Intelligent monitoring methods, systems, and equipment based on dynamic and static balance training

Abstract

This invention discloses an intelligent monitoring method, system, and device based on dynamic and static balance training, belonging to the field of dynamic and static balance monitoring technology. The method includes: synchronous multi-dimensional data acquisition during dynamic and static balance training to construct a multimodal balance dataset; joint feature analysis to obtain a set of balance state feature vectors; using the balance state feature vectors to identify the balance state of a target user, setting a balance state level, performing training compensation based on the balance state level, and generating a balance stability index; performing adaptive analysis to generate training parameter adjustment instructions, and executing a mechanism for intelligent monitoring of dynamic and static balance training. This invention solves the technical problems of existing technologies, such as single-dimensional monitoring of dynamic and static balance training, lack of specificity in parameter adjustment, and insufficient control precision, leading to poor training results. It achieves accurate monitoring and intelligent control of dynamic and static balance training, improving the targeting and effectiveness of training.

Description

Technical Field

[0001] This invention relates to the field of dynamic and static balance monitoring technology, specifically to intelligent monitoring methods, systems, and equipment based on dynamic and static balance training. Background Technology

[0002] In the field of dynamic and static balance training, traditional training methods rely heavily on manual observation and experience-based judgment. They offer limited monitoring of the user's balance during training, relying solely on visual observation of posture or simple equipment to collect single data points. This makes it difficult to comprehensively capture the user's balance characteristics during static maintenance, dynamic adjustment, and combined dynamic and static training, resulting in insufficient accuracy in assessing the user's balance. Furthermore, traditional training parameters are often fixed settings or manually adjusted, failing to adaptively optimize based on the user's real-time balance status. This leads to a lack of targeted training, making it difficult to adapt to differences in balance abilities and training progression needs among users, thus impacting training effectiveness.

[0003] Existing technologies suffer from problems such as limited monitoring dimensions for dynamic and static balance training, lack of targeted parameter adjustments, and insufficient control precision, leading to suboptimal training results. Summary of the Invention

[0004] This application provides an intelligent monitoring method, system, and device based on dynamic and static balance training, which addresses the technical problems of poor training results caused by the single monitoring dimension, lack of targeted parameter adjustment, and insufficient control precision in existing dynamic and static balance training.

[0005] In view of the above problems, this application provides an intelligent monitoring method, system and equipment based on dynamic and static balance training.

[0006] The first aspect of this application provides an intelligent monitoring method based on dynamic and static balance training, the method comprising:

[0007] In dynamic and static balance training, target users are simultaneously and multidimensionally collected to construct a multimodal balance dataset. Based on the multimodal balance dataset, joint feature analysis is performed to obtain a set of balance state feature vectors. The balance state feature vectors are used to identify the balance state of the target users, and a balance state level is set. Training compensation is performed based on the balance state level to generate a balance stability index. Adaptive analysis is performed according to the dynamic and static balance training objectives and the balance stability index to generate training parameter adjustment instructions. The training parameter adjustment instructions are fed back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training.

[0008] A second aspect of this application provides an intelligent monitoring system based on dynamic and static balance training, the system comprising:

[0009] The system includes a balance dataset construction module for synchronously collecting multidimensional data from target users during dynamic and static balance training to construct a multimodal balance dataset; a feature vector group acquisition module for performing joint feature analysis based on the multimodal balance dataset to obtain a balance state feature vector group; a stability index generation module for identifying the balance state of target users using the balance state feature vector group, setting balance state levels, performing training compensation based on the balance state levels, and generating a balance stability index; and an intelligent monitoring module for adaptive analysis based on the dynamic and static balance training objectives and the balance stability index, generating training parameter adjustment instructions, and feeding these instructions back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training.

[0010] A third aspect of this application provides an electronic device comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the intelligent monitoring method based on dynamic-static balance training provided in this application.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] In dynamic and static balance training, synchronous multi-dimensional data is collected from the target user to construct a multimodal balance dataset. Joint feature analysis is performed to obtain a set of balance state feature vectors. The balance state of the target user is identified, a balance state level is set, and training compensation is performed based on the balance state level to generate a balance stability index. Adaptive analysis is performed according to the dynamic and static balance training objectives and the balance stability index to generate training parameter adjustment instructions. These instructions are then fed back to the balance training platform control actuator for intelligent monitoring of dynamic and static balance training. This achieves the technical effect of accurate monitoring and intelligent control of dynamic and static balance training, improving the training's targeting and effectiveness. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the intelligent monitoring method based on dynamic and static balance training provided in an embodiment of this application.

[0015] Figure 2 This is a schematic diagram of the structure of an intelligent monitoring system based on dynamic and static balance training provided in an embodiment of this application.

[0016] Figure 3 This is a schematic diagram of the structure of an electronic device provided in this application.

[0017] Figure labeling: Balanced dataset construction module 10, feature vector group acquisition module 20, stability index generation module 30, intelligent monitoring module 40, processor 21, memory 22, input device 23, output device 24. Detailed Implementation

[0018] This application provides an intelligent monitoring method, system, and device based on dynamic and static balance training to address the technical problems in existing technologies, such as the single monitoring dimension, lack of targeted parameter adjustment, and insufficient control precision, which lead to poor training results.

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0020] Example 1, as Figure 1 As shown, this application provides an intelligent monitoring method based on dynamic and static balance training, the method comprising:

[0021] Step S100: Simultaneously collect multidimensional data from the target user during dynamic-static balance training to construct a multimodal balanced dataset.

[0022] Specifically, the target dynamic and static balance training area is traversed, and multiple types of sensors, including pressure sensors, inertial sensors, and visual acquisition devices, are networked to build a network topology covering the training area. During training, pressure sensors sense the pressure of the target user, generating dynamic pressure distribution data, while inertial sensors sense the user's inertial data, generating 3D posture data. Simultaneously, visual acquisition devices are activated to collect overall motion video data of the user. A unified time synchronization benchmark is then established, and the aforementioned dynamic pressure distribution data, 3D posture data, and overall motion video data are synchronized in time to obtain multi-source time synchronization data. A spatial 3D coordinate system is constructed based on the training area, and the multi-source time synchronization data is mapped to this coordinate system to complete spatial registration, forming multi-source monitoring signals. Next, bidirectional enhancement processing is performed on the multi-source monitoring signals to generate multi-dimensional anti-interference enhancement signals. Finally, combining dynamic and static balance training scene information, including multiple training scene types such as static balance, dynamic balance, and dynamic-static balance, the multi-dimensional anti-interference enhancement signals are matched with the corresponding scene types. The matched multi-dimensional data is then structured to ultimately construct a multimodal balance dataset.

[0023] Step S200: Perform joint feature analysis based on the multimodal equilibrium dataset to obtain a set of equilibrium state feature vectors.

[0024] Specifically, dynamic pressure distribution data is extracted from the multimodal equilibrium dataset, and pressure change trend analysis is performed to plot the user's pressure trajectory during training. The center position of this trajectory is marked to determine the pressure center trajectory information. Then, the pressure center trajectory information is processed by a time-frequency analysis algorithm to extract time-frequency dynamic features reflecting the pressure change pattern. Next, three-dimensional posture data is extracted from the dataset, and the changes in the user's body posture during training are analyzed using a three-dimensional tilt calculation model to calculate posture stability features that reflect the degree of posture stability. Subsequently, overall motion video data is extracted, and the video frames are processed by center-of-gravity projection. Combined with motion trajectory tracking technology, overall motion stability features reflecting the overall motion smoothness of the user are calculated. Then, a recursive elimination method is used to analyze the time-frequency dynamic features, posture stability features, and overall motion stability features to calculate the feature contribution of each feature to the equilibrium state assessment. The three types of features are arranged in descending order of feature contribution to generate a feature sequence. Finally, based on the contribution of each feature, corresponding weights are set, and the feature sequence is weighted and fused. The fused feature data is integrated into a structured vector form to finally obtain the equilibrium state feature vector group.

