Balance training data processing method and system

Abstract

The invention discloses a balance training data processing method and system, and relates to the technical field of data processing, and the method comprises the steps: obtaining multi-modal perception data of a target user during balance function training, and constructing a training perception multi-dimensional vector; evaluating the training perception multi-dimensional vector through a balance training evaluation model to obtain a balance training evaluation sequence, and carrying out balance function training bottleneck tracing to determine balance training bottleneck factors; performing attention association optimization on a real-time balance training scheme of the target user according to the balance training bottleneck factor to obtain a first training adjustment space, and performing balance training deduction optimization according to a balance training evaluation model to obtain a second training adjustment space; and performing neighborhood migration optimization on the training adjustment second space to obtain a balance training adjustment optimization result. The technical problem that effective evaluation and personalized training optimization cannot be carried out in the prior art is solved, and the technical effect of improving the balance training effect and the training adaptability is achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a method and system for processing balanced training data. Background Technology

[0002] During balance training, users' perceptual information, including postural responses, neuromuscular feedback, and training participation behaviors, is multi-dimensional, diverse, and dynamically changing. During training, this perceptual data is often difficult to collect and quantify accurately, leading to a simplistic evaluation of training effectiveness that fails to fully reflect the user's true training state. Furthermore, the lack of in-depth data analysis and bottleneck identification mechanisms makes it difficult to adjust and optimize training programs in a timely and precise manner based on individual differences. This results in insufficient targeting and adaptability in the training process, hindering the effective improvement of users' balance abilities and training efficiency. Summary of the Invention

[0003] This application provides a method and system for processing balanced training data, which addresses the technical problem in the prior art that it is impossible to effectively evaluate and personalize training optimization based on multi-dimensional user perception data during balanced training.

[0004] In view of the above problems, this application provides a balanced training data processing method and system.

[0005] A first aspect of this application provides a method for processing balanced training data, the method comprising: Multimodal perception data of the target user during balance function training is acquired, and a multidimensional training perception vector is constructed based on the multimodal perception data. This multidimensional training perception vector includes posture response feature vectors, neuromuscular feedback feature vectors, and training participation behavior feature vectors. The multidimensional training perception vector is evaluated using a balance training evaluation model to obtain a balance training evaluation sequence. Bottlenecks in balance function training are traced based on the target user's historical balance training dataset and the balance training evaluation sequence to determine balance training bottleneck factors. Attention association optimization is performed on the target user's real-time balance training scheme based on the balance training bottleneck factors to obtain a first training adjustment space. Balance training deduction and optimization are performed on the first training adjustment space using the balance training evaluation model to obtain a second training adjustment space. Finally, neighborhood transfer optimization is performed on the second training adjustment space using a balance training optimization analysis model to obtain the balance training adjustment optimization result.

[0006] A second aspect of this application provides a balanced training data processing system, the system comprising: The data acquisition module acquires multimodal perception data of the target user during balance function training and constructs a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes a posture response feature vector, a neuromuscular feedback feature vector, and a training participation behavior feature vector. The tracing module evaluates the training perception multidimensional vector using a balance training evaluation model to obtain a balance training evaluation sequence. It also traces balance function training bottlenecks based on the target user's historical balance training dataset and the balance training evaluation sequence to determine balance training bottleneck factors. The attention association optimization module optimizes the real-time balance training scheme of the target user based on the balance training bottleneck factors to obtain a first training adjustment space. The deduction optimization module performs balance training deduction optimization on the first training adjustment space using the balance training evaluation model to obtain a second training adjustment space. The transfer optimization module performs neighborhood transfer optimization on the second training adjustment space using a balance training optimization analysis model to obtain balance training adjustment optimization results.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application acquires multimodal perception data of a target user during balance function training, and constructs a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes a posture response feature vector, a neuromuscular feedback feature vector, and a training participation behavior feature vector. The training perception multidimensional vector is evaluated using a balance training evaluation model to obtain a balance training evaluation sequence. Based on the target user's historical balance training dataset and the balance training evaluation sequence, a balance function training bottleneck is traced to determine the balance training bottleneck factor. Attention association optimization is performed on the target user's real-time balance training scheme based on the balance training bottleneck factor to obtain a first training adjustment space. Balance training deduction and optimization are performed on the first training adjustment space using the balance training evaluation model to obtain a second training adjustment space. Finally, neighborhood transfer optimization is performed on the second training adjustment space using a balance training optimization analysis model to obtain the balance training adjustment optimization result. This invention addresses the technical problem in existing technologies that cannot effectively evaluate and personalize training optimization based on multi-dimensional user perception data during balance training. By combining multi-modal perception data modeling, bottleneck factor tracing, and attention association optimization with deduction and neighborhood transfer optimization, it achieves the technical effect of optimizing individualized and precise training programs, improving balance training effectiveness, and enhancing training adaptability. Attached Figure Description

[0008] 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.

[0009] Figure 1 This is a schematic flowchart of a balanced training data processing method provided in an embodiment of this application; Figure 2 This is a schematic diagram of a balanced training data processing system provided in an embodiment of this application.

[0010] Figure labeling: Data acquisition module 11, tracing module 12, attention association optimization module 13, deduction optimization module 14, transfer optimization module 15. Detailed Implementation

[0011] This application provides a method and system for processing balanced training data, which addresses the technical problem in existing technologies that cannot effectively evaluate and personalize training optimization based on multi-dimensional user perception data during balanced training. By combining multi-modal perception data modeling, bottleneck factor tracing, and attention association optimization with deduction and neighborhood transfer optimization, it achieves the technical effect of optimizing individualized and precise training programs, improving the effectiveness of balanced training, and enhancing training adaptability.

[0012] 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.

[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides a method for processing balanced training data, the method comprising: Step S100: Obtain multimodal perception data of the target user during balance function training, and construct a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes posture response feature vector, neuromuscular feedback feature vector, and training participation behavior feature vector.

