A response data-based cognitive training task parameter management method

Abstract

The application relates to the technical field of cognitive training and discloses a cognitive training task parameter management method based on response data. Based on the frequency band energy distribution attribute, the historical task parameter sequence and the corresponding time stamp in response feature data, a task load response surface calculation model is constructed through a double cubic spline interpolation kernel function, gradient data is calculated along the task parameter and time dimensions of the model respectively, and is fused into a response gradient feature vector; response lag ring feature data is extracted by using the asymmetric deviation of the task parameter increment and decrement stages, and a parameter adjustment coefficient is solved based on the closed region measurement value; then, a task parameter optimization interval and a threshold value set are calibrated; finally, a task efficiency index is solved according to the ratio of the object behavior performance and the spectrum resource consumption degree, and task effect evaluation data is generated. The application can realize fine predictive adjustment of the task parameter and individualized effect evaluation.

Description

Technical Field

[0001] This invention relates to the field of cognitive training technology, and more specifically to a method for managing cognitive training task parameters based on response data. Background Technology

[0002] Cognitive training enhances an individual's cognitive abilities through systematic task practice, and its effectiveness is highly dependent on the dynamic matching between task difficulty and the individual's current cognitive state. Existing cognitive training systems primarily adjust difficulty based on user behavioral feedback, such as accuracy and reaction time. However, behavioral feedback is the final output of cognitive processing and inherently lags behind; the system often reacts only after the user has been cognitively overloaded for some time, missing the optimal adjustment window. Furthermore, relying solely on behavioral indicators fails to reflect the degree of neural resource consumption in completing the task, resulting in a simplistic adjustment strategy and difficulty in distinguishing between "efficient learning" and "inefficient effort." Although some research has attempted to introduce signals for cognitive state assessment, such as using deep learning models to map features to macroscopic cognitive states like emotion and fatigue and adjust task parameters accordingly, these typically treat the signals as a "black box," resulting in coarse-grained cognitive states that fail to accurately characterize the user's dynamic response to changes in task difficulty. Summary of the Invention

[0003] In view of the above problems, the present invention provides a cognitive training task parameter management method based on response data, which realizes refined perception and proactive adjustment of the training state by constructing a dynamic surface of the response and extracting its gradient and hysteresis loop features.

[0004] To achieve the above objectives, this application provides a method for managing cognitive training task parameters based on response data, comprising:

[0005] The object training process data record is obtained from the data interface of the task management system. The object training process data record includes historical task parameter sequence, behavior response log data and response feature data. The response feature data is a multi-band energy distribution feature vector sequence embedded after frequency domain transformation.

[0006] Based on the frequency band energy distribution attributes, historical task parameter sequences and corresponding timestamps in the response feature data, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function.

[0007] The parameter gradient data and time gradient data along the task parameter dimension and time dimension of the task load response surface calculation model are obtained respectively, and then fused to obtain the response gradient feature vector.

[0008] Based on the asymmetric deviation characteristics of the task load response surface calculation model between the increasing and decreasing stages of task parameters, the response hysteresis loop characteristic data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value.

[0009] Based on the response gradient feature vector, response hysteresis loop feature data and parameter adjustment coefficient, the task parameter optimization range and task state discrimination threshold set are determined by the task state discrimination model in the step-difficulty calibration task.

[0010] During task execution, based on the ratio of the spectrum resource consumption data between the object behavior performance index data and the response feature data, the task efficiency index data is solved through the neural efficiency calculation model, and task effect evaluation data is generated and output through the task effect evaluation model according to its time series change trend.

[0011] In some embodiments, the method further includes:

[0012] During the task execution phase, a real-time state discrimination vector is constructed based on the response gradient feature vector and response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output.

[0013] Based on the deviation between the current task status judgment result and the task parameter optimization range, task parameter update instruction data is generated and output through the task parameter update model.

[0014] In some embodiments, a real-time state discrimination vector is constructed based on the response gradient feature vector and response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output, including:

[0015] The modulus data of the response gradient feature vector at the current moment, the sign of the gradient data along the time dimension, the deviation data between the current response feature data and the first response curve corresponding to the task load response surface calculation model in the task parameter increment stage, and the parameter adjustment coefficient are obtained and combined into a four-dimensional real-time state discrimination vector.

[0016] The task state discrimination model pre-stores a first threshold, a second threshold, a third threshold, and a fourth threshold. The first to fourth thresholds are adaptively determined based on the statistical distribution of boundary data of the task parameter optimization interval and response lag loop characteristic data. Among them, the first threshold corresponds to the deviation boundary between the resource underutilization state and the resource optimization allocation state, the second threshold corresponds to the deviation boundary between the resource optimization allocation state and the resource shortage allocation state, the third threshold corresponds to the positive and negative sign conversion of the time gradient data to assist in the judgment value, and the fourth threshold corresponds to the deviation boundary of the resource overload state after weighted correction by the parameter adjustment coefficient.

[0017] The task state discrimination model calculates the Mahalanobis distance between the statistical distribution centers of the four-dimensional real-time state discrimination vector and the response gradient feature vector modulus data corresponding to each task parameter during the step-difficulty calibration task within the task parameter optimization interval, and uses it as an indicator of the current state deviation.

[0018] The task status discrimination model compares the current state deviation index with a first threshold. If the current state deviation index is less than the first threshold, the current task status is determined to be a state of insufficient resource mobilization.

[0019] If the current state deviation index is greater than or equal to the first threshold and less than or equal to the second threshold, then the current task state is determined to be a resource optimization and allocation state.

[0020] If the current state deviation index is greater than the second threshold, then a comprehensive judgment is made by combining the positive and negative signs of the gradient data along the time dimension with the deviation data:

[0021] When the gradient data along the time dimension is positive and the deviation data is less than the third threshold, it is determined to be a state of tight resource allocation.

[0022] When the gradient data along the time dimension is positive and the deviation data is greater than or equal to the third threshold, a weighted correction is performed in combination with the parameter adjustment coefficient. If the weighted deviation index is greater than the fourth threshold, it is determined to be a resource overload state.

[0023] In some embodiments, based on the deviation between the current task state determination result and the task parameter optimization range, task parameter update instruction data is generated and output through a task parameter update model, including:

[0024] The task parameter update model reads the current task status judgment result, which is one of the following: resource underutilization, resource optimization, resource shortage, or resource overload.

[0025] If the current task status is determined to be a state where resources are not fully mobilized and the duration exceeds the first preset duration, the task parameter update model generates task parameter improvement instruction data.

[0026] If the current task status is determined to be resource optimization and allocation status, the task parameter update model generates task parameter maintenance instruction data.

[0027] If the current task status is determined to be a resource shortage and the duration exceeds the second preset duration, the task parameter update model generates task parameter reduction instruction data.

[0028] If the current task status is determined to be resource overload, the task parameter update model generates emergency task parameter reduction instruction data and attaches pause instruction data.

[0029] When generating task parameter increase instruction data, task parameter decrease instruction data, task parameter emergency decrease instruction data, and pause instruction data, the task parameter update model adopts a gradual change strategy, adjusting task parameters in a preset step size unit to avoid the impact of sudden changes in task parameters on the task status.

[0030] After completing each behavior statistics unit, the task parameter update model reads the object behavior performance index data corresponding to the behavior statistics unit. If the object behavior performance index data is higher than the first behavior threshold, task parameter improvement correction instruction data is generated. If the object behavior performance index data is lower than the second behavior threshold, task parameter reduction correction instruction data is generated. The weight of task parameter improvement correction instruction data and task parameter reduction correction instruction data is lower than the weight of task parameter improvement instruction data and task parameter reduction instruction data. Furthermore, task parameter improvement instruction data and task parameter reduction instruction data are neural feedback driven regulation, while task parameter improvement correction instruction data and task parameter reduction correction instruction data are behavior feedback assisted correction.

[0031] The task parameter update model records the number of task parameter adjustments within the most recent third preset time period. If the number of task parameter adjustments exceeds the first adjustment frequency threshold, a damping mechanism is triggered, which relaxes the trigger threshold of subsequent adjustments by a preset ratio. If the number of task parameter adjustments exceeds the second adjustment frequency threshold, task parameter adjustments are paused for a fourth preset time period.

[0032] The task parameter update model outputs the final task parameter update instruction data, processed by the damping mechanism, to the task parameter control interface of the task management system.

[0033] In some embodiments, based on the frequency band energy distribution attributes, historical task parameter sequences, and corresponding timestamps in the response feature data, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function, including:

[0034] Extract energy distribution attribute data of the target frequency band from the response feature data. The target frequency band is at least one of the θ band, α band, and low β band.

[0035] Each task parameter value in the historical task parameter sequence is mapped to a discretized task parameter coordinate axis, and the corresponding timestamp is mapped to a continuous time coordinate axis. The energy distribution attribute data of the target frequency band is used as the response value to construct a set of discrete sampling points in the three-dimensional data space. Each sampling point in the set of discrete sampling points contains task parameter coordinates, time coordinates and energy response value.

[0036] The bicubic spline interpolation kernel function is performed on the discrete sampling point set. The bicubic spline interpolation kernel function generates a continuously differentiable interpolation surface covering the entire two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis by solving the second-order partial derivative continuity constraint equations at each sampling point.

[0037] The continuously differentiable interpolation surface is used as the task load response surface calculation model corresponding to the target frequency band. The task load response surface calculation model has a first-order continuous partial derivative at any point in the two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis, which makes it possible to calculate the gradient along the task parameter dimension and along the time dimension.

[0038] If there are multiple target frequency bands, the above-mentioned discrete sampling point set construction and bicubic spline interpolation kernel function operation are performed independently for each frequency band to generate the task load response surface calculation model corresponding to each frequency band, which serves as a set of multi-band task load response surface calculation models.

[0039] In some embodiments, parameter gradient data and time gradient data are obtained along the task parameter dimension and time dimension of the task load response surface calculation model, respectively, and fused to obtain a response gradient feature vector, including:

[0040] Read the task load response surface calculation model. The task load response surface calculation model is a continuously differentiable interpolation surface function in a two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis.

[0041] Locate the task parameter coordinates at the current moment on the task load response surface calculation model, and obtain the first-order partial derivative of the task load response surface calculation model with respect to the task parameter coordinates to obtain the parameter gradient data at the current moment. The parameter gradient data represents the sensitivity of the current response characteristic data to changes in task parameters.

[0042] Locate the time coordinate point corresponding to the current moment on the task load response surface calculation model, obtain the first-order partial derivative of the task load response surface calculation model with respect to the time coordinate, and obtain the time gradient data of the current moment. The time gradient data represents the temporal change rate of the current response characteristic data.

[0043] The parameter gradient data and the time gradient data are used as two components of a two-dimensional vector. The magnitude of the two-dimensional vector is calculated, and the magnitude represents the overall change intensity of the current response feature data.

[0044] The direction angle of the two-dimensional vector is calculated by the arctangent function of the ratio of the temporal gradient data to the parameter gradient data. The direction angle represents the dominant dimension of the current response feature data change. When the direction angle is close to zero, it indicates that the change is dominated by the change of task parameters. When the direction angle is close to ninety degrees, it indicates that the change is dominated by the time process.

