Ankle pump motion data analysis method and system with memory function

Abstract

The invention discloses an ankle pump motion data analysis method and system with a memory function, and the method comprises the steps: collecting the motion data of a user in real time through a wearable device in an ankle pump motion process, and obtaining multi-mode motion data; according to the multi-modal motion data, a motion mode recognition model based on deep learning is adopted, features of ankle pump motion are extracted, and feature representation of a motion mode is obtained; dynamically updating a personalized exercise memory bank of the user by adopting a memory bank construction algorithm based on incremental learning according to the feature representation of the exercise mode to obtain a personalized exercise memory bank; and according to historical data in the personalized exercise memory bank, in combination with real-time exercise data of the user, an exercise optimization suggestion is generated, and through a multi-objective optimization technology, exercise parameters are dynamically adjusted to obtain an optimized exercise feedback suggestion. By means of the embodiment of the invention, the effect of ankle pump exercise can be improved, and health management and exercise performance improvement of a user are promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of ankle pumps, and in particular to an ankle pump motion data analysis method and system with a memory function. Background Art

[0002] Ankle pump exercises play an important role in rehabilitation therapy, sports training and health management, and can promote blood circulation and enhance lower limb muscle strength. However, traditional ankle pump exercise guidance mostly relies on coaching experience or fixed plans, and lacks personalized guidance. With the development of wearable devices, real-time monitoring of motion data has become possible, especially dynamic data such as acceleration, angular velocity and pressure distribution. However, these data are easily affected by noise and environmental interference, affecting accuracy. Therefore, how to remove noise and improve data quality has become an important challenge. At the same time, although deep learning technology has shown powerful capabilities in pattern recognition, existing motion analysis methods are still insufficient in terms of personalization and real-time feedback, and cannot meet the dynamic needs of users. Summary of the Invention

[0003] The purpose of the present invention is to provide an ankle pump exercise data analysis method and system with a memory function to address the deficiencies in the prior art, thereby improving the effect of ankle pump exercise and promoting the user's health management and improvement of exercise performance.

[0004] One embodiment of the present application provides a method for analyzing ankle pump motion data with a memory function, the method comprising:

[0005] During ankle pump exercise, wearable devices are used to collect user motion data in real time. The motion data includes acceleration, angular velocity, and pressure distribution. Multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data.

[0006] Based on multimodal motion data, a deep learning-based motion pattern recognition model is used to extract the characteristics of ankle pump motion. The key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a characteristic representation of the motion pattern.

[0007] According to the characteristic representation of the movement pattern, a memory library construction algorithm based on incremental learning is adopted to dynamically update the user's personalized movement memory library. Among them, the storage and retrieval efficiency of the memory library is optimized through adaptive weight adjustment technology to obtain a personalized movement memory library;

[0008] Based on the historical data in the personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm is used, combined with the user's real-time exercise data, to generate exercise optimization suggestions. Among them, through multi-objective optimization technology, the exercise parameters are dynamically adjusted to obtain optimized exercise feedback suggestions.

[0009] Optionally, during the ankle pump exercise, the user's motion data is collected in real time through a wearable device, and the motion data includes acceleration, angular velocity, and pressure distribution. Multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data, including:

[0010] During ankle pump exercise, the wearable device collects user motion data in real time through the acceleration sensor, angular velocity sensor and pressure sensor. Lightweight data caching technology is used to ensure the real-time and continuity of data collection.

[0011] For the collected multi-sensor data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources. An adaptive data cleaning algorithm is used to filter noise and fill missing values in the data to generate a preliminary standardized data set.

[0012] For the preliminary standardized data set, a dynamic data calibration algorithm based on Kalman filtering is used, combining the temporal characteristics of multi-sensor data to remove noise interference in real time. Dynamic threshold adjustment technology is used to ensure the accuracy and stability of the calibration process, generating a preliminary denoised data set.

[0013] For the preliminary denoised data set, a data integration method based on multi-sensor fusion technology is used to map the acceleration, angular velocity and pressure distribution data into unified multimodal motion data. The missing data are supplemented by the interpolation filling method to generate the final multimodal motion data.

[0014] Optionally, the method uses a deep learning-based motion pattern recognition model to extract ankle pump motion features based on multimodal motion data, wherein the key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern, including:

[0015] For multimodal motion data, we extract preliminary features of acceleration, angular velocity, and pressure distribution based on deep learning. Through a multi-head attention mechanism, we capture the correlation between different modal features and generate a preliminary multimodal feature representation.

[0016] For the preliminary multimodal feature representation, based on the spatiotemporal attention mechanism and combined with the spatiotemporal characteristics of ankle pump movement, the key spatiotemporal features of the movement are captured. The accuracy and stability of the feature representation are optimized through the dynamic weight allocation technology to generate a preliminary spatiotemporal feature representation.

[0017] For the preliminary spatiotemporal feature representation, the spatiotemporal features of different modalities are weighted and fused. Through the cross-modal attention mechanism, the correlation between the features of different modalities is captured to generate a preliminary motion pattern feature representation.

[0018] For the preliminary motion pattern feature representation, based on error feedback technology, combined with real-time motion data and historical data distribution, the feature weights are dynamically adjusted. Regularization constraints are used to prevent overfitting and generate the final motion pattern feature representation.

[0019] Optionally, the user's personalized exercise memory bank is dynamically updated based on the feature representation of the exercise pattern using a memory bank construction algorithm based on incremental learning, wherein the storage and retrieval efficiency of the memory bank is optimized through adaptive weight adjustment technology to obtain a personalized exercise memory bank, including:

[0020] Based on the feature representation of motion patterns, an incremental learning algorithm is used to combine the user's initial motion data to initialize a personalized motion memory library. Dynamic weight allocation technology is used to ensure the rationality and accuracy of memory library initialization and generate a preliminary memory library.

[0021] For the preliminary memory bank, based on the incremental learning algorithm, combined with the user's real-time motion data, the memory bank is dynamically updated. Through the adaptive weight adjustment technology, the storage and retrieval efficiency of the memory bank is optimized to generate a preliminary updated memory bank;

[0022] For the preliminary updated memory library, based on simulation, combined with the user's real-time motion data and historical data distribution, the accuracy and stability of the memory library are verified. Through feedback correction technology, the memory library parameters are dynamically adjusted to generate a preliminary optimized memory library;

[0023] For the preliminary optimized memory library, based on visualization technology, the memory library content and update results are integrated into the final personalized sports memory library.

[0024] Optionally, the method generates exercise optimization suggestions based on historical data in the personalized exercise memory library, employing a reinforcement learning-based exercise data analysis algorithm in combination with the user's real-time exercise data, wherein the multi-objective optimization technology is used to dynamically adjust exercise parameters to obtain optimized exercise feedback suggestions, including:

[0025] The historical data in the personalized exercise memory library is analyzed using a reinforcement learning-based exercise data analysis algorithm combined with the user's real-time exercise data. Through a multi-objective optimization framework, the exercise optimization goal is defined and a preliminary optimization problem model is generated.

[0026] For the preliminary optimization problem model, based on reinforcement learning, combined with the user's real-time motion data, the motion parameters are dynamically adjusted. Through adaptive learning rate adjustment technology, the accuracy and stability of the optimization strategy are optimized to generate a preliminary motion optimization strategy;

[0027] For the preliminary motion optimization strategy, based on multi-objective optimization technology, combined with motion efficiency and motion damage, the motion parameters are dynamically adjusted. Through dynamic weight distribution technology, the importance of different optimization goals is balanced to generate preliminary motion optimization suggestions;

[0028] For the preliminary motion optimization suggestions, based on user feedback, combined with the user's real-time motion data and historical data distribution, the feasibility and effectiveness of the optimization suggestions are verified. Through feedback correction technology, the optimization suggestions are dynamically adjusted to generate the final optimized motion feedback suggestions.

[0029] Another embodiment of the present application provides an ankle pump motion data analysis system with a memory function, the system comprising:

[0030] The acquisition module is used to collect the user's motion data in real time through the wearable device during the ankle pump exercise. The motion data includes acceleration, angular velocity, and pressure distribution. The multi-sensor fusion technology is combined with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data.

[0031] An extraction module is used to extract the characteristics of ankle pump movement based on multimodal motion data using a deep learning-based motion pattern recognition model. The module uses a spatiotemporal attention mechanism to capture key spatiotemporal features in the movement and obtain a feature representation of the movement pattern.

[0032] An update module is used to dynamically update the user's personalized motion memory library based on the characteristic representation of the motion pattern and adopt a memory library construction algorithm based on incremental learning. The adaptive weight adjustment technology is used to optimize the storage and retrieval efficiency of the memory library to obtain a personalized motion memory library;

[0033] The generation module is used to generate exercise optimization suggestions based on the historical data in the personalized exercise memory library, using a reinforcement learning-based exercise data analysis algorithm and combining it with the user's real-time exercise data. Among them, through multi-objective optimization technology, the exercise parameters are dynamically adjusted to obtain optimized exercise feedback suggestions.