[0025] Step S300: Use the equilibrium state feature vector group to identify the equilibrium state of the target user, set the equilibrium state level, perform training compensation based on the equilibrium state level, and generate an equilibrium stability index.

[0026] Specifically, the obtained equilibrium state feature vector set is input into a preset equilibrium state recognition model. This model compares the feature vectors with pre-stored standard feature libraries for different equilibrium states, such as stable state, slightly unbalanced state, moderately unbalanced state, and severely unbalanced state, and uses a feature matching degree algorithm to determine the target user's current equilibrium state. Based on the matching results, a corresponding equilibrium state level is set, clarifying the quantification range of the user's equilibrium ability. For example, a matching degree ≥90% corresponds to a stable level, 70%~89% corresponds to a slightly unbalanced level, 50%~69% corresponds to a moderately unbalanced level, and <50% corresponds to a severely unbalanced level. Subsequently, based on the set equilibrium state... The system assesses the user's balance level and invokes a pre-stored training compensation strategy library. For example, a mild imbalance level corresponds to a compensation scheme that increases the duration of basic static balance training and reduces the difficulty gradient of dynamic training; a moderate imbalance level corresponds to a compensation scheme that introduces an auxiliary balance training module and refines posture correction guidance; and a severe imbalance level corresponds to a compensation scheme that initiates low-intensity adaptive training and increases the frequency of real-time posture monitoring. Through compensation strategies, the system conducts targeted assessments of the user's weak dimensions of balance ability, such as shift of the center of pressure, insufficient posture stability, and poor overall motion coordination. The compensation effect is combined with the characteristics of the initial balance state, and a balance stability index reflecting the user's current overall balance ability level is generated through quantitative calculation.

[0027] Step S400: Based on the dynamic and static balance training objectives and the balance stability index, perform adaptive analysis to generate training parameter adjustment instructions, and feed the training parameter adjustment instructions back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training.

[0028] Specifically, the system extracts information about the target user's current training stage from the balance training platform, such as the basic training stage or the advanced comprehensive training stage. Based on this training stage information, it analyzes the balance ability level the user should currently achieve and defines an appropriate target stability index range. Simultaneously, it sets clear static and dynamic balance training goals based on the user's training needs, such as focusing on static balance improvement in rehabilitation training and dynamic balance strengthening in exercise training. Next, the generated balance stability index is dynamically compared with the defined target stability index range to generate an index comparison result. Based on the result, the stability deviation value between the current index and the target range is analyzed. If the index is below the lower limit of the target range, a negative deviation is calculated; if it is above the upper limit, a positive deviation is calculated. Then, based on the stability deviation value, the adjustment range of training parameters is determined. If the deviation is large, the platform tilt angle adjustment step length needs to be increased, and the movement speed adjustment range needs to be widened; if the deviation is small, the complexity of the movement trajectory is fine-tuned, and the posture correction frequency is optimized. The deviation value and adjustment range are integrated to generate training parameter adjustment instructions. Finally, the instruction is fed back to the balance training platform: the platform tilt angle parameter is converted into a pulse control signal, the motion trajectory parameter is converted into a subdivision control signal, and the motion speed parameter is converted into a voltage regulation signal, which is transmitted to the actuator to drive it to adjust the training conditions; at the same time, the adjusted training control effect, such as changes in the user's balance state and parameter adaptability, is traced back to the platform to perform closed-loop optimization of the training parameter adjustment instruction, and continuously realize intelligent monitoring of dynamic and static balance training.

[0029] In one possible implementation, step S100 further includes:

[0030] Step S110: Traverse the target dynamic and static balance training area to network multiple sensors and construct a network topology.

[0031] Step S120: Perform multi-dimensional monitoring of target users within the target dynamic and static balance training area according to the network topology, and generate multi-dimensional monitoring signals.

[0032] Step S130: The multidimensional monitoring signal is bidirectionally enhanced to generate a multidimensional anti-interference enhancement signal.

[0033] Step S140: Based on the multidimensional anti-interference enhancement signal, match it according to the dynamic and static balance training scenario information to construct the multimodal balance dataset.

[0034] Specifically, a spatial survey of the target dynamic and static balance training area is first conducted to clarify the core scope of the training activities, equipment layout, and possible movement trajectories of users. Then, various suitable sensors are selected, including pressure sensors for collecting pressure data, inertial sensors for recording posture information, and visual acquisition devices for capturing motion images. Following the principles of no blind spots and no data overlap or redundancy, the sensors are deployed in key locations in the training area, such as the surface of the balance training platform and the surrounding supports. All sensors are then networked together via wired or wireless means to construct a network topology that clearly reflects the sensor locations, communication links, and data transmission paths, ensuring that each sensor can collaboratively complete data acquisition.

[0035] Next, based on the established network topology, when the target user conducts dynamic and static balance training, all sensors are activated to work synchronously: pressure sensors collect dynamic pressure distribution data of the user's feet or body contact points with the platform in real time, recording the temporal and spatial distribution of pressure changes; inertial sensors are attached to key parts of the user's body, such as the waist and ankles, to capture the user's posture changes in three-dimensional space and generate three-dimensional posture data including angle, angular velocity, and acceleration; visual acquisition devices, such as high-definition cameras, capture the user's training process from preset angles to obtain complete overall motion video data; then, a unified time synchronization benchmark is set, such as using the activation time of a certain sensor as the origin, to align the dynamic pressure distribution data, three-dimensional posture data, and overall motion video data in time, and then construct a spatial three-dimensional coordinate system based on the actual size of the training area, mapping various types of data to the coordinate system to complete spatial registration, and integrating them to form a multi-dimensional monitoring signal containing multi-dimensional information.

[0036] The generated multidimensional monitoring signals are then processed using a bidirectional enhancement algorithm: on the one hand, a noise filtering module removes redundant noise caused by environmental interference, such as electromagnetic interference, light changes, and equipment errors, such as small sensor offsets, while retaining effective data features; on the other hand, a signal amplification module appropriately amplifies weak key signals, such as small-amplitude inertial signals generated by slight user posture movements and subtle changes in pressure distribution, ensuring that signal features are clear and identifiable, ultimately generating a multidimensional anti-interference enhanced signal with strong anti-interference capabilities and high data accuracy.

[0037] Finally, the training content of the target static and dynamic balance training area is first decomposed into scenarios to determine the multiple training scenario types included in the training scenario information. Specifically, they are divided into static balance training types, such as standing on one leg and standing with eyes closed; dynamic balance training types, such as platform movement and obstacle crossing; and static and dynamic balance training types, such as limb dynamic activities in a standing state. Then, according to the characteristics of different training scenario types, multidimensional anti-interference enhancement signals are matched with corresponding scenarios. For example, stable data of pressure distribution and small fluctuation data of posture are matched to static balance training types, motion trajectory data and continuous change data of posture are matched to dynamic balance training types, and data combining pressure changes and posture activities are matched to static and dynamic balance training types, thus obtaining multidimensional matching data adapted to each scenario. Finally, the multidimensional matching data is structured and standardized according to the format of scenario type-data dimension-timestamp-data value-data feature label, and finally a multimodal balance dataset is constructed.

[0038] In one possible implementation, step S120 further includes:

[0039] Step S121: In the dynamic and static balance training, pressure sensing is performed on the target users of the balance training platform to generate dynamic pressure distribution data.

[0040] Step S122: In the dynamic and static balance training, inertial sensing is performed on the target user to generate three-dimensional attitude data.

[0041] Step S123: Activate the visual acquisition device to acquire motion data of the target user during dynamic and static balance training, and obtain overall motion video data.