[0045] The magnitude and orientation angle of a two-dimensional vector are combined to form the response gradient feature vector.

[0046] In some embodiments, based on the asymmetric deviation characteristics of the task load response surface calculation model between the increasing and decreasing stages of the task parameters, response hysteresis loop feature data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value, including:

[0047] The first response curve corresponding to the increasing stage of the task parameters is extracted from the task load response surface calculation model. The first response curve describes the response value of the response feature data corresponding to each task parameter coordinate point as the task parameters gradually increase from the minimum value to the maximum value.

[0048] The second response curve corresponding to the decreasing stage of the task parameters is extracted from the task load response surface calculation model. The second response curve describes the response value of the response feature data corresponding to each task parameter coordinate point as the task parameters gradually decrease from the maximum value to the minimum value.

[0049] The first response curve and the second response curve are plotted on the same coordinate system, with the task parameter coordinates as the horizontal axis and the response values ​​of the response feature data as the vertical axis. The first response curve and the second response curve have a deviation in response values ​​at the same task parameter coordinate points. The response value of the first response curve is higher than that of the second response curve, forming a closed region.

[0050] The response hysteresis loop calculation model is used to integrate the response value deviation between the first and second response curves along the task parameter coordinate axis to obtain the closed region metric. The closed region metric represents the degree of asymmetry of the response path of an individual task system during the process of increasing and decreasing task parameters.

[0051] The closed region metric is normalized by using a parameter adjustment coefficient calculation model. The normalized closed region metric is then substituted into an inverse proportional function to obtain the parameter adjustment coefficient. The parameter adjustment coefficient is inversely proportional to the closed region metric. The higher the parameter adjustment coefficient, the better the individual task system's adjustment elasticity to changes in task parameters.

[0052] In some embodiments, based on the response gradient feature vector, response hysteresis loop feature data, and parameter adjustment coefficients, the task parameter optimization range and task state discrimination threshold set are calibrated through a task state discrimination model during a stepped difficulty calibration task, including:

[0053] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to each task parameter coordinate point during the step-difficulty calibration task, as well as the local contribution value of the response hysteresis loop feature data corresponding to each task parameter coordinate point.

[0054] The task state discrimination model marks the task parameter coordinate point corresponding to the first time the magnitude data of the response gradient feature vector reaches the preset activation threshold as the lower boundary of the task parameter optimization interval.

[0055] The task state discrimination model marks the maximum task parameter coordinate point corresponding to the condition that the magnitude of the response gradient feature vector does not exceed the preset upper limit threshold and the local contribution value of the response hysteresis loop feature data does not exceed the preset deviation threshold as the upper boundary of the task parameter optimization interval.

[0056] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the lower boundary of the task parameter optimization interval, multiplies it by the first preset scaling factor, and obtains the first threshold.

[0057] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the upper boundary of the task parameter optimization interval, multiplies it by the second preset scaling factor, and obtains the second threshold.

[0058] The task status discrimination model reads the median of the absolute values ​​of all time gradient data during the step-difficulty calibration task, multiplies it by the third preset scaling factor, and obtains the third threshold.

[0059] The task status discrimination model reads the median of all deviation data during the step-difficulty calibration task, multiplies it by the fourth preset proportional coefficient, and obtains the fourth threshold.

[0060] The task state discrimination model reads the parameter adjustment coefficients and multiplies the first threshold, second threshold, third threshold and fourth threshold by the weighting factor corresponding to the parameter adjustment coefficients to obtain the task state discrimination threshold set after individualized correction. Among them, the first threshold and second threshold are used for deviation index judgment, the third threshold is used for auxiliary judgment of positive and negative signs of time gradient data, and the fourth threshold is used for overload state deviation judgment after weighting by parameter adjustment coefficients.

[0061] In some embodiments, based on the ratio of spectral resource consumption data between object behavior performance index data and response feature data, the task efficiency index data is solved using a neural efficiency calculation model, including:

[0062] After each behavior statistics unit ends, read the object behavior response log data corresponding to the behavior statistics unit, and extract the accuracy data and average reaction time data from the object behavior response log data;

[0063] Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline reaction time data from the historical response characteristic data;

[0064] The accuracy data is divided by the ratio of the average reaction time data to the resting baseline reaction time data to obtain the behavioral performance index data. The behavioral performance index data comprehensively reflects the task completion quality and response speed of the subject within the behavioral statistical unit.

[0065] Read the response feature data within the time window corresponding to the behavior statistics unit, and extract the power increase data of the θ band and the power increase data of the low β band from the response feature data;

[0066] Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline power data from the historical response characteristic data;

[0067] The summation of the power increase data in the θ band and the power increase data in the low β band is divided by the resting baseline power data to obtain the spectrum resource consumption data. The spectrum resource consumption data represents the total amount of neural resources consumed by the object to complete the task within the statistical unit of this behavior.

[0068] By dividing the behavioral performance index data by the spectrum resource consumption data through the neural efficiency calculation model, the task efficiency index data is obtained. The task efficiency index data represents the level of behavioral performance supported by a unit of neural resource consumption.

[0069] In some embodiments, task performance evaluation data is generated and output using a task performance evaluation model based on its temporal variation trend, including:

[0070] Read the task efficiency index data corresponding to all behavioral statistical units within the task cycle, and construct a time series sequence of the task efficiency index.

[0071] Read the average value of a preset number of task efficiency index data at the end of the task cycle, and use it as the final task efficiency index data.

[0072] Read the average value of a preset number of task efficiency index data at the beginning of the task cycle, and use it as the initial task efficiency index data.

[0073] Divide the difference between the final task efficiency index data and the initial task efficiency index data by the initial task efficiency index data to obtain the task efficiency index improvement rate data. The task efficiency index improvement rate data represents the extent of improvement in the neural efficiency of the object throughout the entire task cycle.

[0074] The ratio of the standard deviation to the mean of the task efficiency index time series is calculated to obtain the coefficient of variation of the task efficiency index. The coefficient of variation of the task efficiency index characterizes the stability of the neural efficiency of the object during the task.

[0075] Read the parameter adjustment coefficient data calibrated at the end of the task cycle as the final parameter adjustment coefficient data, and read the parameter adjustment coefficient data calibrated at the beginning of the task cycle as the initial parameter adjustment coefficient data. Divide the difference between the final parameter adjustment coefficient data and the initial parameter adjustment coefficient data by the initial parameter adjustment coefficient data to obtain the parameter adjustment coefficient improvement rate data. The parameter adjustment coefficient improvement rate data characterizes the change range of the object's task adjustment elasticity within the task cycle.

[0076] The task effectiveness evaluation model combines the task efficiency index improvement rate data, the task efficiency index variation coefficient data, and the parameter adjustment coefficient improvement rate data into multidimensional task effectiveness evaluation data, and outputs the multidimensional task effectiveness evaluation data to the report generation interface of the task management system.

[0077] Unlike existing technologies, the above-mentioned technical solution provides a task parameter management method based on response data. Based on the frequency band energy distribution attributes, historical task parameter sequences, and corresponding timestamps in the response feature data, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function. Gradient data is calculated along the task parameter and time dimensions of this model and fused into a response gradient feature vector. The asymmetric deviation of the task parameter increment and decrement stages is used to extract response hysteresis loop feature data, and the parameter adjustment coefficient is solved based on its closed region metric. Furthermore, the task parameter optimization interval and discrimination threshold set are calibrated. Finally, the task efficiency index is solved based on the ratio of object behavior performance to spectrum resource consumption, and task effect evaluation data is generated. This invention enables refined predictive adjustment of task parameters and individualized effect evaluation.

[0078] The above description of the invention is merely an overview of the technical solution of this application. In order to enable those skilled in the art to better understand the technical solution of this application and to implement it based on the description and drawings, and to make the above-mentioned objectives and other objectives, features and advantages of this application easier to understand, the following description is provided in conjunction with the specific embodiments and drawings of this application. Attached Figure Description

[0079] The accompanying drawings are only used to illustrate the principles, implementation methods, applications, features, and effects of specific embodiments of the present invention and other related contents, and should not be considered as limitations on this application.

[0080] In the accompanying drawings of the instruction manual:

[0081] Figure 1 This is a schematic diagram illustrating steps S101 to S106 of the method described in the specific implementation embodiment;

[0082] Figure 2This is a schematic diagram illustrating steps S201 to S205 of the method described in a specific implementation.

[0083] Figure 3 This is a schematic diagram illustrating steps S301 to S306 of the method described in a specific implementation.

[0084] Figure 4 This is a schematic diagram illustrating steps S401 to S405 of the method described in a specific embodiment;

[0085] Figure 5 The diagram illustrates steps S501 to S505 of the method described in the specific implementation. Detailed Implementation

[0086] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended to limit the scope of protection of this application.

[0087] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0088] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0089] In the description of this application, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" logical relationship.

[0090] In this application, terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy or order relationship between these entities or operations.

[0091] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar open-ended expressions in this application is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0092] The processor described in the embodiments of this application can be implemented by hardware, firmware, software, or a combination thereof. It can be a circuit, one or more of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field-programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, or a microprocessor. It also includes other physical, biological, or chemical structures that can implement the same or equivalent functions as the processors listed above, such as biological neurons, quantum computing units, DNA computing units, etc., so that the processor can execute some or all of the steps in the computer program or method involved in the various embodiments of this application, or any combination of the steps mentioned therein.

[0093] The computer program involved in the embodiments can be stored in a computer device readable storage medium, which includes, but is not limited to, disks, magnetic tapes, magnetic cards, floppy disks, flash memory, optical disks, optical cards, read-only memory (ROM), random access memory (RAM), erasable programmable ROM (EPROM), and electrically erasable programmable ROM (EEPROM), etc., and also includes other biological, physical, or chemical structures that can achieve the same or equivalent functions as the storage media listed above, such as DNA, RNA, proteins, and other units with information storage capabilities. In specific embodiments, the storage medium involved can be one of the above-mentioned media types, or a combination of the above-mentioned media types. In different embodiments, the computer program involved in the embodiments can be centrally stored in a single medium, or distributed and stored in multiple media. The memory containing the computer device readable storage medium can be non-volatile memory or random access memory. These computer device readable storage media can be built into the device, or can be connected to the device involved in the embodiments as an external device or part of an external device. In some embodiments, the memory having a computer device readable storage medium is deployed locally; in other embodiments, the memory may be deployed remotely from the processor, for example, as a network-attached memory accessed via RF circuitry or an external port and a communication network, wherein the communication network may be the Internet, one or more intranets, a local area network (LAN), a wide area network (WLAN), a storage area network (SAN), or a suitable combination thereof, as long as computer device access to the memory is enabled. Furthermore, the computer program involved in the embodiments may be stored in plaintext / ciphertext form, or it may be designed as training data, integrated and recombined through model training and implicitly stored in the parameter states of a deep neural network or other machine learning model.