[0015] In this embodiment, the target user's motion state, physiological responses, and interactive behaviors are first synchronously collected using a training monitoring device to form a training monitoring dataset. This dataset is then cleaned and standardized in multiple dimensions to remove noise and invalid data, yielding multimodal perception data. This multimodal perception data includes posture information, physiological feedback information, and behavioral participation information.

[0016] Next, a multi-dimensional training perception vector is constructed based on multimodal perception data. First, the multimodal perception data is synchronized in time, aligning posture information, physiological feedback information, and behavioral participation information along a unified time axis. Then, keypoint trajectory reconstruction and center-of-gravity (COP) swing analysis are performed on the posture information to extract basic posture response indicators such as COP trajectory length, COP swing area, and average swing speed, which are then used to construct a posture response feature vector according to a preset field order. For physiological feedback information, time-frequency analysis and event labeling methods are used to extract neuromuscular feedback indicators such as root mean square (RMS), mean absolute value (MSD), median frequency, average frequency, reaction time delay, and muscle group coordination signal intensity, which are then used to generate a neuromuscular feedback feature vector according to a preset field order. For behavioral participation information, statistical analysis is performed to extract behavioral participation features such as task completion rate, reaction time, pause duration, number of prompts, and operation frequency, generating a training participation behavioral feature vector. Finally, the posture response feature vector, neuromuscular feedback feature vector, and training participation behavioral feature vector are cascaded and scaled according to a field mapping table to form a multi-dimensional training perception vector.

[0017] Furthermore, the method provided in the application embodiments, in order to obtain the multimodal perception data of the target user during balance function training, further includes: When the target user performs balance function training according to the real-time balance training scheme, the training monitoring dataset is acquired synchronously; the training monitoring dataset is cleaned in multiple dimensions to generate the multimodal perception data.

[0018] In this embodiment of the application, when the target user performs balance function training according to the real-time balance training scheme, the training monitoring dataset is obtained by synchronously collecting angular velocity and linear acceleration from the inertial measurement unit (IMU), plantar pressure and center of gravity COP trajectory from the force platform, electromyography (sEMG) signals, body key point coordinates estimated from video pose estimation, and operation events from human-computer interaction logs based on unified timestamp alignment.

[0019] Subsequently, the training and monitoring dataset was cleaned in multiple dimensions, including low-pass filtering of IMU signals, baseline correction of force platform signals, band-pass filtering of sEMG signals, screening and time resampling of video key points according to confidence thresholds, removal of invalid and duplicate records from interaction logs, and resampling and scale normalization under a unified time axis to generate multimodal perception data containing posture information, physiological feedback information, and behavioral participation information.

[0020] Step S200: Evaluate the training-aware multidimensional vector using the balanced training evaluation model to obtain a balanced training evaluation sequence, and trace the balanced function training bottleneck based on the target user's balanced training historical dataset and the balanced training evaluation sequence to determine the balanced training bottleneck factor.

[0021] In this embodiment, when evaluating the training perception multidimensional vector using the balanced training evaluation model, the balanced training evaluation model is first activated. Various features from the training perception multidimensional vector are then input into their respective evaluation sub-models. This sequentially completes the posture response evaluation of the posture response feature vector, the neuromuscular feedback evaluation of the neuromuscular feedback feature vector, and the training behavior evaluation of the training participation behavior feature vector. Each sub-model outputs posture response evaluation coefficients, neuromuscular feedback evaluation coefficients, and training behavior evaluation coefficients, which are combined to form a balanced training evaluation sequence.

[0022] Subsequently, bottlenecks in balance function training are traced based on the target user's historical balance training dataset and balance training evaluation sequences. This process begins by predicting balance training evaluation trends based on the historical dataset, extracting trend characteristics. Then, anomaly detection is performed on the balance training evaluation sequences using these trend characteristics, identifying anomalies in posture response evaluation, neuromuscular feedback evaluation, and training behavior evaluation. Next, anomaly tracing is performed using posture response feature vectors, neuromuscular feedback feature vectors, and training participation behavior feature vectors, respectively, to obtain the first, second, and third training bottleneck factors. Finally, the three types of training bottleneck factors are fused to generate the balance training bottleneck factor.

[0023] Furthermore, in the method provided in the application embodiments, the method of evaluating the training-aware multidimensional vector through a balanced training evaluation model to obtain a balanced training evaluation sequence further includes: The balance training evaluation model is activated, which includes a posture response evaluation model, a neuromuscular feedback evaluation model, and a training behavior evaluation model. The posture response feature vector is input into the posture response evaluation model to obtain posture response evaluation coefficients. The neuromuscular feedback feature vector is input into the neuromuscular feedback evaluation model to obtain neuromuscular feedback evaluation coefficients. The training participation behavior feature vector is input into the training behavior evaluation model to obtain training behavior evaluation coefficients. The balance training evaluation sequence is generated by combining the posture response evaluation coefficients and the neuromuscular feedback evaluation coefficients.

[0024] In this embodiment, a pre-trained balance training evaluation model is first activated. This model comprises a posture response evaluation model, a neuromuscular feedback evaluation model, and a training behavior evaluation model. During the training of the posture response evaluation model, a set of posture response feature vector samples is used as input data, and a set of posture response evaluation coefficients annotated by technical experts is used as supervision labels. Supervised learning methods are employed to train the model, enabling it to automatically output posture response evaluation coefficients corresponding to posture stability, center of gravity sway characteristics, and posture adjustment amplitude based on the posture response features. During the training of the neuromuscular feedback evaluation model, a set of neuromuscular feedback feature vector samples is used as input data, and a set of neuromuscular feedback evaluation coefficients annotated by technical experts is used as supervision labels. The model is trained by combining the time-domain and frequency-domain features of electromyography (EMG) signals with reaction time delay information, ensuring it accurately reflects the level of neuromuscular feedback. During the training of the training behavior evaluation model, a set of training participation behavior feature vector samples is used as input data, and a set of training behavior evaluation coefficients annotated by technical experts is used as supervision labels. By modeling behavioral characteristics such as task completion rate, reaction time, and operation frequency, the model can output a quantitative result of the degree of behavioral participation.