[0042] Step S124: Set a time synchronization benchmark, and synchronize the dynamic pressure distribution data, the three-dimensional posture data, and the overall motion video data according to the time synchronization benchmark to generate multi-source time synchronization data.

[0043] Step S125: Perform spatial three-dimensional analysis based on the target dynamic and static balance training area, construct a regional three-dimensional coordinate system, map the multi-source time synchronization data to the regional three-dimensional coordinate system for spatial registration, and construct multi-source monitoring data.

[0044] Step S126: Add the multi-source monitoring data to the multi-dimensional monitoring signal.

[0045] Specifically, when the target user steps onto the balance training platform to conduct dynamic and static balance training, the pressure sensor array pre-deployed on the platform surface, such as a distributed piezoresistive pressure sensor, is activated to collect the pressure changes in the contact area between the user's feet and the platform in real time, including the pressure value at different contact points, the pressure distribution range, and the dynamic trend of pressure changes with training movements. This data is then organized according to a time series to generate dynamic pressure distribution data, which quantitatively reflects the pressure support status of the user's center of gravity during training.

[0046] Inertial sensors, such as six-axis inertial measurement units (SMUs) containing three-axis accelerometers and three-axis gyroscopes, are worn on key parts of the user's body, such as the waist, ankles, and wrists, which are closely related to balance control. During training, these sensors capture the three-dimensional motion state of various parts of the user's body in real time, including acceleration, angular velocity, and posture angles, such as changes in tilt and rotation angles. The posture data collected by each sensor is summarized and formatted to generate three-dimensional posture data that can fully reflect the dynamic adjustment process of the user's body posture.

[0047] Activate visual acquisition devices, such as high-definition industrial cameras and depth cameras, deployed at preset locations around the dynamic and static balance training area. These devices cover the training area at preset angles and can completely capture all of the user's movements from the start of training to the current moment, including overall body movement, limb extension and contraction, and shift of center of gravity. The captured continuous image frames are integrated in chronological order to obtain overall motion video data, so as to intuitively present the user's training movement trajectory.

[0048] To ensure the time consistency of data from different sources, a unified time synchronization benchmark is set, such as the initial moment of pressure sensor activation as the time origin. Based on this benchmark, timestamp matching is performed on dynamic pressure distribution data, three-dimensional attitude data, and overall motion video data. The acquisition results corresponding to the same moment in the three types of data are associated, eliminating time deviations caused by device startup time differences or data transmission delays, generating multi-source time-synchronized data, and ensuring the time correlation of data in subsequent analysis.

[0049] A three-dimensional spatial analysis of the target dynamic and static balance training area is performed. Combining the actual dimensions of the area, such as length, width, and height, and the equipment deployment location, a three-dimensional coordinate system is constructed with the center of the balance training platform as the origin, the horizontal direction as the X and Y axes, and the vertical direction as the Z axis. Various parameters from multi-source time synchronization data, such as the spatial position coordinates of pressure sensors, the attitude spatial coordinates of inertial sensors, and the motion spatial coordinates captured by visual acquisition devices, are mapped to this three-dimensional coordinate system. The deviation between different equipment coordinate systems is eliminated through coordinate calibration algorithms, and spatial registration is completed. This allows various types of data to be correlated and analyzed in the same spatial dimension, thereby constructing multi-source monitoring data.

[0050] The multi-source monitoring data that has undergone time synchronization and spatial registration, including dynamic pressure distribution data, three-dimensional posture data, and overall motion video data, are integrated and added to the multi-dimensional monitoring signal in a preset data format, such as structured data tables and time-series data files. This enables the multi-dimensional monitoring signal to comprehensively cover multi-dimensional information such as pressure, posture, and motion vision during user training.

[0051] In one possible implementation, step S140 further includes:

[0052] Step S141: Perform dynamic balance training analysis based on the target dynamic and static balance training area to obtain dynamic and static balance training scenario information, which includes multiple training scenario types.

[0053] Step S142: Match the multidimensional anti-interference enhancement signal according to the multiple training scenario types to obtain multidimensional matching data.

[0054] Step S143: Perform structured processing on the multidimensional matching data to construct the multimodal balanced dataset.

[0055] Specifically, a comprehensive dynamic balance training analysis is conducted on the training content, equipment functions, and preset training programs of the target static and dynamic balance training area. This analysis combines the adjustable functions of the balance training platform, such as whether it supports tilt angle changes and motion trajectory simulation, training objectives (e.g., improving static stability, enhancing dynamic coordination), and common user training movements (e.g., standing on one leg, platform following movement, obstacle crossing), to clearly define multiple training scenario types. Typical types include static balance training (e.g., standing on one leg on a fixed platform, standing with eyes closed), dynamic balance training (e.g., platform tilting at a constant speed, random trajectory movement), and combined static and dynamic balance training (e.g., dynamic limb extension while standing, maintaining balance while the platform moves slowly). These scenario types and their corresponding characteristic parameters, such as the platform state in static scenarios and the movement speed in dynamic scenarios, are then integrated to form complete static and dynamic balance training scenario information.

[0056] Based on the identified training scenario types, the core data requirement features of each scenario type are extracted. For example, static balance training needs to focus on the stability data of pressure distribution and the data of minute fluctuations in posture; dynamic balance training needs to focus on the data of continuous changes in motion trajectory and the data of posture adjustment with platform movement; and the combined static and dynamic balance training needs to simultaneously correlate the coordinated data of pressure changes, posture adjustment, and limb movement. Subsequently, the generated multidimensional anti-interference enhancement signals, including dynamic pressure distribution enhancement signals, three-dimensional posture enhancement signals, and overall motion video enhancement signals, are matched with the core data requirement features of each scenario type: pressure stability signals and posture micro-motion signals are matched to static balance training, motion trajectory signals and dynamic posture adjustment signals are matched to dynamic balance training, and pressure-posture-motion coordination signals are matched to combined static and dynamic balance training, ultimately obtaining multidimensional matching data that corresponds one-to-one with each scenario type.

[0057] The obtained multidimensional matching data is structured: the data is standardized according to a unified format of training scene type-data dimension-timestamp-data value-data feature label. The training scene type indicates whether the data belongs to a static, dynamic, or combined static / dynamic scene; the data dimension distinguishes data categories such as pressure distribution, 3D pose, and overall motion video; the timestamp is associated with the specific time of data acquisition, consistent with the time synchronization benchmark mentioned earlier; the data value records the quantization result of the original enhanced signal; and the data feature label marks key data features, such as the pressure center offset in static scenes and the trajectory deviation in dynamic scenes. Through this structured processing, the scattered multidimensional matching data is integrated into a logically clear and easily accessible structured dataset for subsequent feature analysis and retrieval, ultimately constructing a multimodal balanced dataset.

[0058] In one possible implementation, step S142 further includes:

[0059] Based on the multiple training scenario types, static and dynamic analysis is performed to determine the static balance training type, dynamic balance training type, and dynamic-static balance training type.

[0060] The multidimensional anti-interference enhancement signal is matched according to the static balance training type to obtain the first pressure distribution signal and the first attitude micro-motion signal.

[0061] The multidimensional anti-interference enhancement signal is matched according to the dynamic balance training type to obtain the first motion trajectory signal and the first posture change signal.

[0062] The multidimensional anti-interference enhancement signal is matched according to the dynamic and static balance training type to obtain the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal.

[0063] The training evaluation is carried out by iterating through the static balance training type, the dynamic balance training type, and the dynamic-static balance training type, and multiple training difficulty levels are set.

[0064] Based on the multiple training difficulty levels and the static balance training type, the first pressure distribution signal and the first posture micro-motion signal are integrated and transformed to obtain the first matching data.