[0034] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when run.

[0035] Yet another embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the above methods.

[0036] Compared with the existing technology, the present invention provides an ankle pump motion data analysis method with a memory function. During the ankle pump motion process, the user's motion data is collected in real time through a wearable device to obtain multimodal motion data; based on the multimodal motion data, a motion pattern recognition model based on deep learning is used to extract the characteristics of the ankle pump motion to obtain a characteristic representation of the motion pattern; based on the characteristic representation of the motion pattern, a memory library construction algorithm based on incremental learning is used to dynamically update the user's personalized motion memory library to obtain a personalized motion memory library; based on the historical data in the personalized motion memory library and combined with the user's real-time motion data, motion optimization suggestions are generated, and through multi-objective optimization technology, the motion parameters are dynamically adjusted to obtain optimized motion feedback suggestions, thereby improving the effect of ankle pump motion and promoting the user's health management and improvement of sports performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A hardware structure block diagram of a computer terminal for an ankle pump motion data analysis method with a memory function provided by an embodiment of the present invention;

[0038] Figure 2 A schematic flow chart of a method for analyzing ankle pump exercise data with a memory function provided by an embodiment of the present invention;

[0039] Figure 3 A structural schematic diagram of an ankle pump motion data analysis system with memory function provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0041] The embodiment of the present invention first provides an ankle pump motion data analysis method with a memory function, which can be applied to electronic devices such as computer terminals, specifically ordinary computers.

[0042] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for an ankle pump motion data analysis method with memory function provided by an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0043] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the ankle pump motion data analysis methods with a memory function.

[0044] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0045] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any ankle pump motion data analysis method with memory function.

[0046] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0047] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0048] See also Figure 2 , an embodiment of the present invention provides an ankle pump motion data analysis method with a memory function, which may include the following steps:

[0049] S201, collecting motion data of the user in real time through a wearable device during an ankle pump exercise, the motion data including acceleration, angular velocity, and pressure distribution, wherein multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data;

[0050] During this phase, users collect their motion data in real time through wearable devices (such as smart bracelets or smart shoes) while performing ankle pump exercises. This data includes acceleration (speed and direction of movement), angular velocity (rotation rate of joints during movement), and pressure distribution (changes in pressure between the foot and the ground). To improve the accuracy of data collection, multi-sensor fusion technology is used, which integrates data from different sensors to obtain a more comprehensive and accurate motion status. These sensors may be distributed in different parts of the user's body. Through real-time collection and analysis, multimodal motion data is formed, providing a basis for subsequent motion analysis.

[0051] Ultimately, the multimodal motion data generated in this step can provide an accurate basis for subsequent movement pattern recognition and personalized motion feedback. This will help athletes or users better understand their performance during exercise and effectively improve exercise results through scientific data support. At the same time, the real-time and efficient nature of this data is also of great significance to users' daily exercise guidance and health management.

[0052] Specifically, during the ankle pump exercise, the user's motion data can be collected in real time through the acceleration sensor, angular velocity sensor and pressure sensor in the wearable device, and the real-time and continuity of data collection can be ensured through lightweight data caching technology;

[0053] In this step, the wearable device uses multiple sensors (accelerometers, angular velocity sensors, and pressure sensors) to collect the user's motion data during the ankle pump exercise in real time. The accelerometer captures the user's overall acceleration and deceleration, the angular velocity sensor monitors the rotation of the user's ankle joint, and the pressure sensor records the contact pressure between the foot and the ground. The data from these sensors will be transmitted to the device's processing unit in real time for subsequent analysis. At the same time, lightweight data caching technology is used to ensure that the data can be cached continuously to ensure the consistency and real-time nature of data collection.

[0054] By collecting user motion data in real time, the device can accurately capture an individual's physical condition during exercise. For example, the device can record the user's body balance response caused by changes in acceleration during ankle pump exercises, providing a basis for exercise optimization. At the same time, real-time caching and processing can reduce latency, allowing data to be quickly fed back to the user and providing immediate exercise adjustment suggestions. This allows users to better understand their physical condition during exercise and adjust their exercise methods in a timely manner to reduce the risk of injury and improve exercise results.

[0055] In its implementation, the wearable device incorporates multiple sensors, including accelerometers, angular velocity sensors, and pressure sensors, which activate data collection during the user's ankle pump exercise. The accelerometer records changes in acceleration during exercise, the angular velocity sensor monitors the rotational velocity of the ankle joint, and the pressure sensor measures the contact pressure between the foot and the ground. The real-time data collected by these sensors records the dynamic changes in the user's movements for subsequent analysis and optimization.

[0056] To ensure real-time and continuous data collection, the system uses lightweight data caching technology. This technology allows sensor data to be first cached in local memory at each moment of data collection, rather than immediately transmitted to the cloud or other storage devices. This approach reduces data transmission latency and ensures rapid and stable recording of motion data. For example, when a user performs a rapid ankle pump exercise, the accelerometer may generate up to 200 data points per second, which are promptly written to memory to ensure no data is lost during analysis.

[0057] When the user starts ankle pump exercise, the device will immediately start all sensors to collect real-time data. For example, when the user is doing ankle pump, as the rhythm of the exercise speeds up, the pressure sensor may record instantaneous pressure changes, while the acceleration and angular velocity sensors are also capturing the dynamic state of the user's body at the same time. All this information will be batched and uploaded to the back-end server for in-depth analysis and research at the end of each exercise cycle. In addition, through technical means, the system can still ensure the integrity of the data even when the network connection is unstable, because all data will be retained in the cache until it is safely uploaded.

[0058] For the collected multi-sensor data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources. An adaptive data cleaning algorithm is used to filter noise and fill missing values in the data to generate a preliminary standardized data set.

[0059] In this step, the device analyzes the collected multi-sensor data to determine its format type. A deep learning-based format recognition model automatically identifies the data characteristics of different sensor outputs, ensuring the system's adaptability and processing capabilities for diverse sensor data. For example, if an accelerometer outputs data at a specific frequency, while a pressure sensor's output may have different timestamps, the deep learning model ensures that all data can be correctly parsed and organized into a consistent format.

[0060] An adaptive data cleaning algorithm filters noise and fills in missing values, further improving data quality. This is crucial because, during actual exercise, various interferences and sensor errors can lead to inaccurate data. For example, if the pressure sensor fails to capture changes in foot pressure at a given moment, the system uses an adaptive algorithm to infer and fill in missing values based on surrounding data to maintain data integrity. This process provides a standardized base dataset for subsequent analysis, significantly improving data quality.

[0061] In this step, the system applies a deep learning-based format recognition model to the collected multi-sensor data. Specifically, the model first analyzes the data characteristics of each sensor's output and automatically identifies its format and data type. For example, an accelerometer may output data in the form of a three-dimensional vector (i.e., acceleration along the x, y, and z axes), while a pressure sensor outputs a single pressure value. This process ensures that the formats of different data sources are accurately parsed, providing a consistent data structure for subsequent data processing.

[0062] The system then uses an adaptive data cleaning algorithm to filter noise and fill in missing values. Noise during data collection can be caused by environmental interference, equipment failure, or the intensity of user movement. For example, during a vigorous exercise, an accelerometer may record an abnormally high acceleration value. The system will use a pre-defined algorithm to identify and remove these outliers. Furthermore, for missing values, the system will use interpolation and other techniques to reasonably infer data from adjacent time points to ensure data integrity and consistency.

[0063] Ultimately, this process generates a preliminary standardized dataset. After noise filtering and missing value filling, this dataset possesses high accuracy and reliability, providing strong support for subsequent analysis and feature extraction. For example, in subsequent motion pattern recognition, scientifically standardized data can significantly improve model training results, enabling the system to more accurately identify users' motion patterns and performance.

[0064] For the preliminary standardized data set, a dynamic data calibration algorithm based on Kalman filtering is used, combining the temporal characteristics of multi-sensor data to remove noise interference in real time. Dynamic threshold adjustment technology is used to ensure the accuracy and stability of the calibration process, generating a preliminary denoised data set.

[0065] The core of this step is the dynamic calibration of the initially standardized dataset using the Kalman filter algorithm. Kalman filtering is a recursive algorithm for linear dynamic systems, suitable for estimating system states from incomplete observational data. During this process, the system synchronously processes data from multiple sensors based on time series characteristics, removing noise in real time. This process effectively improves data reliability and accurately reflects the user's true state during ankle pump exercise.

[0066] Dynamic threshold adjustment technology ensures the accuracy and stability of the calibration process. For example, based on changes in acceleration and angular velocity during motion, the system adjusts the Kalman filter threshold in real time to better match the motion characteristics. This real-time dynamic adjustment capability enables the filter to adapt to rapid changes in motion, thereby improving the stability and responsiveness of data calibration and ensuring the quality of the final dataset.