[0094] Please see Figure 1 This embodiment provides a method for managing cognitive training task parameters based on response data, including:

[0095] S101. Obtain object training process data records from the data interface of the task management system. The object training process data records include historical task parameter sequences, behavior response log data, and response feature data. The response feature data is a multi-band energy distribution feature vector sequence embedded after frequency domain transformation.

[0096] S102. Based on the frequency band energy distribution attributes, historical task parameter sequences and corresponding timestamps in the response feature data, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function.

[0097] S103. For the task load response surface calculation model, obtain parameter gradient data and time gradient data along the task parameter dimension and time dimension respectively, and fuse them to obtain the response gradient feature vector.

[0098] S104. Based on the asymmetric deviation characteristics between the increasing and decreasing stages of task parameters in the task load response surface calculation model, the response hysteresis loop characteristic data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value.

[0099] S105. Based on the response gradient feature vector, response hysteresis loop feature data and parameter adjustment coefficient, the task parameter optimization range and task state discrimination threshold set are determined by the task state discrimination model.

[0100] S106. During task execution, based on the ratio of the spectrum resource consumption data of the object behavior performance index data and the response characteristic data, the task efficiency index data is solved by the neural efficiency calculation model, and the task effect evaluation data is generated and output by the task effect evaluation model according to its time series change trend.

[0101] In step S101, the data management interface connects the task execution module and the acquisition module of the task management system, aligns and structures the acquired heterogeneous data according to a unified timestamp, and then outputs it. The historical task parameter sequence records the time and value of each task parameter adjustment; the behavior response log data contains the correctness mark and reaction time of each trial; the multi-band energy distribution feature vector sequence in the response feature data is obtained by mapping the frequency domain features within the sliding window to a fixed-dimensional vector through a linear embedding layer, so that the multi-band information at each time point is compressed into a compact numerical representation, which is convenient for subsequent model processing.

[0102] In step S102, the frequency band energy distribution attribute at the same timestamp is used as the response value, historical task parameters are used as the horizontal axis coordinate, and the timestamp is used as the vertical axis coordinate, forming a set of discrete three-dimensional sampling points. The bicubic spline interpolation kernel function defines a bicubic polynomial on each grid cell defined by adjacent sampling points. By forcing the function value, first-order partial derivative, and second-order mixed partial derivative to be continuous at the grid boundary, these local polynomials are spliced ​​into a globally smooth continuous surface. This surface is the task load response surface calculation model. This model has continuous first-order partial derivatives at any point in the domain, providing an analytical basis for subsequent gradient calculation.

[0103] In step S103, the coordinate point corresponding to the current moment is located on the task load response surface calculation model, and the first-order partial derivatives along the task parameter coordinate axis and the time coordinate axis at this point are calculated respectively to obtain parameter gradient data and time gradient data. These two gradient data are combined as two components of a two-dimensional vector to obtain the response gradient feature vector. The two components of this vector quantify the degree to which the response is driven by changes in task parameters and the rate of natural evolution over time, respectively. Their magnitude reflects the overall intensity of change, and the ratio between the components reveals the dominant dimension of the change.

[0104] In step S104, the response value sequences corresponding to each parameter point in the increasing phase and the response value sequences corresponding to the decreasing phase are extracted from the task load response surface calculation model. The response hysteresis loop calculation model plots these two sequences on the same coordinate system and calculates the area of ​​the region enclosed between them to obtain a closed region metric, which quantifies the asymmetry of the increasing and decreasing paths. The parameter adjustment coefficient calculation model converts this metric into a parameter adjustment coefficient through a monotonically decreasing mapping function, so that the larger the closed region, i.e., the stronger the adjustment inertia of the task system, the smaller the adjustment coefficient.

[0105] In step S105, the task state discrimination model utilizes the distribution characteristics of the magnitude of the response gradient feature vector as a function of task parameters, combined with the degree of individual response asymmetry reflected by the response hysteresis loop feature data, and the individual adjustment elasticity characterized by the parameter adjustment coefficient, to jointly define a range of task parameters, which is the task parameter optimization interval. Preferably, the boundary of the task parameter optimization interval can be determined by analyzing the inflection points of the gradient magnitude and hysteresis loop contribution values ​​as a function of parameters. The task state discrimination threshold set is calculated based on the feature statistics and parameter adjustment coefficients within the optimization interval. For example, it is calculated based on the gradient magnitude statistics at the boundary of the optimization interval, the distribution characteristics of the time gradient data during calibration, and the parameter adjustment coefficients, through a preset proportional relationship. This threshold is used to divide the continuous task state space into different levels, providing a basis for subsequent real-time adjustment.

[0106] In step S106, the spectrum resource consumption data is obtained by calculating and summing the power increase of a specific frequency band relative to the resting-state baseline during the task, reflecting the total amount of neural resources invested to maintain the current behavioral performance. The neural efficiency calculation model divides the behavioral performance index data, which integrates accuracy and reaction time, by the spectrum resource consumption data to obtain the task efficiency index data. The higher the index, the better the behavioral performance supported by the unit of neural resource consumption. The task effect evaluation model generates structured task effect evaluation data that includes the efficiency improvement and stability by analyzing the linear change trend, fluctuation degree, and difference between the task efficiency index data and the initial baseline throughout the task cycle.

[0107] This embodiment constructs a continuously differentiable task load response surface, providing an analytical mathematical representation of the dynamic response. The response gradient feature vector and response hysteresis loop feature data extracted from this surface quantify the changing trend of the task state and the individual's regulatory resilience, respectively. Based on these features, individualized optimization intervals and thresholds are defined, achieving a shift from a unified group standard to a precise individual interval. The introduction of a neural efficiency index provides a physiological dimension for evaluating task performance, distinguishing between improved neural efficiency and simply increased resource input. This embodiment achieves integrated collaboration between refined perception, predictive regulation, and multi-dimensional evaluation in task parameter management.

[0108] In some embodiments, the method further includes:

[0109] During the task execution phase, a real-time state discrimination vector is constructed based on the response gradient feature vector and response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output.

[0110] Based on the deviation between the current task status judgment result and the task parameter optimization range, task parameter update instruction data is generated and output through the task parameter update model.

[0111] In this embodiment, the real-time state discrimination vector is formed by fusing the response gradient feature vector and the response hysteresis loop feature data at the current moment through vector concatenation or weighted summation, and its dimension is determined by the sum of the dimensions of the two. The task state discrimination model internally pre-sets the task parameter optimization interval and task state discrimination threshold set calibrated in the initialization phase. After receiving the real-time state discrimination vector, the model calculates the similarity or distance metric between the vector and the reference point within the optimization interval, such as Euclidean distance, Mahalanobis distance, or cosine similarity, and combines it with the threshold set to classify the levels, outputting the current task state discrimination result. This result can be a classification label corresponding to a state such as insufficient resource mobilization, optimized resource allocation, tight resource allocation, or resource overload.

[0112] The task parameter update model calculates the numerical deviation between the current state label's corresponding state level and the target state level corresponding to the optimization interval. This deviation can be a scalar value used to quantify the distance and direction of the current task state's deviation from the optimal interval. Preferably, the task parameter update model internally has a pre-set adjustment strategy mapping table corresponding to different deviation levels. Based on the calculated deviation level, the model looks up the table or uses preset decision rules. For example, if the deviation level exceeds a positive threshold, an increase instruction is generated; if it exceeds a negative threshold, a decrease instruction is generated. This generates corresponding task parameter update instruction data, which includes specific instructions for increasing, decreasing, or maintaining task parameters, as well as the adjustment step size. This data is then output to the task parameter control interface of the task management system to execute parameter adjustments.

[0113] This embodiment achieves real-time perception and classification of the current task state of an object by constructing a real-time state discrimination vector and using a task state discrimination model for state inference. By quantifying the deviation between the current state and the optimization range through a task parameter update model and generating adjustment instructions accordingly, a real-time closed-loop control link from task state perception to task parameter adjustment is established. This can dynamically adjust task parameters according to the real-time task response state of the object, thereby maintaining the task load within an individualized optimal range and improving the dynamic adaptability of the task execution process.

[0114] Please see Figure 2 In some embodiments, a real-time state discrimination vector is constructed based on the response gradient feature vector and response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output, including:

[0115] S201. Obtain the magnitude data of the response gradient feature vector at the current moment, the positive and negative signs of the gradient data along the time dimension, the deviation data between the current response feature data and the first response curve corresponding to the task load response surface calculation model in the task parameter increment stage, and the parameter adjustment coefficient, and combine them into a four-dimensional real-time state discrimination vector.

[0116] S202. The task state discrimination model pre-stores a first threshold, a second threshold, a third threshold, and a fourth threshold. The first to fourth thresholds are adaptively determined based on the statistical distribution of boundary data of the task parameter optimization interval and response lag loop characteristic data. Among them, the first threshold corresponds to the deviation boundary between the resource underutilization state and the resource optimization allocation state, the second threshold corresponds to the deviation boundary between the resource optimization allocation state and the resource shortage allocation state, the third threshold corresponds to the positive and negative sign conversion auxiliary judgment value of the time gradient data, and the fourth threshold corresponds to the deviation boundary of the resource overload state after weighted correction by the parameter adjustment coefficient.

[0117] S203. The task state discrimination model calculates the Mahalanobis distance between the statistical distribution center of the four-dimensional real-time state discrimination vector and the response gradient feature vector modulus data corresponding to each task parameter coordinate point during the step-by-step difficulty calibration task within the task parameter optimization interval, and uses it as the current state deviation index.

[0118] S204. The task status discrimination model compares the current status deviation index with the first threshold. If the current status deviation index is less than the first threshold, the current task status is determined to be a state where resources are not fully mobilized.

[0119] If the current state deviation index is greater than or equal to the first threshold and less than or equal to the second threshold, then the current task state is determined to be a resource optimization and allocation state.

[0120] If the current state deviation index is greater than the second threshold, then a comprehensive judgment is made by combining the positive and negative signs of the gradient data along the time dimension with the deviation data:

[0121] S205. When the gradient data along the time dimension is positive and the deviation data is less than the third threshold, it is determined to be a state of tight resource allocation.

[0122] When the gradient data along the time dimension is positive and the deviation data is greater than or equal to the third threshold, a weighted correction is performed in combination with the parameter adjustment coefficient. If the weighted deviation index is greater than the fourth threshold, it is determined to be a resource overload state.

[0123] In step S201, the magnitude of the response gradient feature vector is obtained by performing a square root operation on the sum of the squares of the two components of the vector. In the gradient data along the time dimension, a positive value indicates that the response continuously increases over time, while a negative value indicates that it continuously decreases. The deviation between the current response feature data and the first response curve is obtained by calculating the difference or normalized distance between the current response value and the response value at the same task parameter coordinate point on the first response curve, used to quantify the degree to which the current response deviates from the typical response pattern of the historical increasing phase. The parameter adjustment coefficient, as a quantitative indicator of the adjustment elasticity of the individual task system, is used to weight and correct the deviation index in subsequent judgments. Concatenating the above four components in a fixed order yields the four-dimensional real-time state discrimination vector.