[0065] Based on the multiple training difficulty levels and the dynamic balance training type, the first motion trajectory signal and the first posture change signal are integrated and transformed to obtain the second matching data.

[0066] Based on the multiple training difficulty levels and the dynamic-static balance training type, the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal are integrated and transformed to obtain the third matching data.

[0067] Specifically, based on the identified training scenario types, dynamic and static analysis is conducted. Combining the characteristics of training movements and differences in balance requirements, three core training types are clearly defined: static balance training, which focuses on the user's body remaining relatively still and maintaining a fixed posture, such as single-leg standing and standing with eyes closed; dynamic balance training, which focuses on the continuous movement of the user's body or training platform and the need for dynamic balance adjustment, such as platform tilting and obstacle crossing training; and dynamic-static balance training, which integrates static posture maintenance and dynamic movement execution, such as limb extension and turning training in a standing state.

[0068] Next, the three training types were matched with multidimensional anti-interference enhancement signals respectively: For the static balance training type, the data reflecting the pressure support and subtle posture adjustment of the user in a static state were extracted from the signal, and the first pressure distribution signal was obtained by matching. The stable value and slight fluctuation of the pressure distribution on the sole of the foot and the first posture micro-motion signal were recorded, that is, the small tilt and sway data of the body in three-dimensional space were captured.

[0069] For dynamic balance training, the focus is on extracting data from the signal that reflects the dynamic changes in motion trajectory and posture, matching to obtain the first motion trajectory signal, and recording the motion path of the user's center of gravity or key parts, and the first posture change signal, that is, recording the continuous changes in body angle and angular velocity during the movement.

[0070] For static and dynamic balance training, it is necessary to comprehensively extract four types of data: pressure, micro-motion of posture, movement trajectory, and posture change, and match them to obtain the second pressure distribution signal, which is the pressure change based on the static posture; the second micro-motion signal, which is the subtle posture adjustment in the dynamic movement; the second movement trajectory signal, which is the execution path of the dynamic movement; and the second posture change signal, which is the continuous posture adjustment in the dynamic movement.

[0071] For static balance training types, such as standing on one leg or standing with eyes closed, which focus on maintaining a relatively static posture, the assessment indicators include the duration of balance maintenance, the magnitude of pressure center shift, and the frequency of micro-movements to determine different users' control over static balance. For dynamic balance training types, such as platform tilting and obstacle crossing, which require dynamic balance adjustments, the assessment indicators include the user's deviation from the trajectory of the movement, the speed of posture adjustment response, and the continuity of movement to evaluate their dynamic balance coordination ability. For static-dynamic balance training types, such as standing arm raises and standing turns, which combine static postures and dynamic movements, the assessment indicators include multi-dimensional indicators such as pressure distribution stability, posture micro-movement control, movement trajectory accuracy, and the smoothness of movement completion to evaluate the user's level of balance control in both static and dynamic states. Subsequently, based on the evaluation results of the three training types, and combined with the balance ability foundation of different user groups, such as rehabilitation patients, sports enthusiasts, and professional athletes, and their initial ability differences and training goals, such as basic balance correction, intermediate balance improvement, and advanced balance reinforcement, multiple training difficulty levels were set for each training type. For example, static balance training was set as follows: "Basic level: Stand on one leg for 10 seconds, allowing small pressure deviations", "Intermediate level: Stand on one leg for 30 seconds, close eyes, limit the magnitude of pressure deviations", and "Advanced level: Stand on one leg for 60 seconds, with slight disturbances, strictly control micro-movements of posture"; dynamic balance training was set as follows: "Basic level: Platform tilts slowly, simple trajectory", "Intermediate level: Platform moves at a constant speed, increasing trajectory complexity", and "Advanced level: Platform moves at random speeds, adding disturbances"; and static-dynamic balance training was set as follows: "Basic level: Stand and raise arms to chest level, slow movement", "Intermediate level: Stand and turn 90 degrees, raise arms overhead, continuous movement", and "Advanced level: Stand and cross small obstacles, dynamic limb extension, precise movement", ensuring that the difficulty level matches the user's ability and meets differentiated training needs.

[0072] Multiple training difficulty levels are clearly defined for static balance training, such as basic, intermediate, and advanced levels. Quantitative standards are set for each difficulty level; for example, the basic level allows a pressure center offset of ≤5cm and a posture tilt angle of ≤3°; the intermediate level allows a pressure center offset of ≤3cm and a posture tilt angle of ≤2°; and the advanced level allows a pressure center offset of ≤1cm and a posture tilt angle of ≤1°. Next, for the first pressure distribution signal, data is filtered according to the quantitative standards for different difficulty levels. Signal segments that meet the pressure fluctuation requirements of the current difficulty level are retained, while abnormal data exceeding the deviation range are removed. The filtered pressure data is then converted into a structured format of pressure center coordinates-timestamp-fluctuation amplitude. For the first posture micro-motion signal, similarly, micro-motion data that meets the posture deviation requirements is extracted according to the difficulty level standards. Parameters such as posture tilt angle and sway frequency at each time point are calculated and converted into a structured format of posture angle-timestamp-sway frequency. Finally, using the time synchronization benchmark as a link, the processed first pressure distribution signal and the first posture micro-motion signal under the same difficulty level are correlated and integrated, and the corresponding training difficulty level labels are added to form the first matching data that can accurately reflect the user's pressure support and posture control state under different difficulties in static balance training.

[0073] Multiple training difficulty levels corresponding to the dynamic balance training type are clearly defined, and differentiated quantitative standards are set for each difficulty level. Different difficulty levels differ in the complexity of the motion trajectory, speed range, path accuracy requirements, and response time and angle change range requirements for attitude adjustment. The higher the difficulty, the more stringent the requirements for trajectory accuracy and response speed, and the stricter the restrictions on attitude angle changes. Next, for the first motion trajectory signal, data is filtered and transformed according to the trajectory standards of different difficulty levels. Signal segments that meet the trajectory type, speed range, and path accuracy requirements under the current difficulty are extracted, and abnormal data with trajectory deviations exceeding the allowable range are removed. The processed trajectory data is converted into a structured format containing trajectory coordinates, timestamps, and trajectory deviation values. For the first attitude change signal, according to the attitude adjustment standards of different difficulty levels, attitude data that meets the requirements for response time and angle change range are extracted. Parameters such as attitude adjustment rate and angle stability corresponding to each time node are calculated and converted into a structured format containing attitude angle, timestamps, and adjustment rate. Finally, using a preset time synchronization benchmark as a link, the first motion trajectory signal and the first posture change signal processed under the same difficulty level are associated and integrated, and labels corresponding to the training difficulty level are added to form a second matching data that can accurately reflect the user's motion trajectory following state and posture dynamic adjustment state under different difficulties in dynamic balance training.

[0074] Multiple training difficulty levels are clearly defined for the dynamic and static balance training types, and multi-dimensional quantitative standards are set for each difficulty level. Different difficulty levels differ in the stable range of pressure distribution, the amplitude limit of micro-motion, the complexity and accuracy requirements of the motion trajectory, the response speed of posture changes, and the angle control. The higher the difficulty, the stricter the requirements for pressure stability and micro-motion control, and the higher the standards for motion trajectory accuracy and posture adjustment response speed. Next, for the second pressure distribution signal, signal segments meeting the stability range requirements are selected according to pressure standards of different difficulty levels, data with excessive pressure fluctuations are removed, and the signal is converted into a structured format containing pressure values, timestamps, and pressure fluctuation degrees. For the second attitude micro-motion signal, signals meeting amplitude limits are extracted according to attitude micro-motion standards of different difficulty levels, and the attitude micro-motion frequency and deviation value at each time node are calculated, and the signal is converted into a structured format containing attitude micro-motion angles, timestamps, and micro-motion frequencies. For the second motion trajectory signal, signals meeting complexity and accuracy requirements are selected according to trajectory standards of different difficulty levels, data with excessive trajectory deviations are removed, and the signal is converted into a structured format containing trajectory coordinates, timestamps, and trajectory deviation rates. For the second attitude change signal, signals meeting response speed and angle control requirements are extracted according to attitude adjustment standards of different difficulty levels, and the attitude adjustment rate and stability at each time node are calculated, and the signal is converted into a structured format containing attitude change angles, timestamps, and adjustment rates. Finally, based on the preset time synchronization benchmark, the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal processed under the same difficulty level are correlated and integrated, and labels corresponding to the training difficulty level are added to form the third matching data that can comprehensively reflect the user's pressure support state, posture micro-motion control state, motion trajectory following state, and posture dynamic adjustment state under different difficulties in dynamic and static balance training.