[0025] In actual evaluation, the posture response feature vector, neuromuscular feedback feature vector, and training participation behavior feature vector are sequentially input into the posture response evaluation model, neuromuscular feedback evaluation model, and training behavior evaluation model to obtain the posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient, respectively.

[0026] Finally, the training behavior evaluation coefficient, posture response evaluation coefficient, and neuromuscular feedback evaluation coefficient are combined according to a preset field structure to obtain a balanced training evaluation sequence.

[0027] Furthermore, in the method provided in the application embodiments, the method for tracing the bottleneck of balance function training and determining the balance training bottleneck factor based on the target user's balance training history dataset and the balance training evaluation sequence further includes: Based on the historical dataset of balanced training, a balanced training evaluation trend is predicted to obtain training evaluation trend characteristics. Anomaly detection is performed on the balanced training evaluation sequence based on these trend characteristics to identify anomalies in posture response evaluation, neuromuscular feedback evaluation, and training behavior evaluation. Anomalies in posture response evaluation are traced back to obtain a first training bottleneck factor based on the posture response feature vector. Anomalies in neuromuscular feedback evaluation are traced back to obtain a second training bottleneck factor based on the neuromuscular feedback feature vector. Anomalies in training behavior evaluation are traced back to obtain a third training bottleneck factor based on the training participation behavior feature vector. The first and second training bottleneck factors are combined to generate the balanced training bottleneck factor.

[0028] In this embodiment, the balanced training historical dataset refers to the set of posture response evaluation coefficients, neuromuscular feedback evaluation coefficients, and training behavior evaluation coefficients stored in chronological order during long-term training. This dataset corresponds to the complete training records of the target user across multiple training cycles. When predicting the balanced training evaluation trend based on the balanced training historical dataset, the posture response evaluation coefficients, neuromuscular feedback evaluation coefficients, and training behavior evaluation coefficients from the historical training phases are arranged in chronological order to form a balanced training evaluation time series. The time series is smoothed using a moving average method to eliminate short-term fluctuations. Then, a univariate linear regression method is used to fit the trend curve, and the trend slope, trend stability interval, and average residual are calculated to form the training evaluation trend characteristics.

[0029] Next, when performing anomaly detection on the balanced training evaluation sequence based on the training evaluation trend characteristics, the difference analysis between the balanced training evaluation sequence of the current training period and the trend baseline curve is performed. The z-score method is used to standardize the calculation of the posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient. Based on a fixed threshold, the deviation is judged, and the components exceeding the threshold are marked as posture response evaluation anomalies, neuromuscular feedback evaluation anomalies, and training behavior evaluation anomalies, respectively, to form anomaly detection results.

[0030] Subsequently, when tracing anomalies in the attitude response evaluation based on the attitude response feature vector, the attitude response feature vector corresponding to the time window is extracted, and the relative rate of change of key indicators such as center of gravity trajectory length, center of gravity swing area and average swing speed is calculated. The dominant indicator with the largest change amplitude is selected as the direct source of attitude anomaly, and the first training bottleneck factor is obtained.

[0031] Subsequently, when tracing anomalies in neuromuscular feedback assessment based on neuromuscular feedback feature vectors, the neuromuscular feedback feature vectors corresponding to the time window are extracted. Threshold determination is performed on neuromuscular signal features such as root mean square, average frequency, and reaction delay. The feature with the greatest deviation is identified as the source of the anomaly, thus obtaining the second training bottleneck factor.

[0032] Next, when tracing anomalies in training behavior evaluation based on the training participation behavior feature vector, the training participation behavior feature vector corresponding to the time window is extracted. Thresholds and statistical judgments are made on features such as task completion rate, reaction time and pause time to identify key indicators that trigger anomalies and obtain the third training bottleneck factor.

[0033] Finally, the first training bottleneck factor, the second training bottleneck factor, and the third training bottleneck factor are integrated to obtain the balanced training bottleneck factor.

[0034] Step S300: Optimize the real-time balance training scheme of the target user based on the balance training bottleneck factor to obtain the first space for training adjustment.

[0035] In this embodiment, when optimizing the real-time balance training scheme for a target user based on the balance training bottleneck factor, the process first involves analyzing the attentional association of training parameters according to the first training bottleneck factor to obtain a first balance training adjustment set. Then, based on the second training bottleneck factor, attentional association is locally adjusted for training parameters related to neuromuscular regulation to form a second balance training adjustment set. Next, attentional association is locally adjusted for training parameters related to behavioral participation according to the third training bottleneck factor to form a third balance training adjustment set. Finally, the three local adjustment sets are globally combined and structurally fused to generate a first training adjustment space.

[0036] Furthermore, in the method provided in the application embodiments, optimizing the attention association of the real-time balance training scheme of the target user based on the balance training bottleneck factor to obtain the first training adjustment space, further includes: Based on the first training bottleneck factor, the real-time balanced training scheme is analyzed for attentional correlation of training parameters to obtain the first bottleneck training correlation analysis result; based on the first bottleneck training correlation analysis result, the real-time balanced training scheme is locally optimized and adjusted to obtain the first balanced training adjustment set; based on the second training bottleneck factor, the real-time balanced training scheme is locally adjusted for attentional correlation to obtain the second balanced training adjustment set; based on the third training bottleneck factor, the real-time balanced training scheme is locally adjusted for attentional correlation to obtain the third balanced training adjustment set; based on the first balanced training adjustment set, the second balanced training adjustment set, and the third balanced training adjustment set, the first training adjustment space is generated by global combination.

[0037] In this embodiment, when performing attention correlation analysis on the training parameters of the real-time balance training scheme based on the first training bottleneck factor, a one-to-one correspondence is first established between the first training bottleneck factor and the posture target angle, training support surface width, and posture adjustment frequency included in the real-time balance training scheme. By constructing an attention weight calculation matrix, a normalized attention weight allocation method is used to calculate the correlation strength between each parameter and the first training bottleneck factor, quantifying the correlation between the parameters and the bottleneck factor into weight values. Subsequently, a weight ranking method is used to determine the adjustment priority according to the weight values ​​from high to low, obtaining the first bottleneck training correlation analysis result. Based on this, a parameter linear weighted adjustment method is used to correct the amplitude and rhythm of the posture target angle, training support surface width, and posture adjustment frequency with higher priority, forming the first balance training adjustment set.

[0038] Subsequently, when adjusting the attention-related localization of the real-time balance training program based on the second training bottleneck factor, a correspondence was established between the second training bottleneck factor and the neuromuscular activation intensity, training feedback delay, and movement response rhythm included in the real-time balance training program. An attention weight calculation matrix was constructed, and the correlation weight of each parameter was calculated using a normalized attention weight allocation method. A threshold screening method was then used to determine the priority adjustment terms. Afterwards, a gradient step size adjustment method was used to perform stepwise numerical adjustments and rhythm corrections on the neuromuscular activation intensity, training feedback delay, and movement response rhythm, resulting in the second balance training adjustment set.

[0039] Next, when adjusting the attention-related localization of the real-time balanced training scheme based on the third training bottleneck factor, a correspondence is established between the third training bottleneck factor and the interaction cue frequency, feedback presentation intensity, and task duration included in the real-time balanced training scheme. By constructing an attention weight calculation matrix, a normalized attention weight allocation method is used to calculate the correlation weight between each parameter and the third training bottleneck factor, and the adjustment focus is determined by weight ranking. Subsequently, a segmented parameter adjustment method is used to adjust the interaction cue frequency, feedback presentation intensity, and task duration by magnitude and time allocation, resulting in the third balanced training adjustment set.

[0040] Finally, when performing a global combination based on the first, second, and third balanced training and adjustment sets, the parameter vectors in all three sets are first normalized. Then, a linear combination method is used to multidimensionally superimpose various adjustment parameters according to a preset weight matrix, generating multiple training and adjustment schemes. Next, parameter clustering methods are used to group and structurally manage the training and adjustment schemes, ultimately forming the first training and adjustment space.

[0041] Step S400: Based on the balanced training evaluation model, perform balanced training deduction and optimization on the first training adjustment space to obtain the second training adjustment space.

[0042] In this embodiment, when performing balanced training deduction and optimization on the first training adjustment space according to the balanced training evaluation model, multiple training adjustment schemes in the first training adjustment space are first extracted sequentially, and virtual verification and performance evaluation are performed on each training adjustment scheme. Then, a twin model is established based on the target user to generate a user 3D model. Next, simulation training is performed on the user 3D model according to different training adjustment schemes to generate a corresponding training simulation dataset, which is then input into the balanced training evaluation model to obtain the corresponding training evaluation sequence. Then, the training evaluation sequence is compared and judged according to preset training evaluation constraints, including posture response evaluation constraints, neuromuscular feedback evaluation constraints, and training behavior evaluation constraints. Finally, training adjustment schemes that satisfy all evaluation constraints are screened and aggregated to form a second training adjustment space.

[0043] Furthermore, in the method provided in the application embodiments, the method of performing balanced training deduction and optimization on the first training adjustment space according to the balanced training evaluation model to obtain the second training adjustment space further includes: Based on the first training adjustment space, the Kth training adjustment scheme is extracted, where K is a positive integer; twin modeling is performed on the target user to obtain a user 3D model; the user 3D model is simulated and trained according to the Kth training adjustment scheme to obtain the Kth training simulation dataset; the Kth training simulation dataset is input into the balance training evaluation model to obtain the Kth training evaluation sequence; it is determined whether the Kth training evaluation sequence satisfies the training evaluation constraints, which include posture response evaluation constraints, neuromuscular feedback evaluation constraints, and training behavior evaluation constraints; if the Kth training evaluation sequence satisfies the training evaluation constraints, the Kth training adjustment scheme is added to the second training adjustment space.

[0044] In this embodiment, the Kth training adjustment scheme is first extracted sequentially in the training adjustment first space using a method of sequential iteration and index selection. The Kth training adjustment scheme corresponds to a specific combination of training parameters, including training variables such as the target posture angle, muscle activation intensity, and behavioral control frequency.

[0045] Next, twin modeling is performed based on the target user. In this process, a 3D user model is established using a combination of 3D pose reconstruction and muscle dynamics modeling, based on the target user's physiological characteristics and previous multimodal perception data. Specifically, this involves extracting key skeletal node coordinates and muscle activation features from inertial sensors, electromyography (EMG) signal acquisition devices, and video pose recognition systems. Then, a parameter fitting algorithm is used to ensure that the dynamic response of the virtual model matches the user's actual physiological characteristics, ultimately obtaining the user's 3D model.

[0046] Subsequently, when simulating the user's 3D model according to the Kth training adjustment scheme, the training parameters in the Kth training adjustment scheme are input into the user's 3D model for virtual training. During this process, a physics engine-based simulation method is used to perform time-series calculations on the dynamic characteristics of the user's 3D model during posture adjustment, center of gravity changes, muscle activation, and behavioral feedback. By sampling the posture change trajectory, muscle stress response, and behavioral reaction delay at each training stage, complete training simulation data records are obtained, forming the Kth training simulation dataset.

[0047] Then, the Kth training simulation dataset is input into the balance training evaluation model. The same processing method used to obtain the balance training evaluation sequence is applied. First, a corresponding training perception multidimensional vector is constructed based on the Kth training simulation dataset. Then, the corresponding multidimensional vector is decomposed and input into the posture response evaluation model, neuromuscular feedback evaluation model, and training behavior evaluation model in the balance training evaluation model to obtain the corresponding posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient. These coefficients are then integrated to obtain the Kth training evaluation sequence.

[0048] Next, it is determined whether the Kth training evaluation sequence meets the training evaluation constraints. These constraints include posture response evaluation constraints, neuromuscular feedback evaluation constraints, and training behavior evaluation constraints. These constraints are pre-set coefficient thresholds by technical experts, including posture response evaluation coefficient thresholds, neuromuscular feedback evaluation coefficient thresholds, and training behavior evaluation coefficient thresholds. When the posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient in the Kth training evaluation sequence are all greater than the pre-set coefficient thresholds, it is determined that the training evaluation constraints are met, and the Kth training adjustment scheme is added to the second training adjustment space.

