A multi-dimensional cognitive ability detection system for civil aviation pilots

Abstract

The present application belongs to the technical field of cognitive ability detection, and particularly relates to a multi-dimensional cognitive ability detection system for civil aviation pilots, aiming to solve the problem of low detection accuracy of existing cognitive ability detection systems. The system comprises: a detection result acquisition module configured to acquire the final detection results of each detection index of the pilot to be detected for cognitive ability after passing through each detection module of the cognitive ability detection unit; a consistency checking module configured to construct a de-mentalization judgment matrix based on the final detection results, calculate a consistency ratio CR, and determine whether the index judgment matrix needs to be modified; and a first weight value updating module configured to normalize each element in the de-mentalization judgment matrix to obtain normalized elements, calculate the arithmetic mean of each row of the de-mentalization judgment matrix based on the normalized elements, and update the final weight values of each detection module. The present application improves the accuracy of cognitive ability detection.

Description

Technical Field

[0001] This invention belongs to the field of cognitive ability detection technology, specifically relating to a detection system for multi-dimensional cognitive abilities of civil aviation pilots. Background Technology

[0002] With the rapid development of the global aviation industry, the improvement of aircraft hardware manufacturing and development levels, and the perfection of supporting infrastructure such as airports in various regions, civil aviation is becoming increasingly accepted for people facing medium- and long-distance travel. However, the increasing demand for air travel also means increasingly higher operational risks, making the need for aviation safety and reliability increasingly urgent. During flight, human factors are the primary cause of aviation accidents, and the pilot's operational decisions in emergency situations directly determine the accident risk. Therefore, there is an urgent need for a system that can reasonably test pilots' cognitive abilities, which would help select pilots with strong flight decision-making abilities and reduce the risk and probability of aviation accidents to a certain extent. It is worth noting that a complete flight involves multiple flight stages, each of which requires cognitive testing of the pilot, and the results need to be comprehensive. Current technology can only test pilots' cognitive abilities from a single perspective, which will lead to a significant discrepancy between the test results and the pilot's actual abilities, making it unsuitable as a basis for pilot selection.

[0003] The present invention aims to develop a multi-dimensional cognitive ability detection system for civil aviation pilots. By comprehensively detecting cognitive abilities in multiple stages throughout the flight process, the system can effectively and reasonably select pilots, avoiding biases caused by single-dimensional detection results, thereby supporting and assisting the pilot selection process. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, specifically the low accuracy of current cognitive ability detection systems, this invention proposes a multi-dimensional cognitive ability detection system for civil aviation pilots. This system includes a cognitive ability detection unit, comprising modules for spatiotemporal rotation ability detection, perceptual speed ability detection, professional ability detection, attention ability detection, memory ability detection, and thinking ability detection. The system also includes:

[0005] The test result acquisition module is configured to acquire the test score data obtained by the pilot to be tested after passing each test index in each test module of the cognitive ability test unit, and use it as the first data.

[0006] Normalize each first data point to obtain normalized data; calculate the information entropy based on each first data point, and multiply the information entropy by the corresponding normalized data, and use the result of the multiplication as the second data point; convert each second data point into a percentage system, and use it as the final detection result corresponding to each detection indicator of the pilot.

[0007] The consistency verification module is configured to form an index judgment matrix based on each final detection result, and then construct an interval intelligence matrix. It then uses a de-intelligentization method to transform the true, false, and uncertain intelligence intervals contained in the interval intelligence matrix into single calculated values, constructing a de-intelligentization judgment matrix. Here, the true, false, and uncertain intelligence intervals correspond to the membership degree T. A Hesitation level I A Membership degree F A Together they represent a set A in an interval;

[0008] Based on the decentralization judgment matrix, the consistency index CI is calculated; the average random consistency index RI corresponding to CI is found according to the interval central intelligence evaluation language scale, and the consistency ratio CR is obtained by dividing the RI and RI; it is determined that the CR is less than the set consistency ratio threshold. If not, the index judgment matrix is ​​modified.

[0009] The first weight value update module is configured to normalize each element in the decentralization judgment matrix to obtain normalized elements; and calculate the arithmetic mean of each row of the decentralization judgment matrix based on each normalized element, and update it as the final weight value of each detection module.

[0010] The interval-based intelligence evaluation language scale represents a pre-defined mapping relationship between the relative importance of two detection modules.

[0011] In some preferred embodiments, the information entropy is calculated as follows:

[0012]

[0013] Among them, e j This represents the information entropy corresponding to detection index j. x ij Let k represent the test score data of pilot i corresponding to test index j for the cognitive ability test, k = 1 / ln(n) > 0, and n represent the number of pilots participating in the test.

[0014] In some preferred embodiments, the final test results corresponding to each test indicator of the pilot are as follows: Where x″ represents the second data;

[0015] The final test results for two or more test indicators are as follows:

[0016]

[0017] Where, x″ ij S represents the normalized test score of pilot i corresponding to test indicator j, i.e., the second data corresponding to the j-th test indicator of pilot i, j = 1, 2, ..., m, where m is the number of test indicators in the test questions. test The test score is the final test result corresponding to two or more test indicators.

[0018] In some preferred embodiments, the method for constructing the interval intelligence matrix is ​​as follows:

[0019] Based on the final test results, specific importance scores are obtained, and the importance scores are matrixed to form an indicator judgment matrix.

[0020] Each dimension in the indicator judgment matrix is ​​determined to have a corresponding importance level. By referring to the pre-constructed interval intelligence set evaluation language scale, the importance level in the indicator judgment matrix is ​​replaced with the corresponding interval intelligence set, thereby forming an interval intelligence matrix.

[0021] In some preferred embodiments, the elements in the decentralization judgment matrix are:

[0022]

[0023] Where U represents an element in the decentralization judgment matrix. This represents the lower limit of the membership degree T. Indicates the membership degree T A The upper limit, Indicates the degree of hesitation I A The upper limit, F represents the degree of non-membership. A The lower limit value, F represents the degree of non-membership. A The upper limit.

[0024] In some preferred embodiments, the consistency index is obtained by:

[0025]

[0026] Where CI represents the consistency index, and λ max This represents the largest eigenvalue, which is the eigenvalue of the matrix with the largest absolute value, and n represents the order of the matrix.

[0027] In some preferred embodiments, the arithmetic mean of each row of the decentralization judgment matrix is ​​calculated based on each normalized element, and used as the final weight value of each detection module. The method is as follows:

[0028]

[0029]

[0030] Among them, W i U represents the final weight value of the i-th detection module. ij This represents the element in the i-th row and j-th column of the decentralization judgment matrix.

[0031] In some preferred embodiments, the detection system for the multi-dimensional cognitive ability of civil aviation pilots further includes a second weight update module;

[0032] The second weight update module is configured to obtain the visual features of the detection links corresponding to the sub-modules in each detection module and the visual features of each sub-link in the simulated flight scenario. The eye movement indicators corresponding to the visual features include fixation duration ratio, average fixation duration, saccade amplitude, left eye pupil diameter, and right eye pupil diameter.

[0033] Based on the visual features of each sub-stage in the simulated flight scenario, the importance parameters of the detection stages corresponding to the sub-modules in each detection module are predicted by a pre-constructed multiple linear regression analysis model.

[0034] Based on the importance parameters, obtain the weight increase list and weight not adjusted list corresponding to the sub-modules of each detection module, and then adjust the weights corresponding to the sub-modules of each detection module.

[0035] Calculate the average of the updated weight values ​​of the sub-modules of each detection module after the (N-1)th iteration, the unupdated weights of the sub-modules of each detection module after the Nth iteration, and the adjusted weights of the sub-modules of each detection module after the Nth iteration. Use this average as the updated weight value for the Nth iteration, and then update the weights accordingly. Here, N represents the current iteration.

[0036] In some preferred embodiments, based on the visual features of each sub-stage in a simulated flight scenario, the importance parameters of the detection stages corresponding to the sub-modules in each detection module are predicted using a pre-constructed multiple linear regression analysis model. The method is as follows:

[0037] a i = w1a1 + w2a2 + w3a3 + ... + w 22 a 22 +ε,i=23,24,…,30

[0038] Among them, a iThe vector representing the i-th detection stage, i.e., the visual features of each sub-stage in the simulated flight scenario, w i ε represents the importance parameter of the i-th vector, which is the importance parameter of the detection link corresponding to the submodule in each detection module, and ε represents the set error value.

[0039] In some preferred embodiments, based on the importance parameter, a weight increase list and a weight not adjusted list corresponding to the sub-modules of each detection module are obtained, and then the weights corresponding to the sub-modules of each detection module are adjusted. The method is as follows:

[0040] In any sub-stage of a simulated flight scenario, if w i If the number of non-zero values ​​is greater than the set first threshold, the first M sub-modules are selected as significant cognitive ability modules; otherwise, all sub-modules with non-zero importance coefficients are selected as significant cognitive ability modules; sub-modules with zero importance coefficients are selected as non-significant cognitive ability modules.

[0041] For significant cognitive ability modules that appear in sub-stages of more than a set number of simulated flight scenarios, add them to the weight-enhanced list; for insignificant cognitive ability modules that appear in sub-stages of more than a set number of simulated flight scenarios, add them to the weight-unadjusted list.

[0042] Based on the weight increase list and the weight not adjusted list, the weights corresponding to the sub-modules of each detection module are adjusted.

[0043] The adjustment strategy is as follows: all sub-modules of the detection module start adjusting simultaneously, the weight adjustment step size is set to α, and the number of people in a unit batch is X people. That is, the default number of newly added detection people in each batch is X people. If the number is less than or more than this number, the step size is determined according to α*(actual number of people / X). The weight adjustment direction is determined by its importance value and increases. An update iteration is performed according to the step size.

[0044] The beneficial effects of this invention are:

[0045] This invention improves the accuracy of cognitive ability testing.

[0046] This invention aims to provide a more scientific, accurate, and fair testing method for the pilot selection process. First, the system employs multiple testing dimensions, such as professional skills, memory, thinking ability, attention, spatial-temporal rotation ability, and speed perception, to assess candidate pilots' performance in different cognitive areas. Compared to traditional selection methods, this system can comprehensively and meticulously assess pilots' cognitive abilities, thus better determining their suitability for flight work. Second, the system utilizes intelligent and digital testing technology to improve the scientific rigor, accuracy, and objectivity of the testing. It uses the results of comparative analysis of pilots' eye-tracking signals during cognitive ability system testing and simulated flight as a reference, and optimizes and adjusts the system based on visual eye-tracking characteristics. Furthermore, it uses normalization and entropy weighting methods to score different testing indicators to better identify outstanding pilots. Finally, the system can automate the selection process and meet the requirements of multiple personnel, multiple companies, and multiple levels, thereby significantly reducing labor and time costs and improving selection efficiency. Simultaneously, this system can also be used in conjunction with other selection methods to enhance selection effectiveness and reliability. Therefore, the multi-dimensional cognitive ability detection system of the present invention has broad application prospects and great promotion and application value in pilot selection and related fields. Attached Figure Description

[0047] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0048] Figure 1 This is a simplified framework diagram of a detection system for the multi-dimensional cognitive abilities of civil aviation pilots according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the test result data according to an embodiment of the present invention;

[0050] Figure 3 This is a simplified schematic diagram of the overall detection process of a detection system for multi-dimensional cognitive abilities of civil aviation pilots according to an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram illustrating the process of multiple pilots simultaneously detecting data according to an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of the specific framework of a cognitive ability detection unit according to an embodiment of the present invention;

[0053] Figure 6 This is a schematic diagram of the sub-modules corresponding to each detection module in one embodiment of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0056] A first embodiment of the present invention provides a multi-dimensional cognitive ability detection system for civil aviation pilots. The system includes a cognitive ability detection unit, which comprises a spatiotemporal rotation ability detection module, a perceptual speed ability detection module, a professional ability detection module, an attention ability detection module, a memory ability detection module, and a thinking ability detection module. The system also includes:

[0057] The test result acquisition module is configured to acquire the test score data obtained by the pilot to be tested after passing each test index in each test module of the cognitive ability test unit, and use it as the first data.

[0058] