Cognitive load prediction method, electronic device, and storage medium

Abstract

The application discloses a cognitive load prediction method, an electronic device and a storage medium, comprising: inputting to-be-recognized data into a cognitive load prediction model to obtain a target cognitive load prediction result, wherein the cognitive load prediction model comprises a target learning model and a linear regression model, the target learning model is used to obtain a transition cognitive load prediction result according to the to-be-recognized data, and the linear regression model is used to output the target cognitive load prediction result according to the transition cognitive load prediction result; the target learning model is obtained through training of a training strategy, and the training strategy comprises: obtaining training sample data, the training sample data comprises a plurality of first time nodes, a plurality of first eye movement feature data, a plurality of first differential entropy feature data, a plurality of first power spectral density data and a plurality of cognitive load true value labels, the plurality of first eye movement feature data and the plurality of first differential entropy feature data; and the initial learning model is trained by using the training sample data to obtain the target learning model.

Description

Technical Field

[0001] This application relates to the field of cognitive neuroscience, and in particular to a method for predicting cognitive load, an electronic device, and a storage medium. Background Technology

[0002] Cognitive load prediction is a key issue in the field of human-computer interaction, helping to promote cognitive intelligence in machines and enabling computers to identify, interpret, and respond to an individual's cognitive load level. Cognitive load refers to the portion of an individual's limited cognitive resources actually occupied by task demands. Changes in cognitive load levels are a physiological process influenced by multiple factors, and experts in psychology, neuroscience, and computer science have long conducted in-depth research on cognitive load prediction. However, the accuracy of current cognitive load prediction techniques is not high. Summary of the Invention

[0003] This application aims to propose a cognitive load prediction method, electronic device, and storage medium that can solve the problem of low accuracy in traditional cognitive load prediction technologies.

[0004] In a first aspect, embodiments of this application provide a cognitive load prediction method, including:

[0005] The data to be identified is input into the cognitive load prediction model to obtain the target cognitive load prediction result, where...

[0006] The cognitive load prediction model includes a target learning model and a linear regression model. The target learning model is used to obtain a transitional cognitive load prediction result based on the data to be identified, and the linear regression model is used to output the target cognitive load prediction result based on the transitional cognitive load prediction result.

[0007] The target learning model is trained using a training strategy, which includes:

[0008] Training sample data is acquired, which includes multiple first time points, multiple first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels. Each of the first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels corresponds to a first time point. The first eye-tracking feature data is obtained from eye-tracking signals, and the first differential entropy feature data and the first power spectral density data are obtained from electroencephalogram (EEG) signals. The eye-tracking signals and the EEG signals are obtained by monitoring participants in the cognitive load induced experiment. The cognitive load truth labels characterize the cognitive load truth value of the participants in the cognitive load induced experiment.

[0009] The target learning model is obtained by training the initial learning model using the training sample data.

[0010] According to some embodiments of this application, the data to be identified includes multiple second time nodes, multiple second eye-tracking feature data, multiple second differential entropy feature data, and multiple second power spectral density data, wherein the multiple second eye-tracking feature data, the multiple second differential entropy feature data, and the multiple second power spectral density data all correspond to the multiple second time nodes; the transitional cognitive load prediction result is obtained through the following steps:

[0011] Multiple second time points and multiple second eye movement feature data are input into the target learning model to obtain the first cognitive load prediction result;

[0012] Multiple second time points and multiple second differential entropy feature data are input into the target learning model to obtain the second cognitive load prediction result;

[0013] Multiple second time nodes and multiple second power spectral density data are input into the target learning model to obtain the third cognitive load prediction result;

[0014] The transitional cognitive load prediction result is obtained by integrating the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result.

[0015] According to some embodiments of this application, the initial learning model includes an XGBoost model, a CatBoost model, and a random forest model. The step of training the initial learning model using the training sample data to obtain the target learning model includes:

[0016] The target learning model is obtained by training the XGBoost model, the CatBoost model, and the random forest model using the training sample data.

[0017] According to some embodiments of this application, the step of inputting multiple second time points and multiple second eye-tracking feature data into the target learning model to obtain a first cognitive load prediction result includes:

[0018] Multiple second time points and multiple second eye-tracking feature data are input into the XGBoost model to obtain a first intermediate prediction result;

[0019] Multiple second time points and multiple second eye-tracking feature data are input into the CatBoost model to obtain a second intermediate prediction result;

[0020] Multiple second time points and multiple second eye-tracking feature data are input into the random forest model to obtain a third intermediate prediction result;

[0021] The first cognitive load prediction result is obtained based on the first intermediate prediction result, the second intermediate prediction result, and the third intermediate prediction result.

