(Layer):
[0270] """
[0271] The autoencoder layers are used for noise reduction and normalization of the data.
[0272] """
[0273] def __init__(self, encoding_dim, kwargs):
[0274] self.encoding_dim = encoding_dim
[0275] super(NoiseReductionAutoencoder, self).__init__( kwargs
[0276] def build(self, input_shape):
[0277] # Encoding layer
[0278] self.encoder = Dense(self.encoding_dim, activation='relu')
[0279] # Decoding layer
[0280] self.decoder = Dense(input_shape[1], activation='sigmoid')
[0281] super(NoiseReductionAutoencoder, self).build(input_shape)
[0282] def call(self, x):
[0283] encoded = self.encoder(x)
[0284] decoded = self.decoder(encoded)
[0285] return decoded
[0286] def compute_output_shape(self, input_shape):
[0287] return input_shape
[0288] def preprocess_signals(signals):
[0289] """
[0290] Perform fast Fourier transform (FFT) and wavelet transform on the input time series data to extract key frequency and time features.
[0291] signals: input multimodal time series data, assumed to be a two-dimensional array.
[0292] """
[0293] fft_features = scipy.fftpack.fft(signals) # Fast Fourier Transform
[0294] wavelet_features, _ = pywt.cwt(signals, scales=np.arange(1, 31),wavelet='mexh') # Continuous wavelet transform
[0295] return np.hstack([fft_features.real, fft_features.imag, wavelet_features])
[0296] def build_deep_learning_model(input_shape):
[0297] """
[0298] Build a deep learning model that combines a convolutional neural network (CNN) and a long short-term memory network (LSTM), including an autoencoder preprocessing layer.
[0299] input_shape: The shape of the input data, e.g. (time steps, number of features).
[0300] """
[0301] inputs = Input(shape=input_shape)
[0302] # Preprocessing layer - Autoencoder
[0303] preprocessed = NoiseReductionAutoencoder(128)(inputs)
[0304] # Convolutional layer
[0305] conv = Conv1D(64, kernel_size=3, padding='same')(preprocessed)
[0306] norm = BatchNormalization()(conv)
[0307] relu = Activation('relu')(norm)
[0308] # LSTM layer
[0309] lstm = LSTM(128, return_sequences=True)(relu)
[0310] lstm2 = LSTM(128)(lstm)
[0311] # Output layer
[0312] output = Dense(1, activation='sigmoid')(lstm2)
[0313] model = Model(inputs=inputs, outputs=output)
[0314] model.compile(optimizer='adam', loss='binary_crossentropy',metrics=['accuracy'])
[0315] return model
[0316] # Example data
[0317] time_series_data = np.random.rand(1000, 10) # Assume there are 1000 time points and 10 types of sensor data
[0318] processed_data = preprocess_signals(time_series_data)
[0319] # Build and train the model
[0320] model = build_deep_learning_model(input_shape=(1000, 30)) # Assume that the length of the processed time series is 1000 and the number of features is 30
[0321] model.summary() # Output model structure
[0322] # Assume there are labels and training data
[0323] labels = np.random.randint(0, 2, 1000)
[0324] model.fit(processed_data.reshape(1, processed_data.shape), labels,epochs=10, batch_size=32)
[0325] The feedback generating unit 105 is used to generate a posture correction instruction according to the posture analysis result provided by the posture comparing and analyzing unit.
[0326] The feedback generation unit 105 plays a vital role in the interactive LED display system of smart sports. It generates specific posture correction instructions based on the data analysis results provided by the posture comparison and analysis unit 104. The design and implementation of this unit ensure that athletes receive real-time and specific training adjustment suggestions, thereby optimizing their sports performance and reducing the risk of injury.
[0327] The unit is mainly composed of a high-performance computing platform and dedicated software that can receive and process data from the posture comparison and analysis unit. Based on this data, the software applies predetermined logic and algorithms to determine the difference between the athlete's current movement posture and the expected standard posture, and generates targeted correction instructions accordingly. These instructions include not only basic adjustments such as changing the angle of the arms or legs, but may also include more complex suggestions, such as adjusting the power chain of the entire body.