[0075] In one possible implementation, step S200 further includes:

[0076] Step S210: Perform pressure change analysis based on the dynamic pressure distribution data and draw a pressure trajectory diagram.

[0077] Step S220: Calibrate the center position of the pressure trajectory diagram and determine the pressure center trajectory information.

[0078] Step S230: Perform time-frequency analysis based on the pressure center trajectory information to extract time-frequency dynamic features.

[0079] Step S240: Perform three-dimensional tilt analysis based on the three-dimensional attitude data to calculate attitude stability characteristics.

[0080] Step S250: Project the center of gravity based on the overall motion video data and calculate the overall motion stability characteristics.

[0081] Step S260: Perform recursive elimination based on the time-frequency dynamic features, the attitude stability features, and the overall motion stability features, and calculate the feature contribution.

[0082] Step S270: Arrange the time-frequency dynamic features, the attitude stability features, and the overall motion stability features in descending order according to the feature contribution to generate a feature sequence.

[0083] Step S280: Perform weighted fusion based on the feature sequence to construct the equilibrium state feature vector group.

[0084] Specifically, based on the dynamic pressure distribution data in the multimodal balance dataset, the pressure changes of the target user during the dynamic-static balance training process are continuously analyzed. The pressure value distribution and the movement trend of the pressure center of gravity in the contact area between the user's feet or body and the training platform are tracked at different time points. The pressure change data is visualized according to the time series, and a pressure trajectory map that can intuitively show the movement path of the pressure center of gravity is drawn.

[0085] Next, on the completed pressure trajectory map, the geometric center of the pressure distribution is marked by a coordinate positioning algorithm. Based on this center position, and combined with the movement data of the pressure center of gravity over time, the complete path of the pressure center changing with the training process is sorted out, and the pressure center trajectory information containing information such as the pressure center coordinates, movement distance, and movement direction is determined.

[0086] Time-frequency analysis algorithms, such as wavelet transform, are used to process the trajectory information of the pressure center, converting the time-domain movement data of the pressure center into time-frequency domain data. The frequency characteristics, amplitude variation patterns, and energy distribution of the pressure center's movement in different time periods are analyzed, and time-frequency dynamic features that can reflect the dynamic changes of pressure are extracted.

[0087] Three-dimensional posture data is extracted from the multimodal balance dataset. This data contains the posture information of key body parts of the target user during dynamic and static balance training, such as the waist and ankles, which are closely related to balance control, in three-dimensional space. It covers the angle, angular velocity, and acceleration data of each part at different time points. Next, using a preset spatial coordinate system as a reference, three-dimensional tilt analysis is performed on the three-dimensional posture data. The tilt angle of each key part in the X, Y, and Z axes is calculated one by one, and the trend of the tilt angle in each dimension is tracked with training time, recording the fluctuation range and rate of change of the tilt angle. Subsequently, statistical analysis algorithms are used to process the three-dimensional tilt data of each key part, eliminating abnormal fluctuations caused by accidental movements, and calculating the mean, standard deviation, and coefficient of variation of the tilt angle in each dimension. These parameters reflect the concentration and dispersion of the tilt angle. Finally, the statistical parameters of the tilt angle of each key part in the three dimensions are integrated, combined with the influence weight of different parts on the overall balance, and a weighted calculation is used to generate a posture stability feature that comprehensively reflects the stability of the target user's body posture.

[0088] The video frame parsing tool processes the overall motion video data, breaking down the continuous video stream into independent static image frames at fixed time intervals, such as every 0.05 seconds, to ensure coverage of every motion stage throughout the training process. Next, a human joint point recognition algorithm is run in each frame to accurately locate key joints such as the head, shoulders, hips, knees, and ankles. A pixel coordinate transformation model converts the pixel positions of each joint point in the image into actual spatial coordinates. Subsequently, considering the body weight percentage corresponding to each joint point (e.g., hips approximately 25%, shoulders approximately 20%, head approximately 8%), the spatial coordinates of the user's center of gravity in each frame are calculated through weighted summation. These coordinates are then mapped to a pre-constructed 3D coordinate system of the training area, completing the center of gravity projection and recording the 3D coordinate values ​​of the projection points. Next, a trajectory tracking algorithm is used to concatenate the coordinates of the centroid projection points of all image frames to generate the centroid projection trajectory for the entire training process. Spatial distance calculation tools are used to calculate the maximum displacement difference of the trajectory in the X, Y, and Z axes, yielding the average moving speed and instantaneous maximum speed of the projection points. Frequency analysis tools are then used to count the number of swaying events of the projection points per unit time. Finally, the parameters, including the maximum displacement difference, average moving speed, instantaneous maximum speed, and number of swaying events, are weighted to obtain a comprehensive quantized value. This quantized value represents the overall motion stability characteristic; a smaller value indicates more stable overall motion.

[0089] The time-frequency dynamic features, attitude stability features, and overall motion stability features are decomposed into multiple basic feature components to construct an initial feature set. Then, a recursive feature elimination framework is introduced, using the accuracy of equilibrium state recognition as the evaluation metric. In each iteration, the feature component that contributes the least to the improvement of recognition accuracy in the current feature set is removed, while the impact of the remaining feature components on the accuracy during the iteration is recorded. Through multiple iterations until a preset number of core feature components are retained, the feature contribution of each original feature, namely the time-frequency dynamic features, attitude stability features, and overall motion stability features, is calculated based on the cumulative contribution of each feature component to the recognition accuracy during the iteration.

[0090] The calculated feature contribution values ​​are compared and sorted from high to low to ensure that the feature with the highest contribution is at the beginning of the sequence and the feature with the lowest contribution is at the end of the sequence, thus forming an ordered feature sequence.

[0091] The corresponding weight coefficients are set according to the contribution of each feature in the feature sequence. The higher the contribution, the greater the weight ratio of the feature. Then, the quantized value of each feature is multiplied by the corresponding weight coefficient and summed to obtain a comprehensive feature value. Then, the comprehensive feature value is integrated with the original quantized value of each feature according to the preset format and converted into a vector form. Finally, a balance state feature vector group that can comprehensively reflect the balance state of the target user is constructed.

[0092] In one possible implementation, step S400 further includes:

[0093] Step S410: Extract the training stage information of the target user based on the dynamic and static training platform, perform training stability analysis based on the training stage information, and define the target stability index range.

[0094] Step S420: Based on the training phase information, perform expected control on the target user and set a dynamic-static balance training target.

[0095] Step S430: Dynamically compare the balance stability index with the target stability index range according to the dynamic and static balance training target, and generate index comparison results.

[0096] Step S440: Perform stability deviation analysis based on the index comparison results to generate a stability deviation value.

[0097] Step S450: Determine the parameter adjustment range based on the stable deviation value, integrate the stable deviation value and the parameter adjustment range, and generate the training parameter adjustment instruction.