[0124] In step S202, the statistical distribution of the magnitude data of the response gradient feature vector corresponding to each task parameter coordinate point during the step-difficulty calibration task is preferably utilized. Combined with the local contribution value of the response hysteresis loop feature data at each parameter point, the distribution characteristics of these statistics are analyzed, such as calculating the mean, standard deviation, or specific percentiles, to determine each threshold. The first threshold can be the mean of the magnitude data of the response gradient feature vector at the lower boundary of the task parameter optimization interval multiplied by a preset scaling factor less than 1; the second threshold can be the mean of the magnitude data at the upper boundary of the optimization interval multiplied by a preset scaling factor greater than 1; the third threshold can be the median of the absolute values ​​of all time gradient data during the calibration process multiplied by a preset scaling factor; the fourth threshold can be the median of all deviation data during the calibration process multiplied by a preset scaling factor and then weighted by a parameter adjustment coefficient.

[0125] In step S203, the task state discrimination model first reads the response gradient feature vector magnitude data corresponding to the coordinate points of each task parameter within the optimization interval during the step-difficulty calibration task. It then calculates the mean vector and covariance matrix of these magnitude data, using them as the statistical distribution center of the typical state distribution within the optimization interval. The Mahalanobis distance between the current four-dimensional real-time state discrimination vector and the coordinate vector of this statistical distribution center is calculated. By introducing the inverse of the covariance matrix to eliminate the dimensional differences and consider the correlation between dimensions, the model more accurately measures the degree to which the current state deviates from the typical state distribution within the optimization interval. This distance is the current state deviation index.

[0126] In step S204, the task state discrimination model compares the current state deviation index with a first threshold and a second threshold sequentially. If the current state deviation index is less than the first threshold, it indicates that the current state deviates very little from the optimal allocation state, the response change is weak, and task resources are not fully mobilized. If the current state deviation index is greater than or equal to the first threshold and less than or equal to the second threshold, the current state is within the typical fluctuation range of the optimal allocation state, and task resources are appropriately allocated. If the current state deviation index is greater than the second threshold, it indicates that the current state has significantly deviated from the optimal allocation range, and it is necessary to further distinguish whether it is resource shortage or resource overload.

[0127] In step S205, when the gradient data along the time dimension is positive, it indicates that the response is continuously increasing over time and the task load is accumulating. If the deviation data is less than the third threshold at this time, it indicates that although the response is increasing, the degree of deviation from the typical response pattern of the historical increasing phase is still within an acceptable range, and it is determined to be a state of tight resource allocation.

[0128] If the deviation data is greater than or equal to the third threshold, it indicates that the current response has significantly deviated from the response pattern of the historical increasing phase. At this time, a parameter adjustment coefficient is introduced to weight and correct the deviation index of the current state. The weighting method can be to multiply the deviation index by the reciprocal of the parameter adjustment coefficient or divide it by the coefficient, so that individuals with poor adjustment elasticity (small parameter adjustment coefficient) receive a larger weighted deviation index. If the weighted deviation index is greater than the fourth threshold, it indicates that the task system is in an overloaded state and immediate intervention is required.

[0129] This embodiment constructs a four-dimensional real-time state discrimination vector, integrating gradient magnitude, temporal gradient sign, deviation data, and adjustment coefficient into a unified state representation; it adaptively determines four thresholds based on the statistical distribution of individual calibration data, thus achieving individualization of the state judgment criteria; it introduces Mahalanobis distance to calculate the degree of deviation between the current state and the typical distribution of the optimization interval, and uses a multi-level decision tree structure combined with a weighted correction mechanism for state judgment, achieving fine differentiation of four task states: insufficient resource mobilization, optimized resource allocation, tight resource allocation, and resource overload.

[0130] Please see Figure 3 In some embodiments, based on the deviation between the current task state determination result and the task parameter optimization range, task parameter update instruction data is generated and output through a task parameter update model, including:

[0131] S301. The task parameter update model reads the current task status judgment result. The current task status judgment result is one of the following: resource underutilization, resource optimization, resource shortage, or resource overload.

[0132] S302. If the current task status is determined to be a state where resources are not fully mobilized and the duration exceeds the first preset duration, the task parameter update model generates task parameter improvement instruction data.

[0133] If the current task status is determined to be resource optimization and allocation status, the task parameter update model generates task parameter maintenance instruction data.

[0134] If the current task status is determined to be a resource shortage and the duration exceeds the second preset duration, the task parameter update model generates task parameter reduction instruction data.

[0135] If the current task status is determined to be resource overload, the task parameter update model generates emergency task parameter reduction instruction data and attaches pause instruction data.

[0136] S303. When generating task parameter increase instruction data, task parameter decrease instruction data, task parameter emergency decrease instruction data, and pause instruction data, the task parameter update model adopts a gradual change strategy, adjusting the task parameters in a preset step size unit to avoid the impact of sudden changes in task parameters on the task status.

[0137] S304. After completing each behavior statistics unit, the task parameter update model reads the object behavior performance index data corresponding to the behavior statistics unit. If the object behavior performance index data is higher than the first behavior threshold, task parameter improvement correction instruction data is generated. If the object behavior performance index data is lower than the second behavior threshold, task parameter reduction correction instruction data is generated. The weight of task parameter improvement correction instruction data and task parameter reduction correction instruction data is lower than the weight of task parameter improvement instruction data and task parameter reduction instruction data. Furthermore, task parameter improvement instruction data and task parameter reduction instruction data are neural feedback driven regulation, while task parameter improvement correction instruction data and task parameter reduction correction instruction data are behavior feedback assisted correction.

[0138] S305. The task parameter update model records the number of task parameter adjustments within the most recent third preset time period. If the number of task parameter adjustments exceeds the first adjustment frequency threshold, a damping mechanism is triggered to relax the trigger threshold of subsequent adjustments by a preset ratio. If the number of task parameter adjustments exceeds the second adjustment frequency threshold, task parameter adjustments are paused for a fourth preset time period.

[0139] S306. The task parameter update model outputs the final task parameter update instruction data, processed by the damping mechanism, to the task parameter control interface of the task management system.

[0140] In steps S301 to S302, the current task status determination result is stored in the form of a preset status code or classification label. Different code values ​​correspond to one of the following states: insufficient resource mobilization, optimized resource allocation, tight resource allocation, and resource overload. The task parameter update model executes the corresponding instruction generation logic based on the status category of the current task status determination result and the duration of that status. The first and second preset durations are determined based on the statistical characteristics of individual state transition time in the stepped difficulty calibration task, or by using group experience values ​​as initial settings. Task parameter increase instruction data includes the increase step size and target parameter identifier; task parameter decrease instruction data includes the decrease step size and target parameter identifier; task parameter emergency decrease instruction data includes a larger step size than the normal decrease; and pause instruction data includes a pause duration parameter.

[0141] In step S303, the gradual change strategy requires that each adjustment only changes a preset step size unit. This step size unit is preset according to the type of task parameter; for example, for discrete parameters, it increases or decreases by one level each time, while for continuous parameters, it increases or decreases by a fixed value each time. The step size remains consistent during multiple consecutive adjustments, ensuring that the task parameter changes smoothly in a single direction.

[0142] In step S304, the behavior statistics unit is divided into fixed number of trials or fixed duration. After each unit is completed, the corresponding object behavior performance index data is read. The first behavior threshold and the second behavior threshold are determined based on the statistical distribution of the individual's behavior performance within the optimization interval. The weight of task parameter improvement correction instruction data and task parameter reduction correction instruction data is lower than that of neural feedback driven instruction data. Specifically, the adjustment step size of the correction instruction is set to half the step size of the neural feedback driven instruction, or the correction instruction only takes effect when the neural feedback driven instruction is not triggered.

[0143] In step S305, the third preset duration is set as a fixed time window, used to count the number of times the task parameters are adjusted within this window. The first adjustment frequency threshold is set as the theoretically maximum number of stable adjustments allowed within this time window, for example, calculated based on the window length divided by the minimum adjustment interval. When the number of adjustments exceeds this threshold, it indicates that the adjustment frequency is too high, and the damping mechanism relaxes the trigger threshold for subsequent adjustments by a preset ratio. The second adjustment frequency threshold is set as a multiple of the first adjustment frequency threshold. When the number of adjustments exceeds the second adjustment frequency threshold, it indicates that the adjustment frequency is too high, and the adjustment needs to be completely paused for a fourth preset duration to stabilize the task state of the object. The difference between the first and second adjustment frequency thresholds allows the damping mechanism to respond to adjustment frequency anomalies of different severity in stages.

[0144] This embodiment establishes a multi-level instruction generation logic based on state category and duration, enabling differentiated responses to four states: insufficient resource mobilization, optimized resource allocation, resource strain, and resource overload. A gradual change strategy is introduced to ensure smooth parameter adjustment and avoid abrupt shocks. A behavioral feedback-assisted correction mechanism is employed to supplement and correct neural feedback-driven adjustments with lower weights, forming a dual-mode closed-loop control. A damping mechanism limits high-frequency adjustments to prevent parameter oscillations. This embodiment achieves stable, smooth, and adaptive adjustment of task parameters.

[0145] Please see Figure 4 In some embodiments, based on the frequency band energy distribution attributes, historical task parameter sequences, and corresponding timestamps in the response feature data, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function, including:

[0146] S401. Extract energy distribution attribute data of the target frequency band from the response feature data. The target frequency band is at least one of the θ band, α band, and low β band.

[0147] S402. Map each task parameter value in the historical task parameter sequence to the discretized task parameter coordinate axis, and map the corresponding timestamp to the continuous time coordinate axis. Using the energy distribution attribute data of the target frequency band as the response value, construct a set of discrete sampling points in the three-dimensional data space. Each sampling point in the set of discrete sampling points contains task parameter coordinates, time coordinates and energy response value.

[0148] S403. Perform bicubic spline interpolation kernel function operation on the discrete sampling point set. The bicubic spline interpolation kernel function generates a continuously differentiable interpolation surface covering the entire two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis by solving the second-order partial derivative continuity constraint equations at each sampling point.

[0149] S404. The continuously differentiable interpolation surface is used as the task load response surface calculation model corresponding to the target frequency band. The task load response surface calculation model has a first-order continuous partial derivative at any point in the two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis, which makes it possible to calculate the gradient along the task parameter dimension and along the time dimension.

[0150] S405. If the target frequency band is multiple frequency bands, then for each frequency band, the above-mentioned discrete sampling point set construction and bicubic spline interpolation kernel function operation are performed independently to generate the task load response surface calculation model corresponding to each frequency band, which is used as a set of multi-frequency band task load response surface calculation models.

[0151] In step S401, the energy values ​​corresponding to the target frequency band are filtered from the multi-band energy distribution feature vector sequence in the response feature data according to the preset frequency band index. The selection of the target frequency band is determined according to the task type. For example, tasks involving working memory preferentially select the θ frequency band, tasks involving attention maintenance preferentially select the α frequency band, and tasks involving execution functions preferentially select the low β frequency band. Multiple frequency bands can also be selected simultaneously to obtain more comprehensive task response information.