[0328] To achieve these functions, the software of the feedback generation unit needs to have powerful data processing capabilities to ensure that the data received from the posture comparison and analysis unit can be quickly and accurately converted into specific training suggestions. In addition, the software needs to be highly configurable so that it can be personalized according to the specific needs and training conditions of different athletes.
[0329] In terms of implementation, it is necessary to ensure that the unit's algorithm can be flexibly adjusted according to different sports types and the specific performance of athletes. This includes programming knowledge and a deep understanding of kinematics and biomechanics in order to accurately interpret posture data and develop effective correction strategies. At the same time, the hardware configuration of the unit should be powerful enough to support complex calculations and data processing, ensuring that the system response time is short and athletes can get instant feedback during training.
[0330] In addition, the unit should be designed with a user-friendly interface so that coaches can adjust the feedback parameters as needed, such as adjusting the sensitivity of the feedback or modifying the specific content of the feedback. This design not only enhances the flexibility of the system, but also improves its practicality, allowing coaches and athletes to better utilize the data and suggestions provided by the system to maximize the training effect.
[0331] Furthermore, the feedback generating unit is specifically used for:
[0332] Obtaining key deviation parameters according to the posture analysis results provided by the posture comparison and analysis unit, wherein the key deviation parameters include angle, position and force;
[0333] Determining posture parameters that need to be adjusted based on the key deviation parameters, the posture parameters including changes in joint angles, body alignment, or balance points;
[0334] Generate a posture correction instruction according to the posture parameters that need to be adjusted.
[0335] In the interactive LED display system of smart sports, the feedback generation unit plays a key role. Its main task is to generate real-time posture correction instructions based on the analysis results provided by the posture comparison and analysis unit to help athletes adjust their posture to achieve the ideal sports effect. This process involves not only data analysis, but also the generation of specific instructions to ensure that athletes can understand and execute these instructions to optimize their performance.
[0336] First, the feedback generation unit receives detailed posture analysis results from the posture comparison and analysis unit. These results contain key deviation parameters of the athlete when performing the movement, which are mainly about the deviation of the athlete's angle, position and force. For example, the analysis results may point out that when the athlete performs a certain action, the angle of the arm deviates from the standard model, or the position of the body's center of gravity is not correctly aligned.
[0337] Then, based on these key deviation parameters, the feedback generation unit will determine the specific posture parameters that need to be adjusted. These posture parameters include but are not limited to joint angle adjustment, changes in body alignment, or corrections to balance points. By accurately adjusting these parameters, athletes can improve their sports skills, reduce the risk of injury, and improve sports efficiency.
[0338] Finally, once the posture parameters that need to be adjusted are determined, the feedback generation unit will generate specific posture correction instructions based on these parameters.
[0339] The LED display unit 106 is used to display the image of the athlete's current posture, the image corresponding to the adjusted standard posture data, specific posture adjustment suggestions and the simulation animation of the athlete's posture adjustment according to the posture correction instruction provided by the feedback generation unit.
[0340] The LED display unit 106 plays a vital role in the interactive LED display system of smart sports, and is responsible for presenting the output of the posture comparison and analysis unit 104 to athletes and coaches in a visual form. The core function of this unit is to display the athlete's current sports posture image, the corresponding standard posture image, specific posture adjustment suggestions, and display the athlete's posture adjustment process through simulation animation.
[0341] To achieve these functions, the LED display unit includes a high-resolution LED screen, image processing hardware, and specially designed software. The LED screen must be able to provide clear visual output under various lighting conditions, including outdoor environments under direct sunlight. The image processing hardware is used to process the data received from the feedback generation unit 105 in real time and convert it into high-quality images and animations. In addition, the software component is responsible for controlling the layout, update frequency, and user interaction of the displayed content.
[0342] Specifically, the software needs to have the ability to dynamically render images, which includes receiving and parsing posture data in real time, converting this data into graphics and text information, and managing how these contents are displayed on the LED screen. The software should also be able to process user input, allowing coaches or athletes to adjust display settings as needed, such as changing the image zoom ratio, adjusting brightness, or modifying the display position of information.