[0067] The system will apply the Kalman filter algorithm to the preliminary standardized data set for dynamic data calibration. In the specific implementation process, the prediction and update mechanism of the Kalman filter allows the system to infer the current state based on the data of the previous moment. During calibration, the system will integrate data from different sensors and remove noise in real time. For example, when a user performs an ankle pump exercise, interference may be introduced due to uneven ground. The Kalman filter will use the previous data state to predict the current state, thereby identifying and filtering out these outliers.

[0068] By combining the temporal characteristics of multi-sensor data, the system can account for the influence of time during data calibration. For example, data from accelerometers and pressure sensors are not independent but rather influence each other over time. During calibration, the Kalman filter dynamically adjusts the threshold, enabling flexible adaptation to different stages of motion. For example, during slow motion, the system might set a lower noise threshold, while during fast motion, it might increase it to prevent incorrectly filtering out valid signals.

[0069] Ultimately, this process generates a preliminary denoised dataset. After Kalman filtering, the data becomes more accurate, truly reflecting the user's actual state during the ankle pump exercise. This lays the foundation for subsequent analyses such as motion pattern recognition and data mining. This processing significantly improves the accuracy and stability of the data, enabling the system to accurately and timely reflect the user's performance during analysis and feedback.

[0070] For the preliminary denoised data set, a data integration method based on multi-sensor fusion technology is used to map the acceleration, angular velocity and pressure distribution data into unified multimodal motion data. The missing data are supplemented by the interpolation filling method to generate the final multimodal motion data.

[0071] In this step, the hospital applies multi-sensor fusion technology to integrate the processed denoised data into unified multimodal motion data. This process involves fusing data from acceleration, angular velocity, and pressure sensors to generate an integrated motion data set. For example, by combining data from acceleration sensors and pressure sensors, the system can fully describe the user's physical state during ankle pump exercise, including speed, angle, and force on the foot. This integration can provide a more comprehensive view of sports performance and provide strong data support for subsequent analysis.

[0072] To compensate for potential missing data, the system also incorporates interpolation. For example, if the pressure sensor fails to capture data during a certain time period, interpolation allows the system to infer the pressure value for that period based on pressure data from adjacent time points, thus ensuring the integrity of the final dataset. This supplementation makes the resulting multimodal motion data more accurate and reliable, avoiding the negative impact of missing values on subsequent analysis.

[0073] The system applies a data integration method based on multi-sensor fusion technology to the preliminary denoised dataset. Specifically, the system maps data from acceleration, angular velocity, and pressure distribution sensors into a unified format. For example, acceleration data can be combined with angular velocity and pressure data, and a fusion algorithm can be used to generate a comprehensive description of the motion state, thus presenting information that cannot be provided by a single data source. This multimodal data can provide a broader perspective in the analysis, for example, understanding the speed, angle, and force under a specific motion state.

[0074] To address potential missing values in the data, the system also employs interpolation. If the pressure sensor fails to provide data for a certain period, the system can interpolate the missing values using linear or polynomial interpolation based on the data from adjacent time periods. This ensures the integrity of the final dataset and prevents the negative impact of missing data on subsequent motion analysis. By utilizing surrounding data points for reasonable inference, the system can provide a more coherent user's motion data, presenting a complete picture of the movement process.

[0075] Ultimately, a unified multimodal motion dataset is generated. This integrated and supplemented dataset not only comprehensively reflects the user's ankle pump performance metrics but also provides a solid foundation for subsequent feature extraction, model training, and personalized recommendations. Combined with the characteristics of real-time motion data, the system will be able to provide users with more precise exercise guidance, helping them better understand their performance and more effectively adjust their exercise strategies.

[0076] S202, extracting ankle pump motion features using a deep learning-based motion pattern recognition model based on the multimodal motion data, wherein the key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern;

[0077] In this step, the system analyzes and extracts features from the collected multimodal motion data through a motion pattern recognition model based on deep learning. This analysis not only focuses on a single dimension of the data, but also captures the key spatiotemporal features displayed by multiple data modalities (such as acceleration, angular velocity, and pressure distribution) during the ankle pump movement through a spatiotemporal attention mechanism. The spatiotemporal attention mechanism can dynamically adjust the focus of attention according to the weight of each data point, so that the model can more sensitively capture important changes in the moment of movement. This mechanism allows the system to consider not only the data at the current time point, but also the data of the previous and next time periods to form a comprehensive representation of motion features.

[0078] Specifically, for multimodal motion data, we can extract preliminary features of acceleration, angular velocity, and pressure distribution based on deep learning. Through the multi-head attention mechanism, we can capture the correlation between different modal features and generate preliminary multimodal feature representations.

[0079] In this step, the multimodal motion data collected from the wearable device is first analyzed, including information such as acceleration, angular velocity, and pressure distribution. Through a deep learning-based model (such as a convolutional neural network or a recurrent neural network), the system will extract preliminary features for each modality. In addition, using a multi-head attention mechanism, the model can pay attention to the correlation between each modal feature and other modal features when processing it. This process will lay the foundation for subsequent feature fusion. Using deep learning and attention mechanisms, key information in multimodal data can be effectively extracted, thereby improving the accuracy of motion pattern recognition. By fully understanding the relationship between different modal features, the system can better capture the complexity of motion and provide users with more accurate motion analysis services.

[0080] In this step, the user's multimodal motion data must first be acquired from the wearable device. This data includes acceleration, angular velocity, and pressure distribution information. To ensure the real-time and integrity of the data, the data acquisition process uses lightweight data caching technology to save the collected sensor data to a temporary storage area. After data acquisition is completed, a deep learning model is used to perform preliminary feature extraction on the data of different modalities. Specifically, a convolutional neural network (CNN) can be used for acceleration and pressure distribution data, while a recurrent neural network (RNN) is used for angular velocity data to fully exploit the spatial and temporal features in the data.

[0081] Next, after preliminary feature extraction, a multi-head attention mechanism is used to further enhance the quality of the feature representation. The key to the multi-head attention mechanism is its ability to compare and correlate features from different modalities, thereby extracting the features that have the greatest impact on the overall motion pattern. For example, through analysis, the system may discover a significant correlation between changes in acceleration and pressure distribution in certain specific scenarios. Among the extracted features, the attention mechanism automatically identifies which modal features are most important in the current motion phase and accordingly enhances the influence of these features in the final feature representation.

[0082] Finally, after processing through the multi-head attention mechanism, a preliminary multimodal feature representation is generated, which will lay a solid foundation for the in-depth analysis in the subsequent steps. These preliminary features not only contain the independent information of each modality, but also reflect the inherent connections between them, making the subsequent spatiotemporal feature extraction more accurate and comprehensive.

[0083] For the preliminary multimodal feature representation, based on the spatiotemporal attention mechanism and combined with the spatiotemporal characteristics of ankle pump movement, the key spatiotemporal features of the movement are captured. The accuracy and stability of the feature representation are optimized through the dynamic weight allocation technology to generate a preliminary spatiotemporal feature representation.

[0084] In this step, the system inputs the preliminarily extracted multimodal features into a spatiotemporal attention mechanism to better capture the important spatiotemporal features of the exercise process. The system combines the characteristics of ankle pump movement and optimizes the representation accuracy of features through dynamic weight distribution technology to ensure that key features can be effectively focused on in both time and space dimensions. This mechanism can help identify important patterns at different time points and different stages of exercise, such as the relationship between acceleration and pressure in a specific period. Through the spatiotemporal attention mechanism, the motion features of key moments can be accurately captured, thereby improving the depth and breadth of motion data analysis. By optimizing the accuracy and stability of feature representation, the system can better predict and analyze the user's exercise performance, providing a reliable basis for personalized exercise guidance.

[0085] In this step, the system introduces the preliminary multimodal feature representation into a spatiotemporal attention network to enhance the ability to capture key spatiotemporal features in the movement. First, the system analyzes each time point and spatial position to identify which features have an important impact on the performance of the ankle pump movement at a specific moment. Through the spatiotemporal attention mechanism, the system can dynamically adjust the weight of each feature. For example, if the acceleration changes significantly within a certain period of time, the system can increase the weight of the feature, thereby emphasizing its importance in the analysis of movement patterns.

[0086] Next, combined with the spatiotemporal characteristics of ankle pump movements, the system will consider the periodicity and continuity of the movement. By introducing dynamic weight allocation technology, the system can evaluate and adjust the features of different time periods. This means that in different stages of movement, there may be different feature combinations with higher weights. For example, in the acceleration phase, the system will emphasize features related to speed changes; while in the deceleration phase, it may focus on features related to force output. Through this dynamic adjustment, the system not only improves the accuracy and stability of feature representation, but also ensures comprehensive coverage of features in different motion states.

