A method for determining training maturity based on neural efficiency dynamic tracking
By using a neural efficiency dynamic tracking model that combines EEG signals and manipulation performance data, the problems of behavioral performance dependence and subjective scoring in existing training maturity assessments have been solved, enabling objective evaluation of training maturity and the development of personalized training programs.
Patent Information
- Application Number
- CN202611147039.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-30
- Publication Date
- 2026-11-10
AI Technical Summary
Existing training maturity assessment methods rely too heavily on behavioral performance and subjective ratings, making it difficult to distinguish between "excessive effort to maintain high performance" and "truly internalized skillful control." Furthermore, retesting at fixed calendar times leads to incomparable training accumulation, resulting in a lack of comparability in assessment results.
By collecting EEG signals and control performance data from operators, a dynamic tracking model of neural efficiency based on cumulative effective training time is established. The model combines regression slope and significance test to determine the neural efficiency plateau period and matches it with behavioral performance to output the training maturity status.
It achieves objective and reproducible training maturity assessment, can distinguish different training states, improves the stability and safety of training admission judgment for high cognitive load positions, and supports the development of personalized training programs.
Smart Images

Figure CN123058730A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of neural engineering and personnel training assessment technology, and in particular to a method for determining training maturity based on dynamic tracking of neural efficiency. Background Technology
[0002] Manipulation training for high cognitive load positions typically requires assessing whether operators have truly mastered a skill and are qualified for promotion or employment. Existing training graduation standards rely heavily on behavioral and managerial indicators such as exam scores, task completion rates, instructor ratings, or number of training days. While these indicators can reflect the explicit outcomes of manipulation tasks, they fail to directly reflect the neural resource consumption and skill automation levels of operators during task completion.
[0003] Some existing methods attempt to assess training effectiveness using the difference in neural efficiency between two time points before and after training. However, this two-point static comparison cannot reflect the gradual changes during skill internalization and is difficult to identify high-performance states resulting from short-term effort, fatigue compensation, or strategic test-taking. In practice, some operators can maintain high behavioral performance in assessment scenarios through sustained high-intensity attention, but their neural resource consumption remains high, and their skills are not yet automated. Judging them as qualified solely based on behavioral performance may lead to decreased stability after they are put on the job.
[0004] Furthermore, inconsistencies in subjective evaluation criteria exist between different instructors, training batches, or training courses; the fixed-calendar-time retesting method also fails to consider individual differences in training intensity, leading to significant variations in the cumulative practical training volume of different operators on the same retest date, thus reducing the comparability of evaluation results. Therefore, it is necessary to propose a training maturity evaluation scheme that can longitudinally track cumulative effective training time and comprehensively assess neural efficiency plateaus and behavioral performance achievement. Summary of the Invention
[0005] To address the problems existing in the prior art, the present invention aims to provide a method and system for determining training maturity based on dynamic tracking of neural efficiency, in order to solve the problems that existing training maturity evaluation relies too much on behavioral performance and subjective scoring, has difficulty distinguishing between "excessive effort to maintain high performance" and "truly internalized skillful control", and has incomparable training accumulation due to retesting at fixed calendar times.
[0006] To achieve the above objectives, this invention provides a method for determining training maturity based on dynamic tracking of neural efficiency, comprising the following steps: Step S1: Before the operator begins system training, EEG signals and operation performance data are collected synchronously through standardized simulated operation tasks. The EEG signals are preprocessed, and the mean power spectral density of the θ band of the preset electrodes in the prefrontal cortex is extracted. And based on the comprehensive score of control performance and Calculate the initial neural efficiency baseline ; Step S2: Using the cumulative effective practical training time of operators reaching a preset threshold of X hours as the retest trigger condition, whenever the newly added cumulative effective practical training time reaches the preset threshold of X hours, a retest task consistent with the baseline test is executed, and the neural efficiency at each retest time is calculated. and record with the above The corresponding cumulative effective practical training time t; Step S3: When there are at least four measurement time points, including the baseline, establish a regression model with the cumulative effective practical training time t as the independent variable and neural efficiency NE as the dependent variable, and obtain the regression slope β, the confidence interval of β, the significance test p-value, and the model fit. ; Step S4: When the confidence interval of β contains zero and the p value is greater than the preset significance level α, it is determined that the operator has entered a neural efficiency plateau in the corresponding operation subject. Step S5: Obtain the average of the comprehensive control performance scores from at least the three most recent retests. and the With the preset performance target The matching process is performed, and the training maturity status is output based on the plateau period determination result and the performance matching result.