[0098] Specifically, the training stage information of the target user is extracted from the database of the dynamic and static training platform. This information includes the user's current training stage, such as the basic training stage, the intermediate comprehensive training stage, the advanced intensive training stage, the completed training time, and historical training results. Based on this training stage information, the user's balance stability performance at the corresponding stage is analyzed. For example, the basic stage focuses on evaluating basic static balance ability, the intermediate stage focuses on analyzing dynamic balance coordination ability, and the advanced stage focuses on balance control ability under complex disturbances. Combining the balance level standards that users should achieve at each stage, a target stability index range that is suitable for the current training stage is defined. For example, the target range for the basic stage is 60-75, for the intermediate stage it is 76-90, and for the advanced stage it is 91-100.

[0099] Based on the extracted training stage information, the expected control of the target user is carried out, and dynamic and static balance training objectives are set in combination with the training focus of different stages: the initial stage objective focuses on improving the duration of static balance maintenance and the stability of pressure distribution, ensuring that the user masters the basic balance control methods; the intermediate stage objective shifts to strengthening the dynamic balance response speed and motion trajectory following accuracy, enhancing the user's ability to adjust balance in dynamic scenarios; the advanced stage objective is set to improve balance stability in complex interference environments, achieving adaptation to high-difficulty training scenarios, so that the training objectives are accurately matched with the user's current stage ability and advanced needs.

[0100] Based on the set dynamic and static balance training target, the generated balance stability index is dynamically compared with the defined target stability index range: if the balance stability index is within the target range, it is determined that the index meets the target; if the index is lower than the lower limit of the target range, it is determined that the index is too low; if the index is higher than the upper limit of the target range, it is determined that the index exceeds the target. A clear index comparison result is generated through this comparison logic.

[0101] Subsequently, stability deviation analysis is conducted based on the index comparison results: when the result is that the index is too low, the difference between the balance stability index and the lower limit of the target interval is calculated, and this difference is a positive deviation value; when the result is that the index exceeds the target, the difference between the index and the upper limit of the target interval is calculated, and this difference is a negative deviation value; when the result is that the index meets the target, the deviation value is recorded as 0. Through quantitative calculation, a stable deviation value that can reflect the gap between the current balance stability and the target requirement is generated.

[0102] The adjustment range of training parameters is determined based on the magnitude of the stable deviation value: the larger the deviation value, the larger the parameter adjustment range. For example, when the positive deviation value is 10, the adjustment range of the platform tilt angle is set to 5° and the adjustment range of the movement speed is set to 0.2m / s; when the positive deviation value is 5, the adjustment ranges are set to 2° and 0.1m / s respectively; when the deviation value is 0, the adjustment range is set to 0. The specific adjustment direction corresponding to the stable deviation value, such as increasing / decreasing the difficulty, is integrated with the determined parameter adjustment range to clarify the types of parameters that need to be adjusted on the training platform, such as tilt angle, movement trajectory, movement speed and specific values, and finally, an executable training parameter adjustment instruction is generated.

[0103] In one possible implementation, step S400 further includes:

[0104] Step S410: Feed back the training parameter adjustment command to the static and dynamic training platform for tilt recognition, obtain the platform tilt angle adjustment parameters, and convert the platform tilt angle adjustment parameters into pulse control signals.

[0105] Step S420: Feed back the training parameter adjustment command to the dynamic and static training platform for motion trajectory recognition, obtain the platform motion trajectory adjustment parameters, and convert the platform motion trajectory adjustment parameters into subdivided control signals.

[0106] Step S430: Feed back the training parameter adjustment command to the dynamic and static training platform to identify the motion speed, obtain the platform motion speed adjustment parameters, and convert the platform motion speed adjustment parameters into a voltage regulation signal.

[0107] Step S440: Transmit the pulse control signal, the subdivision control signal, and the voltage adjustment signal to the actuator to drive the actuator to perform training control and generate a dynamic and static balance control effect.

[0108] Step S450: The dynamic-static balance control effect is traced back to the dynamic-static training platform to perform closed-loop optimization of the training parameter adjustment instructions.

[0109] Specifically, the generated training parameter adjustment instructions are fed back to the dynamic and static training platform. The instructions are analyzed to extract parameters related to platform tilt control, including the tilt direction to be adjusted (e.g., forward / backward tilt along the X-axis, left / right tilt along the Y-axis) and the specific tilt angle value. After parameter verification and calibration, the final executable platform tilt angle adjustment parameters are determined. Subsequently, the platform's built-in signal conversion algorithm is invoked to convert the determined platform tilt angle adjustment parameters into pulse control signals that can be directly recognized by the actuator platform drive motor. The frequency of the pulse signal corresponds to the tilt angle adjustment rate, and the total number of pulse signals corresponds to the tilt angle to be adjusted. This ensures that the actuator can accurately complete the platform tilt angle adjustment action according to the pulse control signal, providing a foundation for parameter adaptation in subsequent dynamic and static balance training.

[0110] The generated training parameter adjustment instructions are fed back to the dynamic and static training platform. The instructions are analyzed in detail, focusing on extracting control parameters related to the platform's motion trajectory. These parameters include the trajectory type (e.g., straight line, arc, polyline, or custom curve), key coordinate points on the trajectory path (including spatial coordinates in the X, Y, and Z axes), the start and end points of the trajectory, and the required accuracy. By verifying and integrating these parameters, the final executable platform motion trajectory adjustment parameters are determined. Next, the platform's preset signal conversion module is invoked, and a subdivision control algorithm is used to convert the determined platform motion trajectory adjustment parameters into subdivision control signals that can be accurately recognized by the actuator platform drive motor. These signals refine the trajectory path according to preset step sizes, ensuring that the actuator strictly follows the set trajectory when driving the platform, reducing trajectory deviation, and meeting the precise control requirements for the platform's motion trajectory in different dynamic and static balance training scenarios. This provides a suitable dynamic training environment for the target user.

[0111] The training parameter adjustment command is fed back to the dynamic and static training platform. The speed-related parameters in the command are analyzed to extract the direction of the motion speed to be adjusted, such as acceleration, deceleration, or constant speed maintenance, the target speed value, and the smoothness requirements of speed changes. By comparing and verifying with the current operating speed of the platform, the final platform motion speed adjustment parameters are determined. Then, a signal conversion algorithm is called to convert the speed adjustment parameters into a voltage regulation signal. The amplitude of the voltage signal corresponds to the target speed, and the rate of voltage change corresponds to the smoothness of the speed adjustment, ensuring that the actuator can accurately control the platform motion speed according to the voltage signal to adapt to the user's current balance training rhythm.

[0112] Next, the generated pulse control signal, the generated subdivision control signal, and the generated voltage regulation signal are synchronously integrated and transmitted to the actuator of the dynamic and static training platform through a dedicated signal transmission channel. After receiving the three types of signals, the actuator synchronously adjusts the tilt angle, motion trajectory, and motion speed of the platform according to the signal instructions, driving the platform to operate with optimized parameters, providing the target user with a training environment that matches their current ability. During this process, the user's balance state feedback and the platform parameter execution accuracy are recorded in real time, generating a dynamic and static balance control effect that reflects the training effect after parameter adjustment.