[0343] In terms of technical implementation, it is necessary to ensure that the LED display units are designed with ease of use and reliability in mind. This means that all hardware components such as screens, processors, and connectivity devices should be high-quality products that meet industry standards. In addition, software development should focus on user experience, ensuring that the interface is intuitive and easy to understand, and that it is responsive and can withstand long periods of continuous use without delays or glitches.
[0344] Furthermore, the LED display unit is specifically used for:
[0345] receiving a posture correction instruction from a feedback generation unit;
[0346] According to the posture correction instruction, the image of the athlete's current posture is adjusted to obtain an image corresponding to the adjusted standard posture data, a specific posture adjustment suggestion, and a simulation animation of the athlete's posture adjustment.
[0347] In the interactive LED display system of smart sports, the LED display unit is a key component that directly interacts with athletes. Its function is to intuitively display posture correction instructions and help athletes understand and implement these instructions in a visually visible way. The workflow and implementation details of this unit are crucial to ensure that athletes can effectively receive and apply correction suggestions.
[0348] The main task of the LED display unit begins with receiving posture correction instructions from the feedback generation unit. These instructions include specific movement adjustment requirements, such as changing the arm raising angle, adjusting the leg swing amplitude, or correcting the upright back posture. These correction instructions are based on the analysis of the posture comparison and analysis unit, reflecting the specific deviation between the athlete's current posture and the ideal standard posture.
[0349] After receiving the posture correction instructions, the next task of the LED display unit is to convert these abstract instructions into specific visual outputs. This conversion process includes three main visual display contents: the image corresponding to the adjusted standard posture data, the specific posture adjustment suggestions, and the simulation animation of the athlete's posture adjustment.
[0350] First, the display unit will display the adjusted standard posture image, which is usually presented in a more ideal sports posture, providing an intuitive target reference for athletes. By comparing with the current posture, athletes can clearly see the difference between themselves and the standard posture.
[0351] Secondly, specific posture adjustment suggestions are displayed on the LED screen through text descriptions, arrows or other graphic marks, directly pointing out the specific parts and movements that need to be improved. For example, if the arm position needs to be raised, the screen may display an upward arrow accompanied by a brief text description, such as "Raise your arms at least 15 degrees."
[0352] Finally, to help athletes better understand and simulate these adjustments, the LED display unit also displays a series of animations that simulate the athlete's process of executing the correct posture. This dynamic display helps athletes visualize the ideal movement flow and can be imitated and practiced in real time.
[0353] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.
Claims
1. An interactive LED display system for smart sports, characterized in that: include: A multimodal motion posture capture unit configured to capture the athlete's motion posture data and contact force data with the ground or equipment in real time through wearable sensors, high-speed cameras and pressure sensors; An environmental sensing unit, which uses environmental sensors to capture environmental data of the training environment, wherein the environmental data includes temperature, humidity, wind speed, and light intensity; The adaptive sports standard database is used to store standard posture data of various sports actions; and dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level and environmental data provided by the environmental perception unit to obtain the adjusted standard posture data; a posture comparison and analysis unit configured to compare and analyze the data captured by the multimodal motion posture capture unit with the adjusted standard posture data provided by the adaptive motion standard database in real time to obtain a posture analysis result; A feedback generating unit, used for generating a posture correction instruction according to the posture analysis result provided by the posture comparing and analyzing unit; An LED display unit is used to display an image of the athlete's current posture, an image corresponding to the adjusted standard posture data, specific posture adjustment suggestions, and a simulation animation of the athlete's posture adjustment according to the posture correction instruction provided by the feedback generation unit; Wherein, the multimodal motion posture capture unit further includes an electromyography sensor and an electroencephalography sensor; the electromyography sensor is configured to monitor the muscle activity intensity of the athlete in real time to provide muscle activity data on the athlete's muscle use efficiency and fatigue level; the electroencephalography sensor is configured to monitor the athlete's brain electrical activity to provide brain electrical activity data on the athlete's mental concentration and psychological state; The adaptive sports standard database is further configured to dynamically adjust the stored standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data to obtain adjusted standard posture data; The adaptive sports standard database uses a support vector machine model to obtain adjustment parameters for standard posture data according to the athlete's physical characteristics, skill level, environmental data provided by the environmental perception unit, muscle activity data and brain electrical activity data; and dynamically adjusts the stored standard posture data according to the adjustment parameters to obtain adjusted standard posture data; The kernel function used by the support vector machine model is shown in the following formula 1: ; in, and Represent the new input feature vector and the feature vector in the training data respectively; The first feature vector representing the new input elements; Represents the feature vector in the training data The elements; the first part of the kernel function in formula 1 uses the environmental data feature vector, where Represent temperature, humidity, wind speed and light intensity respectively; the kernel function in the second part of formula 1 uses the athlete's physical characteristics, skill level, muscle activity data and EEG activity data, where, They represent physical characteristics, skill levels, muscle activity data, and EEG activity data respectively; and It is the parameter obtained through data training; is the regularization parameter.