[0007] Furthermore, the EEG signal mentioned in step S1 is acquired by an EEG acquisition device with no less than 32 electrodes, the sampling rate is no less than 250Hz, and the electrode impedance is controlled below 10kΩ during acquisition; the preprocessing includes at least one of whole-brain average rereference, 0.5-45Hz bandpass filtering, 50Hz power frequency notch filtering, bad conduction detection and interpolation, and independent component analysis for artifact removal.
[0008] Furthermore, the prefrontal cortex preset electrodes include Fz, F3, F4, and optional FCz; the θ band is a rhythmic component within the frequency range of 4-8 Hz; and the neural efficiency NE is determined according to... Calculate, where P is the overall performance score. The average power spectral density of the preset electrode in the θ band during the corresponding test period.
[0009] Furthermore, the comprehensive control performance score P is based on... Calculation, where , , These represent the error indicators generated by the control task. , , These are the corresponding weights. , , These represent the normalized values of track deviation, response delay, and error rate generated by the control task, respectively. The normalized value of track deviation is calculated from the root mean square deviation between the actual trajectory and the target trajectory during the task. The normalized value of response delay is calculated from the average delay from the issuance of the command to the effective date of the control action. The normalized value of error rate is calculated as the ratio of the number of safety boundary violations or collision risk events during the task to the total number of operations in the task. , , Pre-set according to the task priority or security level of the training subject, and meet the requirements. + + =1.
[0010] Furthermore, the preset threshold X is 10-30 hours; when using drone operation training, the preset threshold X is 15 hours; the cumulative effective practical training time only counts the practical training time that meets the validity conditions of the training record, excluding theoretical lectures, equipment waiting, invalid interruptions, or periods of unqualified data quality. The timing of the cumulative effective practical training time starts from the operator receiving the standardized simulated operation task start signal and ends with the task completion signal or reaching the preset single training time limit. When any of the following conditions are met, the corresponding training period is determined to be invalid and is not included in the cumulative effective practical training time: the effective channel ratio of EEG signals is less than 80%; the duration of a single communication interruption or task pause exceeds 30 seconds; or there are missing command response data in the operation performance record.
[0011] Furthermore, the regression model is The least squares method is used for fitting; the confidence interval is a 95% confidence interval, and the preset significance level α is 0.05-0.10.
[0012] Furthermore, when there are fewer than four measurement time points, an insufficient data accumulation prompt will be displayed; when... and When the output data fluctuates significantly, a warning message is displayed; and At that time, it was determined that neural efficiency was in a phase of continuous improvement.
[0013] Furthermore, the training maturity states in step S5 include: mature state, in progress state, skill bottleneck state, and neural adaptation period state; among them, the mature state corresponds to having entered the neural efficiency plateau period and The ongoing state corresponds to the period before the neural efficiency plateau. and The skill bottleneck state corresponds to entering a period of plateau in neural efficiency, but... The neural adaptation period corresponds to the period before entering the neural efficiency plateau and... Or the trend is unstable.
[0014] Furthermore, steps S1-S5 are executed independently for multiple control subjects performed by the same operator, and the cumulative effective practical training time, neural efficiency sequence, regression judgment result and maturity status of each control subject are maintained respectively. The control subjects include at least one of hovering, flight path and emergency obstacle avoidance.
[0015] Furthermore, when outputting the training maturity status, a training maturity assessment report is generated simultaneously. The training maturity assessment report includes the maturity level of each control subject, the assessment basis, a longitudinal trend chart of neural efficiency changes, a multi-dimensional skill maturity map, and promotion or job placement suggestions.
[0016] The present invention also provides a neural efficiency tracking training maturity assessment system, comprising: a baseline acquisition module, a training duration recording and retest triggering module, a signal preprocessing module, a neural efficiency calculation module, a longitudinal data management module, a regression analysis and statistical assessment module, a behavioral performance matching verification module, a maturity comprehensive assessment module, and a report and graph generation module; the system is configured to execute the above-described method.
[0017] Furthermore, the training duration recording and retest triggering module is configured to receive training records sent by the training management system, simulation control platform or manual confirmation terminal, and trigger a retest when the cumulative effective practical training duration reaches a preset threshold of X hours; the report and graph generation module is configured to output a training maturity assessment report and a skill maturity graph on the display terminal.
