A human body training management system based on a brain-computer interface

Abstract

The present application relates to the technical field of human training management, and particularly relates to a human training management system based on a brain-computer interface, which comprises a brain-computer interface signal front end, a neural signal acquisition module, a neural feature extraction module, an individualized baseline construction module, a central state evaluation module, an adaptive load mapping module and an execution feedback module. The brain-computer interface signal front end acquires multi-channel original electroencephalogram signals in real time through a distributed electrode array. The modules are sequentially cascaded, forming a closed-loop data processing flow of signal acquisition, feature extraction, state evaluation, adaptive mapping and execution feedback. The individualized baseline construction module provides a personalized reference benchmark for the central state evaluation module. The nonlinear mapping model built in the adaptive load mapping module is continuously updated in each training cycle, so that the system can track the nonlinear evolution law of the central nervous state of the trainee with the training process, and realize accurate determination and differentiated output of the training intensity adjustment opportunity.

Description

Technical Field

[0001] This invention relates to the field of human training management technology, and in particular to a human training management system based on a brain-computer interface. Background Technology

[0002] Human training management systems typically refer to comprehensive platforms that combine sensor technology, data analysis algorithms, and feedback mechanisms to monitor and regulate the physiological state, movement posture, and training load of trainees. Traditional systems often rely on peripheral physiological signals such as heart rate, electromyography, and acceleration, combined with video motion capture, to assess training effectiveness and prevent sports injuries, and are widely used in sports competitions, rehabilitation medicine, and physical training. However, such systems often focus on quantifying external behavioral performance and struggle to directly capture the driving and regulatory information of the central nervous system during training, resulting in a lag and indirectness in the assessment of training fatigue, neural adaptation, and movement intention.

[0003] With the evolution of brain-computer interface (BCI) technology, its application in human training management has become a research hotspot. BCIs can acquire electroencephalogram (EEG) or functional near-infrared spectroscopy (fNIRS) signals from the cerebral cortex in real time during motor imagery, action execution, and feedback cognition processes, providing a quantitative window directly reaching the central nervous system for training management. Current technologies attempt to determine the trainee's attention span, the starting point of motor intention, and the degree of autonomic neural involvement by decoding motor cortical potentials (MRCP) or sensorimotor rhythms (SMR), thereby achieving dynamic adjustment of training intensity or real-time correction of movement patterns. This closed-loop combination of neural information and motor behavior can theoretically significantly improve the accuracy and efficiency of training.

[0004] However, existing brain-computer interface-based human training management systems suffer from a lack of dynamic adaptive mapping mechanism between EEG characteristics and training load status during actual use. This makes it difficult for the system to distinguish between central nervous system adaptive optimization and physiological fatigue, thus making it unable to accurately determine when to adjust training intensity. Summary of the Invention

[0005] The purpose of this invention is to provide a human training management system based on brain-computer interface, which solves the technical problem that existing human training management systems based on brain-computer interface lack a dynamic adaptive mapping mechanism between EEG characteristics and training load status during actual use, making it difficult for the system to distinguish between central nervous system adaptive optimization and physiological fatigue, and thus unable to accurately determine the timing of training intensity adjustment.

[0006] To achieve the above objectives, the present invention provides a human training management system based on a brain-computer interface. The human training management system based on a brain-computer interface includes a brain-computer interface signal front end, a neural signal acquisition module, a neural feature extraction module, an individualized baseline construction module, a central state assessment module, an adaptive load mapping module, and an execution feedback module. The brain-computer interface signal front end is worn on the head of the trainee and is used to acquire multi-channel raw EEG signals in real time through a distributed electrode array. The input end of the neural signal acquisition module is connected to the output end of the brain-computer interface signal front end, and is used to preprocess the raw EEG signal to generate a conditioned EEG digital signal. The input end of the neural feature extraction module is connected to the output end of the neural signal acquisition module, and is used to extract multidimensional neural feature parameters of the brain-modulated electro-digital signal and output them to the central state assessment module. The individualized baseline construction module is used to establish and store the individualized neural response baseline library of trainees and output it to the central state assessment module; The first input terminal of the central state assessment module is connected to the output terminal of the neural feature extraction module, and the second input terminal of the central state assessment module is connected to the output terminal of the individualized baseline construction module. The central state assessment module is used to dynamically compare the multidimensional neural feature parameters with the baseline data in the individualized neural response baseline library, generate a central state vector, and output it to the adaptive load mapping module. The input of the adaptive load mapping module is connected to the output of the central state evaluation module. It is used to generate a training intensity adjustment instruction based on the central state vector, the preset training target and historical training load data, through a nonlinear mapping model, and output it to the execution feedback module. The nonlinear mapping model is a dynamic update model that continuously adjusts the model parameters based on the correlation data between the central state vector and the training effect during the historical training process. The input of the execution feedback module is connected to the output of the adaptive load mapping module, and is used to convert the training intensity adjustment command into a physical stimulus signal or a task difficulty adjustment signal, and output it to the training execution device.