2. The interactive LED display system for smart sports according to claim 1, characterized in that: The posture comparison and analysis unit is specifically used for: Fast Fourier transform and wavelet transform are used to extract key frequency and time features from the time series data captured by the multimodal motion posture capture unit to obtain a structured feature vector; Applying dynamic time warping algorithm to synchronize structured feature vectors and adjusted standard posture data, obtaining feature vector pairs of real-time data and standard posture data after time synchronization; According to the time-synchronized real-time data and standard posture data feature vector pairs, a pre-trained deep learning model combining a convolutional neural network and a long short-term memory network is used to obtain a posture deviation analysis result.
3. The interactive LED display system for smart sports according to claim 2, characterized in that: The dynamic time warping algorithm calculates the minimum distance between two time series using the following formula 2: ; in, Indicates The structured feature vector and The Euclidean distance between the adjusted standard posture data; Indicates arrival The cumulative minimum distance of points; is a weighting function adjusted based on the athlete's fatigue and environmental factors, expressed as follows: ; in, Represents at a point in time Athlete fatigue; Represents at a point in time Ambient temperature; The ideal temperature for athletes to perform at their best; Represents at a point in time Relative humidity; ideal humidity for optimal performance of athletes; and It is a parameter that adjusts the influence of each factor and is obtained through experimental data.
4. The interactive LED display system for smart sports according to claim 2, characterized in that: The deep learning model includes a preprocessing layer, a convolutional neural network layer, a long short-term memory network layer, and a feature fusion and analysis layer; The preprocessing layer is used to process the time-synchronized real-time data and standard posture data feature vector pairs to obtain preprocessed normalized feature vectors; the preprocessing layer uses an autoencoder to reduce noise and standardize input data to enhance the generalization ability of the model; The convolutional neural network layer is used to process the preprocessed normalized feature vector to obtain a feature map; the convolutional neural network layer adopts a combination of multiple convolutional layers and pooling layers, and each convolution layer is followed by batch normalization and a ReLU activation function; The long short-term memory network layer is used to process the feature map to obtain a high-level abstraction of time-dependent features; The feature fusion and analysis layer is used to process the high-level abstraction of the time-dependent features to obtain a posture deviation classification result.
5. The interactive LED display system for smart sports according to claim 1, characterized in that: The feedback generation unit is specifically used for: Obtaining key deviation parameters according to the posture analysis results provided by the posture comparison and analysis unit, wherein the key deviation parameters include angle, position and force; Determining posture parameters that need to be adjusted based on the key deviation parameters, the posture parameters including changes in joint angles, body alignment, or balance points; Generate a posture correction instruction according to the posture parameters that need to be adjusted.
6. The interactive LED display system for smart sports according to claim 1, characterized in that: The LED display unit is specifically used for: receiving a posture correction instruction from a feedback generation unit; According to the posture correction instruction, the image of the athlete's current posture is adjusted to obtain an image corresponding to the adjusted standard posture data, a specific posture adjustment suggestion, and a simulation animation of the athlete's posture adjustment.
7. The interactive LED display system for smart sports according to claim 1, characterized in that: The wearable sensor includes a wearable accelerometer and a gyroscope.
Citation Information
Cited By
Leg force training image-based trainer adaptive detection method and system
CN120708288B