[0018] This invention replaces fixed calendar time with cumulative effective practical training time as the trigger for retesting, making the evaluation results more comparable among operators with different training intensities. Through dual criteria of regression slope confidence interval and significance level, it forms an objective and reproducible rule for determining neural efficiency plateaus, reducing subjective differences among instructors. By matching neural efficiency plateaus with behavioral performance achievement lines, it can distinguish between different training states such as true maturity, continued improvement, skill bottleneck, and neural adaptation. It supports independent tracking of multiple subjects and generation of skill maturity maps, facilitating the development of targeted follow-up training plans and promotion suggestions. It can identify situations where "excessive effort is used to maintain high performance" but neural efficiency has not yet plateaued, improving the stability and safety of training admission judgments for high cognitive load positions. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the training maturity assessment method based on neural efficiency dynamic tracking. Figure 2 A schematic diagram illustrating the longitudinal trend of neural efficiency and the statistical determination of plateau periods; Figure 3 A schematic diagram illustrating the four states of training maturity classification rules; Figure 4 The structural block diagram for training maturity determination system. Detailed Implementation
[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.
[0021] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Where there is no conflict, the technical features in the following embodiments can be combined with each other.
[0022] This embodiment provides a training maturity assessment method based on dynamic tracking of neural efficiency. This method can be applied to assess the training maturity of high-cognitive-load positions such as UAV pilots, remotely operated equipment operators, and personnel training in complex equipment simulation. See also... Figure 1 The method includes the following steps S1-S5.
[0023] Step S1: Acquire initial neural efficiency baseline Before operators begin system training, standardized simulated control tasks are presented to them through a standardized simulation control platform. These standardized simulated control tasks can be drone hovering and maintaining tasks, flight path tracking tasks, emergency obstacle avoidance tasks, or other simulated tasks requiring continuous attention, spatial judgment, and precise control.
[0024] During the standardized simulated control task performed by the operator, the system simultaneously collects two types of data: the first type is electroencephalogram (EEG) signals, and the second type is control performance data. EEG signals characterize the level of cortical activation when the operator completes the task, while control performance data characterizes the quality of task completion.
[0025] After the EEG signal acquisition is completed, the EEG signal is preprocessed, and the mean power spectral density of the prefrontal cortex preset electrode theta band is extracted from the preprocessed EEG signal. At the same time, a comprehensive control performance score is calculated based on the error indicators generated by the standardized simulated control task. The initial neural efficiency baseline was then calculated according to the following relationship. :
[0026] in, This represents the initial neural efficiency baseline before operator training begins. This represents the overall performance score during the baseline testing period. This represents the average power spectral density of the preset electrode in the prefrontal cortex during the baseline test.
[0027] Through the above step S1, the initial level of neural efficiency of the operator before systematic training can be obtained, providing a benchmark for subsequent longitudinal tracking.
[0028] Step S2: Trigger a retest by accumulating effective practical training time. After the operator enters the system for training, the system continuously records the operator's cumulative effective practical training time. The cumulative effective practical training time refers to the training time during which the operator actually performs the target control task, and the corresponding training records, EEG data quality, and performance data all meet the effectiveness requirements.
[0029] When the operator's cumulative effective practical training time reaches a preset threshold of X hours, the system triggers a retest. The retest uses the same task flow, EEG acquisition parameters, performance recording rules, and neural efficiency calculation rules as the baseline test in step S1 to ensure comparability between the measurement results.
[0030] After each retest, the system calculates the neural efficiency at that retest time. And record the neural efficiency. The corresponding cumulative effective practical training time t. Specifically, Calculated according to the following relationship:
[0031] in, This represents the comprehensive performance score for operation when the cumulative effective practical training time is t. This represents the average power spectral density of the prefrontal cortex preset electrode in the θ band when the cumulative effective practical training time is t.
[0032] Through step S2 above, the present invention no longer uses a fixed calendar date as the basis for retesting, but uses the effective training amount actually accumulated by the operator as the basis for retesting, thereby improving the comparability of training maturity assessment results among different operators.
[0033] Step S3: Establish a longitudinal regression model for neural efficiency When there are no fewer than four measurement time points, including the baseline measurement point, the system uses the cumulative effective practical training time t as the independent variable and the neural efficiency NE as the dependent variable to establish a regression model of the change of neural efficiency with the cumulative effective practical training time.
[0034] In a preferred embodiment, the regression model is a linear regression model:
[0035] Where β represents the regression slope of neural efficiency as a function of cumulative effective practical training time, and b represents the intercept. The system can use the least squares method to fit the above model and obtain the regression slope β, the confidence interval of β, the significance test p-value, and the model fit goodness of fit. .
[0036] β is used to reflect the direction and speed of change in neural efficiency as training time increases; when β is greater than 0 and statistically significant, it indicates that neural efficiency is still in a state of continuous improvement; when β is close to 0 and not statistically significant, it indicates that neural efficiency tends to stabilize.