[0152] In step S402, each parameter value in the historical task parameter sequence is mapped to an integer coordinate point on the discretized task parameter coordinate axis in ascending order. For example, N values ​​1 to 6 in the N-back task are mapped to coordinate points 1 to 6, or the stimulus presentation duration is mapped to equally spaced integer coordinates with a fixed step size. The corresponding timestamp is converted into consecutive seconds or minutes as the time coordinate, with the task start time as the origin. Using the energy distribution attribute data of the target frequency band as the response value, each sampling point consists of a triplet of task parameter coordinates, time coordinates, and energy response value. All sampling points form a sparsely distributed set of discrete sampling points in a two-dimensional domain.

[0153] In step S403, the bicubic spline interpolation kernel function operation includes: dividing the two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis with discrete sampling points as nodes, and each grid cell is defined by four adjacent nodes; defining a bicubic polynomial containing sixteen undetermined coefficients on each grid cell; establishing continuity constraints at each grid node, requiring that the function values, first-order partial derivatives, and second-order mixed partial derivatives of the polynomials of adjacent grid cells are equal at that node, forming a linear constraint equation system, and supplementing natural boundary conditions or clamping boundary conditions at the boundary of the two-dimensional domain to make the equation system have a unique solution; obtaining the polynomial coefficients of all grid cells by solving the linear equation system, and these coefficients together define a continuously differentiable interpolation surface covering the entire two-dimensional domain.

[0154] Furthermore, the detailed calculation process involved in step S403 is as follows:

[0155] (1) Standard form of bicubic polynomial

[0156] In the coordinate axis of the task parameters and time axis Within the defined two-dimensional domain, assume there exists a region consisting of four adjacent discrete sampling points. , , , Define a rectangular network element, within which a bicubic polynomial is defined. It is an interpolation function, and its standard form contains 16 undetermined coefficients. The expression is as follows:

[0157] ;

[0158] (2) Specific form of the continuity constraint equation system

[0159] To ensure the smooth continuity of the entire interpolation surface at the grid nodes, the function value, first-order partial derivative, and second-order mixed partial derivative at each grid node must satisfy the continuity condition. For any internal grid node... The polynomials corresponding to its four adjacent grid cells (top left, top right, bottom left, bottom right) must satisfy the following set of constraint equations:

[0160] Function value continuity constraint: The function values ​​of four adjacent polynomials at a node are equal, and all are equal to the energy response value at that sampling point. .

[0161]

[0162] in, Indicates the first A bicubic polynomial of adjacent grid cells.

[0163] First-order partial derivative continuity constraint: The four adjacent polynomials at this node have the following relationship: and The first-order partial derivatives are equal:

[0164] ;

[0165] Second-order mixed partial derivative continuity constraint: The second-order mixed partial derivatives of the four adjacent polynomials at this node are equal.

[0166] ;

[0167] The above constraints can be used to establish a set of linear equations at each internal grid node. By simultaneously solving the constraint equations for all grid nodes, a set of undetermined coefficients for all bicubic polynomials can be formed. A large system of linear equations.

[0168] (3) Methods for handling boundary conditions

[0169] To ensure a unique solution to the above linear equations, boundary conditions need to be added at the boundaries of the two-dimensional domain. This embodiment uses a natural boundary condition, which assumes that the second-order partial derivatives of the interpolation surface are zero at the boundaries of the domain. Specifically, for the boundary... or Additional constraints for the nodes:

[0170] ;

[0171] Specifically, for those located at the boundary or Additional constraints for the nodes:

[0172] ;

[0173] For the four corner points, constraints are simultaneously applied such that the second-order partial derivatives in both directions are zero. Natural boundary conditions ensure a smooth transition of the surface at the boundaries and do not rely on prior knowledge, making them suitable for scenarios where there is no explicit boundary derivative information. In other embodiments, if the trend at the boundaries is known, a clamped boundary condition can be used, specifying the exact value of the first-order partial derivative at the boundaries.

[0174] (4) Algorithm for solving linear equation systems

[0175] The linear equation system constituted by the aforementioned continuity constraints and boundary conditions has a coefficient matrix that is a large, sparse, banded matrix. To efficiently solve this system of equations, this embodiment employs a two-dimensional generalization of the Thomas Algorithm. The specific steps are as follows:

[0176] Matrix rearrangement: rearranging the 16 coefficients to be solved Arrange them into a column vector and reorganize the coefficient matrix according to the constraint equations.

[0177] Step-by-step solution: Utilizing the sparse banded nature of the coefficient matrix, first, along the task parameter coordinate axes... The direction is solved using a catching-up method to obtain an intermediate result; then, along the time axis... The direction is then used to solve the problem a second time using the chasing method, which finally yields the bicubic polynomial coefficients of all grid elements.

[0178] By following the above steps, the polynomial coefficients of all grid cells can be solved, thereby obtaining a continuously differentiable interpolation surface covering the entire two-dimensional domain with continuous first-order partial derivatives, which is the training load response surface calculation model.

[0179] In steps S404 and S405, the task load response surface calculation model is stored in piecewise polynomial form. When it is necessary to calculate the response value or partial derivative at any coordinate point, the result can be quickly obtained by locating the grid cell where the point is located and calling the evaluation function of the corresponding polynomial, without having to repeatedly perform interpolation calculations. The task load response surface calculation models for all frequency bands are organized into a multi-frequency band task load response surface calculation model set, with the frequency band identifier as the key and the model data structure as the value, for subsequent steps to call as needed.

[0180] This embodiment extracts the energy distribution attributes of the target frequency band from the response feature data, transforming discrete task execution process data into a set of sampling points with a clear spatial structure. It then employs a bicubic spline interpolation algorithm to construct a continuously differentiable task load response surface calculation model, enabling the dynamic characteristics of the response to change with task parameters and time to be accurately described in analytical function form. This provides a mathematical foundation for subsequent gradient calculation and hysteresis loop feature extraction. The strategy of constructing multiple frequency bands independently preserves the independence of the task response in each frequency band, providing data support for multi-dimensional task status analysis.

[0181] In some embodiments, parameter gradient data and time gradient data are obtained along the task parameter dimension and time dimension of the task load response surface calculation model, respectively, and fused to obtain a response gradient feature vector, including:

[0182] Read the task load response surface calculation model. The task load response surface calculation model is a continuously differentiable interpolation surface function in a two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis.

[0183] Locate the task parameter coordinates at the current moment on the task load response surface calculation model, and obtain the first-order partial derivative of the task load response surface calculation model with respect to the task parameter coordinates to obtain the parameter gradient data at the current moment. The parameter gradient data represents the sensitivity of the current response characteristic data to changes in task parameters.

[0184] Locate the time coordinate point corresponding to the current moment on the task load response surface calculation model, obtain the first-order partial derivative of the task load response surface calculation model with respect to the time coordinate, and obtain the time gradient data of the current moment. The time gradient data represents the temporal change rate of the current response characteristic data.

[0185] The parameter gradient data and the time gradient data are used as two components of a two-dimensional vector. The magnitude of the two-dimensional vector is calculated, and the magnitude represents the overall change intensity of the current response feature data.

[0186] The direction angle of the two-dimensional vector is calculated by the arctangent function of the ratio of the temporal gradient data to the parameter gradient data. The direction angle represents the dominant dimension of the current response feature data change. When the direction angle is close to zero, it indicates that the change is dominated by the change of task parameters. When the direction angle is close to ninety degrees, it indicates that the change is dominated by the time process.

[0187] The magnitude and orientation angle of a two-dimensional vector are combined to form the response gradient feature vector.

[0188] In this embodiment, the data structure of the task load response surface calculation model includes a grid node coordinate array, a bicubic polynomial coefficient matrix for each grid cell, and a domain range parameter. When reading this model, this data is loaded into the computing environment and a fast positioning index is constructed, enabling the grid cell to be quickly determined and the corresponding polynomial expression to be obtained for any given coordinate point.

[0189] When locating the task parameter coordinates at the current moment, the coordinates corresponding to that moment are found in the two-dimensional domain of the model based on the actual parameter values ​​and timestamp of the current task. The first-order partial derivative of the polynomial expression of the grid cell to which the coordinates belong is obtained with respect to the task parameter coordinates. Substituting the coordinate values ​​into the partial derivative expression yields the parameter gradient data. The larger the absolute value of the parameter gradient data, the more significant the change in response can be caused by fine-tuning the task parameters.

[0190] The method for obtaining time gradient data is similar to that for parametric gradient data. The first-order partial derivative of the polynomial with respect to the time coordinate is obtained for the same grid cell and substituted into the coordinate value for calculation. The positive or negative sign of the time gradient data indicates whether the response increases or decreases with time, and the magnitude of the absolute value indicates the rate of change.

[0191] The magnitude of a two-dimensional vector is a non-negative scalar value, obtained by squaring the two components, summing them, and then taking the square root. The direction angle is calculated using the four-quadrant arctangent function. The quadrant of the angle is determined by the ratio of the time gradient data to the parameter gradient data and the signs of the two components. This outputs the direction angle value in the range of -180 degrees to +180 degrees or 0 degrees to 360 degrees. When it is close to zero degrees, it indicates that the response change is mainly driven by the change of task parameters. When it is close to ±90 degrees, it indicates that the change is mainly driven by the time process. When it is in between, it indicates that both factors work together.

[0192] The response gradient feature vector, in polar coordinates, simultaneously encodes the intensity and dominant dimension of the response change, providing a feature input for subsequent state discrimination that includes information on both the magnitude and direction of the change.

[0193] This embodiment utilizes the continuous differentiability of the task load response surface calculation model to quickly obtain the first-order partial derivatives along the task parameter dimension and time dimension at any coordinate point, and fuses them into a response gradient feature vector represented by the magnitude and direction angle. This achieves a feature upgrade from static response values ​​to dynamic change trends, providing a quantitative indicator with clear physical meaning for finely perceiving changes in task state.

[0194] Please see Figure 5In some embodiments, based on the asymmetric deviation characteristics of the task load response surface calculation model between the increasing and decreasing stages of the task parameters, response hysteresis loop feature data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value, including:

[0195] S501. Extract the first response curve corresponding to the increasing stage of task parameters from the task load response surface calculation model. The first response curve describes the response value of the response characteristic data corresponding to each task parameter coordinate point as the task parameters gradually increase from the minimum value to the maximum value.

[0196] S502. Extract the second response curve corresponding to the decreasing stage of the task parameters from the task load response surface calculation model. The second response curve describes the response value of the response characteristic data corresponding to each task parameter coordinate point as the task parameters gradually decrease from the maximum value to the minimum value.

[0197] S503. Plot the first response curve and the second response curve in the same coordinate system. The coordinate system has the task parameter coordinates as the horizontal axis and the response value of the response feature data as the vertical axis. The first response curve and the second response curve have a response value deviation at the same task parameter coordinate point. The response value of the first response curve is higher than the response value of the second response curve, forming a closed region.

[0198] S504. The response value deviation between the first response curve and the second response curve is integrated along the task parameter coordinate axis by the response hysteresis loop calculation model to obtain the closed region metric value. The closed region metric value represents the degree of asymmetry of the response path of an individual task system during the process of increasing and decreasing task parameters.