[0022] According to some embodiments of this application, the step of inputting multiple second time nodes and multiple second differential entropy feature data into the target learning model to obtain a second cognitive load prediction result includes:

[0023] Multiple second time nodes and multiple second differential entropy feature data are input into the XGBoost model to obtain a fourth intermediate prediction result;

[0024] Multiple second time nodes and multiple second differential entropy feature data are input into the CatBoost model to obtain the fifth intermediate prediction result;

[0025] Multiple second time nodes and multiple second differential entropy feature data are input into the random forest model to obtain the sixth intermediate prediction result;

[0026] The second cognitive load prediction result is obtained based on the fourth intermediate prediction result, the fifth intermediate prediction result, and the sixth intermediate prediction result.

[0027] According to some embodiments of this application, the step of inputting multiple second time nodes and multiple second power spectral density data into the target learning model to obtain a third cognitive load prediction result includes:

[0028] Multiple second time nodes and multiple second power spectral density data are input into the XGBoost model to obtain the seventh intermediate prediction result;

[0029] Multiple second time nodes and multiple second power spectral density data are input into the CatBoost model to obtain the eighth intermediate prediction result;

[0030] Multiple second time points and multiple second power spectral density data are input into the random forest model to obtain the ninth intermediate prediction result;

[0031] The third cognitive load prediction result is obtained based on the seventh intermediate prediction result, the eighth intermediate prediction result, and the ninth intermediate prediction result.

[0032] According to some embodiments of this application, the step of fusing the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result to obtain the transitional cognitive load prediction result includes:

[0033] The first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result are concatenated along their vector dimensions to obtain the transitional cognitive load prediction result.

[0034] According to some embodiments of this application, the cognitive load truth label is obtained through the following steps:

[0035] Obtain multiple variables and corresponding response quantities from the cognitive load induced experiment;

[0036] Calculate the partial correlation coefficients between the multiple response quantities and the multiple variables to obtain multiple target partial correlation coefficients;

[0037] The true labels of the cognitive load are generated based on multiple target partial correlation coefficients.

[0038] Secondly, embodiments of this application provide an electronic device, including:

[0039] At least one processor;

[0040] At least one memory for storing at least one program;

[0041] The cognitive load prediction method described above is implemented when at least one of the programs is executed by at least one of the processors.

[0042] Thirdly, embodiments of this application provide a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the cognitive load prediction method described above.

[0043] In this embodiment, electroencephalogram (EEG) signals reflect brain electrical activity during the cognitive process, and eye movement (EMG) signals reflect the processing of visual information. Combining EEG and EMG signals can reflect cognitive states at different levels. Therefore, the initial learning model is trained using first eye movement feature data obtained from EMG signals and first differential entropy feature data and first power spectral density data obtained from EEG signals. This enables the target learning model to analyze the data to be identified at different levels, obtaining a prediction result for transitional cognitive load. A linear regression model performs regression analysis based on the transitional cognitive load prediction result and outputs a prediction result for the target cognitive load. Compared to traditional cognitive load prediction techniques, this improves the accuracy of cognitive load prediction.

[0044] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0045] The present application will be further described below with reference to the accompanying drawings and embodiments, wherein:

[0046] Figure 1 This is a flowchart of the training strategy in the embodiments of this application;

[0047] Figure 2 This is a flowchart illustrating the results of predicting transitional cognitive load in the embodiments of this application;

[0048] Figure 3 This is a flowchart illustrating the process of obtaining the first cognitive load prediction result in the embodiments of this application;

[0049] Figure 4 This is a flowchart illustrating the second cognitive load prediction result obtained in an embodiment of this application;

[0050] Figure 5 This is a flowchart illustrating the process of obtaining the third cognitive load prediction result in the embodiments of this application;

[0051] Figure 6 A schematic diagram of the structure of the electronic device provided in this application.

[0052] Figure label:

[0053] Electronic device 100, processor 110, memory 120. Detailed Implementation

[0054] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0055] In the description of this application, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0056] In the description of this application, "multiple" refers to two or more. The use of "first" and "second" is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance, or implicitly indicating the number of technical features indicated, or the order in which the technical features are indicated.

