A method for evaluating EEG indicators of air traffic controllers based on mRMR algorithm

Abstract

The present invention discloses a method for evaluating the electroencephalogram (EEG) index of air traffic controllers based on an mRMR algorithm, comprising the following steps: (1) selecting an EEG index to quantitatively characterize the state of the air traffic controller; (2) obtaining the EEG data and NASA-TXL scale data of air traffic controllers at civil aviation airports in different working scenes, and preprocessing the collected EEG data; (3) using a frequency domain analysis method to calculate the workload index, and performing a Pearson correlation analysis on the workload; (4) calculating the mutual information between 59 channels corresponding to each index and the workload, and the sum of the redundancy within the 59 channels, finding the maximum correlation-minimum redundancy values ​​of all the indexes, and sorting them to evaluate the performance of the EEG index proposed above; (5) introducing an SVM classifier to evaluate the classifiability of the index, and verifying whether the sorting result is accurate. The present invention is conducive to solving the problem that the current workload evaluation relies on a scale, and frequent filling of the scale interferes with actual work, causing additional workload.

Description

Technical Field

[0001] The present invention relates to the field of civil aviation airport operation safety technology, and in particular to a method for evaluating electroencephalogram (EEG) indicators of controllers based on mRMR algorithm. Background Art

[0002] People are prone to mental fatigue when they engage in high-intensity mental activities for a long time. For air traffic controllers, the degree of mental fatigue will seriously affect their cognitive ability, reaction speed and alertness level, which will in turn affect their work performance, and in severe cases may even cause safety accidents. As a key position to ensure the safe and efficient operation of civil aviation, the controller's good cognitive process and appropriate workload are the basis for ensuring the safety of aircraft operations in the sector. As the number of flights increases, the workload of controllers is getting heavier. Therefore, it is particularly important to detect and analyze the workload of controllers.

[0003] At present, the workload detection of air traffic controllers is mainly divided into two categories: subjective detection methods and objective detection methods. Among them, subjective detection methods include KSS scale, NASA-TLX scale, etc. However, frequent filling of scales will interfere with actual work or experimental subjects, and filling of scales will also cause additional workload. Objective detection is mainly divided into EEG signal detection, ECG signal detection, eye information detection, facial feature detection and voice feature detection. It uses physiological data as a sign, can accurately obtain physiological indicators, has high reliability and validity, and reflects real-time status. Among the existing workload detection methods for air traffic controllers, ECG signals, eye and facial information, and EEG signals can reflect the real-time status of air traffic controllers and have the basis for implementation of evaluation. Among these methods, EEG signals have the highest accuracy and are known as the "golden indicator" for workload evaluation. In recent years, with the rapid development of EEG equipment, more and more EEG indicators for workload detection have been reported, but it is still unknown which EEG indicator is more suitable for workload detection of air traffic controllers. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method for evaluating the EEG indicators of controllers based on the mRMR algorithm, which deeply analyzes the correlation between the indicators and the workload in each EEG channel and the redundancy of information between different channels, and evaluates the performance of the indicators through efficient and concise mathematical calculation formulas, so as to screen indicators with better performance, detect the workload of the controller more accurately, and prevent the occurrence of unsafe incidents.

[0005] Technical solution: The method for evaluating the EEG index of a controller based on the mRMR algorithm of the present invention comprises the following steps:

[0006] (1) Select EEG indicators to quantitatively characterize the controller's status;

[0007] (2) Obtain EEG data and NASA-TXL scale data of civil aviation airport air traffic controllers in different working scenarios, and pre-process the collected EEG data;

[0008] (3) Using the frequency domain analysis method, the values ​​of 25 indicators in the 59-lead channel of each subject in different scenarios were calculated and Pearson correlation analysis was performed on the workload indicators and workload;

[0009] (4) Calculate the mutual information between the 59 channels and the workload corresponding to each indicator, and the sum of the redundancy within the 59 channels, find the maximum correlation-minimum redundancy value of all indicators, and sort them to evaluate the performance of the EEG indicators proposed above;

[0010] (5) Introduce the classifiability of the SVM classifier evaluation indicator to verify whether the above sorting results are accurate.