[0007] The neural signal acquisition module performs preprocessing on the raw EEG signals, including signal amplification, filtering and noise reduction, and artifact removal.

[0008] The neural feature extraction module extracts multidimensional neural feature parameters including time-domain feature parameters, frequency-domain feature parameters, and spatial-domain feature parameters. The time-domain feature parameters include the amplitude change slope and latency of motion-related cortical potentials. The frequency-domain feature parameters include the power spectral density and ratio of the theta wave, alpha wave, beta wave, and gamma wave frequency bands. The spatial-domain feature parameters include the coherence and synchronicity characteristics at each lead position.

[0009] The individualized baseline construction module includes a resting-state acquisition unit, a standard-state acquisition unit, a baseline database construction unit, and a baseline update unit; the resting-state acquisition unit is used to record the baseline EEG signals of the trainee in the closed-eye relaxation state and the open-eye sitting state. The standard state acquisition unit is used to record the induced electroencephalogram (EEG) signals of trainees when performing preset standard action paradigms. The baseline library building unit is used to generate an individualized neural response baseline library containing a mean vector and a covariance matrix based on the baseline EEG signal and the evoked EEG signal. The baseline update unit is used to dynamically update the individualized neural response baseline library according to the trainee's training progress at a preset cycle.

[0010] The central state vector output by the central state assessment module includes a neural participation component and a central fatigue index component. The neural participation component reflects the current activation level of the trainee's motor cortex, and the central fatigue index component reflects the functional response decay trend of the trainee's central nervous system under continuous load.

[0011] The central state assessment module includes a deviation calculation unit, a trend prediction unit, and a state quantification unit. The deviation calculation unit is used to calculate the deviation vector of multidimensional neural feature parameters relative to the individualized neural response baseline library. The trend prediction unit is used to predict the changing trend of the central state within a preset time window based on time series analysis. The state quantization unit is used to weight and fuse the deviation vector and the change trend to generate a central state vector that includes a neural participation component and a central fatigue index component.

[0012] The adaptive load mapping module includes a target parsing unit, a model matching unit, an instruction generation unit, and a model online update unit. The target parsing unit is used to parse the target type of the current training task. The target types include strength growth, endurance improvement, and skill proficiency. The model matching unit is used to select a matching nonlinear mapping model based on the target type; The instruction generation unit is used to input the central state vector into the nonlinear mapping model and output training intensity adjustment instructions including intensity level, duration and interval ratio. The online model update unit is used to update the parameters of the nonlinear mapping model after each training session, based on the central state vector sequence, actual execution intensity, and training effect score of this training session, using an online incremental learning algorithm.

[0013] The execution feedback module includes a physical stimulation interface unit, a task adjustment interface unit, and a feedback presentation unit. The physical stimulation interface unit is used to convert training intensity adjustment instructions into electrical or vibration stimulation signals, which are applied to the peripheral nerves or muscles of the trainee. The task adjustment interface unit is used to convert training intensity adjustment commands into parameters such as the running speed, resistance value, or difficulty level of the training equipment. The feedback presentation unit is used to present the current central state and adjusted training task parameters to the trainee in real time through visual or auditory means.

[0014] The brain-computer interface-based human training management system further includes a training effect evaluation module. The input of the training effect evaluation module is connected to the output of the neural feature extraction module and the output of the execution feedback module. The training effect evaluation module is used to generate a training benefit score by performing a comprehensive weighted calculation based on the changes in neural features before and after training and the actual training intensity parameters after each training session.