[0057] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0058] The following reference Figures 1 to 6 This application describes a cognitive load prediction method, electronic device, and storage medium according to embodiments of the present application.

[0059] This application provides a cognitive load prediction method, including:

[0060] Step S100: Input the data to be identified into the cognitive load prediction model to obtain the target cognitive load prediction result, wherein,

[0061] The cognitive load prediction model includes a target learning model and a linear regression model. The target learning model is used to obtain the transitional cognitive load prediction result based on the data to be identified, and the linear regression model is used to output the target cognitive load prediction result based on the transitional cognitive load prediction result.

[0062] The target learning model is trained through a training strategy, such as... Figure 1 As shown, the training strategy includes:

[0063] Step S200: Obtain training sample data. The training sample data includes multiple first time nodes, multiple first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels. The multiple first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels all correspond to the first time nodes. The first eye-tracking feature data is obtained from eye-tracking signals, and the first differential entropy feature data and first power spectral density data are obtained from electroencephalogram (EEG) signals. The eye-tracking signals and EEG signals are obtained by monitoring the participants in the cognitive load induced experiment. The cognitive load truth labels represent the cognitive load truth values ​​of the participants in the cognitive load induced experiment.

[0064] Step S300: Train the initial learning model using the training sample data to obtain the target learning model.

[0065] In this embodiment, electroencephalogram (EEG) signals reflect brain electrical activity during the cognitive process, and eye movement (EMG) signals reflect the processing of visual information. Combining EEG and EMG signals can reflect cognitive states at different levels. Therefore, the initial learning model is trained using first eye movement feature data obtained from EMG signals and first differential entropy feature data and first power spectral density data obtained from EEG signals. This enables the target learning model to analyze the data to be identified at different levels, obtaining a prediction result for transitional cognitive load. A linear regression model performs regression analysis based on the transitional cognitive load prediction result and outputs a prediction result for the target cognitive load. Compared to traditional cognitive load prediction techniques, this improves the accuracy of cognitive load prediction.

[0066] It is understandable that the cognitive load induced experiment in step S200 can induce continuous changes in the cognitive load level of the participants. The target learning model obtained by training the initial learning model with the training sample data obtained from the cognitive load induced experiment can continuously predict the cognitive load.

[0067] Understandably, the cognitive load elicitation experiment in step S200 uses a multi-object tracking task as the objective, aiming to investigate the mechanism of cognitive load level changes caused by differences in task difficulty during visual attention processing. The multi-object tracking task includes three phases:

[0068] Phase 1: The prompt phase presents several objects with similar surface features (e.g., an airplane) on the screen. Some of these objects are marked as targets by flashing red, while the others are distractions.

[0069] Phase 2: The visual markers from the cue phase disappear, and all objects begin to move randomly and independently. Participants are asked to track the targets marked in the cue phase.

[0070] Phase 3: After a preset tracking time, when the object's movement stops, the participants are asked to identify which objects are the target. The system simultaneously records their reaction time and recognition accuracy.

[0071] During the cognitive load evoked experiment, EEG and eye movement (EMG) signals from participants were collected using equipment. These signals were then preprocessed to remove artifacts and interference as much as possible. EEG signal preprocessing utilized the EEGLAB toolbox in MATLAB. First, a 0.1-40Hz bandpass filter was applied to eliminate high-frequency noise. Then, a 48-52Hz bandpass filter was used to eliminate power frequency interference. Finally, the Automatic Artifact Removal (AAR) toolbox was used to automatically remove EEG artifacts. The EEG and EMG signals were synchronized using the computer system's timestamps.

[0072] It is understandable that the multiple first time nodes in step S200 represent the trials of multiple consecutive multi-target tracking tasks, and the multiple first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels are all obtained through multiple consecutive multi-target tracking tasks.

[0073] It is understandable that in step S200, the first power spectral density data is obtained from the electroencephalogram (EEG) signal, specifically as follows:

[0074] The EEG signal is converted into a frequency domain signal, which is then divided into four frequency bands: δ (1-3Hz); θ (4-7Hz); α (8-12Hz); β (13-30Hz). Fourier transform is then used to convert the preprocessed EEG signal back into a frequency domain signal, as shown in the formula below.