[0011] Furthermore, the step (1) is specifically as follows: The EEG indicators are shown in the following table:

[0012] 1 δ absolute energy 10 θ / β 19 γ / β 2 θ absolute energy 11 α / β 20 γ / (α + β) 3 α absolute energy 12 (α + θ) / β 21 γ / (θ + β) 4 β absolute energy 13 (α + θ) / (α + β) 22 γ / (θ + α) 5 δ + β + θ + α 14 γ absolute energy 23 γ / (θ + β + α) 6 δ / (δ + β + θ + α) 15 δ + β + θ + α + γ 24 γ / (δ + β + θ + α) 7 θ / (δ + β + θ + α) 16 γ / δ 25 γ / (δ + β + θ + α + γ) 8 α / (δ + β + θ + α) 17 γ / θ 9 β / (δ + β + θ + α) 18 γ / α

[0013] Furthermore, the step (2) is specifically as follows: using the NASA-TXL scale to evaluate the workload on each exercise, and reflecting the medium and low incremental values ​​of each point in the scale; using the scoring results as a representation of the workload, collecting the workload data after the subjects complete each exercise.

[0014] Furthermore, in step (3), the formula for performing Pearson correlation analysis on workload index and workload is as follows:

[0015]

[0016] Among them, X is the corresponding value of the subject's indicator in a certain scenario, and Y is the value of the NASA-TXL scale of the subject in the scenario. The Pearson coefficients of the 25 indicators in different channels and different scenarios and the corresponding workload are calculated in this way. When the Pearson coefficients are all greater than 0.4, it is considered that there is a correlation between the indicator and the workload. Among them, 25 indicators include: δ absolute energy; θ absolute energy; α absolute energy; β absolute energy; δ+β+θ+α; δ / (δ+β+θ+α); θ / (δ+β+θ+α); α / (δ+β+θ+α); β / (δ+β+θ+α); θ / β; α / β; (α+θ) / β; (α+θ) / (α+β); γ absolute energy; δ+β+θ+α+γ; γ / δ; γ / θ; γ / α; γ / β; γ / (α+β); γ / (θ+β); γ / (θ+α); γ / (θ+β+α); γ / (δ+β+θ+α); γ / (δ+β+θ+α+γ).

[0017] Furthermore, the step (4) is specifically as follows: First, let S represent the channel x of a certain indicator i , remove 5 useless channels, and the remaining 59 channels are |S|=59; regard the 59 channels as a whole, calculate the mutual information between the 59 channels and the workload c, so that S satisfies the formula Obtain the maximum correlation D between the index value and the workload; secondly, the Min-Redundancy method can be used to eliminate the redundant information between channels, that is, calculate the sum of the redundancy R within the 59 channels, the formula is Finally, the values ​​of maximum correlation - minimum redundancy Φ of all indicators are sorted to evaluate the performance of EEG indicators; the formula is as follows: maxφ(D,R),φ=DR.

[0018] Furthermore, in step (5), the specific steps are as follows: considering the resting state, the controller index data is divided into two categories: resting and working; 0 load, high load, medium and low load; 0 load, medium and high load, low load; 0 load, low load, medium load, high load; without considering the resting state, only considering the working scene: the workload is divided into three categories: medium, low, and high; the workload is divided into two categories: medium and high load and low load; the workload is divided into two categories: high load and medium and low load. The accuracy of each indicator under different classification methods is standardized and the indicators are ranked, and then compared with the indicator evaluation obtained in step (4) to verify the effectiveness of the above evaluation method.

[0019] The present invention provides a controller electroencephalogram index evaluation system based on mRMR algorithm, comprising:

[0020] EEG index module: used to select EEG indexes to quantitatively represent the controller's status;

[0021] Preprocessing module: used to obtain EEG data and NASA-TXL scale data of civil aviation airport air traffic controllers in different working scenarios, and preprocess the collected EEG data;

[0022] Correlation analysis module: used to calculate the values ​​of 25 indicators in 59-lead channels for each subject in different scenarios using frequency domain analysis method and perform Pearson correlation analysis on workload indicators and workload;

[0023] Evaluation module: used to calculate the mutual information between the 59 channels corresponding to each indicator and the workload, the sum of the redundancy within the 59 channels, find the value of the maximum correlation-minimum redundancy of all indicators, sort them, and evaluate the performance of the EEG indicators proposed above;

[0024] Verification module: used to introduce the classifiability of the SVM classifier evaluation indicators and verify whether the above sorting results are accurate.

[0025] A device described in the present invention includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, any one of the methods for evaluating the EEG indicators of controllers based on the mRMR algorithm is implemented.

[0026] A storage medium described in the present invention stores a computer program, and the computer program is designed to implement any one of the methods for evaluating the EEG indicators of controllers based on the mRMR algorithm when running.

[0027] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: it realizes the evaluation and screening of EEG indicators, and the indicators related to gamma waves proposed in the invention also perform better in the evaluation than the existing indicators, which can achieve more accurate detection of workload, and in the future, it can achieve a major breakthrough in the accuracy of workload detection of controllers. Compared with the separability evaluation method, the indicator evaluation method proposed in the present invention is fast and simple, which can effectively avoid the increase in workload caused by multiple classifications and achieve a more accurate evaluation of workload. The present invention can process the redundancy between channels while measuring the correlation, and screen out indicators with better performance for the establishment of a workload evaluation model for controllers, which is conducive to solving the problem that the current workload evaluation relies on scales, and frequent filling of scales interferes with actual work and causes additional workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of the present invention;

[0029] Figure 2 A flowchart of the experiment of the present invention;

[0030] Figure 3 It is a result verification diagram of the present invention. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0032] like Figure 1 - 2 As shown, an embodiment of the present invention provides a method for evaluating EEG indicators of a controller based on an mRMR algorithm, comprising the following steps:

[0033] (1) Select EEG indicators to quantitatively characterize the state of the controller; the details are as follows: Traditional EEG power detection indicators are generally related to the four wave bands of delta wave, theta wave, alpha wave, and beta wave, ignoring the role of gamma wave in mental state analysis. Gamma wave represents high-level cognitive activities and should be used for workload detection. Therefore, in order to better analyze the workload of controllers in different scenarios, the present invention first considers the common traditional indicators (indicators 1-13), and on this basis, proposes new workload detection indicators related to gamma wave (indicators 14-25):

[0034]

[0035]

[0036] (2) Obtain EEG data and NASA-TXL scale data from civil aviation airport air traffic controllers in different work scenarios, and preprocess the collected EEG data; specifically, use the NASA-TXL scale to evaluate the workload on each exercise, and reflect the medium and low incremental values ​​of each point in the scale; use the scoring results as a representation of the workload, and collect the workload data after the subjects complete each exercise.

[0037] Before the experiment began, four highly distinguishable and representative exercises were selected on the control simulator through pre-experiments: low load, high load, overload, and special situation scenario. Among them, the special situation scenario refers to the unexpected situation that the controller may encounter during work. In this experiment, the unexpected situation is the intrusion of foreign objects on the runway.

[0038] The experiment selected 8 mature controllers as subjects. Each scene took half an hour, and the scene interval was five minutes to fill in the scale. All subjects were required to work for more than 2 hours to collect EEG data and NASA-TXL subjective scale data under simulated control. The EEG acquisition equipment is the NeuSen W series wireless EEG acquisition system of Bo Ruikang, with a total of 64 electrodes and a sampling frequency of 1000Hz. EEG data preprocessing includes multiple processes such as removing useless electrodes, re-referencing, filtering, ICA, and removing artifacts. Since the sample size of each scene is less than 5000, the NASA-TXL scale data is subjected to the Shapiro-Wilk normality test. A P value greater than 0.05 indicates that the NASA-TXL value of the subject in this scene conforms to the normal distribution and the scene is valid. If the P value is less than 0.05, it means that the scene is invalid and the scene needs to be reselected.

[0039] (3) Using the frequency domain analysis method, the values ​​of the 25 indicators in the 59-lead channel of each subject in different scenarios were calculated, and the Pearson correlation analysis of the workload indicator and workload was performed; the formula for the Pearson correlation analysis of the workload indicator and workload is as follows:

[0040]

[0041] Among them, X is the corresponding value of the subject's indicator in a certain scenario, and Y is the value of the NASA-TXL scale of the subject in the scenario. The Pearson coefficients of the 25 indicators in different channels and different scenarios and the corresponding workload are calculated in this way. When the Pearson coefficients are all greater than 0.4, it is considered that there is a correlation between the indicator and the workload. Among them, 25 indicators include: δ absolute energy; θ absolute energy; α absolute energy; β absolute energy; δ+β+θ+α; δ / (δ+β+θ+α); θ / (δ+β+θ+α); α / (δ+β+θ+α); β / (δ+β+θ+α); θ / β; α / β; (α+θ) / β; (α+θ) / (α+β); γ absolute energy; δ+β+θ+α+γ; γ / δ; γ / θ; γ / α; γ / β; γ / (α+β); γ / (θ+β); γ / (θ+α); γ / (θ+β+α); γ / (δ+β+θ+α); γ / (δ+β+θ+α+γ).

[0042] (4) Calculate the mutual information between the 59 channels corresponding to each indicator and the workload, and the sum of the redundancy within the 59 channels, find the value of the maximum correlation-minimum redundancy of all indicators, and sort them to evaluate the performance of the EEG indicators proposed above; the details are as follows: First, let S represent the set of channels xi of a certain indicator, remove 5 useless channels, and the remaining 59 channels are |S| = 59; regard the 59 channels as a whole, calculate the mutual information between the 59 channels and the workload c, so that S satisfies the formula Obtain the maximum correlation D between the index value and the workload; secondly, the Min-Redundancy method can be used to eliminate the redundant information between channels, that is, calculate the sum of the redundancy R within the 59 channels, the formula is Finally, the values ​​of maximum correlation - minimum redundancy Φ of all indicators are sorted to evaluate the performance of EEG indicators; the formula is as follows: maxφ(D,R),φ=DR.

[0043] (5) Introduce the SVM classifier to evaluate the classifiability of the indicators and verify whether the above sorting results are accurate. Specifically, when considering the resting state, the controller indicator data is divided into two categories: resting and working; three categories: 0 load, high load, and medium and low load; three categories: 0 load, medium and high load, and low load; 0 load, low load, medium load, and high load; when not considering the resting state and only considering the working scene: the workload is divided into three categories: medium, low, and high; the workload is divided into two categories: medium and high load and low load; the workload is divided into two categories: high load and medium and low load. The accuracy of each indicator under different classification methods is standardized and the indicators are sorted, and then compared with the indicator evaluation obtained in step (4) to verify the effectiveness of the above evaluation method.

[0044] like Figure 3 As shown in the figure, the example results show that the SVM and MRMR ranking results are basically consistent. Among the 25 indicators proposed in this paper, both ranking methods show that indicators 15 and 5 rank best, and indicators 1, 2, 3, 4, 14, 18, and 19 rank relatively well. However, indicators 6, 7, 8, 9, 16, 24, and 25 have poor separability and low rankings. Therefore, it can be concluded that the EEG index evaluation method based on the mRMR algorithm is reliable.

[0045] The embodiment of the present invention also provides a controller electroencephalogram index evaluation system based on the mRMR algorithm, comprising:

[0046] EEG index module: used to select EEG indexes to quantitatively represent the controller's status;

[0047] Preprocessing module: used to obtain EEG data and NASA-TXL scale data of civil aviation airport air traffic controllers in different working scenarios, and preprocess the collected EEG data;

[0048] Correlation analysis module: used to calculate the values ​​of 25 indicators in 59-lead channels for each subject in different scenarios using frequency domain analysis method and perform Pearson correlation analysis on workload indicators and workload;

[0049] Evaluation module: used to calculate the mutual information between the 59 channels corresponding to each indicator and the workload, the sum of the redundancy within the 59 channels, find the value of the maximum correlation-minimum redundancy of all indicators, sort them, and evaluate the performance of the EEG indicators proposed above;

[0050] Verification module: used to introduce the classifiability of the SVM classifier evaluation indicators and verify whether the above sorting results are accurate.

[0051] An embodiment of the present invention also provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the methods for evaluating EEG indicators of controllers based on the mRMR algorithm is implemented.

[0052] An embodiment of the present invention further provides a storage medium storing a computer program, wherein the computer program is designed to implement any one of the methods for evaluating EEG indicators of controllers based on the mRMR algorithm when running.

Claims

1. A method for evaluating EEG indicators of controllers based on mRMR algorithm, characterized in that: The following steps are involved: (1) EEG indicators are selected to quantitatively characterize the controller's status; the EEG indicators are shown in the following table: (2) Obtain EEG data and NASA-TLX scale data of civil aviation airport air traffic controllers in different work scenarios, and preprocess the collected EEG data; specifically, use the NASA-TLX scale to evaluate the workload on each exercise, and reflect the medium and low incremental values ​​of each point in the scale; use the scoring results as a representation of the workload, and collect the workload data after the subjects complete each exercise; (3) Using the frequency domain analysis method, the values ​​of 25 indicators in the 59-lead channel of each subject in different scenarios were calculated and Pearson correlation analysis was performed on the values ​​and workload; (4) Calculate the mutual information between the 59 channels corresponding to each indicator and the workload, and the sum of the redundancy within the 59 channels, find the value of the maximum correlation-minimum redundancy of all indicators, and sort them to evaluate the performance of the 25 EEG indicators; (5) Introduce the SVM classifier to evaluate the classifiability of the indicators and verify whether the above sorting results are accurate; specifically, when the resting state is considered, the controller indicator data is divided into two categories: resting or working; and three categories: zero load, high load, and medium and low load; when the resting state is not considered and only the working scene is considered: the workload is divided into three categories: medium, low, and high; or the workload is divided into two categories: medium-high load and low load; or the workload is divided into two categories: high load and medium-low load; the accuracy of each indicator under different classification methods is standardized and the indicators are sorted, and then compared with the indicator evaluation obtained in step (4) to verify the effectiveness of the indicator evaluation.