[0087] Finally, after this process, the system generates a preliminary spatiotemporal feature representation that effectively reflects the changes in motion across both time and space, providing a more accurate data foundation for subsequent motion pattern feature generation. This representation not only condenses modal information such as acceleration, angular velocity, and pressure, but also profoundly reflects the spatiotemporal dynamic relationships between these elements during motion.

[0088] For the preliminary spatiotemporal feature representation, the spatiotemporal features of different modalities are weighted and fused. Through the cross-modal attention mechanism, the correlation between the features of different modalities is captured to generate a preliminary motion pattern feature representation.

[0089] In this step, the system weightedly fuses the spatiotemporal features of each modality to form a more comprehensive motion pattern feature representation. Through the cross-modal attention mechanism, the correlation between different modal features can be effectively captured, ensuring that the final generated motion pattern feature representation comprehensively considers the joint effects of acceleration, angular velocity, and pressure distribution. This means that the system not only pays attention to the independent information of each modality, but also recognizes the interaction between modalities. Through the weighted fusion of spatiotemporal features, the system can generate a more comprehensive and accurate motion pattern feature representation, thereby improving the comprehensiveness and accuracy of motion analysis. Such a feature representation is crucial for subsequent motion optimization and personalized recommendations because it can reflect the complex dynamic changes of the user during exercise.

[0090] At this stage, the system will perform weighted fusion on the preliminary spatiotemporal feature representations to generate a comprehensive motion pattern feature representation. First, the system will use a cross-modal attention mechanism to integrate and weight the spatiotemporal features of different modalities. Through this mechanism, the system can identify the mutual influence and correlation of the features of each modality within a specific time period. For example, in a characteristic stage of an ankle pump movement, changes in pressure distribution may significantly affect the performance of acceleration. The system will adjust the weights of each modality based on this correlation to ensure that the final generated motion pattern feature representation can truly reflect the important dynamic changes during the movement.

[0091] Next, the system will introduce dynamic weighting technology to adjust the proportion of different modalities in the feature representation based on their importance within a specific time period. For example, when acceleration and angular velocity change simultaneously, the system will evaluate the combined influence of these two modalities, determine which modality has the greater impact on the motion pattern, and prioritize the features of that modality. This weighted fusion ensures a comprehensive representation of motion pattern features while also capturing subtle changes and dynamic characteristics.

[0092] Finally, the resulting preliminary motion pattern feature representation serves as the basis for subsequent analysis. This representation encompasses the dynamic changes of all important modalities, providing reliable information for subsequent feature optimization and personalized recommendations. This comprehensive representation not only improves motion pattern recognition accuracy but also provides users with a clear perspective on motion analysis, enabling more effective training and adjustment strategies.

[0093] For the preliminary motion pattern feature representation, based on error feedback technology, combined with real-time motion data and historical data distribution, the feature weights are dynamically adjusted. Regularization constraints are used to prevent overfitting and generate the final motion pattern feature representation.

[0094] Error feedback technology is used to dynamically adjust the preliminary motion pattern feature representation. By comparing the real-time motion data with the historical data distribution, the system can detect deviations in the feature weights and make corresponding corrections. In addition, to enhance the generalization ability of the model, the system also applies regularization technology to prevent overfitting and ensure that the feature representation can adapt to various motion states. The core of this step is to improve the accuracy and reliability of the motion pattern feature representation by dynamically adjusting the feature weights to adapt to the user's changing motion needs. Through an effective feedback mechanism and regularization control, the system can ensure that the final generated feature representation has both high accuracy and good generalization ability, thereby providing strong support for subsequent motion analysis and optimization.

[0095] In the final step, the system optimizes the preliminary motion pattern feature representation to ensure its accuracy and reliability. First, the system compares the real-time motion data with the historical data distribution, using error feedback techniques to identify any deviations in feature weights. The core of this process is to dynamically adjust feature weights to ensure that the most important features always dominate under different motion states. For example, if a feature shows no significant relevance in recent motion, the system will reduce its weight, and vice versa.

[0096] Next, combined with regularization constraints, the system also introduces compensation mechanisms to prevent overfitting. Overfitting can cause a model to perform well on the training set but poorly on new data or unseen motion scenarios. Therefore, the system introduces penalty terms during feature training to ensure that the model maintains good generalization capabilities when understanding motion patterns. For example, through L2 regularization, the system limits the magnitude of changes in certain feature weights, making the model more stable and adaptable to diverse motion data.

[0097] Finally, through these steps, the system generates a final motion pattern feature representation. This representation not only accurately captures the user's motion state but also takes into account the reference and dynamic nature of historical data. Based on this feature representation, users can receive precise motion feedback and optimization suggestions, providing data support for personalized exercise plans, helping them more effectively achieve their exercise goals.

[0098] S203, dynamically updating the user's personalized exercise memory library using a memory library construction algorithm based on incremental learning based on the characteristic representation of the exercise pattern, wherein the storage and retrieval efficiency of the memory library is optimized through adaptive weight adjustment technology to obtain a personalized exercise memory library;

[0099] Specifically, the algorithm first initializes the memory bank using the user's initial motion data. As the user performs further ankle pump exercises, the real-time collected motion data will continuously update the memory bank. The core of incremental learning is the ability to continuously learn new knowledge and update previous knowledge at any time without retraining the entire model, which is especially important for rapidly changing movement behaviors. Each user's motion data will be processed through adaptive weight adjustment technology to optimize storage and retrieval efficiency in the memory bank, ensuring that the information in the memory bank is always consistent with the user's actual movement status and reflects the user's personalized needs.

[0100] The Movement Memory Library is a specially designed database that stores a variety of data collected while users perform specific exercises, such as ankle pumps. This data includes multi-dimensional motion parameters such as acceleration, angular velocity, and pressure distribution, designed to provide customized feedback and guidance to each user. The Movement Memory Library is not static. Through incremental learning and real-time data input, its content is dynamically updated and optimized to reflect the user's performance over time and under different training conditions. This dynamic nature ensures that the library can adapt to the user's progress and changing needs. Using deep learning and data analysis techniques, the Movement Memory Library extracts the user's movement pattern characteristics from the stored data. This means that the system not only records data but also analyzes, identifies, and understands the user's movement behavior, thereby generating personalized training plans and optimization recommendations. The Movement Memory Library combines the user's real-time motion data with historical data distribution to generate personalized optimization recommendations. These recommendations are dynamically adjusted based on the user's performance, progress, and personal goals, thereby improving exercise effectiveness and reducing the risk of sports injuries. The Movement Memory Library is often integrated with visualization tools to help users intuitively understand their motion data and optimization recommendations. Through charts and dashboards, users can easily view their exercise progress, making it easier to make appropriate adjustments and optimizations. The Sports Memory Library is not only a data storage tool, but also a system that promotes user self-feedback and reflection. By providing detailed data analysis and optimization suggestions, users can gain a deeper understanding of their exercise habits, identify areas for improvement, and continuously improve their athletic performance.

[0101] The Sports Memory Bank acts as an information storage system, recording various data generated during exercise, such as acceleration, angular velocity, and pressure distribution. This storage function is similar to human memory, allowing users to recall past experiences and information when needed, helping them recall and analyze their performance. The Sports Memory Bank's memory function goes beyond simple data storage. Through incremental learning and real-time data updates, it continuously updates and optimizes its stored content to ensure it reflects the user's latest exercise status and performance. This dynamic memory capability is similar to the human learning process, adapting and optimizing existing memories based on newly acquired information. The Sports Memory Bank uses deep learning algorithms to extract and identify user movement patterns and generate personalized exercise recommendations. This capability is similar to how humans recall past experiences, integrating and reconstructing relevant memory fragments into meaningful information to help us make better decisions. The Sports Memory Bank combines historical data with real-time feedback to generate optimization recommendations and training plans. This feedback mechanism allows users to make adjustments based on past exercise experiences, strengthening their memory and enabling them to better apply these experiences in future training. Through visualization tools, the Sports Memory Bank helps users more clearly understand and recall their exercise trajectory, progress, and achievements. This visual memory function enhances users' awareness of their own exercise status to a certain extent, promoting their self-reflection and optimization.

[0102] Specifically, the feature representation of the motion pattern can be used to initialize a personalized motion memory library based on the incremental learning algorithm and the user's initial motion data. The rationality and accuracy of the memory library initialization can be ensured through the dynamic weight allocation technology to generate a preliminary memory library.

[0103] In this step, the system combines the characteristic representation of the movement pattern with the user's initial movement data to initialize the personalized movement memory library. First, the system will use an incremental learning algorithm to analyze the historical movement data to determine the user's movement characteristics and behavior patterns. In this process, for example, if the user shows a certain ankle pump movement frequency in the initial movement data, the system will use this frequency as an important indicator in the memory library. Then, through dynamic weight allocation technology, the system determines the weight value of the feature, so that during the initialization process, important features are given higher weights, thereby ensuring the rationality and accuracy of the memory library.