[0015] The brain-computer interface-based human training management system further includes a data recording module and an offline optimization module. The input end of the data recording module is connected to the output end of the central state assessment module, the output end of the adaptive load mapping module, and the output end of the training effect assessment module. The data recording module is used to associate and store the central state vector sequence, training intensity adjustment instructions, and training benefit scores in each training process as a structured training log. The input of the offline optimization module is connected to the output of the data recording module, and the output of the offline optimization module is connected to the input of the adaptive load mapping module. The offline optimization module is used to retrain the nonlinear mapping model offline using a batch learning method based on multiple accumulated structured training logs, and pushes the optimized model parameters to the adaptive load mapping module to update the nonlinear mapping model.

[0016] This invention discloses a brain-computer interface-based human training management system, comprising a brain-computer interface signal front-end, a neural signal acquisition module, a neural feature extraction module, an individualized baseline construction module, a central nervous system state assessment module, an adaptive load mapping module, and an execution feedback module. The brain-computer interface signal front-end serves as the neural signal entry point of the system, covering the trainee's motor cortex, prefrontal cortex, and parietal cortex regions through a distributed electrode array. The modules are cascaded sequentially to form a closed-loop data processing flow of signal acquisition, feature extraction, state assessment, adaptive mapping, and execution feedback. The individualized baseline construction module provides a personalized reference benchmark for the central nervous system state assessment module. The nonlinear mapping model built into the adaptive load mapping module is continuously updated in each training cycle, enabling the system to track the nonlinear evolution of the trainee's central nervous system state as the training progresses, achieving precise determination and differentiated output of training intensity adjustment timing. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0018] Figure 1 This is a schematic diagram of the human training management system based on brain-computer interface provided by the present invention.

[0019] 101-Brain-Computer Interface Signal Front End, 102-Neural Signal Acquisition Module, 103-Neural Feature Extraction Module, 104-Individualized Baseline Construction Module, 105-Central State Assessment Module, 106-Adaptive Load Mapping Module, 107-Execution Feedback Module, 108-Rest-State Acquisition Unit, 109-Standard State Acquisition Unit, 110-Baseline Library Construction Unit, 111-Baseline Update Unit, 112-Deviance Calculation Unit, 113-Trend Prediction Unit, 114-State Quantization Unit, 115-Target Resolution Unit, 116-Model Matching Unit, 117-Instruction Generation Unit, 118-Online Model Update Unit, 119-Physical Stimulation Interface Unit, 120-Task Adjustment Interface Unit, 121-Feedback Presentation Unit, 122-Training Effect Evaluation Module, 123-Data Recording Module, 124-Offline Optimization Module. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0021] Please see Figure 1 The present invention provides a human training management system based on brain-computer interface. The human training management system based on brain-computer interface includes a brain-computer interface signal front end 101, a neural signal acquisition module 102, a neural feature extraction module 103, an individualized baseline construction module 104, a central state assessment module 105, an adaptive load mapping module 106, and an execution feedback module 107. The brain-computer interface signal front end 101 is worn on the head of the trainee and is used to acquire multi-channel raw EEG signals in real time through a distributed electrode array. The input end of the neural signal acquisition module 102 is connected to the output end of the brain-computer interface signal front end 101, and is used to preprocess the raw EEG signal to generate a conditioned EEG digital signal. The input end of the neural feature extraction module 103 is connected to the output end of the neural signal acquisition module 102, and is used to extract multidimensional neural feature parameters of the brain-modulated digital signal and output them to the central state assessment module 105. The individualized baseline construction module 104 is used to establish and store the individualized neural response baseline library of the trainee and output it to the central state assessment module 105. The first input terminal of the central state assessment module 105 is connected to the output terminal of the neural feature extraction module 103, and the second input terminal of the central state assessment module 105 is connected to the output terminal of the individualized baseline construction module 104. The central state assessment module 105 is used to dynamically compare the multidimensional neural feature parameters with the baseline data in the individualized neural response baseline library, generate a central state vector, and output it to the adaptive load mapping module 106. The input of the adaptive load mapping module 106 is connected to the output of the central state evaluation module 105. It is used to generate a training intensity adjustment instruction by combining the central state vector with the preset training target and historical training load data through a nonlinear mapping model, and output it to the execution feedback module 107. The nonlinear mapping model is a dynamic update model that continuously adjusts the model parameters based on the correlation data between the central state vector and the training effect during the historical training process. The input terminal of the execution feedback module 107 is connected to the output terminal of the adaptive load mapping module 106, and is used to convert the training intensity adjustment command into a physical stimulus signal or a task difficulty adjustment signal, and output it to the training execution device.