[0037] Step S4: Determine the plateau period of neural efficiency The system determines whether the operator has entered a plateau in neural efficiency based on the regression slope β, the confidence interval of β, and the significance test p-value obtained in step S3.
[0038] In a preferred embodiment, the plateau period determination requires that the following two conditions be met simultaneously: Condition 1: The confidence interval for β includes zero. Condition 2: The p-value is greater than the preset significance level α.
[0039] When both of the above conditions are met, it indicates that neural efficiency no longer shows a significant increasing trend with the change of cumulative effective practical training time, and the system determines that the operator has entered a neural efficiency plateau period in the corresponding operation subject.
[0040] Through step S4, this invention does not judge the training effect based solely on the results of a single test or the difference between two points before and after training, but rather judges whether skill automation tends to stabilize based on the longitudinal trend of neural efficiency and statistical significance.
[0041] Step S5: Combine behavioral performance output with training maturity status After determining the plateau period of neural efficiency, the system further extracts the average of the comprehensive control performance scores from at least three recent retests. and will With the preset performance target Compare them.
[0042] The system integrates the plateau determination result from step S4 with the behavioral performance matching result to output the training maturity status. Specifically, if the operator has entered a neural efficiency plateau and If the operator has not entered the neural efficiency plateau but the neural efficiency is still significantly improved, then the output will show a mature state; if the operator has entered the neural efficiency plateau but the neural efficiency is still significantly improved, then the output will show a progress .... If the operator has not entered the neural efficiency plateau period and the neural efficiency trend is not positive or unstable, then the neural adaptation period state will be output.
[0043] Through step S5, the present invention can distinguish between the training state of "high behavioral performance but still relying on high neural resource input" and the truly mature state of "behavioral performance meeting the standard and neural efficiency tending to be stable", thereby providing an objective basis for training advancement and on-the-job qualification certification.
[0044] In this embodiment, the EEG signal in step S1 is acquired using an EEG acquisition device with no fewer than 32 electrodes. The sampling rate of the EEG acquisition device is no less than 250Hz, and the electrode impedance is controlled below 10kΩ during the acquisition process to ensure that the EEG signal has sufficient time resolution and signal-to-noise ratio.
[0045] After the EEG acquisition is completed, the signal preprocessing module performs at least one of the following processing on the EEG signal: whole brain average rereference, 0.5-45Hz bandpass filtering, 50Hz power frequency notch filtering, bad conduction detection and interpolation, and independent component analysis to remove artifacts.
[0046] Among them, whole-brain average rereference is used to reduce the impact of reference electrode selection on signal analysis results; 0.5-45Hz bandpass filtering is used to retain the main EEG frequency bands related to cognitive load and task operation; 50Hz power frequency notch filtering is used to suppress power frequency interference; bad conduction detection and interpolation are used to process electrode channels with poor contact or abnormal noise; independent component analysis artifact removal is used to remove non-brain-derived components such as blinking, eye movement, electromyography and motor artifacts.
[0047] Through the above processing, EEG data suitable for calculating the theta band power spectral density can be obtained, providing reliable input for subsequent neural efficiency calculations.
[0048] In this embodiment, the prefrontal cortex preset electrodes include Fz, F3, F4, and optional FCz. These electrode sites can reflect the level of prefrontal cortex activity related to attention control, working memory, and task monitoring during maneuvering tasks.
[0049] The theta band comprises rhythmic components within the 4-8 Hz frequency range. The system performs spectral analysis on the preprocessed EEG signal, calculates the power spectral density of the prefrontal cortex preset electrodes in the theta band, and averages the theta band power spectral density of each prefrontal cortex preset electrode to obtain... .
[0050] Neural efficiency (NE) is calculated according to the following relationship:
[0051] Where P represents the overall performance score for manipulation. This represents the average power spectral density of the preset electrode in the θ band of the prefrontal cortex during the corresponding test period.
[0052] Under this calculation method, neural efficiency NE is high when the operator achieves high control performance with relatively low prefrontal theta power input; conversely, neural efficiency NE is low when the operator requires higher neural resource input to maintain the same control performance. Therefore, this indicator can reflect the changing trend of the degree of automation of control skills.
[0053] In this embodiment, the overall performance score P is calculated according to the following relationship:
[0054] in, , , These represent the error indicators generated by the control task. , , These are the corresponding weights.
[0055] Taking drone operation training as an example, It can represent the normalized value of track deviation or position deviation. It can represent a normalized value of response latency or instruction execution delay. It can represent the task error rate, the number of times the safety boundary is crossed, or the normalized value of the collision risk. , , The settings can be configured based on the training outline, job importance, or historical sample statistics, and each weight and error indicator should reflect the overall quality of task completion.