[0104] The purpose of this phase is to lay the foundation for the user's personalized exercise memory library, ensuring it reflects their actual exercise behavior. By assigning appropriate weights, the system accurately captures the user's movement characteristics during the initial phase, providing a crucial basis for subsequent dynamic updates. Furthermore, this personalized memory library helps users identify changes in their performance, motivating them to continuously refine their training regimen and ultimately achieve improved performance.

[0105] At this stage, the system uses the user's initial motion data to initialize the personalized motion memory library. First, the system needs to collect the user's motion data in the initial stage. The data types include acceleration, angular velocity, and pressure distribution. These data will be used to build a basic model. The system will analyze these data through an incremental learning algorithm to extract the characteristics of the motion pattern. In order to ensure the rationality of the memory library initialization, the system will adopt dynamic weight allocation technology, which means that during the initialization process, important features (such as motion intensity or frequency) will be given greater weights. This ensures that in the subsequent learning and updating process, the memory library can more accurately reflect the user's motion status.

[0106] Next, the system uses machine learning techniques to generate a preliminary memory library. Specifically, the system stores the extracted motion features in a database and categorizes them based on how they affect the user's performance. For example, if a user's acceleration plays a key role in a particular sport, this feature will be stored as a high-priority feature for subsequent data analysis and retrieval. During this process, the system also maintains a record of the user's performance under different exercise conditions to provide a reference for future exercise improvements.

[0107] Finally, after initial memory library construction and feature classification, the system generates a motion memory library that is not just a collection of raw data, but an optimized and enhanced data model. During subsequent training, the system can provide feedback and adjustments based on this memory library, promoting a personalized exercise experience and efficiency. Through dynamic weighting, the system ensures that important features dominate the memory library, laying the foundation for the user's continued improvement.

[0108] For the preliminary memory bank, based on the incremental learning algorithm, combined with the user's real-time motion data, the memory bank is dynamically updated. Through the adaptive weight adjustment technology, the storage and retrieval efficiency of the memory bank is optimized to generate a preliminary updated memory bank;

[0109] In this step, the system will dynamically update the preliminary memory bank based on the user's real-time motion data using an incremental learning algorithm. Whenever the user performs a new ankle pump exercise, the system will collect their motion data in real time and analyze changes in motion patterns. The advantage of incremental learning is that it can locally update the memory bank based on new data without retraining the entire model. For example, if the user accelerates or decelerates during a specific motion cycle, the system will recognize this change and map it to the memory bank. During this process, adaptive weight adjustment technology ensures that the impact of the latest motion data on the memory bank is reflected in the weight adjustment, thereby dynamically optimizing the storage and retrieval efficiency of information.

[0110] Through this dynamic update mechanism, the personalized exercise memory library maintains the latest user activity status, improving the accuracy and timeliness of exercise pattern analysis. The system can adjust the information in the memory library based on the user's real-time data at any time, ensuring that all stored data effectively reflects the user's current exercise needs. This not only makes exercise optimization recommendations more personalized and targeted, but also responds to user changes in a timely manner, improving exercise effectiveness and safety.

[0111] During this phase, the system will dynamically update the preliminary memory library to ensure that it always reflects the user's latest exercise status. The collection of real-time exercise data is key to the update process. The system will use wearable devices to monitor the user's acceleration, angular velocity, pressure and other data during training, and compare this real-time data with the contents of the preliminary memory library. Through incremental learning algorithms, the system can identify new patterns or changes in the user's training and adjust the memory library in a timely manner. For example, if the user has increased the intensity of exercise in recent training, the system will update the weight of the relevant features in the memory library based on this change.

[0112] Adaptive weight adjustment technology is particularly important during this update process. The system automatically adjusts the weights of various features in the memory based on the user's real-time data performance. For example, if a user's performance in a particular training exercise is significantly improved, the system will automatically prioritize the features of that exercise, allowing the memory to better focus on the user's current needs. This dynamic update of real-time data ensures that the memory maintains the latest and most relevant information, optimizing storage and retrieval efficiency.

[0113] The system also performs a data consistency check to ensure the update process doesn't result in data distortion or loss. By comparing old and new data, the system verifies the validity of the update and makes necessary corrections if any inconsistencies are found. Ultimately, this process generates a preliminary update memory that not only accurately reflects the user's training changes but also provides an authoritative basis for subsequent exercise recommendations, helping users continuously improve their training.

[0114] For the preliminary updated memory library, based on simulation, combined with the user's real-time motion data and historical data distribution, the accuracy and stability of the memory library are verified. Through feedback correction technology, the memory library parameters are dynamically adjusted to generate a preliminary optimized memory library;

[0115] In this step, the system will use simulation technology to combine the user's real-time motion data and historical data distribution to verify the initially updated memory library. Specifically, the system will build a simulation environment to reproduce the user's motion characteristics in different training scenarios, and combine these scenarios with the user's historical motion data to verify the accuracy and stability of the memory library. For example, the system may simulate the user's performance during acceleration and deceleration, and then compare these simulation data with the actual collected user data to find out whether the stored characteristics are reasonable. On this basis, the parameters of the memory library are dynamically adjusted through feedback correction technology to ensure that it always remains accurate and continuously effective.

[0116] This process provides a strong guarantee for the reliability of the memory bank. By comparing simulation data with real-world data, the system can promptly identify potential errors or deficiencies in the memory bank. This verification mechanism improves the accuracy of the memory bank. Furthermore, the dynamic adjustment process ensures that the memory bank remains effective across different motion scenarios, providing more accurate data support for subsequent motion optimization.

[0117] In this step, the system uses simulation technology to verify the updated memory library to ensure its accuracy and stability. First, the system will create a virtual environment to simulate the user's performance under different training conditions, such as ankle pump exercises at different intensity levels. By combining the user's real-time motion data with historical data, the system is able to analyze the differences between the simulation results and the actual training data. This comparison can help identify potential errors in the memory library or features that need improvement. For example, if the simulation results show that the user performed poorly in high-intensity training, but the actual data shows the opposite, the system will mark these features for further analysis.

[0118] Next, the system applies feedback correction technology to dynamically adjust the memory bank to address any issues encountered. Based on feedback from simulations, the system adjusts the parameters in the memory bank. For example, if certain features are found to be irrelevant to a user's performance, the system can reduce their weight or, if necessary, remove them. Through continuous feedback iteration, the system refines the contents of the memory bank to ensure that it effectively predicts and reflects the user's actual performance.

[0119] Ultimately, through this series of verification and adjustments, the system will generate a preliminary optimized memory library. This library will more accurately reflect the user's movement patterns and provide a solid data foundation for subsequent exercise optimization recommendations. This verified and revised memory library will significantly improve the effectiveness of exercise recommendations, enabling users to obtain better feedback and results during training, thereby achieving greater athletic progress.

[0120] For the preliminary optimized memory library, based on visualization technology, the memory library content and update results are integrated into the final personalized sports memory library.

[0121] In the final step, the system will use visualization technology to integrate the optimized memory library content and update results to generate the final personalized sports memory library. This process not only involves the comprehensive organization of the data in the memory library, but also includes the graphical display of the changing trends and status of various sports characteristics, so that users can intuitively understand their own sports performance. For example, the system may generate a chart showing the changes in the user's acceleration, angular velocity and pressure distribution in different time periods to help users identify the effects of different training stages. At the same time, the system will also highlight the dynamic changes of various key sports characteristics to improve the readability and practicality of the memory library.

[0122] Through visualization technology, users can effectively understand their own exercise data and characteristics, making complex exercise data easy to digest. This intuitive presentation not only increases user trust in the exercise memory library, but also enhances user engagement and motivation, making them more willing to adjust and improve based on optimization results. Furthermore, the integrated personalized exercise memory library provides intuitive data support for subsequent exercise optimization suggestions and training plans, helping users achieve higher exercise goals.

[0123] In the final step, the system will use visualization technology to integrate the optimized memory library contents to generate the final personalized sports memory library. This process first requires visualizing the data in the optimized memory library so that users can intuitively understand and analyze their sports performance. The system may use a variety of chart forms, such as line charts, bar charts, and heat maps, to display information such as the user's acceleration, angular velocity, and pressure distribution at different training stages. Through these intuitive graphical representations, users can quickly identify their strengths and areas for improvement, thereby adjusting their training plans more effectively.

[0124] Next, the system will integrate all updated results, including real-time collected data and feedback from simulation verification. During the integration process, the system will generate a dynamic dashboard through visualization technology, allowing users to monitor their athletic performance in real time. This dashboard will display all key indicators, including real-time acceleration, angular velocity, and pressure distribution, while also providing a comparison of historical data to help users clearly understand exercise trends. For example, users can view performance changes within a specific training cycle and understand performance changes under different training intensities. This intuitive display method allows users to find the gap between results and expectations at a glance, thereby effectively adjusting their training strategies.