[0075]

[0076] Where X(w) is the Fourier transform of the EEG signal, j is the imaginary prefix, and w represents the quantity of 2π;

[0077] Analyzing signals in different frequency domains yields corresponding frequency domain information. Based on this information, power spectral density data for different frequency bands are obtained, specifically using the following formula:

[0078]

[0079] Where Δt, N and fs represent the sampling period, the number of samples within the time period and the sampling frequency, respectively, and S(w) represents the power spectral density of the corresponding frequency band.

[0080] Understandably, differential entropy features are considered one of the most effective EEG features for estimating cognitive states. Their calculation method involves first defining [x, x ~ G(μ, σ)]. 2 Let f(x|μ, σ) represent a random time series that follows a Gaussian distribution. 2 () represents its probability density function. The calculation formula is as follows:

[0081]

[0082] The first differential entropy feature data of the EEG signal is obtained according to the following formula:

[0083]

[0084] Where h(x|μ,σ) 2 ) is a characteristic of differential entropy.

[0085] For eye-tracking signals, eye-tracking feature data was exported using Tobii Pro Lab software, including the average X-coordinate of the gaze point (gaze point X), the average Y-coordinate of the gaze point (gaze point Y), the X-axis gaze direction (gaze direction X), the Y-axis gaze direction (gaze direction Y), and the average pupil diameter (pupil diameter). Subsequently, static gaze and saccades in the eye-tracking signals were analyzed from an information theory perspective. Specifically, differential entropy features were calculated for the variables of gaze point X, gaze point Y, gaze direction X, and gaze direction Y. The obtained differential entropy features were then concatenated with the pupil diameter features to form a 5-dimensional feature vector, which served as the first eye-tracking feature data.

[0086] In one embodiment of this application, the data to be identified in step S100 includes multiple second time nodes, multiple second eye-tracking feature data, multiple second differential entropy feature data, and multiple second power spectral density data, all of which correspond to multiple second time nodes; for example... Figure 2 As shown, the transitional cognitive load prediction results in step S100 are obtained through the following steps:

[0087] Step S110: Input multiple second time points and multiple second eye-tracking feature data into the target learning model to obtain the first cognitive load prediction result;

[0088] Step S120: Input multiple second time nodes and multiple second differential entropy feature data into the target learning model to obtain the second cognitive load prediction results;

[0089] Step S130: Input multiple second time nodes and multiple second power spectral density data into the target learning model to obtain the third cognitive load prediction result;

[0090] Step S140: Integrate the first cognitive load prediction results, the second cognitive load prediction results, and the third cognitive load prediction results to obtain the transitional cognitive load prediction results.

[0091] In this embodiment, a target learning model is used to analyze multiple second eye-tracking feature data, multiple second differential entropy feature data, and multiple second power spectral density data. Based on these data, a first cognitive load prediction result, a second cognitive load prediction result, and a third cognitive load prediction result corresponding to a second time node are output. The first, second, and third cognitive load prediction results are then fused to obtain a transitional cognitive load prediction result, which accurately reflects the cognitive load level.

[0092] In one embodiment of this application, the initial learning model includes an XGBoost model, a CatBoost model, and a random forest model. The target learning model is obtained by training the initial learning model using training sample data, including:

[0093] The XGBoost model, CatBoost model, and Random Forest model were trained using training sample data to obtain the target learning model.

[0094] In this embodiment, the XGBoost model is a gradient-enhanced decision tree algorithm. It utilizes the second derivative to improve the accuracy of the loss function, a regularization term to prevent overfitting, and block storage for parallel computation, making it efficient, flexible, and lightweight. The XGBoost model has been widely used in data mining and feature engineering. The CatBoost model uses a forgetting tree as the base learner, has fewer parameters, supports categorical variables, and achieves higher accuracy, effectively addressing the problem of handling categorical characteristics. CatBoost solves problems such as gradient bias and prediction shift, thereby improving the accuracy and generalization ability of the target learning model. The Random Forest model is built upon decision trees as the base learner, introducing random attribute selection during decision tree training to enhance the generalization ability of the target learning model.

[0095] The target learning model, which is trained on XGBoost, CatBoost and Random Forest models using training sample data, enables ensemble learning, has strong generalization ability and high accuracy, and improves the predictive ability of cognitive load level.