[0056] In the specific calculation, the system can first normalize each error index to make them fall within a uniform dimension, then sum them according to their weights, and finally subtract the total weighted error from 1 to obtain the comprehensive control performance score P. The higher the P value, the better the control performance.
[0057] In a preferred embodiment, Calculated by normalizing the root mean square deviation between the actual trajectory and the target trajectory during the mission. Calculated based on the normalized average delay from the issuance of the command to the effective implementation of the control action. The calculation is normalized according to the ratio of the number of times the safety boundary is crossed or the number of collision risk events to the total number of mission operations. , , The value can be preset according to the safety level of the training subject; in one example, when the safety boundary is crossed or the collision risk is given a higher attention weight, it can be set to... =0.4、 =0.3、 =0.3, but the present invention is not limited to this specific value.
[0058] In this embodiment, the preset threshold X is 10-30 hours. Taking UAV operation training as an example, X is preferably 15 hours. That is, the system triggers a retest for every additional 15 hours of effective practical training time accumulated by the operator.
[0059] The cumulative effective practical training time only includes practical training time that meets the validity criteria of the training record. Time periods spent in theoretical instruction, equipment waiting, invalid interruptions, communication abnormalities, incomplete tasks, substandard EEG signal quality, or missing performance records are not included in the cumulative effective practical training time.
[0060] For example, if a pilot accumulates 18 hours of training in a training cycle, including 15 hours of effective practical training, 2 hours of theoretical debriefing, and 1 hour of equipment setup and waiting time, the system will only count the 15 hours as the cumulative effective practical training time. When the cumulative effective practical training time reaches a threshold X, the system will automatically generate a retest task and prompt the instructor or training administrator to arrange a retest.
[0061] In this way, the present invention uses the actual amount of training rather than the natural date as the basis for retesting, which can avoid the problem of incomparable evaluation results caused by different operators having different training frequencies.
[0062] In this embodiment, the system uses the following linear regression model to perform longitudinal trend analysis of neural efficiency:
[0063] Where t represents the cumulative effective practical training time, NE represents the neural efficiency at the corresponding measurement time point, β represents the regression slope, and b represents the intercept.
[0064] The system uses the least squares method to fit longitudinal data, including baseline measurement points and multiple remeasurement points, and calculates the 95% confidence interval of β, the p-value, and... The preset significance level α is 0.05-0.10, preferably 0.05.
[0065] When the 95% confidence interval of β includes zero, and At this point, the system determines that neural efficiency has reached a plateau. This determination rule indicates that although the operator continues to train, neural efficiency no longer increases significantly with training duration, and the level of neural resource consumption for control skills tends to stabilize.
[0066] For example, if an operator's β is 0.00005 / h, the 95% confidence interval is [-0.00009, 0.00019], p=0.38, and α=0.05, then since the confidence interval includes 0 and p>0.05, the system determines that the operator has entered a neural efficiency plateau.
[0067] In this embodiment, the system provides prompts for different data states to avoid making unreliable judgments when data is insufficient or fluctuates excessively.
[0068] When there are fewer than four measurement time points, including the baseline, the system does not perform a plateau period statistical judgment. Instead, it outputs a "Insufficient data accumulation" message and suggests continuing training and retesting. This rule is used to prevent drawing unstable conclusions based on only a small number of measurement points.
[0069] when and At this point, the system outputs a "significant data fluctuation" message. This message indicates that the linear relationship between the current neural efficiency sequence and training duration is weak, which may be affected by factors such as fatigue, sleep, device signal quality, changes in task difficulty, or changes in training strategy, requiring further investigation or the addition of measurement points.
[0070] when and At this point, the system determines that neural efficiency is in a phase of continuous improvement and outputs a prompt such as "Continue current training" or "Training in progress." This rule indicates that the operator has not yet entered a plateau phase, and neural efficiency continues to significantly improve with effective training duration.
[0071] In this embodiment, the system divides the training maturity state into mature state, in-process state, skill bottleneck state, and neural adaptation period state based on the neural efficiency plateau determination result and the behavioral performance matching result.
[0072] First, the mature state. If the operator has entered a plateau in neural efficiency, and If the system determines that the operator is in a mature state, it indicates that the operator's performance has not only met the standards but also that their neural efficiency has stabilized, suggesting that their operational skills have become automated, and a promotion or reassignment can be recommended.
[0073] Second, the ongoing state. If the operator has not entered a neural efficiency plateau, and , If the system detects a failure, it indicates that the training is in progress. This status means the operator is still in the skill improvement phase, and it is recommended to continue the current training program.