[0125] In addition, the system provides users with customized data visualization options. Users can select variables of interest for combined display, such as the relationship between acceleration and pressure distribution, or the relationship between exercise intensity and exercise time. This flexible visualization option not only enhances user engagement but also makes data analysis more personalized, allowing users to view relevant data based on their needs and draw more targeted training recommendations.

[0126] Finally, the integrated, personalized workout memory library will contain all historical data, real-time updates, and verified optimization results. This database will serve as a reference for long-term improvement. Users can access this library before and after each training session to gain a deeper understanding of their progress and make adjustments based on the system's feedback. Through visual optimization integration, users will gain an intuitive and efficient workout management tool, allowing them to be more proactive and flexible during exercise, ultimately promoting better health and performance.

[0127] S204, based on the historical data in the personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm is used in combination with the user's real-time exercise data to generate exercise optimization suggestions, wherein the exercise parameters are dynamically adjusted through multi-objective optimization technology to obtain optimized exercise feedback suggestions.

[0128] Based on historical data from a personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm, combined with the user's real-time exercise data, generates exercise optimization recommendations. This process first analyzes the user's accumulated exercise data from past activities to identify effective exercise patterns and key characteristics. The system then compares this historical data with real-time monitored exercise data. Using reinforcement learning, based on the current state, it dynamically adjusts exercise parameters such as intensity, frequency, and duration to achieve optimal results.

[0129] Specifically, we can use a reinforcement learning-based motion data analysis algorithm to analyze the historical data in the personalized motion memory library, combined with the user's real-time motion data, to perform motion data analysis. Through a multi-objective optimization framework, we can define motion optimization goals and generate a preliminary optimization problem model.

[0130] In this step, the system first extracts historical data related to the user from the personalized exercise memory library, such as exercise frequency, duration, and feedback information. These data will be compared and comprehensively analyzed with the user's current motion data. Using reinforcement learning algorithms, the system establishes a multi-objective optimization framework to clarify the priorities between various exercise goals (such as improving strength, endurance, and reducing the risk of injury), and generates a preliminary optimization problem model based on this to prepare for subsequent strategy adjustments. This process ensures that the user's exercise recommendations are not only based on real-time performance, but are also deeply rooted in their exercise history. Through this comprehensive analysis, the system can more accurately identify the user's real needs and formulate targeted training plans, thereby improving the user's training efficiency and effectiveness.

[0131] The system first extracts historical data related to the user from its personalized exercise memory library. This data may include various aspects of the user's past training performance, such as exercise frequency, duration, intensity, heart rate changes, and corresponding feedback. Using data mining techniques, the system can identify performance trends under different conditions. This information is then used to set optimization goals.

[0132] Next, the system analyzes real-time exercise data. For example, suppose a user is currently training, monitoring real-time data such as heart rate, pace, and time. The system compares the user's current status with historical performance in real time, dynamically adjusting optimization goals. This process involves the application of reinforcement learning algorithms. Based on user feedback and changes, the system adjusts exercise optimization goals in real time to ensure they are challenging and achievable.

[0133] Ultimately, the system generates a preliminary optimization problem model based on a multi-objective optimization framework. For example, if the user's goal is to improve cardiorespiratory endurance, increase strength, and reduce injury risk, the system will clarify the importance and priority of each goal and generate a corresponding optimization strategy for each goal. In this way, users can get a clear direction in their new training plan, allowing them to focus on improving cardiorespiratory endurance while also paying attention to strength training and injury risk control.

[0134] For example, let's say user Xiao Li has performed a series of aerobic and strength training exercises over the past six months. His personal exercise memory stores each of his training sessions, including heart rate, training time, weight used, number of repetitions, and user feedback (such as perceived fatigue). The system first extracts Xiao Li's historical data and finds that he maintains good exercise results and notices a significant improvement in his cardiopulmonary function when his heart rate is between 130 and 140 beats per minute.

[0135] When combined with real-time exercise data, let's say a girl is doing an aerobic workout and her heart rate is monitored at 138 beats per minute for 30 minutes. The system will use a reinforcement learning algorithm to compare historical data from that period and analyze her performance within the same heart rate range. The system will also set several optimization goals, such as improving cardiorespiratory endurance, increasing training time, and reducing injury risk.

[0136] Using a multi-objective optimization framework, the system assigns different weights to each objective. Considering that Xiao Li has just resumed training, the primary focus might be on reducing injury risk, followed by improving cardiorespiratory endurance. Ultimately, based on this analysis, the system generates a preliminary optimization model, recommending that Xiao Li maintain a heart rate between 130 and 140 beats per minute for the remainder of his training, while gradually increasing the duration and intensity of his workouts.

[0137] For the preliminary optimization problem model, based on reinforcement learning, combined with the user's real-time motion data, the motion parameters are dynamically adjusted. Through adaptive learning rate adjustment technology, the accuracy and stability of the optimization strategy are optimized to generate a preliminary motion optimization strategy;

[0138] During this phase, the system implements the preliminary optimization model, continuously updating the exercise parameters based on real-time user data. Using adaptive learning rate adjustment technology, the reinforcement learning model flexibly adjusts the learning rate based on system feedback, ensuring that the optimization strategy remains efficient and stable across different training phases and user states. The ability to dynamically adjust exercise parameters enables the system to quickly respond to changing user needs, ensuring the flexibility and adaptability of the training plan. This feature not only increases user satisfaction but also significantly enhances the effectiveness of the exercise plan, helping users maintain optimal performance in a changing training environment.

[0139] In this step, the system dynamically adjusts parameters based on the initial optimization problem model and the user's real-time motion data. For example, during strength training, the system records key metrics such as weights lifted, number of repetitions, and heart rate. As training progresses, this real-time data is continuously fed into the reinforcement learning model, which evaluates the user's performance in real time and adjusts the exercise parameters based on the acquired data.

[0140] At the same time, the system applies adaptive learning rate adjustment technology. This technology allows the model to dynamically change its learning rate during training to adapt to the user's status. For example, if the user performs well in a training project (such as successfully completing a preset number of repetitions), the system can increase the learning rate to accelerate its strategy adjustment; conversely, if it is found that the user is easily fatigued under a certain load, the learning rate will be reduced to ensure the stability of the strategy.

[0141] After generating preliminary exercise optimization strategies, the system communicates these strategies to the user. For example, after integrating user information, the system might recommend increasing the weight by 5% and reducing the number of repetitions by 1-2 in the next strength training session to avoid excessive fatigue and ensure training effectiveness. This real-time feedback and adjustment mechanism not only improves training effectiveness but also ensures user safety and health.

[0142] For example, during Xiao Li's next strength training session, the system analyzes his performance in real time based on the preliminary optimization problem model. Suppose Xiao Li begins squatting with a 50 kg weight. During training, the system monitors Xiao Li's heart rate, performance, and number of lifts in real time. After several rounds of training, Xiao Li performs well with this weight, successfully completing each set of 10 reps.

[0143] As training progressed, the system applied a reinforcement learning algorithm to evaluate Xiao Li's performance. The system noticed that Xiao Li's heart rate rose to 150 beats per minute while completing squats, indicating that she was approaching fatigue. To ensure Xiao Li's training effectiveness and safety, the system adjusted her training parameters, reducing the training load through an adaptive learning rate. It recommended that she reduce the weight by 5 kg and increase her rest time for the next set of squats.

[0144] Through these adjustments, Xiao Li can maintain appropriate intensity and adequate recovery during training, avoiding excessive fatigue and potential injury. The system continuously updates and optimizes strategies based on Xiao Li's real-time data, making the training process both challenging and safe.

[0145] For the preliminary motion optimization strategy, based on multi-objective optimization technology, combined with motion efficiency and motion damage, the motion parameters are dynamically adjusted. Through dynamic weight distribution technology, the importance of different optimization goals is balanced to generate preliminary motion optimization suggestions;

[0146] In this step, the system will use multi-objective optimization technology to comprehensively evaluate the current optimization strategy with the user's exercise efficiency and injury risk. Through dynamic weight allocation technology, the system can flexibly adjust the importance of different goals based on real-time data. For example, when the user's performance reaches a certain standard, it can be more inclined to improve exercise efficiency; when the user approaches the fatigue threshold, the weight of reducing the risk of injury will increase. This dynamic adjustment allows the user's training plan to maintain a reasonable balance in different situations, which not only promotes exercise improvement but also effectively reduces the potential risk of injury. This flexible and scientific training method can enhance the user's overall training experience and effect, and lay a good foundation for long-term exercise goals.

[0147] During this phase, the system applies multi-objective optimization techniques to the initial exercise optimization strategy, aiming to achieve a balance between different training objectives. For example, a user's goals might include improving movement efficiency, increasing muscle strength, and reducing potential injury risk. The system dynamically assesses the importance of each objective based on the user's historical performance and current status, and assigns weights accordingly.

[0148] For example, let's say a user is doing interval running training. Using historical data, the system discovers that the user shows signs of fatigue after reaching a certain heart rate threshold, so it decides to reduce the intensity of the exercise. In this case, the system might lower the weight of the goal to improve exercise efficiency and increase the weight of reducing injury risk to ensure the user can complete the training safely without excessive fatigue.