[0199] S505. The closed region metric is normalized by the parameter adjustment coefficient calculation model. The normalized closed region metric is substituted into the inverse proportional function to obtain the parameter adjustment coefficient. The parameter adjustment coefficient is inversely proportional to the closed region metric. The higher the parameter adjustment coefficient, the better the adjustment elasticity of the individual task system to changes in task parameters.

[0200] In step S501, the task parameter increment phase corresponds to the complete process of the task parameters gradually increasing from the minimum to the maximum value in the step-difficulty calibration task. Its time interval is defined by the start and end times of the calibration task. When extracting response value slices within this time interval from the task load response surface calculation model, each task parameter coordinate point is fixed along the time axis. The response value sequence at that point, changing with task time, is read. Then, the mean or median of this sequence within the increment phase time interval is taken as the response value corresponding to that task parameter coordinate point, thus obtaining a first response curve with the task parameter coordinates as the independent variable and the response value as the dependent variable.

[0201] In step S502, the task parameter decreasing phase corresponds to the complete process of the task parameters gradually decreasing from their maximum value to their minimum value. Its time interval is also defined by the start and end times of the calibration task. Using the same extraction method as the first response curve, slices of response values ​​within the decreasing phase time interval are extracted from the task load response surface calculation model. The mean or median of the response values ​​at each coordinate point is extracted along the task parameter coordinate axis, and arranged in descending order of the task parameter coordinates to form the second response curve.

[0202] In step S503, the first response curve and the second response curve are plotted on a coordinate system with the task parameter coordinates as the horizontal axis and the response values ​​of the response feature data as the vertical axis. Since the task system is in a state of load accumulation and task resources are continuously invested during the increasing phase, and the task system is in a state of load release and task resource recovery is delayed during the decreasing phase, the response value of the first response curve at the same task parameter coordinate point is higher than the response value of the second response curve. The two curves form a closed region in the coordinate system. The size of this closed region intuitively reflects the degree of asymmetry of the task response path.

[0203] In step S504, when the response hysteresis loop calculation model performs integration on the deviation of the response value between the first response curve and the second response curve along the task parameter coordinate axis, the task parameter coordinate axis is discretized into equally spaced sampling points. At each sampling point, the difference between the response values ​​of the first response curve and the second response curve is calculated. Then, numerical integration methods such as the trapezoidal method or Simpson's method are used to accumulate and integrate the difference of all sampling points along the task parameter coordinate axis. The obtained integral value is the closed region metric. The larger the metric value, the more significant the difference between the response paths in the increasing and decreasing stages, and the stronger the inertia of the individual task system adjustment.

[0204] In step S505, the parameter adjustment coefficient calculation model first divides the closed region measurement value by a preset normalized benchmark value, which can be determined based on the upper bound of population statistics or the theoretical maximum possible area, to obtain a normalized closed region measurement value between 0 and 1. The normalized value is then substituted into an inverse proportional function, such as the parameter adjustment coefficient being equal to 1 divided by 1 plus the normalized closed region measurement value, or other monotonically decreasing functions such as a negative exponential function, so that the larger the closed region measurement value, the smaller the parameter adjustment coefficient. The closer the parameter adjustment coefficient is to 1, the better the adjustment elasticity of the individual task system, and the closer it is to 0, the worse the adjustment elasticity.

[0205] This embodiment extracts the response curves of the increasing and decreasing stages from the task load response surface calculation model, quantifies the asymmetric deviation between the two curves using the response hysteresis loop calculation model, and solves the parameter adjustment coefficients based on the closed region metric value through an inverse proportional function. This transforms the abstract physiological characteristic of the individual task system's adjustment elasticity into a quantifiable numerical indicator, providing key parameters reflecting individual differences for subsequent individualized threshold calibration and state discrimination.

[0206] In some embodiments, based on the response gradient feature vector, response hysteresis loop feature data, and parameter adjustment coefficients, the task parameter optimization range and task state discrimination threshold set are calibrated through a task state discrimination model during a stepped difficulty calibration task, including:

[0207] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to each task parameter coordinate point during the step-difficulty calibration task, as well as the local contribution value of the response hysteresis loop feature data corresponding to each task parameter coordinate point.

[0208] The task state discrimination model marks the task parameter coordinate point corresponding to the first time the magnitude data of the response gradient feature vector reaches the preset activation threshold as the lower boundary of the task parameter optimization interval.

[0209] The task state discrimination model marks the maximum task parameter coordinate point corresponding to the condition that the magnitude of the response gradient feature vector does not exceed the preset upper limit threshold and the local contribution value of the response hysteresis loop feature data does not exceed the preset deviation threshold as the upper boundary of the task parameter optimization interval.

[0210] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the lower boundary of the task parameter optimization interval, multiplies it by the first preset scaling factor, and obtains the first threshold.

[0211] The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the upper boundary of the task parameter optimization interval, multiplies it by the second preset scaling factor, and obtains the second threshold.

[0212] The task status discrimination model reads the median of the absolute values ​​of all time gradient data during the step-difficulty calibration task, multiplies it by the third preset scaling factor, and obtains the third threshold.

[0213] The task status discrimination model reads the median of all deviation data during the step-difficulty calibration task, multiplies it by the fourth preset proportional coefficient, and obtains the fourth threshold.

[0214] The task state discrimination model reads the parameter adjustment coefficients and multiplies the first threshold, second threshold, third threshold and fourth threshold by the weighting factor corresponding to the parameter adjustment coefficients to obtain the task state discrimination threshold set after individualized correction. Among them, the first threshold and second threshold are used for deviation index judgment, the third threshold is used for auxiliary judgment of positive and negative signs of time gradient data, and the fourth threshold is used for overload state deviation judgment after weighting by parameter adjustment coefficients.

[0215] In this embodiment, the local contribution value of the response hysteresis loop characteristic data corresponding to each task parameter coordinate point during the stepped difficulty calibration task is obtained by calculating the absolute value of the difference between the response values ​​of the first response curve and the second response curve at that point. The larger the absolute value, the greater the contribution of the parameter point to the hysteresis loop closure region, that is, the more significant the asymmetry of the task response at that difficulty level.

[0216] The preset activation threshold is determined based on the resting-state baseline data. For example, it is twice the mean magnitude of the response gradient feature vector in the resting state. When the magnitude data first exceeds this threshold, it indicates that task resources are being mobilized, and this point is marked as the lower boundary of the task parameter optimization interval. The preset upper limit threshold is determined based on the statistical upper bound of the group or the historical overload data of an individual. The preset deviation threshold is determined based on the statistical distribution of the local contribution values ​​corresponding to all task parameter coordinate points. The maximum task parameter coordinate point corresponding to the moment when the magnitude data does not exceed the preset upper limit threshold and the local contribution value does not exceed the preset deviation threshold is marked as the upper boundary of the task parameter optimization interval.

[0217] Preferably, the first preset proportional coefficient is set to a value less than 1, so that the first threshold obtained by multiplying the modulus data at the lower boundary by this coefficient is lower than the modulus value at the lower boundary, which is used to determine the state of insufficient resource allocation; the second preset proportional coefficient is set to a value greater than 1, so that the second threshold obtained by multiplying the modulus data at the upper boundary by this coefficient is higher than the modulus value at the upper boundary, which is used to determine the state of tight resource allocation. The third and fourth preset proportional coefficients are set based on empirical values ​​and are used to convert the median of the absolute value of the time gradient data and the median of the deviation data into the corresponding third and fourth thresholds.

[0218] The weighting factor corresponding to the parameter adjustment coefficient is either the parameter adjustment coefficient itself or a monotonically increasing function of that coefficient. Individuals with good adjustment flexibility receive a weighting factor close to 1, and the thresholds remain essentially unchanged; individuals with poor adjustment flexibility receive a weighting factor less than 1, and the thresholds decrease accordingly, making the state judgment criteria more stringent and the system more sensitive to task load deviations. In the set of task state discrimination thresholds after individualized correction, the first and second thresholds are used for deviation index judgment, the third threshold is used for auxiliary judgment of the positive and negative signs of time gradient data, and the fourth threshold is used for overload state deviation judgment after weighting by the parameter adjustment coefficient.

[0219] This embodiment calibrates individualized task parameter optimization ranges based on tiered difficulty calibration task data, and calculates the individualized task state discrimination threshold set using preset proportional coefficients and parameter adjustment coefficient weighting factors. This achieves a leap from a unified standard for the group to a precise range for individuals, enabling subsequent real-time state discrimination to adapt to the task response characteristics of different individuals, thereby improving the accuracy of state discrimination and individual adaptability.

[0220] In some embodiments, based on the ratio of spectral resource consumption data between object behavior performance index data and response feature data, the task efficiency index data is solved using a neural efficiency calculation model, including:

[0221] After each behavior statistics unit ends, read the object behavior response log data corresponding to the behavior statistics unit, and extract the accuracy data and average reaction time data from the object behavior response log data;

[0222] Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline reaction time data from the historical response characteristic data;

[0223] The accuracy data is divided by the ratio of the average reaction time data to the resting baseline reaction time data to obtain the behavioral performance index data. The behavioral performance index data comprehensively reflects the task completion quality and response speed of the subject within the behavioral statistical unit.

[0224] Read the response feature data within the time window corresponding to the behavior statistics unit, and extract the power increase data of the θ band and the power increase data of the low β band from the response feature data;

[0225] Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline power data from the historical response characteristic data;

[0226] The summation of the power increase data in the θ band and the power increase data in the low β band is divided by the resting baseline power data to obtain the spectrum resource consumption data. The spectrum resource consumption data represents the total amount of neural resources consumed by the object to complete the task within the statistical unit of this behavior.

[0227] By dividing the behavioral performance index data by the spectrum resource consumption data through the neural efficiency calculation model, the task efficiency index data is obtained. The task efficiency index data represents the level of behavioral performance supported by a unit of neural resource consumption.

[0228] In this embodiment, the behavioral statistics unit is divided based on a fixed number of trials, with each unit containing a preset number of consecutive trials. The accuracy data within this unit is extracted from the object's behavioral response log data by dividing the number of correct trials by the total number of trials. The average reaction time data is obtained by calculating the average reaction time of all correct trials within the unit. The resting-state baseline phase is the period before the task begins when the object is in a relaxed state. During this period, the object's baseline reaction level is obtained through a simple reaction time test, and the resting-state baseline reaction time data is the average of multiple tests. The behavioral performance index data is obtained by dividing the accuracy data by the ratio of the average reaction time data to the resting-state baseline reaction time data. This index reflects both the change in task completion quality and response speed relative to the baseline.

[0229] The response characteristic data within the corresponding time window of the behavioral statistics unit is taken from the time interval of that unit. The θ-band power increase data is obtained by subtracting the average θ-band power during the resting-state baseline phase from the average θ-band power during the time window. The low-β-band power increase data is calculated in the same way. The resting-state baseline power data is the average power of each band during the resting-state baseline phase. The spectrum resource consumption data is obtained by summing the θ-band power increase data and the low-β-band power increase data and then dividing by the resting-state baseline power data. This relative value eliminates the differences in baseline levels between individuals, making the task resource consumption of different individuals comparable.