[0049] Step S500: The second space of the training regulation is optimized by neighborhood transfer using the balanced training optimization analytical model to obtain the balanced training regulation optimization result.

[0050] In this embodiment, when performing neighborhood transfer optimization on the training and regulation second space using the balanced training optimality analytical model, the optimality of each training scheme in the training and regulation second space is first calculated using the balanced training optimality analytical model to obtain the balanced training optimality distribution. Then, based on a predetermined balanced training optimality threshold, the optimality distribution is used to identify the current optimality, selecting multiple current optimal regulation schemes with higher optimality. Next, based on the balanced training evaluation model and the balanced training optimality analytical model, neighborhood transfer optimization is performed on the selected current optimal regulation schemes. By calculating the parameter differences and optimality gradients between adjacent schemes, multiple regulation transfer optimization neighborhoods are generated. Finally, the balanced training optimality analytical model is used again to iteratively optimize the multiple current optimal regulation schemes and multiple regulation transfer optimization neighborhoods to obtain the balanced training regulation optimization result.

[0051] Furthermore, the method provided in the application embodiments also includes: The balance training evaluation model is obtained by weighting the training evaluation multivariate indicators according to the model. The training evaluation multivariate indicators include posture response evaluation indicators, neuromuscular feedback evaluation indicators, and training behavior evaluation indicators.

[0052] In this embodiment, the optimality analysis model for balanced training is a weighted function, which calculates optimality based on the posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient output by the balanced training evaluation model. Specifically, firstly, the posture response evaluation index, neuromuscular feedback evaluation index, and training behavior evaluation index in the multivariate training evaluation indicators are determined, and corresponding preset weights are assigned to each of the three indicators. These preset weights are set by technical experts based on statistical analysis results from a large number of training samples. Subsequently, the weight corresponding to the posture response evaluation index is multiplied by the posture response evaluation coefficient, the weight corresponding to the neuromuscular feedback evaluation index is multiplied by the neuromuscular feedback evaluation coefficient, and the weight corresponding to the training behavior evaluation index is multiplied by the training behavior evaluation coefficient. The resulting products are then weighted and summed to obtain a comprehensive optimality value for balanced training.

[0053] Furthermore, in the method provided in the application embodiments, the method further includes performing neighborhood transfer optimization on the second training regulation space using a balanced training optimization analytical model to obtain the balanced training regulation optimization result, and also includes: The training and adjustment second space is analyzed for balance training optimality and appropriateness according to the aforementioned balanced training optimality and appropriateness analysis model to obtain a balanced training optimality and appropriateness distribution. Based on the balanced training optimality and appropriateness distribution, the training and adjustment second space is identified for current optimization according to a predetermined balanced training optimality and appropriateness to obtain multiple current optimization adjustment schemes. Based on the balanced training evaluation model and the balanced training optimality and appropriateness analysis model, the training and adjustment second space is optimized for neighborhood migration according to the multiple current optimization adjustment schemes to obtain multiple adjustment migration optimization neighborhoods. The balanced training optimality and appropriateness is iteratively optimized for the multiple current optimization adjustment schemes and the multiple adjustment migration optimization neighborhoods according to the balanced training optimality and appropriateness analysis model to generate the balanced training and adjustment optimization result.

[0054] In this embodiment, when performing balance training optimality analysis on the second training adjustment space using the balance training optimality analysis model, the optimality of each training adjustment scheme in the second training adjustment space is calculated using the balance training optimality analysis model. The posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient are multiplied by their respective preset weights and then weighted and summed to obtain the balance training optimality value for each training scheme. By statistically organizing the optimality values ​​of all training schemes, a balance training optimality distribution is formed.

[0055] Next, based on the balanced training optimality distribution and combined with a predetermined balanced training optimality threshold, the current optimization identification is performed in the training adjustment second space. In this process, by comparing the balanced training optimality value of each training scheme with the predetermined balanced training optimality threshold, when the balanced training optimality of a certain training scheme is greater than or equal to the set predetermined balanced training optimality threshold, it is determined to be the current optimization adjustment scheme, thereby obtaining multiple current optimization adjustment schemes.

[0056] Subsequently, based on the balanced training evaluation model and the balanced training optimality analysis model, neighborhood transfer optimization is performed on multiple current optimal regulation schemes. In this process, firstly, a first current optimal regulation scheme is extracted from the multiple schemes, and the differences in training parameters in the second training regulation space are analyzed to form a training regulation difference distribution. Then, based on this distribution, the parameters in the second training regulation space are adjusted in a guided manner to generate a new training regulation variation domain. Next, the balanced training evaluation model is used to perform balanced training deduction optimization on the variation domain to obtain the corresponding regulation transfer neighborhood. Finally, based on a predetermined balanced training optimality, the balanced training optimality analysis model is used to perform balanced training optimality analysis optimization on the regulation transfer neighborhood, selecting regulation schemes that meet the optimality requirements and obtaining multiple regulation transfer optimization neighborhoods.

[0057] Finally, based on the balanced training optimality analytical model, iterative optimization of the balanced training optimality is performed on multiple current optimization schemes and their corresponding adjustment transfer optimization neighborhoods. In each iteration, the balanced training optimality value is recalculated for each adjustment transfer optimization neighborhood scheme, and the balanced training optimality distribution is updated based on the results. After multiple iterations and comparisons, the training scheme with the largest balanced training optimality value is selected from all schemes as the final balanced training adjustment optimization result.