[0149] Ultimately, the system generates preliminary exercise optimization recommendations, taking into account multiple factors. For example, the system might recommend that the user maintain a certain heart rate range during their next run, while also adjusting their speed and duration to achieve optimal results. This strategic balance is crucial, providing users with targeted training plans that maximize efficiency while minimizing the risk of injury.

[0150] For example, after Xiao Li completes several strength training sessions, the system begins evaluating her initial optimization strategy based on her feedback and real-time data. For example, if the system detects that Xiao Li feels a good load when doing free-weight squats, but her heart rate exceeds 160 beats / min during the last set, indicating a high level of fatigue, the system may be unable to detect the load.

[0151] To address this issue, the system applied multi-objective optimization techniques, comprehensively considering Xiao Li's goals of improving exercise efficiency and reducing injury. At this stage, the system, through dynamic weighting, determined that Xiao Li's high heart rate increased his risk of injury. Therefore, in the next training session, the system reduced the weight of improving exercise efficiency and increased the weight of reducing injury risk.

[0152] Based on this weight adjustment, the system provides Xiao Li with specific exercise optimization suggestions, suggesting that she use a lower weight (for example, 45 kg) in the next training session and change the training mode, such as adding more rest intervals or adopting different training movements (such as squats assisted by a chair) to help reduce the burden while effectively improving muscle strength and cardiopulmonary adaptability.

[0153] For the preliminary motion optimization suggestions, based on user feedback, combined with the user's real-time motion data and historical data distribution, the feasibility and effectiveness of the optimization suggestions are verified. Through feedback correction technology, the optimization suggestions are dynamically adjusted to generate the final optimized motion feedback suggestions.

[0154] In the final step, the system leverages user feedback and real-time data to verify the effectiveness of the initial exercise optimization recommendations. By comprehensively analyzing the user's performance after implementing the recommendations, the system can provide feedback and corrections to the optimization suggestions, dynamically adjusting the strategy in a timely manner to ensure that the final recommendations are feasible and effective in real-world situations. Real-time feedback promotes system self-improvement, continuously optimizing exercise recommendations and truly achieving personalized and intelligent training management. Ultimately, users receive fully verified and effective exercise recommendations, thereby improving their exercise results and avoiding unnecessary mistakes and risks.

[0155] In the final step, the system integrates user feedback to validate the effectiveness of the initial exercise optimization recommendations. For example, users will provide feedback after actual training, including their perceived fatigue, post-exercise physical reactions, and satisfaction with the training recommendations. This feedback will serve as an important basis for subsequent optimization.

[0156] The system will combine the user's real-time exercise data to see the actual effect of these suggestions. For example, if the user feels relatively tired after training according to the system's recommendations and their heart rate fails to remain within the predetermined range, the system will analyze this feedback and compare it with the user's historical data to find possible reasons. If it is found that the user's physical condition has changed during this period, such as a decrease in training effect due to external factors (such as diet or lack of rest), the system will make necessary adjustments based on this information.

[0157] Ultimately, through feedback correction technology, the system will dynamically adjust optimization recommendations. This may mean developing a new training plan for the user, such as increasing recovery time, adjusting training intensity, or even switching training programs. Through this feedback loop, users can continuously receive scientific and personalized training recommendations, helping them maintain continuous progress and health over the long term.

[0158] For example, after Alice follows the system's initial exercise optimization suggestions, the system invites her to complete a feedback questionnaire after the training session to evaluate her training experience. Let's assume Alice reports that when squatting with a 45kg weight, the load felt moderate, her heart rate didn't reach the preset limit, and her fatigue was significantly reduced after the session, indicating that these adjustments were very effective for her training.

[0159] The system analyzes Xiao Li's feedback and her real-time training data, which shows that she completed 12 squats at 45 kg and maintained a heart rate of 135-145 beats per minute, which is within the system's set heart rate range. Based on this successful feedback, the system will further verify the effectiveness of the initial optimization suggestions.

[0160] The next step is to dynamically adjust Xiao Li's optimization suggestions through feedback correction technology. Assuming the system decides to further increase Xiao Li's training goals, it recommends that she gradually increase the weight back to 50 kg during her next training session, while increasing the number of repetitions per set to 12. To prevent her from getting tired again, the system also recommends that Xiao Li increase the rest time between sets and reminds her to pay attention to her body's feedback. Through this continuous feedback mechanism, the system can continuously optimize Xiao Li's exercise suggestions, enabling her to achieve improved training results while ensuring safety.

[0161] It can be seen that during the ankle pump exercise, the user's motion data is collected in real time through wearable devices to obtain multimodal motion data; based on the multimodal motion data, a motion pattern recognition model based on deep learning is used to extract the characteristics of the ankle pump motion and obtain the characteristic representation of the motion pattern; based on the characteristic representation of the motion pattern, a memory library construction algorithm based on incremental learning is used to dynamically update the user's personalized motion memory library to obtain a personalized motion memory library; based on the historical data in the personalized motion memory library and combined with the user's real-time motion data, motion optimization suggestions are generated, and through multi-objective optimization technology, the motion parameters are dynamically adjusted to obtain optimized motion feedback suggestions, thereby improving the effect of ankle pump exercise and promoting the user's health management and improvement of sports performance.

[0162] Another embodiment of the present invention provides an ankle pump motion data analysis system with a memory function, see Figure 3 , the system may include:

[0163] The acquisition module 301 is used to collect the user's motion data in real time through the wearable device during the ankle pump exercise. The motion data includes acceleration, angular velocity, and pressure distribution. The multi-sensor fusion technology is combined with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data.

[0164] Extraction module 302 is configured to extract ankle pump motion features based on the multimodal motion data using a deep learning-based motion pattern recognition model, wherein the key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern;

[0165] An updating module 303 is configured to dynamically update the user's personalized exercise memory library based on the characteristic representation of the exercise pattern and a memory library construction algorithm based on incremental learning. The adaptive weight adjustment technology is used to optimize the storage and retrieval efficiency of the memory library to obtain a personalized exercise memory library.

[0166] The generation module 304 is used to generate exercise optimization suggestions based on the historical data in the personalized exercise memory library, using a reinforcement learning-based exercise data analysis algorithm and combining it with the user's real-time exercise data. The exercise parameters are dynamically adjusted through multi-objective optimization technology to obtain optimized exercise feedback suggestions.

[0167] An embodiment of the present invention further provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps of any one of the above method embodiments when running.

[0168] Specifically, in this embodiment, the above-mentioned storage medium may be configured to store a computer program for performing the following steps:

[0169] S201, collecting motion data of the user in real time through a wearable device during an ankle pump exercise, the motion data including acceleration, angular velocity, and pressure distribution, wherein multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data;

[0170] S202, extracting ankle pump motion features using a deep learning-based motion pattern recognition model based on the multimodal motion data, wherein the key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern;

[0171] S203, dynamically updating the user's personalized exercise memory library using a memory library construction algorithm based on incremental learning based on the characteristic representation of the exercise pattern, wherein the storage and retrieval efficiency of the memory library is optimized through adaptive weight adjustment technology to obtain a personalized exercise memory library;

[0172] S204, based on the historical data in the personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm is used in combination with the user's real-time exercise data to generate exercise optimization suggestions, wherein the exercise parameters are dynamically adjusted through multi-objective optimization technology to obtain optimized exercise feedback suggestions.

[0173] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any one of the above method embodiments.

[0174] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0175] Specifically, in this embodiment, the processor may be configured to execute the following steps through a computer program:

[0176] S201, collecting motion data of the user in real time through a wearable device during an ankle pump exercise, the motion data including acceleration, angular velocity, and pressure distribution, wherein multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data;

[0177] S202, extracting ankle pump motion features using a deep learning-based motion pattern recognition model based on the multimodal motion data, wherein the key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern;

[0178] S203, dynamically updating the user's personalized exercise memory library using a memory library construction algorithm based on incremental learning based on the characteristic representation of the exercise pattern, wherein the storage and retrieval efficiency of the memory library is optimized through adaptive weight adjustment technology to obtain a personalized exercise memory library;

[0179] S204, based on the historical data in the personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm is used in combination with the user's real-time exercise data to generate exercise optimization suggestions, wherein the exercise parameters are dynamically adjusted through multi-objective optimization technology to obtain optimized exercise feedback suggestions.

[0180] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the scope of implementation of the present invention is not limited to what is shown in the drawings. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which do not exceed the spirit covered by the description and drawings, should be within the scope of protection of the present invention.