[0096] Understandably, since the target learning model integrates the XGBoost model, CatBoost model, and random forest model, it uses the XGBoost model, CatBoost model, and random forest model to predict the data to be identified to obtain multiple prediction results. Then, it uses a linear regression model to perform regression analysis on the multiple prediction results to obtain an accurate target cognitive load prediction result.

[0097] In one embodiment of this application, step S110, "inputting multiple second time points and multiple second eye-tracking feature data into the target learning model to obtain the first cognitive load prediction result," will be further explained, as follows: Figure 3 As shown, step S110 includes:

[0098] Step S111: Input multiple second time nodes and multiple second eye-tracking feature data into the XGBoost model to obtain the first intermediate prediction result;

[0099] Step S112: Input multiple second time points and multiple second eye-tracking feature data into the CatBoost model to obtain the second intermediate prediction results;

[0100] Step S113: Input multiple second time nodes and multiple second eye-tracking feature data into the random forest model to obtain the third intermediate prediction result;

[0101] Step S114: Obtain the first cognitive load prediction result based on the first intermediate prediction result, the second intermediate prediction result, and the third intermediate prediction result.

[0102] In this embodiment, the XGBoost model, CatBoost model, and Random Forest model in the target learning model are used to analyze the second eye-tracking feature data. The XGBoost model, CatBoost model, and Random Forest model output the first intermediate prediction result, the second intermediate prediction result, and the third intermediate prediction result, respectively. The first cognitive load prediction result is obtained by concatenating the first intermediate prediction result, the second intermediate prediction result, and the third intermediate prediction result. For example, if the first intermediate prediction result is 0.6, the second intermediate prediction result is 0.7, and the third intermediate prediction result is 0.8, then the first cognitive load prediction result is [0.6, 0.7, 0.8].

[0103] In one embodiment of this application, step S120, "inputting multiple second time nodes and multiple second differential entropy feature data into the target learning model to obtain the second cognitive load prediction result," will be further explained as follows: Figure 4 As shown, step S120 includes:

[0104] Step S121: Input multiple second time nodes and multiple second differential entropy feature data into the XGBoost model to obtain the fourth intermediate prediction result;

[0105] Step S122: Input multiple second time nodes and multiple second differential entropy feature data into the CatBoost model to obtain the fifth intermediate prediction result;

[0106] Step S123: Input multiple second time nodes and multiple second differential entropy feature data into the random forest model to obtain the sixth intermediate prediction result;

[0107] Step S124: Obtain the second cognitive load prediction result based on the fourth intermediate prediction result, the fifth intermediate prediction result, and the sixth intermediate prediction result.

[0108] In this embodiment, the XGBoost model, CatBoost model and Random Forest model in the target learning model are used to analyze the second differential entropy feature data respectively. The XGBoost model, CatBoost model and Random Forest model output the fourth intermediate prediction result, the fifth intermediate prediction result and the sixth intermediate prediction result respectively. The fourth intermediate prediction result, the fifth intermediate prediction result and the sixth intermediate prediction result are concatenated to obtain the second cognitive load prediction result corresponding to multiple second time nodes.

[0109] In one embodiment of this application, step S130, "inputting multiple second time nodes and multiple second power spectral density data into the target learning model to obtain a third cognitive load prediction result," is as follows: Figure 5 As shown, it includes:

[0110] Step S131: Input multiple second time nodes and multiple second power spectral density data into the XGBoost model to obtain the seventh intermediate prediction result;

[0111] Step S132: Input multiple second time nodes and multiple second power spectral density data into the CatBoost model to obtain the eighth intermediate prediction result;

[0112] Step S133: Input multiple second time nodes and multiple second power spectral density data into the random forest model to obtain the ninth intermediate prediction result;

[0113] Step S134: Obtain the third cognitive load prediction result based on the seventh intermediate prediction result, the eighth intermediate prediction result, and the ninth intermediate prediction result.

[0114] In this embodiment, the XGBoost model, CatBoost model and random forest model in the target learning model are used to analyze the second power spectral density data respectively. The XGBoost model, CatBoost model and random forest model output the seventh intermediate prediction result, the eighth intermediate prediction result and the ninth intermediate prediction result respectively. The seventh intermediate prediction result, the eighth intermediate prediction result and the ninth intermediate prediction result are spliced ​​together to obtain the third cognitive load prediction result corresponding to multiple second time nodes.

[0115] In one embodiment of this application, step S140, "fusing the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result to obtain the transitional cognitive load prediction result," includes:

[0116] Step S141: Concatenate the vector dimensions of the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result to obtain the transitional cognitive load prediction result.