[0058] Furthermore, in the method provided in the application embodiments, based on the balanced training evaluation model and the balanced training optimality analysis model, neighborhood transfer optimization is performed on the training adjustment second space according to the multiple current optimization adjustment schemes to obtain multiple adjustment transfer optimization neighborhoods, and the method further includes: Based on the multiple current optimization adjustment schemes, a first current optimization adjustment scheme is extracted; based on the first current optimization adjustment scheme, the training adjustment second space is analyzed for differential features to obtain a first training adjustment differential distribution; based on the first training adjustment differential distribution, the training adjustment second space is subjected to mutation-guided adjustment to obtain a first training adjustment variation domain; based on the balanced training evaluation model, the first training adjustment variation domain is optimized through balanced training deduction to obtain a first adjustment transfer neighborhood; based on the predetermined balanced training optimality, the first adjustment transfer neighborhood is optimized through balanced training optimality analysis based on the balanced training optimality analysis model to obtain a first adjustment transfer optimization neighborhood.

[0059] In this embodiment of the application, a first current optimization adjustment scheme is obtained by randomly extracting from multiple current optimization adjustment schemes.

[0060] Next, differential feature analysis is performed on the second training regulation space based on the first current optimization regulation scheme. In this process, the first current optimization regulation scheme is compared with other training regulation schemes in the second training regulation space, and the difference magnitude and direction of training variables such as posture target angle, muscle activation intensity, and behavioral control frequency are calculated to obtain the first training regulation difference distribution.

[0061] Subsequently, the training regulation second space is subjected to variation-guided regulation based on the first training regulation difference distribution. In this process, the training regulation scheme is used as the operational unit, and variation generation is performed on the training regulation scheme according to the direction and magnitude of parameter changes shown in the difference distribution, thereby obtaining the first training regulation variation domain. The first training regulation variation domain represents the set of improveable schemes generated through difference guidance in the multidimensional parameter space of training variables such as posture target angle, muscle activation intensity, and behavioral control frequency.

[0062] Subsequently, the first training regulation variation domain is optimized through balance training deduction based on the balance training evaluation model. This process is the same as the aforementioned balance training deduction optimization process: each training scheme in the first training regulation variation domain is sequentially input into the balance training evaluation model, its posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient are calculated, and the scheme that meets the training evaluation constraints is selected to form the first regulation transfer neighborhood.

[0063] Finally, based on a predetermined balance training optimality, the first regulation-transfer neighborhood is analyzed and optimized according to the balance training optimality analytical model. In this process, the balance training optimality value is calculated by multiplying the posture response evaluation coefficient, neuromuscular feedback evaluation coefficient, and training behavior evaluation coefficient of each training scheme in the first regulation-transfer neighborhood by their corresponding weights and then summing them. A set of schemes with balance training optimality values ​​greater than or equal to a predetermined balance training optimality threshold is then selected, resulting in the first regulation-transfer optimization neighborhood.

[0064] In summary, the embodiments of this application have at least the following technical effects: This application acquires multimodal perception data of a target user during balance function training, and constructs a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes a posture response feature vector, a neuromuscular feedback feature vector, and a training participation behavior feature vector. The training perception multidimensional vector is evaluated using a balance training evaluation model to obtain a balance training evaluation sequence. Based on the target user's historical balance training dataset and the balance training evaluation sequence, a balance function training bottleneck is traced to determine the balance training bottleneck factor. Attention association optimization is performed on the target user's real-time balance training scheme based on the balance training bottleneck factor to obtain a first training adjustment space. Balance training deduction and optimization are performed on the first training adjustment space using the balance training evaluation model to obtain a second training adjustment space. Finally, neighborhood transfer optimization is performed on the second training adjustment space using a balance training optimization analysis model to obtain the balance training adjustment optimization result. This invention addresses the technical problem in existing technologies that cannot effectively evaluate and personalize training optimization based on multi-dimensional user perception data during balance training. By combining multi-modal perception data modeling, bottleneck factor tracing, and attention association optimization with deduction and neighborhood transfer optimization, it achieves the technical effect of optimizing individualized and precise training programs, improving balance training effectiveness, and enhancing training adaptability.

[0065] Example 2, based on the same inventive concept as the balanced training data processing method in the foregoing examples, such as... Figure 2 As shown, this application provides a balanced training data processing system. The system and method embodiments in this application are based on the same inventive concept. The system includes: The data acquisition module 11 is used to acquire multimodal perception data of the target user during balance function training, and construct a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes a posture response feature vector, a neuromuscular feedback feature vector, and a training participation behavior feature vector. The tracing module 12 is used to evaluate the training perception multidimensional vector through a balance training evaluation model to obtain a balance training evaluation sequence, and trace the balance function training bottleneck based on the target user's balance training history dataset and the balance training evaluation sequence to determine the balance training bottleneck factor. The attention association optimization module 13 is used to optimize the attention association of the target user's real-time balance training scheme based on the balance training bottleneck factor to obtain a training adjustment first space. The deduction optimization module 14 is used to perform balance training deduction optimization on the training adjustment first space based on the balance training evaluation model to obtain a training adjustment second space. The transfer optimization module 15 is used to perform neighborhood transfer optimization on the training adjustment second space through a balance training optimization moderation model to obtain a balance training adjustment optimization result.

[0066] Furthermore, the system is also used to implement the following functions: The balance training evaluation model is activated, which includes a posture response evaluation model, a neuromuscular feedback evaluation model, and a training behavior evaluation model. The posture response feature vector is input into the posture response evaluation model to obtain posture response evaluation coefficients. The neuromuscular feedback feature vector is input into the neuromuscular feedback evaluation model to obtain neuromuscular feedback evaluation coefficients. The training participation behavior feature vector is input into the training behavior evaluation model to obtain training behavior evaluation coefficients. The balance training evaluation sequence is generated by combining the posture response evaluation coefficients and the neuromuscular feedback evaluation coefficients.

[0067] Furthermore, the system is also used to implement the following functions: Based on the historical dataset of balanced training, a balanced training evaluation trend is predicted to obtain training evaluation trend characteristics. Anomaly detection is performed on the balanced training evaluation sequence based on these trend characteristics to identify anomalies in posture response evaluation, neuromuscular feedback evaluation, and training behavior evaluation. Anomalies in posture response evaluation are traced back to obtain a first training bottleneck factor based on the posture response feature vector. Anomalies in neuromuscular feedback evaluation are traced back to obtain a second training bottleneck factor based on the neuromuscular feedback feature vector. Anomalies in training behavior evaluation are traced back to obtain a third training bottleneck factor based on the training participation behavior feature vector. The first and second training bottleneck factors are combined to generate the balanced training bottleneck factor.