Claims

1. A method for analyzing ankle pump motion data with a memory function, characterized in that: The method comprises: During ankle pump exercise, wearable devices are used to collect user motion data in real time. The motion data includes acceleration, angular velocity, and pressure distribution. Multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data. Based on multimodal motion data, a deep learning-based motion pattern recognition model is used to extract the characteristics of ankle pump motion. The key spatiotemporal features of the motion are captured through a spatiotemporal attention mechanism to obtain a characteristic representation of the motion pattern. According to the characteristic representation of the movement pattern, a memory library construction algorithm based on incremental learning is adopted to dynamically update the user's personalized movement memory library. Among them, the storage and retrieval efficiency of the memory library is optimized through adaptive weight adjustment technology to obtain a personalized movement memory library; Based on the historical data in the personalized exercise memory library, a reinforcement learning-based exercise data analysis algorithm is used, combined with the user's real-time exercise data, to generate exercise optimization suggestions. Among them, through multi-objective optimization technology, the exercise parameters are dynamically adjusted to obtain optimized exercise feedback suggestions.

2. The method according to claim 1, characterized in that During the ankle pump exercise, the user's motion data is collected in real time through a wearable device. The motion data includes acceleration, angular velocity, and pressure distribution. Multi-sensor fusion technology is used in combination with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data, including: During ankle pump exercise, the wearable device collects user motion data in real time through the acceleration sensor, angular velocity sensor and pressure sensor. Lightweight data caching technology is used to ensure the real-time and continuity of data collection. For the collected multi-sensor data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources. An adaptive data cleaning algorithm is used to filter noise and fill missing values in the data to generate a preliminary standardized data set. For the preliminary standardized data set, a dynamic data calibration algorithm based on Kalman filtering is used, combining the temporal characteristics of multi-sensor data to remove noise interference in real time. Dynamic threshold adjustment technology is used to ensure the accuracy and stability of the calibration process, generating a preliminary denoised data set. For the preliminary denoised data set, a data integration method based on multi-sensor fusion technology is used to map the acceleration, angular velocity and pressure distribution data into unified multimodal motion data. The missing data are supplemented by the interpolation filling method to generate the final multimodal motion data.

3. The method according to claim 2, characterized in that The method uses a deep learning-based motion pattern recognition model to extract the characteristics of ankle pump motion based on multimodal motion data. The method captures the key spatiotemporal features of the motion through a spatiotemporal attention mechanism to obtain a feature representation of the motion pattern, including: For multimodal motion data, we extract preliminary features of acceleration, angular velocity, and pressure distribution based on deep learning. Through a multi-head attention mechanism, we capture the correlation between different modal features and generate a preliminary multimodal feature representation. For the preliminary multimodal feature representation, based on the spatiotemporal attention mechanism and combined with the spatiotemporal characteristics of ankle pump movement, the key spatiotemporal features of the movement are captured. The accuracy and stability of the feature representation are optimized through the dynamic weight allocation technology to generate a preliminary spatiotemporal feature representation. For the preliminary spatiotemporal feature representation, the spatiotemporal features of different modalities are weighted and fused. Through the cross-modal attention mechanism, the correlation between the features of different modalities is captured to generate a preliminary motion pattern feature representation. For the preliminary motion pattern feature representation, based on error feedback technology, combined with real-time motion data and historical data distribution, the feature weights are dynamically adjusted. Regularization constraints are used to prevent overfitting and generate the final motion pattern feature representation.

4. The method according to claim 3, characterized in that The method uses an incremental learning-based memory library construction algorithm based on the characteristic representation of the motion pattern to dynamically update the user's personalized motion memory library, wherein the adaptive weight adjustment technology is used to optimize the storage and retrieval efficiency of the memory library to obtain a personalized motion memory library, including: Based on the feature representation of motion patterns, an incremental learning algorithm is used to combine the user's initial motion data to initialize a personalized motion memory library. Dynamic weight allocation technology is used to ensure the rationality and accuracy of memory library initialization and generate a preliminary memory library. For the preliminary memory bank, based on the incremental learning algorithm, combined with the user's real-time motion data, the memory bank is dynamically updated. Through the adaptive weight adjustment technology, the storage and retrieval efficiency of the memory bank is optimized to generate a preliminary updated memory bank; For the preliminary updated memory library, based on simulation, combined with the user's real-time motion data and historical data distribution, the accuracy and stability of the memory library are verified. Through feedback correction technology, the memory library parameters are dynamically adjusted to generate a preliminary optimized memory library; For the preliminary optimized memory library, based on visualization technology, the memory library content and update results are integrated into the final personalized sports memory library.

5. The method according to claim 4, characterized in that The method generates exercise optimization suggestions based on historical data in the personalized exercise memory library, using a reinforcement learning-based exercise data analysis algorithm and combining it with the user's real-time exercise data. The method dynamically adjusts exercise parameters through multi-objective optimization technology to obtain optimized exercise feedback suggestions, including: The historical data in the personalized exercise memory library is analyzed using a reinforcement learning-based exercise data analysis algorithm combined with the user's real-time exercise data. Through a multi-objective optimization framework, the exercise optimization goal is defined and a preliminary optimization problem model is generated. For the preliminary optimization problem model, based on reinforcement learning, combined with the user's real-time motion data, the motion parameters are dynamically adjusted. Through adaptive learning rate adjustment technology, the accuracy and stability of the optimization strategy are optimized to generate a preliminary motion optimization strategy; For the preliminary motion optimization strategy, based on multi-objective optimization technology, combined with motion efficiency and motion damage, the motion parameters are dynamically adjusted. Through dynamic weight distribution technology, the importance of different optimization goals is balanced to generate preliminary motion optimization suggestions; For the preliminary motion optimization suggestions, based on user feedback, combined with the user's real-time motion data and historical data distribution, the feasibility and effectiveness of the optimization suggestions are verified. Through feedback correction technology, the optimization suggestions are dynamically adjusted to generate the final optimized motion feedback suggestions.

6. An ankle pump exercise data analysis system with memory function, characterized in that: The system comprises: The acquisition module is used to collect the user's motion data in real time through the wearable device during the ankle pump exercise. The motion data includes acceleration, angular velocity, and pressure distribution. The multi-sensor fusion technology is combined with a dynamic data calibration algorithm based on Kalman filtering to remove noise interference and obtain multimodal motion data. An extraction module is used to extract the characteristics of ankle pump movement based on multimodal motion data using a deep learning-based motion pattern recognition model. The module uses a spatiotemporal attention mechanism to capture key spatiotemporal features in the movement and obtain a feature representation of the movement pattern. An update module is used to dynamically update the user's personalized motion memory library based on the characteristic representation of the motion pattern and adopt a memory library construction algorithm based on incremental learning. The adaptive weight adjustment technology is used to optimize the storage and retrieval efficiency of the memory library to obtain a personalized motion memory library; The generation module is used to generate exercise optimization suggestions based on the historical data in the personalized exercise memory library, using a reinforcement learning-based exercise data analysis algorithm and combining it with the user's real-time exercise data. Among them, through multi-objective optimization technology, the exercise parameters are dynamically adjusted to obtain optimized exercise feedback suggestions.

7. The system according to claim 6, characterized in that The acquisition module is specifically used to: During ankle pump exercise, the wearable device collects user motion data in real time through the acceleration sensor, angular velocity sensor and pressure sensor. Lightweight data caching technology is used to ensure the real-time and continuity of data collection. For the collected multi-sensor data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources. An adaptive data cleaning algorithm is used to filter noise and fill missing values in the data to generate a preliminary standardized data set. For the preliminary standardized data set, a dynamic data calibration algorithm based on Kalman filtering is used, combining the temporal characteristics of multi-sensor data to remove noise interference in real time. Dynamic threshold adjustment technology is used to ensure the accuracy and stability of the calibration process, generating a preliminary denoised data set. For the preliminary denoised data set, a data integration method based on multi-sensor fusion technology is used to map the acceleration, angular velocity and pressure distribution data into unified multimodal motion data. The missing data are supplemented by the interpolation filling method to generate the final multimodal motion data.

8. The system according to claim 7, characterized in that The extraction module is specifically used to: For multimodal motion data, we extract preliminary features of acceleration, angular velocity, and pressure distribution based on deep learning. Through a multi-head attention mechanism, we capture the correlation between different modal features and generate a preliminary multimodal feature representation. For the preliminary multimodal feature representation, based on the spatiotemporal attention mechanism and combined with the spatiotemporal characteristics of ankle pump movement, the key spatiotemporal features of the movement are captured. The accuracy and stability of the feature representation are optimized through the dynamic weight allocation technology to generate a preliminary spatiotemporal feature representation. For the preliminary spatiotemporal feature representation, the spatiotemporal features of different modalities are weighted and fused. Through the cross-modal attention mechanism, the correlation between the features of different modalities is captured to generate a preliminary motion pattern feature representation. For the preliminary motion pattern feature representation, based on error feedback technology, combined with real-time motion data and historical data distribution, the feature weights are dynamically adjusted. Regularization constraints are used to prevent overfitting and generate the final motion pattern feature representation.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 5 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 5.

Images