[0022] In this embodiment, the brain-computer interface signal front-end 101 serves as the neural signal entry point of the system. It covers the motor cortex, prefrontal cortex, and parietal cortex of the trainee through a distributed electrode array. The modules are cascaded sequentially to form a closed-loop data processing flow of signal acquisition, feature extraction, state assessment, adaptive mapping, and execution feedback. The individualized baseline construction module 104 provides a personalized reference benchmark for the central nervous system state assessment module 105. The nonlinear mapping model built into the adaptive load mapping module 106 is continuously updated in each training cycle, enabling the system to track the nonlinear evolution of the trainee's central nervous system state as the training progresses, and to achieve accurate determination and differentiated output of the timing for adjusting training intensity.

[0023] Furthermore, the neural signal acquisition module 102 performs preprocessing on the raw EEG signal, including signal amplification, filtering and noise reduction, and artifact removal.

[0024] In this embodiment, the neural signal acquisition module 102 pre-amplifies the microvolt-level raw EEG signal output by the brain-computer interface signal front-end 101 to a suitable amplitude, filters out low-frequency drift and high-frequency interference through a bandpass filter, and uses an independent component analysis algorithm to remove physiological artifacts such as electrooculography and electromyography, ensuring that the EEG digital signal on which subsequent feature extraction is based has a high signal-to-noise ratio and good signal fidelity.

[0025] Furthermore, the multidimensional neural feature parameters extracted by the neural feature extraction module 103 include time-domain feature parameters, frequency-domain feature parameters, and spatial-domain feature parameters. The time-domain feature parameters include the amplitude change slope and latency of the motion-related cortical potentials. The frequency-domain feature parameters include the power spectral density and ratio of the theta wave, alpha wave, beta wave, and gamma wave frequency bands. The spatial-domain feature parameters include the coherence and synchronicity characteristics at each lead position.

[0026] In this embodiment, the neural feature extraction module 103 performs multi-domain parallel analysis on the conditioned EEG digital signal using a sliding time window as the unit. In the time domain, it extracts the amplitude change slope and latency of motor-related cortical potentials by superimposing and averaging to reflect the temporal characteristics of cortical activation during motor preparation and execution. In the frequency domain, it calculates the power spectral density and their ratios in the theta, alpha, beta, and gamma wave bands using fast Fourier transform to characterize the trainee's current alertness, concentration, and motor cortical oscillation rhythm. In the spatial domain, it calculates the amplitude squared coherence coefficient and phase lock value between each lead to characterize the functional connectivity and co-activation patterns of different cortical regions during training. The above three types of features are normalized and fused into a multi-dimensional neural feature parameter vector, providing comprehensive neural information input for central nervous system state assessment.

[0027] Furthermore, the individualized baseline construction module 104 includes a resting-state acquisition unit 108, a standard-state acquisition unit 109, a baseline database construction unit 110, and a baseline update unit 111; the resting-state acquisition unit 108 is used to record the baseline EEG signals of the trainee in the closed-eye relaxation state and the open-eye sitting state. The standard state acquisition unit 109 is used to record the induced EEG signals of the trainee when performing a preset standard action paradigm; The baseline library building unit 110 is used to generate an individualized neural response baseline library containing a mean vector and a covariance matrix based on the baseline EEG signal and the evoked EEG signal. The baseline update unit 111 is used to dynamically update the individualized neural response baseline library according to the trainee's training progress at a preset period.