2. According to claim 1, a method for evaluating the EEG index of a controller based on the mRMR algorithm is characterized in that: In step (3), the values ​​of the 25 indicators in the 59-lead channel in different scenarios and the Pearson correlation analysis formula between the values ​​and the workload are as follows: Among them, X is the corresponding value of the subject's indicator in a certain scene, and Y is the value of the NASA-TLX scale of the subject in the scene. The Pearson coefficients of the 25 indicators in different channels and different scenes and the corresponding workload are calculated in this way. When the Pearson coefficients are all greater than 0.4, it is considered that there is a correlation between the indicator and the workload; among them, the 25 indicators include: δ absolute energy; θ absolute energy; α absolute energy; β absolute energy; δ+β+θ+α; δ / (δ+β+ θ+α); θ / (δ+β+θ+α); α / (δ+β+θ+α); β / (δ+β+θ+α); θ / β; α / β; (α+θ) / β; (α+θ) / (α+β); γ absolute energy; δ+β+θ +α+γ; γ / δ; γ / θ; γ / α; γ / β; γ / (α+β); γ / (θ+β); γ / (θ+α); γ / (θ+β+α); γ / (δ+β+θ+α); γ / (δ+β+θ+α+γ).

3. The method for evaluating the EEG index of a controller based on the mRMR algorithm according to claim 1 is characterized in that: The step (4) is specifically as follows: First, let S represent the channel x of a certain indicator i , remove 5 useless channels, and the remaining 59 channels are |S|=59; regard the 59 channels as a whole, calculate the mutual information between the 59 channels and the workload c, so that S satisfies the formula Obtain the maximum correlation D between the index value and the workload; secondly, select the Min-Redundancy method to eliminate the redundant information between channels, that is, calculate the sum of the redundancy R within the 59 channels, the formula is Finally, the values ​​of maximum correlation - minimum redundancy Φ of all indicators are sorted to evaluate the performance of EEG indicators; the formula is as follows: maxφ(D,R),φ=DR.

4. A controller EEG index evaluation system based on mRMR algorithm, characterized in that: include: EEG index module: used to select EEG indexes to quantitatively characterize the controller's status; the EEG indexes are shown in the following table: Preprocessing module: used to obtain EEG data and NASA-TLX scale data of civil aviation airport air traffic controllers in different work scenarios, and preprocess the collected EEG data; specifically, use the NASA-TLX scale to evaluate the workload on each exercise, and reflect the medium and low incremental values ​​of each point in the scale; use the scoring results as a representation of the workload, and collect the workload data after the subjects complete each exercise; Correlation analysis module: used to calculate the values ​​of 25 indicators in 59-lead channels of each subject in different scenarios using frequency domain analysis method and perform Pearson correlation analysis on the values ​​and workload; Evaluation module: used to calculate the mutual information between the 59 channels corresponding to each indicator and the workload, the sum of the redundancy within the 59 channels, find the value of the maximum correlation-minimum redundancy of all indicators, sort them, and evaluate the performance of 25 EEG indicators; Verification module: used to introduce the classifiability of the SVM classifier evaluation index and verify whether the above sorting results are accurate; specifically as follows: considering the resting state, the controller index data is divided into two categories: resting or working; 0 load, high load, medium and low load; without considering the resting state, only considering the working scene: the workload is divided into three categories: medium, low and high, or the workload is divided into two categories: medium-high load and low load, or the workload is divided into two categories: high load and medium-low load; the accuracy of each indicator under different classification methods is standardized and the indicators are sorted, and then compared with the indicator evaluation obtained by the evaluation module to verify the validity of the indicator evaluation.

5. A device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for evaluating the EEG index of a controller based on the mRMR algorithm as described in claims 1 to 3 is implemented.

6. A storage medium storing a computer program, characterized in that: The computer program is designed to implement, when running, a method for evaluating the EEG indicators of a controller based on the mRMR algorithm according to claims 1-3.

Images