[0230] The neural efficiency calculation model divides the behavioral performance index data by the spectrum resource consumption data to obtain the task efficiency index data. The higher the index, the better the behavioral performance supported by the unit task resource consumption, that is, the higher the task efficiency of the object.

[0231] This embodiment introduces a task efficiency index, which calculates the ratio between behavioral performance and task resource consumption. This provides a quantitative indicator for the task management system to distinguish between efficient execution and inefficient effort, so that task effect evaluation is no longer limited to the improvement of behavioral indicators, but can evaluate the true effect of the task from the physiological dimension of task efficiency improvement.

[0232] In some embodiments, task performance evaluation data is generated and output using a task performance evaluation model based on its temporal variation trend, including:

[0233] Read the task efficiency index data corresponding to all behavioral statistical units within the task cycle, and construct a time series sequence of the task efficiency index.

[0234] Read the average value of a preset number of task efficiency index data at the end of the task cycle, and use it as the final task efficiency index data.

[0235] Read the average value of a preset number of task efficiency index data at the beginning of the task cycle, and use it as the initial task efficiency index data.

[0236] Divide the difference between the final task efficiency index data and the initial task efficiency index data by the initial task efficiency index data to obtain the task efficiency index improvement rate data. The task efficiency index improvement rate data represents the extent of improvement in the neural efficiency of the object throughout the entire task cycle.

[0237] The ratio of the standard deviation to the mean of the task efficiency index time series is calculated to obtain the coefficient of variation of the task efficiency index. The coefficient of variation of the task efficiency index characterizes the stability of the neural efficiency of the object during the task.

[0238] Read the parameter adjustment coefficient data calibrated at the end of the task cycle as the final parameter adjustment coefficient data, and read the parameter adjustment coefficient data calibrated at the beginning of the task cycle as the initial parameter adjustment coefficient data. Divide the difference between the final parameter adjustment coefficient data and the initial parameter adjustment coefficient data by the initial parameter adjustment coefficient data to obtain the parameter adjustment coefficient improvement rate data. The parameter adjustment coefficient improvement rate data characterizes the change range of the object's task adjustment elasticity within the task cycle.

[0239] The task effectiveness evaluation model combines the task efficiency index improvement rate data, the task efficiency index variation coefficient data, and the parameter adjustment coefficient improvement rate data into multidimensional task effectiveness evaluation data, and outputs the multidimensional task effectiveness evaluation data to the report generation interface of the task management system.

[0240] In this embodiment, the task cycle is set as a complete task session or a task phase spanning multiple days. The task efficiency index time series is obtained by arranging the task efficiency index data corresponding to each behavioral statistical unit within the task cycle in chronological order, with each data point accompanied by a corresponding timestamp or unit number, forming a time series reflecting the evolution of task efficiency over time.

[0241] The average value of the task efficiency index data at the end of the task cycle is calculated using data from a fixed number of behavioral statistical units at the end of the task cycle. The preset number is dynamically adjusted according to the total length of the task cycle to ensure that the data at the end of the task can stably reflect the task efficiency level at the end of the task. The average value of the task efficiency index data at the beginning of the task cycle is calculated using data from a fixed number of behavioral statistical units at the beginning of the task cycle. The method for determining the preset number is similar to that at the end of the task cycle, and it is used to establish an initial baseline level for task efficiency.

[0242] The task efficiency index improvement rate is obtained by dividing the difference between the final task efficiency index and the initial task efficiency index by the initial task efficiency index. A positive value indicates that task efficiency has improved throughout the task cycle, while a negative value indicates a decrease. The absolute value reflects the magnitude of improvement or decline. The task efficiency index coefficient of variation is obtained by calculating the ratio of the standard deviation to the mean of the task efficiency index time series. This ratio eliminates the influence of the mean level on the dispersion. The smaller the coefficient of variation, the more stable the task efficiency is during the task.

[0243] The parameter adjustment coefficient data calibrated at the end of the task cycle and the parameter adjustment coefficient data calibrated at the beginning of the task cycle are respectively taken from the parameter adjustment coefficients calculated by the step-difficulty calibration tasks conducted before and after the start and end of the task. The parameter adjustment coefficient improvement rate data is obtained by dividing the difference between the parameter adjustment coefficient data at the end and the initial parameter adjustment coefficient data by the initial parameter adjustment coefficient data. A positive value indicates that the adjustment elasticity of the object's task system has improved during the task cycle.

[0244] The task effectiveness evaluation model combines the task efficiency index improvement rate data, the task efficiency index variation coefficient data, and the parameter adjustment coefficient improvement rate data into a structured data object. This object also contains information on the improvement magnitude, stability, and changes in the task system's adjustment flexibility of task efficiency. After being output to the task management system's report generation interface, the interface generates a comprehensive visualization report that includes trend charts, numerical summaries, and textual interpretations.

[0245] This embodiment constructs a time series of task efficiency index and calculates its improvement rate, coefficient of variation, and parameter adjustment coefficient improvement rate. It comprehensively evaluates the task effect from three dimensions: the magnitude of task efficiency improvement, stability, and the elasticity of task system adjustment. This provides a multi-dimensional evaluation system for task management that goes beyond traditional behavioral indicators.

[0246] By adopting the above technical solutions, the present invention differs from the prior art and has the following beneficial effects:

[0247] By constructing a continuously differentiable task load response surface calculation model, the dynamic response characteristics of the spectrum are transformed into an analytical mathematical representation. The response gradient feature vector extracted from this model quantifies the changing trend and dominant dimension of the task state. The response hysteresis loop feature data and its derived parameter adjustment coefficient quantify the adjustment elasticity of the individual task system. Together, they provide a systematic basis for the individualized calibration of the task parameter optimization range and the task state discrimination threshold set. Based on the above features, the real-time state discrimination vector, combined with multi-level threshold judgment, realizes the fine distinction between four task states: insufficient resource mobilization, optimized resource allocation, tight resource allocation, and resource overload. The task parameter update instruction data generated accordingly is dominated by neural feedback-driven adjustment and supplemented by behavioral feedback-assisted correction. A gradual change strategy and damping mechanism are introduced to ensure the smoothness and stability of parameter adjustment. The task efficiency index quantifies the level of behavioral performance supported by unit task resource consumption through the ratio of behavioral performance to spectrum resource consumption. Its time-series change trend, combined with the parameter adjustment coefficient improvement rate, generates multi-dimensional task effect evaluation data from three dimensions: task efficiency improvement magnitude, stability, and task system adjustment elasticity. This invention achieves integrated management of task parameters, from refined perception to predictive adjustment and multi-dimensional evaluation, significantly improving the dynamic adaptability of task execution and the systematic depth of effect evaluation.

[0248] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. A method for managing parameters of a cognitive training task based on response data, characterized in that, include: The object training process data record is obtained from the data interface of the task management system. The object training process data record includes historical task parameter sequences, behavior response log data, and response feature data. The response feature data is a multi-band energy distribution feature vector sequence embedded after frequency domain transformation. Based on the frequency band energy distribution attributes in the response feature data, the historical task parameter sequence and the corresponding timestamp, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function. For the task load response surface calculation model, parameter gradient data and time gradient data are obtained along the task parameter dimension and time dimension respectively, and fused to obtain the response gradient feature vector; Based on the asymmetric deviation characteristics of the task load response surface calculation model between the increasing and decreasing stages of the task parameters, the response hysteresis loop feature data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value. Based on the response gradient feature vector, the response hysteresis loop feature data and the parameter adjustment coefficient, the task parameter optimization range and the task state discrimination threshold set are determined by the task state discrimination model during the step-difficulty calibration task. During task execution, based on the ratio of the spectrum resource consumption data between the object behavior performance index data and the response feature data, the task efficiency index data is solved through the neural efficiency calculation model, and task effect evaluation data is generated and output through the task effect evaluation model according to its time series change trend.

2. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, The method further includes: During the task execution phase, a real-time state discrimination vector is constructed based on the response gradient feature vector and the response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output. Based on the deviation between the current task status judgment result and the task parameter optimization range, task parameter update instruction data is generated and output through the task parameter update model.

3. The cognitive training task parameter management method based on response data according to claim 2, characterized in that, A real-time state discrimination vector is constructed based on the response gradient feature vector and the response hysteresis loop feature data at the current moment. This vector is then input into the task state discrimination model for state inference, and the current task state discrimination result is output, including: The modulus data of the response gradient feature vector at the current moment, the sign of the gradient data along the time dimension, the deviation data between the current response feature data and the first response curve corresponding to the task load response surface calculation model in the task parameter increment stage, and the parameter adjustment coefficient are obtained and combined into a four-dimensional real-time state discrimination vector. The task state discrimination model pre-stores a first threshold, a second threshold, a third threshold, and a fourth threshold. The first to fourth thresholds are adaptively determined based on the statistical distribution of the boundary data of the task parameter optimization interval and the response lag loop feature data. The first threshold corresponds to the deviation boundary between the resource underutilization state and the resource optimization allocation state, the second threshold corresponds to the deviation boundary between the resource optimization allocation state and the resource shortage allocation state, the third threshold corresponds to the positive and negative sign conversion auxiliary judgment value of the time gradient data, and the fourth threshold corresponds to the deviation boundary of the resource overload state after weighted correction by the parameter adjustment coefficient. The task state discrimination model calculates the Mahalanobis distance between the four-dimensional real-time state discrimination vector and the statistical distribution center of the magnitude data of the response gradient feature vector corresponding to each task parameter in the step-difficulty calibration task within the task parameter optimization interval, and uses it as the current state deviation index. The task status discrimination model compares the current status deviation index with the first threshold. If the current status deviation index is less than the first threshold, the current task status is determined to be a state where resources are not fully mobilized. If the current state deviation index is greater than or equal to the first threshold and less than or equal to the second threshold, then the current task state is determined to be a resource optimization and allocation state. If the current state deviation index is greater than the second threshold, then a comprehensive judgment is made by combining the positive and negative signs of the gradient data along the time dimension with the deviation data: When the gradient data along the time dimension is positive and the deviation data is less than the third threshold, it is determined to be a state of tight resource allocation. When the gradient data along the time dimension is positive and the deviation data is greater than or equal to the third threshold, a weighted correction is performed in conjunction with the parameter adjustment coefficient. If the weighted deviation index is greater than the fourth threshold, it is determined to be a resource overload state.