[0074] Third, the skill bottleneck state. If the operator has entered a period of plateau in neural efficiency, but If this condition is not met, the system will determine that the system is in a skill bottleneck state. This state indicates that neural efficiency is no longer significantly improving, but the control performance still does not meet the job requirements, requiring adjustments to the training program, task difficulty, or training strategy.
[0075] Fourth, the neural adaptation period. If the operator has not entered the neural efficiency plateau period, and If the trend is unstable, the system determines it to be in the neural adaptation period. This state may indicate that the operator has not yet formed a stable skill strategy, or is affected by factors such as fatigue, sleep, equipment adaptation, or insufficient task understanding, requiring increased training or investigation of interfering factors.
[0076] By using a four-state classification, this invention not only determines whether a training is qualified, but also further distinguishes between different states such as continued improvement in training, bottleneck stagnation, and abnormal adaptation, providing a basis for personalized training intervention.
[0077] In this embodiment, the system independently executes the above steps S1-S5 for multiple operation subjects of the same operator, and maintains the cumulative effective practical training time, neural efficiency sequence, regression judgment result and maturity status of each operation subject.
[0078] Taking drone pilot training as an example, the control subjects can include at least one of hovering, flight path, and emergency obstacle avoidance. For hovering, the system independently records the effective training time, NE sequence, and maturity status. For flight path, the system independently records the flight path deviation, response latency, EEG theta band power, and neural efficiency changes. For emergency obstacle avoidance, the system independently records the obstacle avoidance response, number of boundary violations, mission success rate, and corresponding neural efficiency.
[0079] By tracking multiple subjects independently, the system can prevent operators from performing well in one subject and masking their lack of proficiency in another. For example, if a pilot is assessed as proficient in the flight path subject but is still in the emergency obstacle avoidance subject, the system will not simply output overall proficiency. Instead, it will generate a proficiency map for each subject and suggest continued training for the emergency obstacle avoidance subject.
[0080] In this embodiment, the system generates a training maturity assessment report simultaneously when outputting the training maturity status. The training maturity assessment report includes the maturity level of each subject, the assessment criteria, a longitudinal trend chart of neural efficiency, a multi-dimensional skill maturity map, and recommendations for promotion or job placement.
[0081] The maturity level for each subject is used to display the maturity, progress, skill bottleneck, or neural adaptation period status of each subject; the judgment criteria are used to display β, confidence interval, p-value, etc. , and Key indicators include: a longitudinal trend chart of neural efficiency to show the curve of NE changing with cumulative effective practical training time; a multi-dimensional skill maturity map to show the maturity distribution of operators in different subjects; and promotion or job placement recommendations to generate suggestions for continued training, adjustment of plans, postponement of promotion, or permission for promotion based on maturity status.
[0082] By generating reports, training managers can gain a clear understanding of the skill maturity of operators and the statistical basis for this understanding, thereby reducing subjective scoring discrepancies.
[0083] In this embodiment, the present invention also provides a neural efficiency tracking training maturity assessment system. See also Figure 4 The system includes a baseline acquisition module, a training duration recording and retest triggering module, a signal preprocessing module, a neural efficiency calculation module, a longitudinal data management module, a regression analysis and statistical judgment module, a behavioral performance matching and verification module, a maturity comprehensive judgment module, and a report and map generation module.
[0084] The baseline acquisition module is used to perform standardized simulation control tasks before training and simultaneously acquire EEG signals and control performance data.
[0085] The training duration recording and retest triggering module is used to record the cumulative effective practical training time of operators and trigger a retest when the cumulative effective practical training time reaches a preset threshold of X hours.
[0086] The signal preprocessing module is used to perform rereference, filtering, notch filtering, bad conduction detection and interpolation, and ICA artifact removal on EEG signals.
[0087] The neural efficiency calculation module is used for calculation and each retest time .
[0088] The longitudinal data management module is used to maintain neural efficiency sequences and performance data by personnel, training batches, control subjects, and cumulative effective practical training time.
[0089] The regression analysis and statistical judgment module is used to establish The regression model outputs β, β confidence interval, p-value, and... .
[0090] The behavior performance matching verification module is used to calculate and will and Compare them.
[0091] The maturity assessment module is used to output the status of mature, in progress, skill bottleneck, or neural adaptation period based on the plateau assessment results and performance matching results.
[0092] The report and map generation module is used to generate training maturity assessment reports, longitudinal trend charts of neural efficiency, and multi-dimensional skill maturity maps.
[0093] The system described above can execute the methods described above by working together through the collaboration of its various modules.