[0068] Furthermore, the system is also used to implement the following functions: Based on the first training bottleneck factor, the real-time balanced training scheme is analyzed for attentional correlation of training parameters to obtain the first bottleneck training correlation analysis result; based on the first bottleneck training correlation analysis result, the real-time balanced training scheme is locally optimized and adjusted to obtain the first balanced training adjustment set; based on the second training bottleneck factor, the real-time balanced training scheme is locally adjusted for attentional correlation to obtain the second balanced training adjustment set; based on the third training bottleneck factor, the real-time balanced training scheme is locally adjusted for attentional correlation to obtain the third balanced training adjustment set; based on the first balanced training adjustment set, the second balanced training adjustment set, and the third balanced training adjustment set, the first training adjustment space is generated by global combination.

[0069] Furthermore, the system is also used to implement the following functions: Based on the first training adjustment space, the Kth training adjustment scheme is extracted, where K is a positive integer; twin modeling is performed on the target user to obtain a user 3D model; the user 3D model is simulated and trained according to the Kth training adjustment scheme to obtain the Kth training simulation dataset; the Kth training simulation dataset is input into the balance training evaluation model to obtain the Kth training evaluation sequence; it is determined whether the Kth training evaluation sequence satisfies the training evaluation constraints, which include posture response evaluation constraints, neuromuscular feedback evaluation constraints, and training behavior evaluation constraints; if the Kth training evaluation sequence satisfies the training evaluation constraints, the Kth training adjustment scheme is added to the second training adjustment space.

[0070] Furthermore, the system is also used to implement the following functions: The training and adjustment second space is analyzed for balance training optimality and appropriateness according to the aforementioned balanced training optimality and appropriateness analysis model to obtain a balanced training optimality and appropriateness distribution. Based on the balanced training optimality and appropriateness distribution, the training and adjustment second space is identified for current optimization according to a predetermined balanced training optimality and appropriateness to obtain multiple current optimization adjustment schemes. Based on the balanced training evaluation model and the balanced training optimality and appropriateness analysis model, the training and adjustment second space is optimized for neighborhood migration according to the multiple current optimization adjustment schemes to obtain multiple adjustment migration optimization neighborhoods. The balanced training optimality and appropriateness is iteratively optimized for the multiple current optimization adjustment schemes and the multiple adjustment migration optimization neighborhoods according to the balanced training optimality and appropriateness analysis model to generate the balanced training and adjustment optimization result.

[0071] Furthermore, the system is also used to implement the following functions: Based on the multiple current optimization adjustment schemes, a first current optimization adjustment scheme is extracted; based on the first current optimization adjustment scheme, the training adjustment second space is analyzed for differential features to obtain a first training adjustment differential distribution; based on the first training adjustment differential distribution, the training adjustment second space is subjected to mutation-guided adjustment to obtain a first training adjustment variation domain; based on the balanced training evaluation model, the first training adjustment variation domain is optimized through balanced training deduction to obtain a first adjustment transfer neighborhood; based on the predetermined balanced training optimality, the first adjustment transfer neighborhood is optimized through balanced training optimality analysis based on the balanced training optimality analysis model to obtain a first adjustment transfer optimization neighborhood.

[0072] Furthermore, the system is also used to implement the following functions: When the target user performs balance function training according to the real-time balance training scheme, the training monitoring dataset is acquired synchronously; the training monitoring dataset is cleaned in multiple dimensions to generate the multimodal perception data.

[0073] Furthermore, the system is also used to implement the following functions: The balance training evaluation model is obtained by weighting the training evaluation multivariate indicators according to the model. The training evaluation multivariate indicators include posture response evaluation indicators, neuromuscular feedback evaluation indicators, and training behavior evaluation indicators.

[0074] 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. 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.

[0075] 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.

[0076] 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 modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for processing balanced training data, characterized in that, The method includes: Acquire multimodal perception data of the target user during balance function training, and construct a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes posture response feature vector, neuromuscular feedback feature vector, and training participation behavior feature vector. The training perception multidimensional vector is evaluated by the balanced training evaluation model to obtain the balanced training evaluation sequence. The balanced function training bottleneck is traced based on the balanced training history dataset of the target user and the balanced training evaluation sequence to determine the balanced training bottleneck factor. Based on the aforementioned balance training bottleneck factor, the real-time balance training scheme for the target user is optimized by attention association to obtain the first space for training adjustment. Based on the balanced training evaluation model, the first training adjustment space is optimized through balanced training deduction to obtain the second training adjustment space. The balanced training regulation optimization result is obtained by performing neighborhood transfer optimization on the second space of the training regulation through a balanced training optimization analytical model.

2. The balanced training data processing method as described in claim 1, characterized in that, The training-aware multidimensional vector is evaluated using a balanced training evaluation model to obtain a balanced training evaluation sequence, including: Activate the balance training evaluation model, which includes a posture response evaluation model, a neuromuscular feedback evaluation model, and a training behavior evaluation model; The attitude response feature vector is input into the attitude response evaluation model to obtain the attitude response evaluation coefficients; The neuromuscular feedback feature vector is input into the neuromuscular feedback evaluation model to obtain the neuromuscular feedback evaluation coefficient. The training participation behavior feature vector is input into the training behavior evaluation model to obtain the training behavior evaluation coefficient. The posture response evaluation coefficient and the neuromuscular feedback evaluation coefficient are combined to generate the balance training evaluation sequence.