[0117] In one embodiment of this application, the truth label of cognitive load is obtained through the following steps:

[0118] Obtain multiple variables and corresponding response quantities from the cognitive load induced experiment;

[0119] Calculate the partial correlation coefficients between multiple response quantities and multiple variables to obtain multiple target partial correlation coefficients;

[0120] Generate true labels for cognitive load based on partial correlation coefficients of multiple objectives.

[0121] In this embodiment, the change in the response quantity of the cognitive load induced experiment is caused by the change in the variable. By calculating the partial correlation coefficients between multiple response quantities and multiple variables, multiple target partial correlation coefficients are obtained, and the true value label of cognitive load is generated based on the multiple target partial correlation coefficients.

[0122] For example, in a multi-object tracking task, variables related to task difficulty include the number of target objects (t), the total number of objects (o), the horizontal movement speed of the objects (h), and the vertical movement speed of the objects (v). To account for the impact of different variables on participant performance, the number of target objects correctly identified by the participants (C) is introduced into the response phase of the cognitive load truth label generation process.

[0123] The specific steps involved in generating labels include calculating the partial correlation coefficients between the number of correctly identified targets and various variables. The formula for calculating the partial correlation coefficients is as follows:

[0124]

[0125] In the formula, r 12 r represents the correlation coefficient between x1 and x2. 13 r represents the correlation coefficient between x1 and x3. 23 This represents the correlation coefficient between x2 and x3, calculated using the Pearson correlation coefficient. After expansion, x3, x4, ..., x... are excluded. p After considering the effects, the correlation coefficient between x1 and x2 is:

[0126]

[0127] By substituting C, t, o, h, and v into the formula, the partial correlation coefficient r of C relative to t, o, h, and v can be calculated. t r o r h r v The weighting coefficient wt of the number of target objects in the cognitive load truth label is calculated using the following formula:

[0128]

[0129] Where m represents the corresponding experimental participant.

[0130] Similarly, the weighting coefficient w of the total number of objects in the truth label of cognitive load can be calculated using the above formula. o The weighting coefficient w of the horizontal movement speed of an object in the truth label of cognitive load. h The weighting coefficient w of the vertical velocity of an object in the truth label of cognitive load. v .

[0131] The following formula is used to standardize t, o, h, and v, limiting them to the range [0,1]:

[0132]

[0133] The true labels of the cognitive load of the participants were calculated using the following formula:

[0134] Y (m) =w t (m) ×t+w o (m) ×o+w h (m) ×h+m v (m) ×v,

[0135] By iterating through all the participants in the experiment, a matrix of truth labels for cognitive load can be obtained.

[0136] In addition, one embodiment of this application also discloses an electronic device 100, such as... Figure 6 As shown, it includes:

[0137] At least one processor 110;

[0138] At least one memory 120 is used to store at least one program;

[0139] The cognitive load prediction method described above is implemented when at least one program is executed by at least one processor 110.

[0140] The electronic device 100 provided in this application embodiment can achieve... Figures 1 to 5 The various processes implemented in the method implementation examples achieve the same beneficial effects, and will not be described again here to avoid repetition.

[0141] In addition, one embodiment of this application discloses a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the cognitive load prediction method described above.

[0142] The computer-readable storage medium provided in the embodiments of this application can achieve... Figures 1 to 5 The various processes implemented in the method implementation examples achieve the same beneficial effects, and will not be described again here to avoid repetition.

[0143] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0144] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application.