4. The cognitive training task parameter management method based on response data according to claim 2, characterized in that, Based on the deviation between the current task state determination result and the task parameter optimization range, task parameter update instruction data is generated and output through the task parameter update model, including: The task parameter update model reads the current task status judgment result, which is one of the following: resource underutilization, resource optimization, resource shortage, or resource overload. If the current task status determination result is that the resources are not fully mobilized and the duration exceeds the first preset duration, the task parameter update model generates task parameter improvement instruction data; If the current task status determination result is a resource optimization and allocation status, the task parameter update model generates task parameter maintenance instruction data; If the current task status determination result is a resource shortage allocation state and the duration exceeds the second preset duration, the task parameter update model generates task parameter reduction instruction data. If the current task status determination result is a resource overload state, the task parameter update model generates task parameter emergency reduction instruction data and attaches pause instruction data; When generating task parameter increase instruction data, task parameter decrease instruction data, task parameter emergency decrease instruction data, and pause instruction data, the task parameter update model adopts a gradual change strategy, adjusting the task parameters in a preset step size unit to avoid the impact of sudden changes in task parameters on the task status. After completing each behavior statistics unit, the task parameter update model reads the object behavior performance index data corresponding to the behavior statistics unit. If the object behavior performance index data is higher than the first behavior threshold, task parameter improvement correction instruction data is generated. If the object behavior performance index data is lower than the second behavior threshold, task parameter reduction correction instruction data is generated. The weights of the task parameter improvement correction instruction data and the task parameter reduction correction instruction data are lower than the weights of the task parameter improvement instruction data and the task parameter reduction instruction data. Furthermore, the task parameter improvement instruction data and the task parameter reduction instruction data are neural feedback driven regulation, and the task parameter improvement correction instruction data and the task parameter reduction correction instruction data are behavior feedback assisted correction. The task parameter update model records the number of times the task parameters are adjusted within the most recent third preset time period. If the number of times the task parameters are adjusted exceeds the first adjustment frequency threshold, a damping mechanism is triggered to relax the trigger threshold of subsequent adjustments by a preset ratio. If the number of times the task parameters are adjusted exceeds the second adjustment frequency threshold, the task parameter adjustment is paused for a fourth preset time period. The task parameter update model outputs the final task parameter update instruction data, processed by the damping mechanism, to the task parameter control interface of the task management system.

5. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, Based on the frequency band energy distribution attributes in the response feature data, the historical task parameter sequence, and the corresponding timestamp markers, a task load response surface calculation model is constructed using a bicubic spline interpolation kernel function, including: Energy distribution attribute data of the target frequency band is extracted from the response feature data, wherein the target frequency band is at least one of the θ band, α band, and low β band. Each task parameter value in the historical task parameter sequence is mapped to a discretized task parameter coordinate axis, and the corresponding timestamp is mapped to a continuous time coordinate axis. The energy distribution attribute data of the target frequency band is used as the response value to construct a set of discrete sampling points in a three-dimensional data space. Each sampling point in the set of discrete sampling points contains task parameter coordinates, time coordinates and energy response value. The set of discrete sampling points is subjected to a bicubic spline interpolation kernel function operation. The bicubic spline interpolation kernel function generates a continuously differentiable interpolation surface covering the entire two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis by solving the second-order partial derivative continuity constraint equations at each sampling point. The continuously differentiable interpolation surface is used as the task load response surface calculation model corresponding to the target frequency band. The task load response surface calculation model has a first-order continuous partial derivative at any point in the two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis, which makes it possible to calculate the gradient along the task parameter dimension and along the time dimension. If the target frequency band is multiple frequency bands, the above-mentioned discrete sampling point set construction and bicubic spline interpolation kernel function operation are performed independently for each frequency band to generate the task load response surface calculation model corresponding to each frequency band, which serves as a set of multi-frequency band task load response surface calculation models.

6. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, The parameter gradient data and time gradient data are obtained along the task parameter dimension and time dimension of the task load response surface calculation model, respectively, and fused to obtain the response gradient feature vector, including: Read the task load response surface calculation model, which is a continuously differentiable interpolation surface function in a two-dimensional domain defined by the task parameter coordinate axis and the time coordinate axis; Locate the task parameter coordinates at the current moment on the task load response surface calculation model, and obtain the first-order partial derivative of the task load response surface calculation model with respect to the task parameter coordinates to obtain the parameter gradient data at the current moment. The parameter gradient data represents the sensitivity of the current response feature data to changes in task parameters. Locate the time coordinate point corresponding to the current moment on the task load response surface calculation model, and obtain the first-order partial derivative of the task load response surface calculation model with respect to the time coordinate to obtain the time gradient data of the current moment. The time gradient data represents the temporal change rate of the current response characteristic data. The parameter gradient data and the time gradient data are used as two components of a two-dimensional vector. The magnitude of the two-dimensional vector is calculated, and the magnitude represents the overall change intensity of the current response feature data. The direction angle of the two-dimensional vector is calculated. The direction angle is obtained by the arctangent function of the ratio of the time gradient data to the parameter gradient data. The direction angle represents the dominant dimension of the current response feature data change. When the direction angle is close to zero, it indicates that the change is dominated by the change of task parameters. When the direction angle is close to ninety degrees, it indicates that the change is dominated by the time process. The magnitude of the two-dimensional vector and the direction angle are combined to form the response gradient feature vector.

7. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, Based on the asymmetric deviation characteristics of the task load response surface calculation model between the increasing and decreasing stages of the task parameters, the response hysteresis loop feature data is extracted through the response hysteresis loop calculation model, and the parameter adjustment coefficient is solved through the parameter adjustment coefficient calculation model based on its closed region metric value, including: Extract the first response curve corresponding to the increasing stage of the task parameters from the task load response surface calculation model. The first response curve describes the response value of the response feature data corresponding to each task parameter coordinate point as the task parameters gradually increase from the minimum value to the maximum value. The second response curve corresponding to the decreasing stage of the task parameters is extracted from the task load response surface calculation model. The second response curve describes the response value of the response feature data corresponding to each task parameter coordinate point as the task parameters gradually decrease from the maximum value to the minimum value. The first response curve and the second response curve are plotted on the same coordinate system, with the task parameter coordinates as the horizontal axis and the response value of the response feature data as the vertical axis. The first response curve and the second response curve have a response value deviation at the same task parameter coordinate point. The response value of the first response curve is higher than the response value of the second response curve, forming a closed region. The response value deviation between the first response curve and the second response curve is integrated along the task parameter coordinate axis by using a response hysteresis loop calculation model to obtain a closed region metric. The closed region metric represents the degree of asymmetry of the response path of an individual task system during the process of increasing and decreasing task parameters. The closed region metric is normalized using a parameter adjustment coefficient calculation model. The normalized closed region metric is then substituted into an inverse proportional function to obtain the parameter adjustment coefficient. The parameter adjustment coefficient is inversely proportional to the closed region metric. A higher parameter adjustment coefficient indicates better flexibility of the individual task system in adjusting to changes in task parameters.

8. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, Based on the response gradient feature vector, the response hysteresis loop feature data, and the parameter adjustment coefficient, the task parameter optimization range and the task state discrimination threshold set are calibrated through a task state discrimination model during the step-difficulty calibration task, including: The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to each task parameter coordinate point during the step-difficulty calibration task, as well as the local contribution value of the response hysteresis loop feature data corresponding to each task parameter coordinate point. The task state discrimination model marks the task parameter coordinate point corresponding to the first time the magnitude data of the response gradient feature vector reaches the preset activation threshold as the lower boundary of the task parameter optimization interval. The task state discrimination model marks the maximum task parameter coordinate point corresponding to the condition that the magnitude of the response gradient feature vector does not exceed a preset upper limit threshold and the local contribution value of the response hysteresis loop feature data does not exceed a preset deviation threshold as the upper boundary of the task parameter optimization interval. The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the lower boundary of the task parameter optimization interval, multiplies it by a first preset scaling factor, and obtains a first threshold. The task state discrimination model reads the magnitude data of the response gradient feature vector corresponding to the upper boundary of the task parameter optimization interval, multiplies it by the second preset scaling factor, and obtains the second threshold. The task state discrimination model reads the median of the absolute values ​​of all time gradient data during the step-difficulty calibration task, multiplies it by a third preset proportional coefficient, and obtains the third threshold. The task status discrimination model reads the median of all deviation data during the step-difficulty calibration task, multiplies it by the fourth preset proportional coefficient, and obtains the fourth threshold. The task state discrimination model reads the parameter adjustment coefficient, and multiplies the first threshold, the second threshold, the third threshold and the fourth threshold by the weighting factor corresponding to the parameter adjustment coefficient to obtain a set of task state discrimination thresholds after individualized correction. The first threshold and the second threshold are used for deviation index determination, the third threshold is used for auxiliary determination of the positive and negative signs of time gradient data, and the fourth threshold is used for overload state deviation determination after weighting by the parameter adjustment coefficient.

9. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, Based on the ratio of spectral resource consumption data between object behavior performance index data and response feature data, a neural efficiency calculation model is used to solve for task efficiency index data, including: After each behavior statistics unit is completed, read the object behavior response log data corresponding to the behavior statistics unit, and extract the accuracy data and average reaction time data from the object behavior response log data; Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline reaction time data from the historical response characteristic data; Divide the accuracy data by the ratio of the average reaction time data to the resting baseline reaction time data to obtain the behavioral performance index data, which comprehensively reflects the task completion quality and response speed of the object in the behavioral statistical unit. Read the response feature data within the time window corresponding to the behavior statistics unit, and extract the θ band power increase data and the low β band power increase data from the response feature data; Read the historical response characteristic data of the object in the resting-state baseline phase, and extract the resting-state baseline power data from the historical response characteristic data; The summation of the θ-band power increase data and the low β-band power increase data is divided by the resting-state baseline power data to obtain spectrum resource consumption data. The spectrum resource consumption data represents the total amount of neural resources consumed by the object to complete the task within the behavioral statistical unit. The task efficiency index is obtained by dividing the behavioral performance index data by the spectrum resource consumption data using a neural efficiency calculation model. The task efficiency index data represents the level of behavioral performance supported by a unit of neural resource consumption.

10. The cognitive training task parameter management method based on response data according to claim 1, characterized in that, Based on its temporal variation trend, task effectiveness evaluation data is generated and output through the task effectiveness evaluation model, including: Read the task efficiency index data corresponding to all behavioral statistical units within the task cycle, and construct a time series sequence of the task efficiency index. Read the average value of a preset number of task efficiency index data at the end of the task cycle, and use it as the final task efficiency index data. Read the average value of a preset number of task efficiency index data at the beginning of the task cycle, and use it as the initial task efficiency index data. The difference between the final task efficiency index data and the initial task efficiency index data is divided by the initial task efficiency index data to obtain the task efficiency index improvement rate data, which represents the extent of improvement in the neural efficiency of the object throughout the task cycle. The ratio of the standard deviation to the mean of the task efficiency index time series is calculated to obtain the coefficient of variation of the task efficiency index, which characterizes the stability of the neural efficiency of the object during the task. Read the parameter adjustment coefficient data calibrated at the end of the task cycle as the final parameter adjustment coefficient data, read the parameter adjustment coefficient data calibrated at the beginning of the task cycle as the initial parameter adjustment coefficient data, divide the difference between the final parameter adjustment coefficient data and the initial parameter adjustment coefficient data by the initial parameter adjustment coefficient data to obtain the parameter adjustment coefficient improvement rate data. The parameter adjustment coefficient improvement rate data characterizes the change range of the object's task adjustment elasticity within the task cycle. The task effectiveness evaluation model combines the task efficiency index improvement rate data, the task efficiency index variation coefficient data, and the parameter adjustment coefficient improvement rate data into multidimensional task effectiveness evaluation data, and outputs the multidimensional task effectiveness evaluation data to the report generation interface of the task management system.

Images