3. The balanced training data processing method as described in claim 1, characterized in that, Based on the target user's historical balance training dataset and the balance training evaluation sequence, the bottleneck of balance function training is traced to determine the balance training bottleneck factors, including: Based on the historical dataset of balanced training, the balanced training evaluation trend is predicted to obtain the training evaluation trend characteristics. Based on the training evaluation trend characteristics, anomaly detection is performed on the balanced training evaluation sequence to identify abnormalities in posture response evaluation, neuromuscular feedback evaluation, and training behavior evaluation. Based on the posture response feature vector, anomalies in the posture response evaluation are traced to obtain the first training bottleneck factor. Based on the neuromuscular feedback feature vector, anomalies in the neuromuscular feedback evaluation are traced to obtain the second training bottleneck factor. Based on the training participation behavior feature vector, anomalies in the training behavior evaluation are traced to obtain a third training bottleneck factor. The first training bottleneck factor and the second training bottleneck factor are combined to generate the balanced training bottleneck factor.

4. The balanced training data processing method as described in claim 1, characterized in that, Based on the aforementioned bottleneck factor in balance training, the real-time balance training scheme for the target user is optimized for attention association to obtain a first training adjustment space, including: Based on the first training bottleneck factor, the real-time balanced training scheme is analyzed for training parameter attention correlation to obtain the first bottleneck training correlation analysis result. Based on the first bottleneck training correlation analysis result, the real-time balanced training scheme is locally optimized and adjusted to obtain the first balanced training adjustment set. Based on the second training bottleneck factor, the real-time balance training scheme is locally adjusted for attention association to obtain the second balance training adjustment set. Based on the third training bottleneck factor, the real-time balance training scheme is locally adjusted for attention association to obtain the third balance training adjustment set. The training adjustment first space is generated by globally combining the first balanced training adjustment set, the second balanced training adjustment set, and the third balanced training adjustment set.

5. The balanced training data processing method as described in claim 1, characterized in that, Based on the balanced training evaluation model, the first training adjustment space is optimized through balanced training deduction to obtain the second training adjustment space, including: Based on the first training adjustment space, extract the Kth training adjustment scheme, where K is a positive integer; Based on the target user, a twin model is created to obtain a 3D model of the user; The user's 3D model is simulated and trained according to the Kth training adjustment scheme to obtain the Kth training simulation dataset; Input the Kth training simulation dataset into the balanced training evaluation model to obtain the Kth training evaluation sequence; Determine whether the Kth training evaluation sequence satisfies the training evaluation constraints, which include posture response evaluation constraints, neuromuscular feedback evaluation constraints, and training behavior evaluation constraints. If the Kth training evaluation sequence satisfies the training evaluation constraint, the Kth training adjustment scheme is added to the second training adjustment space.

6. The balanced training data processing method as described in claim 1, characterized in that, The balanced training regulation optimization results are obtained by performing neighborhood transfer optimization on the second training regulation space using a balanced training optimization analytical model, including: Based on the aforementioned balanced training optimality analytical model, the second space of training adjustment is analyzed for balanced training optimality to obtain the balanced training optimality distribution. Based on the balanced training optimality distribution, the training adjustment second space is currently optimized according to the predetermined balanced training optimality to obtain multiple current optimization adjustment schemes. Based on the balanced training evaluation model and the balanced training optimality analysis model, the training regulation second space is optimized by neighborhood migration according to the multiple current optimization regulation schemes to obtain multiple regulation migration optimization neighborhoods. Based on the balanced training optimality analytical model, the balanced training optimality is iteratively optimized for the multiple current optimization adjustment schemes and the multiple adjustment migration optimization neighborhoods to generate the balanced training adjustment optimization result.

7. The balanced training data processing method as described in claim 6, characterized in that, Based on the balanced training evaluation model and the balanced training optimality analysis model, neighborhood transfer optimization is performed on the second training regulation space according to the multiple current optimization adjustment schemes to obtain multiple regulation transfer optimization neighborhoods, including: Based on the multiple current optimization adjustment schemes, extract the first current optimization adjustment scheme; Based on the first current optimization adjustment scheme, the training adjustment second space is analyzed for differential features to obtain the first training adjustment differential distribution; Based on the first training regulation difference distribution, the training regulation second space is subjected to mutation-guided regulation to obtain the first training regulation variation domain; Based on the balanced training evaluation model, the first training regulation variation domain is optimized through balanced training deduction to obtain the first regulation transfer neighborhood. Based on the predetermined balance training optimality, the first regulation-transfer neighborhood is optimized by analytical search of the balance training optimality according to the balance training optimality analytical model, thereby obtaining the first regulation-transfer optimization neighborhood.

8. The balanced training data processing method as described in claim 1, characterized in that, Acquire multimodal awareness data of the target user during balance function training, including: When the target user performs balance function training according to the real-time balance training scheme, the training monitoring dataset is acquired synchronously. The training and monitoring dataset is cleaned in multiple dimensions to generate the multimodal sensing data.

9. The balanced training data processing method as described in claim 1, characterized in that, The balance training evaluation model is obtained by weighting the training evaluation multivariate indicators according to the model. The training evaluation multivariate indicators include posture response evaluation indicators, neuromuscular feedback evaluation indicators, and training behavior evaluation indicators.

10. A balanced training data processing system, characterized in that, The system is used to execute a balanced training data processing method as described in any one of claims 1-9, the system comprising: The data acquisition module is used to acquire multimodal perception data of the target user during balance function training, and to construct a training perception multidimensional vector based on the multimodal perception data. The training perception multidimensional vector includes posture response feature vector, neuromuscular feedback feature vector, and training participation behavior feature vector. The tracing module is used to evaluate the training perception multidimensional vector through the balanced training evaluation model, obtain the balanced training evaluation sequence, and trace the balanced function training bottleneck based on the balanced training history dataset of the target user and the balanced training evaluation sequence to determine the balanced training bottleneck factor. The attention association optimization module is used to optimize the attention association of the real-time balance training scheme of the target user based on the balance training bottleneck factor, so as to obtain the first space for training adjustment. The deduction and optimization module is used to perform balanced training deduction and optimization on the first training adjustment space according to the balanced training evaluation model to obtain the second training adjustment space. The migration optimization module is used to perform neighborhood migration optimization on the second training regulation space through the balanced training optimization analytical model to obtain the balanced training regulation optimization result.

Images