[0113] Data related to the generated dynamic and static balance control effects are collected. This data covers the balance stability performance of the target user under adjusted training parameters, such as changes in the balance stability index, deviations between the actual execution parameters and command parameters of the training platform, and key information such as the smoothness of the user's posture adjustment and the stability of pressure distribution during training. Next, this dynamic and static balance control effect data is completely backtracked to the dynamic and static training platform. The collected actual control effects are comprehensively compared with preset training effect evaluation standards. If the actual control effect meets or exceeds the standards, such as the balance stability index increasing to the target range or the parameter deviation value being controlled within the allowable range, the current training parameter adjustment command is considered to have good adaptability, and the existing command generation logic is retained. If the actual control effect does not meet the standards, such as the balance stability index not significantly improving or even decreasing, or the parameter deviation value exceeding the threshold, the causes of the deviation are analyzed in depth, such as unreasonable parameter adjustment ranges or execution deviations caused by signal conversion delays. Finally, based on the analysis results, the generation rules for training parameter adjustment instructions are modified in a targeted manner. For example, the calculation formula for the parameter adjustment range is optimized, and the signal conversion coefficient is adjusted to reduce execution delay. The modified rules will be used to generate subsequent training parameter adjustment instructions, forming a closed-loop mechanism of instruction generation-execution-effect feedback-instruction optimization. This ensures that subsequent training parameter adjustments can more accurately adapt to the balance training needs of the target user and continuously improve the intelligent monitoring and control quality of dynamic and static balance training.

[0114] Example 2, based on the same inventive concept as the intelligent monitoring method based on dynamic and static balance training in the foregoing examples, such as... Figure 2 As shown, this application provides an intelligent monitoring system based on dynamic and static balance training. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0115] The balanced dataset construction module 10 is used to simultaneously collect multidimensional data from target users during dynamic and static balance training to construct a multimodal balanced dataset.

[0116] The feature vector group acquisition module 20 is used to perform joint feature analysis based on the multimodal equilibrium dataset to obtain the equilibrium state feature vector group.

[0117] The stability index generation module 30 is used to identify the balance state of the target user by the balance state feature vector group, set the balance state level, perform training compensation based on the balance state level, and generate a balance stability index.

[0118] The intelligent monitoring module 40 is used to perform adaptive analysis based on the dynamic and static balance training objectives and the balance stability index, generate training parameter adjustment instructions, and feed the training parameter adjustment instructions back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training.

[0119] Furthermore, the system is also used to implement the following functions:

[0120] The target dynamic-static balance training area is traversed to network multiple sensors and construct a network topology; target users within the target dynamic-static balance training area are monitored in multiple dimensions according to the network topology to generate multi-dimensional monitoring signals; the multi-dimensional monitoring signals are bidirectionally enhanced to generate multi-dimensional anti-interference enhancement signals; and the multi-dimensional anti-interference enhancement signals are matched according to the dynamic-static balance training scene information to construct the multimodal balance dataset.

[0121] Furthermore, the system is also used to implement the following functions:

[0122] In dynamic and static balance training, pressure sensing is performed on the target user of the balance training platform to generate dynamic pressure distribution data; inertial sensing is performed on the target user to generate three-dimensional posture data; a visual acquisition device is activated to collect motion data of the target user during dynamic and static balance training, obtaining overall motion video data; a time synchronization benchmark is set, and the dynamic pressure distribution data, the three-dimensional posture data, and the overall motion video data are synchronized in time according to the time synchronization benchmark to generate multi-source time synchronization data; spatial three-dimensional analysis is performed based on the target dynamic and static balance training area to construct a regional three-dimensional coordinate system, and the multi-source time synchronization data is mapped to the regional three-dimensional coordinate system for spatial registration to construct multi-source monitoring data; the multi-source monitoring data is added to the multi-dimensional monitoring signal.

[0123] Furthermore, the system is also used to implement the following functions:

[0124] Dynamic balance training analysis is performed based on the target dynamic and static balance training region to obtain dynamic and static balance training scenario information, which includes multiple training scenario types. The multidimensional anti-interference enhancement signal is matched according to the multiple training scenario types to obtain multidimensional matching data. The multidimensional matching data is then structured to construct the multimodal balance dataset.

[0125] Furthermore, the system is also used to implement the following functions:

[0126] Based on the multiple training scenario types, dynamic and static analysis is performed to determine the static balance training type, dynamic balance training type, and dynamic-static balance training type. The multidimensional anti-interference enhancement signal is matched according to the static balance training type to obtain a first pressure distribution signal and a first posture micro-motion signal. The multidimensional anti-interference enhancement signal is matched according to the dynamic balance training type to obtain a first motion trajectory signal and a first posture change signal. The multidimensional anti-interference enhancement signal is matched according to the dynamic-static balance training type to obtain a second pressure distribution signal, a second posture micro-motion signal, a second motion trajectory signal, and a second posture change signal. The process iterates through the static balance training type, the dynamic balance training type, and the... The training evaluation is conducted using static and dynamic balance training types, setting multiple training difficulty levels. Based on these multiple training difficulty levels and the static balance training type, the first pressure distribution signal and the first posture micro-motion signal are integrated and transformed to obtain first matching data. Based on these multiple training difficulty levels and the dynamic balance training type, the first motion trajectory signal and the first posture change signal are integrated and transformed to obtain second matching data. Based on these multiple training difficulty levels and the static and dynamic balance training type, the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal are integrated and transformed to obtain third matching data.

[0127] Furthermore, the system is also used to implement the following functions:

[0128] Pressure change analysis is performed based on the dynamic pressure distribution data, and a pressure trajectory diagram is drawn. The center position of the pressure trajectory diagram is marked to determine the pressure center trajectory information. Time-frequency analysis is performed based on the pressure center trajectory information to extract time-frequency dynamic features. Three-dimensional tilt analysis is performed based on the three-dimensional posture data to calculate posture stability features. Center-of-gravity projection is performed based on the overall motion video data to calculate overall motion stability features. Recursive elimination is performed based on the time-frequency dynamic features, posture stability features, and overall motion stability features to calculate feature contribution. The time-frequency dynamic features, posture stability features, and overall motion stability features are arranged in descending order according to the feature contribution to generate a feature sequence. Weighted fusion is performed based on the feature sequence to construct the equilibrium state feature vector group.

[0129] Furthermore, the system is also used to implement the following functions:

[0130] Training stage information of the target user is extracted based on the dynamic and static training platform. Training stability analysis is performed based on the training stage information to define the target stability index range. Expectation control is applied to the target user based on the training stage information to set a dynamic and static balance training target. The balance stability index is dynamically compared with the target stability index range according to the dynamic and static balance training target to generate an index comparison result. Stability deviation analysis is performed based on the index comparison result to generate a stability deviation value. The parameter adjustment range is determined based on the stability deviation value, and the stability deviation value and the parameter adjustment range are integrated to generate the training parameter adjustment instruction.

[0131] Furthermore, the system is also used to implement the following functions:

[0132] The training parameter adjustment command is fed back to the static and dynamic training platform for tilt recognition to obtain platform tilt angle adjustment parameters, which are then converted into pulse control signals. The training parameter adjustment command is also fed back to the static and dynamic training platform for motion trajectory recognition to obtain platform motion trajectory adjustment parameters, which are then converted into subdivision control signals. Furthermore, the training parameter adjustment command is fed back to the static and dynamic training platform for motion speed recognition to obtain platform motion speed adjustment parameters, which are then converted into voltage regulation signals. The pulse control signals, subdivision control signals, and voltage regulation signals are transmitted to the actuator to drive it in training control, generating a dynamic and static balance control effect. This dynamic and static balance control effect is then fed back to the static and dynamic training platform for closed-loop optimization of the training parameter adjustment command.

[0133] Example 3, Figure 3 This is a schematic diagram of the electronic device provided by the intelligent monitoring method based on dynamic and static balance training of the present invention, showing an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of the present invention. Figure 3 As shown, the electronic device includes a processor 21, a memory 22, an input device 23, and an output device 24; the number of processors 21 in the electronic device can be one or more. Figure 3 Taking a processor 21 as an example, the processor 21, memory 22, input device 23, and output device 24 in an electronic device can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0134] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0135] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0136] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An intelligent monitoring method based on dynamic and static balance training, characterized in that, The method includes: In dynamic-static balance training, target users are simultaneously collected in multiple dimensions to construct a multimodal balanced dataset; Based on the multimodal equilibrium dataset, joint feature analysis is performed to obtain a set of equilibrium state feature vectors; The equilibrium state feature vector group is used to identify the equilibrium state of the target user, an equilibrium state level is set, and training compensation is performed based on the equilibrium state level to generate an equilibrium stability index. Adaptive analysis is performed based on the dynamic and static balance training objectives and the balance stability index to generate training parameter adjustment instructions. These instructions are then fed back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training. In the static-dynamic balance training, simultaneous multidimensional data collection from target users is performed to construct a multimodal balance dataset. The methods include: The target dynamic and static balance training area is traversed to network multiple sensors and construct a network topology. According to the network topology, multi-dimensional monitoring is performed on target users within the target dynamic and static balance training area to generate multi-dimensional monitoring signals. The multidimensional monitoring signal is bidirectionally enhanced to generate a multidimensional anti-interference enhanced signal; Based on the multidimensional anti-interference enhancement signal, the multimodal balance dataset is constructed by matching it with the dynamic and static balance training scenario information. According to the aforementioned network topology, multi-dimensional monitoring of target users within the target dynamic and static balance training area is performed to generate multi-dimensional monitoring signals. The method includes: In dynamic and static balance training, pressure sensing is performed on the target users of the balance training platform to generate dynamic pressure distribution data. In dynamic and static balance training, inertial sensing is performed on the target user to generate three-dimensional attitude data; Activate the visual acquisition device to capture the motion of the target user during dynamic and static balance training and obtain overall motion video data; A time synchronization benchmark is set, and the dynamic pressure distribution data, the three-dimensional posture data, and the overall motion video data are synchronized in time according to the time synchronization benchmark to generate multi-source time synchronization data; Based on the target dynamic and static balance training area, spatial three-dimensional analysis is performed to construct a three-dimensional coordinate system for the area. The multi-source time synchronization data is then mapped to the three-dimensional coordinate system for spatial registration to construct multi-source monitoring data. Add the multi-source monitoring data to the multi-dimensional monitoring signal; Based on the aforementioned multimodal equilibrium dataset, joint feature analysis is performed to obtain a set of equilibrium state feature vectors. The method includes: Based on the dynamic pressure distribution data, pressure change analysis is performed, and a pressure trajectory diagram is drawn. The center position of the pressure trajectory diagram is determined to identify the pressure center trajectory information; Time-frequency analysis is performed based on the pressure center trajectory information to extract time-frequency dynamic features; Based on the three-dimensional attitude data, a three-dimensional tilt analysis is performed to calculate the attitude stability characteristics. Based on the overall motion video data, the center of gravity is projected, and the overall motion stability characteristics are calculated. Recursive elimination is performed based on the time-frequency dynamic features, the attitude stability features, and the overall motion stability features, and the feature contribution is calculated. The time-frequency dynamic features, the attitude stability features, and the overall motion stability features are sorted in descending order according to their feature contribution to generate a feature sequence. The equilibrium state feature vector group is constructed by weighted fusion based on the feature sequence. Adaptive analysis is performed based on the dynamic and static balance training objectives and the aforementioned balance stability index to generate training parameter adjustment instructions. The method includes: Based on the dynamic and static training platform, the training stage information of the target user is extracted, and the training stability analysis is performed based on the training stage information to define the target stability index range. Based on the training phase information, the target user's expectations are controlled, and a dynamic-static balance training objective is set. According to the dynamic and static balance training objective, the balance stability index is dynamically compared with the target stability index range to generate an index comparison result. Based on the index comparison results, a stability deviation analysis is performed to generate a stability deviation value; The parameter adjustment range is determined based on the stable deviation value, and the stable deviation value and the parameter adjustment range are integrated to generate the training parameter adjustment instruction.

2. The intelligent monitoring method based on dynamic-static balance training as described in claim 1, characterized in that, Based on the multidimensional anti-interference enhancement signal, a multimodal equilibrium dataset is constructed by matching it with dynamic and static equilibrium training scene information. The method includes: Dynamic balance training analysis is performed based on the target dynamic and static balance training area to obtain dynamic and static balance training scenario information, which includes multiple training scenario types. The multidimensional anti-interference enhancement signal is matched according to the multiple training scenario types to obtain multidimensional matching data; The multidimensional matching data is structured to construct the multimodal balanced dataset.

3. The intelligent monitoring method based on dynamic-static balance training as described in claim 2, characterized in that, The method involves matching the multidimensional anti-interference enhancement signal according to the multiple training scenario types to obtain multidimensional matching data, including: Based on the multiple training scenario types, dynamic and static analysis is performed to determine the static balance training type, dynamic balance training type, and dynamic-static balance training type. The multidimensional anti-interference enhancement signal is matched according to the static balance training type to obtain the first pressure distribution signal and the first attitude micro-motion signal; The multidimensional anti-interference enhancement signal is matched according to the dynamic balance training type to obtain the first motion trajectory signal and the first posture change signal; The multidimensional anti-interference enhancement signal is matched according to the dynamic and static balance training type to obtain the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal. The training evaluation is carried out by iterating through the static balance training type, the dynamic balance training type, and the dynamic-static balance training type, and multiple training difficulty levels are set. Based on the multiple training difficulty levels and the static balance training type, the first pressure distribution signal and the first posture micro-motion signal are integrated and transformed to obtain the first matching data. Based on the multiple training difficulty levels and the dynamic balance training type, the first motion trajectory signal and the first posture change signal are integrated and transformed to obtain the second matching data. Based on the multiple training difficulty levels and the dynamic-static balance training type, the second pressure distribution signal, the second posture micro-motion signal, the second motion trajectory signal, and the second posture change signal are integrated and transformed to obtain the third matching data.

4. The intelligent monitoring method based on dynamic-static balance training as described in claim 1, characterized in that, The method for intelligently monitoring dynamic and static balance training by feeding back the training parameter adjustment instructions to the dynamic and static training platform control actuator includes: The training parameter adjustment command is fed back to the static and dynamic training platform for tilt recognition to obtain the platform tilt angle adjustment parameters, and the platform tilt angle adjustment parameters are converted into pulse control signals. The training parameter adjustment command is fed back to the static and dynamic training platform for motion trajectory recognition to obtain the platform motion trajectory adjustment parameters, and the platform motion trajectory adjustment parameters are converted into subdivided control signals; The training parameter adjustment command is fed back to the static and dynamic training platform for motion speed identification, the platform motion speed adjustment parameters are obtained, and the platform motion speed adjustment parameters are converted into voltage regulation signals. The pulse control signal, the subdivision control signal, and the voltage regulation signal are transmitted to the actuator to drive the actuator to perform training control and generate a dynamic and static balance control effect. The dynamic-static balance control effect is traced back to the dynamic-static training platform to perform closed-loop optimization of the training parameter adjustment instructions.

5. An intelligent monitoring system based on dynamic and static balance training, characterized in that, The system is used to implement the intelligent monitoring method based on dynamic-static balance training as described in any one of claims 1-4, and the system comprises: The balanced dataset construction module is used to simultaneously collect multidimensional data from target users during dynamic and static balance training, and to construct a multimodal balanced dataset. The feature vector group acquisition module is used to perform joint feature analysis based on the multimodal equilibrium dataset to obtain the equilibrium state feature vector group; The stability index generation module is used to identify the balance state of the target user by the balance state feature vector group, set the balance state level, perform training compensation based on the balance state level, and generate a balance stability index. The intelligent monitoring module is used to perform adaptive analysis based on the dynamic and static balance training objectives and the balance stability index, generate training parameter adjustment instructions, and feed the training parameter adjustment instructions back to the balance training platform control execution mechanism for intelligent monitoring of dynamic and static balance training.

6. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is used to execute the intelligent monitoring method based on dynamic and static balance training as described in any one of claims 1 to 4.

Images