[0094] In this embodiment, the training duration recording and retest triggering module can receive training records sent by the training management system, the simulation control platform, or the manual confirmation terminal. The training records may include trainee number, training subject, training start time, training end time, effective training duration, training task completion status, abnormal interruption records, and data quality identifiers.
[0095] When the training duration recording and retest triggering module determines that an operator's cumulative effective practical training time in a certain operation subject has reached a preset threshold of X hours, the system automatically triggers the retest process and sends a retest prompt to the display terminal, instructor terminal, or training management platform.
[0096] The report and map generation module can output training maturity assessment reports and skill maturity maps to the display terminal. The display terminal can be a training management workstation, an instructor's computer, a training evaluation server display interface, or other terminal devices with graphical display capabilities.
[0097] In one implementation, the EEG acquisition device, the simulation control platform, and the training management system are connected to the training maturity assessment system via a wired network or local area network. In another implementation, the EEG acquisition device and the simulation control platform can upload EEG data and performance data to the training management server, respectively, and the training maturity assessment system can perform data fusion, neural efficiency calculation, statistical assessment, and report generation on the server side.
[0098] Example 1: Pilot D Training Tracking Pilot D completed three retests, with a total of four measurement points including the baseline. The neural efficiency sequences were 0.032, 0.038, 0.043, and 0.049, corresponding to cumulative effective practical training times of 0h, 15h, 30h, and 45h. The system fit yielded β=0.00038 / h, 95%CI=[0.00019,0.00054], where the confidence interval does not contain 0, and p=0.014<0.05.
[0099] Since β>0 and p<0.05, the neural efficiency of pilot D still increases significantly with training time. The system determines that it has not yet entered a plateau period and outputs the "in progress" status, suggesting that the current training program be continued.
[0100] Example 2: Pilot E-Training Tracking Pilot E completed 5 retests, with a total of 6 measurement points including the baseline. The neural efficiency sequence was 0.050, 0.052, 0.055, 0.054, 0.053, and 0.055, corresponding to a cumulative effective practical training time of 0 to 75 hours. The system fit yielded β = 0.00005 / h, 95% CI = [-0.00009, 0.00019], with the confidence interval including 0, and p = 0.38 > 0.05. The mean of the most recent performance tests is also shown. =0.82, preset performance target =0.75.
[0101] Since the pilot E has entered a period of plateau in neural efficiency, and The system determines that pilot E is in a mature state and provides suggestions for promotion or employment.
[0102] In an alternative implementation, neural activity indicators in neural efficiency are not limited to prefrontal theta power, but can also include indicators that characterize task-related neural resource input, such as parietal alpha wave suppression, prefrontal theta / alpha ratio, or event-related desynchronization amplitude.
[0103] In another alternative implementation, the retest triggering mechanism is not limited to the cumulative effective practical training time, but can also adopt a hybrid triggering mechanism that combines the cumulative effective training times, the cumulative number of completed tasks, or the cumulative training time with the number of training times.
[0104] In another alternative implementation, the statistical determination of the plateau period is not limited to linear regression. Piecewise linear regression can be used to automatically detect inflection points, exponential decay model fitting, or Bayesian change point detection, as long as it can determine whether the trend of skill automation has entered a stable stage based on the longitudinal neural efficiency sequence.
[0105] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, improvements, or combinations made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for determining training maturity based on dynamic tracking of neural efficiency, characterized in that, Includes the following steps: Step S1: Before the operator begins system training, EEG signals and operation performance data are collected synchronously through standardized simulated operation tasks. The EEG signals are preprocessed, and the mean power spectral density of the θ band of the preset electrodes in the prefrontal cortex is extracted. And based on the comprehensive score of control performance. and Calculate the initial neural efficiency baseline ; Step S2: Using the cumulative effective practical training time of operators reaching a preset threshold of X hours as the retest trigger condition, whenever the newly added cumulative effective practical training time reaches the preset threshold of X hours, a retest task consistent with the baseline test is executed, and the neural efficiency at each retest time is calculated. and record with the above The corresponding cumulative effective practical training time t; Step S3: When there are at least four measurement time points, including the baseline, establish a regression model with the cumulative effective practical training time t as the independent variable and neural efficiency NE as the dependent variable, and obtain the regression slope β, the confidence interval of β, the significance test p-value, and the model fit. ; Step S4: When the confidence interval of β contains zero and the p value is greater than the preset significance level α, it is determined that the operator has entered a neural efficiency plateau in the corresponding operation subject. Step S5: Obtain the average of the comprehensive control performance scores from at least the three most recent retests. and the With the preset performance target The matching process is performed, and the training maturity status is output based on the plateau period determination result and the performance matching result.