[0028] In this embodiment, the individualized baseline construction module 104 initiates the initialization process when the trainee uses the system for the first time: the resting-state acquisition unit 108 first acquires the background EEG activity of the trainee in two states: closed-eye relaxation and open-eye sitting, to establish an individualized basic neural oscillation level; the standard state acquisition unit 109 then acquires the evoked EEG activity of the trainee when performing standard limb movements at a preset speed and amplitude, to establish an individualized task-state neural response pattern; the baseline library construction unit 110 performs statistical modeling on the neural features extracted in the above two states, calculates the mean vector and covariance matrix of each feature dimension, and forms a multi-condition individualized neural response baseline library covering the resting state and the standard state; the baseline update unit 111, in the subsequent training process, according to the preset training day interval or training stage transition node, weights and integrates the newly added stable state neural feature data into the baseline library, so that the baseline library can slowly drift and update with the long-term changes in the trainee's neural adaptability, avoiding evaluation bias in long-term training of the fixed baseline.

[0029] Furthermore, the central state vector output by the central state assessment module 105 includes a neural participation component and a central fatigue index component. The neural participation component is used to reflect the current activation level of the trainee's motor cortex, and the central fatigue index component is used to reflect the functional response decay trend of the trainee's current central nervous system under continuous load.

[0030] In this embodiment, the central state vector uses a two-dimensional quantification to comprehensively describe the trainee's current neural functional state: the neural participation component quantifies the deviation between real-time multidimensional neural features and the standard baseline in the individualized baseline library, and combines the current time-domain motor cortical potential amplitude and frequency-domain β-wave power ratio to comprehensively assess the trainee's motor cortex activation intensity and attention focus at the current moment; the central fatigue index component quantifies the degree of functional response attenuation of the central nervous system under continuous training load by analyzing the power spectrum shift direction of real-time neural features relative to the resting baseline, the upward trend of α-wave power, and the dynamic changes in the θ / β ratio. The two components are output in independent numerical form, providing differentiated decision-making basis for the adaptive load mapping module 106.

[0031] Furthermore, the central state assessment module 105 includes a deviation calculation unit 112, a trend prediction unit 113, and a state quantification unit 114. The deviation calculation unit 112 is used to calculate the deviation vector of the multidimensional neural feature parameters relative to the individualized neural response baseline library. The trend prediction unit 113 is used to predict the changing trend of the central state within a future preset time window based on time series analysis. The state quantization unit 114 is used to weight and fuse the deviation vector and the change trend to generate a central state vector that includes a neural participation component and a central fatigue index component.

[0032] In this embodiment, the deviation calculation unit 112 uses a combination of Mahalanobis distance and Euclidean distance to calculate the multidimensional deviation vector between the multidimensional neural feature parameter vector within the current sliding window and the mean vector of the corresponding state (resting state or standard state) in the individualized baseline library. This deviation vector reflects both the deviation magnitude of each feature dimension and the correlation structure between the dimensions. The trend prediction unit 113 performs autoregressive moving average modeling on the deviation vector sequence of multiple consecutive time windows to predict the changing trend of the central state vector within the next 5 to 10 seconds time window, thus identifying early signs of fatigue accumulation or activation level decay. The state quantization unit 114 adaptively weights and fuses the instantaneous state corresponding to the current deviation vector with the predicted trend output by the trend prediction unit 113. When the trend prediction shows rapid decay, the weight of the trend term is increased to improve response sensitivity; when the trend prediction shows stability, the weight of the deviation term is increased to maintain stability. Finally, a two-dimensional central state vector is output to the adaptive load mapping module 106.

[0033] Furthermore, the adaptive load mapping module 106 includes a target parsing unit 115, a model matching unit 116, an instruction generation unit 117, and a model online update unit 118. The target parsing unit 115 is used to parse the target type of the current training task. The target types include strength growth, endurance improvement, and skill proficiency. The model matching unit 116 is used to select a matching nonlinear mapping model according to the target type; The instruction generation unit 117 is used to input the central state vector into the nonlinear mapping model and output training intensity adjustment instructions including intensity level, duration and interval ratio. The online model update unit 118 is used to update the parameters of the nonlinear mapping model using an online incremental learning algorithm after each training session, based on the central state vector sequence, actual execution intensity, and training effect score of the current training session.