Claims

1. A method for predicting cognitive load, characterized in that, include: The data to be identified is input into the cognitive load prediction model to obtain the target cognitive load prediction result, where... The cognitive load prediction model includes a target learning model and a linear regression model. The target learning model is used to obtain a transitional cognitive load prediction result based on the data to be identified, and the linear regression model is used to output the target cognitive load prediction result based on the transitional cognitive load prediction result. The target learning model is trained using a training strategy, which includes: Training sample data is acquired, which includes multiple first time points, multiple first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels. Each of the first eye-tracking feature data, multiple first differential entropy feature data, multiple first power spectral density data, and multiple cognitive load truth labels corresponds to a first time point. The first eye-tracking feature data is obtained from eye-tracking signals, and the first differential entropy feature data and the first power spectral density data are obtained from electroencephalogram (EEG) signals. The eye-tracking signals and the EEG signals are obtained by monitoring participants in the cognitive load induced experiment. The cognitive load truth labels characterize the cognitive load truth value of the participants in the cognitive load induced experiment. The target learning model is obtained by training the initial learning model using the training sample data. The data to be identified includes multiple second time nodes, multiple second eye-tracking feature data, multiple second differential entropy feature data, and multiple second power spectral density data, wherein each of the multiple second eye-tracking feature data, multiple second differential entropy feature data, and multiple second power spectral density data corresponds to a multiple second time node; the transitional cognitive load prediction result is obtained through the following steps: Multiple second time points and multiple second eye movement feature data are input into the target learning model to obtain the first cognitive load prediction result; Multiple second time points and multiple second differential entropy feature data are input into the target learning model to obtain the second cognitive load prediction result; Multiple second time nodes and multiple second power spectral density data are input into the target learning model to obtain the third cognitive load prediction result; The transitional cognitive load prediction result is obtained by integrating the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result.

2. The cognitive load prediction method according to claim 1, characterized in that, The initial learning model includes an XGBoost model, a CatBoost model, and a random forest model. The step of training the initial learning model using the training sample data to obtain the target learning model includes: The target learning model is obtained by training the XGBoost model, the CatBoost model, and the random forest model using the training sample data.

3. The cognitive load prediction method according to claim 2, characterized in that, The step of inputting multiple second time points and multiple second eye-tracking feature data into the target learning model to obtain a first cognitive load prediction result includes: Multiple second time points and multiple second eye-tracking feature data are input into the XGBoost model to obtain a first intermediate prediction result; Multiple second time points and multiple second eye-tracking feature data are input into the CatBoost model to obtain a second intermediate prediction result; Multiple second time points and multiple second eye-tracking feature data are input into the random forest model to obtain a third intermediate prediction result; The first cognitive load prediction result is obtained based on the first intermediate prediction result, the second intermediate prediction result, and the third intermediate prediction result.

4. The cognitive load prediction method according to claim 2, characterized in that, The step of inputting multiple second time points and multiple second differential entropy feature data into the target learning model to obtain the second cognitive load prediction result includes: Multiple second time nodes and multiple second differential entropy feature data are input into the XGBoost model to obtain a fourth intermediate prediction result; Multiple second time nodes and multiple second differential entropy feature data are input into the CatBoost model to obtain the fifth intermediate prediction result; Multiple second time nodes and multiple second differential entropy feature data are input into the random forest model to obtain the sixth intermediate prediction result; The second cognitive load prediction result is obtained based on the fourth intermediate prediction result, the fifth intermediate prediction result, and the sixth intermediate prediction result.

5. The cognitive load prediction method according to claim 2, characterized in that, The step of inputting multiple second time points and multiple second power spectral density data into the target learning model to obtain a third cognitive load prediction result includes: Multiple second time nodes and multiple second power spectral density data are input into the XGBoost model to obtain the seventh intermediate prediction result; Multiple second time nodes and multiple second power spectral density data are input into the CatBoost model to obtain the eighth intermediate prediction result; Multiple second time points and multiple second power spectral density data are input into the random forest model to obtain the ninth intermediate prediction result; The third cognitive load prediction result is obtained based on the seventh intermediate prediction result, the eighth intermediate prediction result, and the ninth intermediate prediction result.

6. The cognitive load prediction method according to claim 2, characterized in that, The process of fusing the first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result to obtain the transitional cognitive load prediction result includes: The first cognitive load prediction result, the second cognitive load prediction result, and the third cognitive load prediction result are concatenated along their vector dimensions to obtain the transitional cognitive load prediction result.

7. The cognitive load prediction method according to claim 1, characterized in that, The truth labels for cognitive load are obtained through the following steps: Obtain multiple variables and corresponding response quantities from the cognitive load induced experiment; Calculate the partial correlation coefficients between the multiple response quantities and the multiple variables to obtain multiple target partial correlation coefficients; The true labels of the cognitive load are generated based on multiple target partial correlation coefficients.

8. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; The cognitive load prediction method as described in any one of claims 1 to 7 is implemented when at least one of the programs is executed by at least one of the processors.

9. A computer-readable storage medium, characterized in that, It contains a processor-executable program, which, when executed by a processor, is used to implement the cognitive load prediction method as described in any one of claims 1 to 7.

Images