2. The method according to claim 1, characterized in that, The EEG signal in step S1 is acquired by an EEG acquisition device with no less than 32 electrodes, a sampling rate of no less than 250Hz, and the electrode impedance is controlled below 10kΩ during acquisition. The preprocessing includes at least one of whole-brain average rereference, 0.5-45Hz bandpass filtering, 50Hz power frequency notch filtering, bad conduction detection and interpolation, and independent component analysis for artifact removal.
3. The method according to claim 1, characterized in that, The prefrontal cortex preset electrodes include Fz, F3, F4, and optional FCz; the theta band is a rhythmic component within the frequency range of 4-8 Hz; and the neural efficiency NE is determined according to... Calculate, where P is the overall performance score. The average power spectral density of the preset electrode in the θ band during the corresponding test period.
4. The method according to claim 3, characterized in that, The overall performance score P of the control is based on Calculation, where , , These represent the error indicators generated by the control task. , , These are the corresponding weights. , , These represent the normalized values of track deviation, response delay, and error rate generated by the control task, respectively. The normalized value of track deviation is calculated from the root mean square deviation between the actual trajectory and the target trajectory during the task. The normalized value of response delay is calculated from the average delay from the issuance of the command to the effective date of the control action. The normalized value of error rate is calculated as the ratio of the number of safety boundary violations or collision risk events during the task to the total number of operations in the task. , , Pre-set according to the task priority or security level of the training subject, and meet the requirements. + + =1.
5. The method according to claim 1, characterized in that, The preset threshold X is 10-30 hours; when using drone operation training, the preset threshold X is 15 hours; the cumulative effective practical training time only counts the practical training time that meets the conditions for the validity of the training record, excluding theoretical lectures, equipment waiting, invalid interruptions or unqualified data quality periods; The cumulative effective practical training time is calculated from the moment when the operator enters the standardized simulated operation task and receives the task start signal, and ends when the task completion signal is triggered or the preset single training time limit for the subject is reached. When the proportion of effective channels of EEG signals is lower than the preset channel efficiency threshold, the duration of a single task interruption exceeds the preset interruption threshold, or there are missing control performance data, the corresponding time period will not be included in the cumulative effective practical training time.
6. The method according to claim 1, characterized in that, The regression model is as follows: The least squares method is used for fitting; the confidence interval is a 95% confidence interval, and the preset significance level α is 0.05-0.
10.
7. The method according to claim 1, characterized in that, When there are fewer than 4 measurement time points, an insufficient data accumulation prompt will be displayed; when and When the output data fluctuates significantly, a warning message is displayed; and At that time, it was determined that neural efficiency was in a phase of continuous improvement.
8. The method according to claim 1, characterized in that, The training maturity states in step S5 include: mature state, in progress state, skill bottleneck state, and neural adaptation period state; among them, the mature state corresponds to having entered the neural efficiency plateau period and The ongoing state corresponds to the period before the neural efficiency plateau. and The skill bottleneck state corresponds to entering a period of plateau in neural efficiency, but... The neural adaptation period corresponds to the period before entering the neural efficiency plateau and... Or the trend is unstable.
9. The method according to claim 1, characterized in that, Steps S1-S5 are executed independently for multiple control subjects performed by the same operator, and the cumulative effective practical training time, neural efficiency sequence, regression judgment result and maturity status of each control subject are maintained separately. The control subjects include at least one of hovering, flight path and emergency obstacle avoidance.
10. The method according to claim 1, characterized in that, When outputting the training maturity status, a training maturity assessment report is generated simultaneously. The training maturity assessment report includes the maturity level of each control subject, the assessment basis, the longitudinal trend chart of neural efficiency, the multi-dimensional skill maturity map, and the promotion or job placement suggestions.
11. A neural efficiency tracking training maturity assessment system, characterized in that, include: The system comprises a baseline acquisition module, a training duration recording and retest triggering module, a signal preprocessing module, a neural efficiency calculation module, a longitudinal data management module, a regression analysis and statistical judgment module, a behavioral performance matching verification module, a maturity comprehensive judgment module, and a report and map generation module; the system is configured to perform the method described in any one of claims 1-10.
12. The system according to claim 11, characterized in that, The training duration recording and retest triggering module is configured to receive training records sent by the training management system, simulation control platform or manual confirmation terminal, and trigger a retest when the cumulative effective practical training duration reaches a preset threshold of X hours; the report and map generation module is configured to output a training maturity assessment report and a skill maturity map on the display terminal.