[0034] In this embodiment, the target parsing unit 115 categorizes the current training task into three types—strength growth, endurance improvement, and skill proficiency—by reading the system's preset training plan or the user's manually set training target. Different target types correspond to different training intensity control strategy preferences. The model matching unit 116 selects a pre-trained and matched nonlinear mapping model from a preset model library based on the parsed target type: a radial basis function neural network model emphasizing high-intensity intermittent response for strength growth, a Gaussian process regression model emphasizing continuous load tracking for endurance improvement, and a support vector regression model emphasizing precision control for skill proficiency. The instruction generation unit 117 then... The two components of the state vector (neural engagement and central fatigue index) are used as input features. After forward computation by the selected nonlinear mapping model, the output includes training intensity adjustment instructions containing intensity levels (such as low / medium / high / extreme), duration, and interval ratio. After each training session, the online model update unit 118 uses the central state vector sequence at each time point during the training process, the actual training intensity parameters, and the training benefit score fed back by the training effect evaluation module 122 as new samples. It uses an online incremental learning algorithm of recursive least squares or stochastic gradient descent to incrementally update the weight parameters of the nonlinear mapping model, so that the model can absorb the latest individual response experience after each training cycle.

[0035] Furthermore, the execution feedback module 107 includes a physical stimulation interface unit 119, a task adjustment interface unit 120, and a feedback presentation unit 121. The physical stimulation interface unit 119 is used to convert training intensity adjustment instructions into electrical stimulation or vibration stimulation signals, which are applied to the peripheral nerves or muscles of the trainee. The task adjustment interface unit 120 is used to convert training intensity adjustment commands into the running speed, resistance value, or difficulty level parameters of the training equipment; The feedback presentation unit 121 is used to present the current central state and the adjusted training task parameters to the trainee in real time through visual or auditory means.

[0036] In this embodiment, the physical stimulation interface unit 119, for rehabilitation training or neuromuscular electrical stimulation training scenarios, maps the intensity level parameter in the training intensity adjustment command to the amplitude, frequency, and pulse width parameters of the electrical stimulation pulse, or to the amplitude and frequency parameters of the vibration stimulation. This is achieved by applying stimulation electrodes or vibration motors worn on the trainee's limbs to the peripheral nerves or target muscle groups, enabling a rapid transition from neural decision-making to physical intervention. The task adjustment interface unit 120, for physical training or motor skill training scenarios, converts the intensity level, duration, and interval ratio parameters in the training intensity adjustment command into the operating speed, resistance value, or difficulty level parameters of the training equipment (such as a treadmill, stationary bike, impedance trainer, or human-computer interaction game terminal) via a communication protocol, thus achieving physical adjustment of the training load. The feedback presentation unit 121, through a display screen or bone conduction headphones, presents the current central nervous system state assessment results and the parameters of the training task to be executed or adjusted in real time to the trainee in the form of a digital dashboard, trend curve, or voice prompts, enabling the trainee to perceive the matching relationship between their own neural state and the training load.

[0037] Furthermore, the brain-computer interface-based human training management system also includes a training effect evaluation module 122. The input end of the training effect evaluation module 122 is connected to the output end of the neural feature extraction module 103 and the output end of the execution feedback module 107. The training effect evaluation module 122 is used to perform a comprehensive weighted calculation based on the changes in neural features before and after training and the actual training intensity parameters after each training session to generate a training benefit score.

[0038] In this embodiment, the training effect evaluation module 122 triggers an evaluation process after each training task: First, it calculates the changes in multidimensional neural characteristic parameters relative to the resting baseline before training, within a preset time period after training, including neural plasticity indicators such as the increase in amplitude of motor-related cortical potentials, the degree of alpha wave desynchronization, and the β wave power recovery rate; simultaneously, it acquires the training intensity parameters actually output by the execution feedback module 107 during this training process (including the actual average intensity level achieved, the effective training duration, and the actual interval ratio); it normalizes and weights the above-mentioned neural characteristic changes and training intensity parameters according to preset weights, and generates a comprehensive training benefit score presented in percentage form. This score reflects both the adaptive changes of the trainee at the neural level and the actual completion of the training load execution.

[0039] Furthermore, the brain-computer interface-based human training management system also includes a data recording module 123 and an offline optimization module 124. The input end of the data recording module 123 is connected to the output end of the central state assessment module 105, the output end of the adaptive load mapping module 106, and the output end of the training effect assessment module 122. The data recording module 123 is used to associate and store the central state vector sequence, training intensity adjustment instructions, and training benefit scores in each training process as a structured training log. The input of the offline optimization module 124 is connected to the output of the data recording module 123, and the output of the offline optimization module 124 is connected to the input of the adaptive load mapping module 106. The offline optimization module 124 is used to retrain the nonlinear mapping model offline using a batch learning method based on accumulated multiple structured training logs, and push the optimized model parameters to the adaptive load mapping module 106 to update the nonlinear mapping model.

[0040] In this embodiment, the data recording module 123 continuously receives and caches the central state vector sequence (including timestamps) output by the central state evaluation module 105 and the training intensity adjustment instructions output by the adaptive load mapping module 106 at each time point during each training process. After training, it receives the comprehensive training benefit score output by the training effect evaluation module 122. The above data is stored in a structured format (such as JSON or time-series database entries) as a structured training log for a single training session, accumulating to form a longitudinal training dataset for each trainee. The offline optimization module 124, in the accumulated structured... When the training logs reach a preset threshold (e.g., 50 or 100 logs), a batch retraining process is automatically triggered. Using batch gradient descent or Bayesian optimization methods, the central state vector sequence in the historical training logs is used as the input feature, and the comprehensive training benefit score is used as the supervision label to perform offline global optimization on all parameters of the nonlinear mapping model. The optimized model parameters are then pushed to the adaptive load mapping module 106 through the data interface to replace or merge and update the currently used model parameters. Thus, through a two-layer model update mechanism, the nonlinear mapping model continuously approaches the optimal mapping relationship during long-term individual training.

[0041] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A human training management system based on a brain-computer interface, characterized in that, It includes a brain-computer interface signal front end, a neural signal acquisition module, a neural feature extraction module, an individualized baseline construction module, a central state assessment module, an adaptive load mapping module, and an execution feedback module. The brain-computer interface signal front end is worn on the trainee's head and is used to acquire multi-channel raw EEG signals in real time through a distributed electrode array. The input end of the neural signal acquisition module is connected to the output end of the brain-computer interface signal front end, and is used to preprocess the raw EEG signal to generate a conditioned EEG digital signal. The input end of the neural feature extraction module is connected to the output end of the neural signal acquisition module, and is used to extract multidimensional neural feature parameters of the brain-modulated electro-digital signal and output them to the central state assessment module. The individualized baseline construction module is used to establish and store the individualized neural response baseline library of trainees and output it to the central state assessment module; The first input terminal of the central state assessment module is connected to the output terminal of the neural feature extraction module, and the second input terminal of the central state assessment module is connected to the output terminal of the individualized baseline construction module. The central state assessment module is used to dynamically compare the multidimensional neural feature parameters with the baseline data in the individualized neural response baseline library, generate a central state vector, and output it to the adaptive load mapping module. The input of the adaptive load mapping module is connected to the output of the central state evaluation module. It is used to generate a training intensity adjustment instruction based on the central state vector, the preset training target and historical training load data, through a nonlinear mapping model, and output it to the execution feedback module. The nonlinear mapping model is a dynamic update model that continuously adjusts the model parameters based on the correlation data between the central state vector and the training effect during the historical training process. The input of the execution feedback module is connected to the output of the adaptive load mapping module, and is used to convert the training intensity adjustment command into a physical stimulus signal or a task difficulty adjustment signal, and output it to the training execution device.

2. The human training management system based on brain-computer interface as described in claim 1, characterized in that, The neural signal acquisition module performs preprocessing on the raw EEG signals, including signal amplification, filtering and noise reduction, and artifact removal.

3. The human training management system based on brain-computer interface as described in claim 2, characterized in that, The neural feature extraction module extracts multidimensional neural feature parameters including time-domain feature parameters, frequency-domain feature parameters, and spatial-domain feature parameters. The time-domain feature parameters include the amplitude change slope and latency of motion-related cortical potentials. The frequency-domain feature parameters include the power spectral density and ratio of the theta wave, alpha wave, beta wave, and gamma wave frequency bands. The spatial-domain feature parameters include the coherence and synchronicity characteristics at each lead position.

4. The human training management system based on brain-computer interface as described in claim 3, characterized in that, The individualized baseline construction module includes a resting-state acquisition unit, a standard-state acquisition unit, a baseline database construction unit, and a baseline update unit; the resting-state acquisition unit is used to record the baseline EEG signals of the trainee in the closed-eye relaxation state and the open-eye sitting state. The standard state acquisition unit is used to record the induced electroencephalogram (EEG) signals of trainees when performing preset standard action paradigms. The baseline library building unit is used to generate an individualized neural response baseline library containing a mean vector and a covariance matrix based on the baseline EEG signal and the evoked EEG signal. The baseline update unit is used to dynamically update the individualized neural response baseline library according to the trainee's training progress at a preset cycle.

5. The human training management system based on a brain-computer interface as described in claim 4, characterized in that, The central state vector output by the central state assessment module includes a neural participation component and a central fatigue index component. The neural participation component is used to reflect the current activation level of the trainee's motor cortex, and the central fatigue index component is used to reflect the functional response decay trend of the trainee's central nervous system under continuous load.

6. The human training management system based on brain-computer interface as described in claim 5, characterized in that, The central state assessment module includes a deviation calculation unit, a trend prediction unit, and a state quantification unit. The deviation calculation unit is used to calculate the deviation vector of multidimensional neural feature parameters relative to the individualized neural response baseline library. The trend prediction unit is used to predict the changing trend of the central state within a preset time window based on time series analysis. The state quantization unit is used to weight and fuse the deviation vector and the change trend to generate a central state vector that includes a neural participation component and a central fatigue index component.

7. The human training management system based on a brain-computer interface as described in claim 6, characterized in that, The adaptive load mapping module includes a target parsing unit, a model matching unit, an instruction generation unit, and a model online update unit. The target parsing unit is used to parse the target type of the current training task. The target types include strength growth, endurance improvement, and skill proficiency. The model matching unit is used to select a matching nonlinear mapping model based on the target type; The instruction generation unit is used to input the central state vector into the nonlinear mapping model and output training intensity adjustment instructions including intensity level, duration and interval ratio. The online model update unit is used to update the parameters of the nonlinear mapping model after each training session, based on the central state vector sequence, actual execution intensity, and training effect score of this training session, using an online incremental learning algorithm.

8. The human training management system based on brain-computer interface as described in claim 7, characterized in that, The execution feedback module includes a physical stimulation interface unit, a task adjustment interface unit, and a feedback presentation unit. The physical stimulation interface unit is used to convert training intensity adjustment instructions into electrical or vibration stimulation signals, which are applied to the peripheral nerves or muscles of the trainee. The task adjustment interface unit is used to convert training intensity adjustment commands into parameters such as the running speed, resistance value, or difficulty level of the training equipment. The feedback presentation unit is used to present the current central state and adjusted training task parameters to the trainee in real time through visual or auditory means.

9. The human training management system based on a brain-computer interface as described in claim 8, characterized in that, The brain-computer interface-based human training management system also includes a training effect evaluation module. The input of the training effect evaluation module is connected to the output of the neural feature extraction module and the output of the execution feedback module. The training effect evaluation module is used to generate a training benefit score by performing a comprehensive weighted calculation based on the changes in neural features before and after training and the actual training intensity parameters after each training session.

10. The human training management system based on a brain-computer interface as described in claim 9, characterized in that, The brain-computer interface-based human training management system further includes a data recording module and an offline optimization module. The input end of the data recording module is connected to the output end of the central state assessment module, the output end of the adaptive load mapping module, and the output end of the training effect assessment module. The data recording module is used to associate and store the central state vector sequence, training intensity adjustment instructions, and training benefit scores in each training process as a structured training log. The input of the offline optimization module is connected to the output of the data recording module, and the output of the offline optimization module is connected to the input of the adaptive load mapping module. The offline optimization module is used to retrain the nonlinear mapping model offline using a batch learning method based on multiple accumulated structured training logs, and pushes the optimized model parameters to the adaptive load mapping module to update the nonlinear mapping model.

Images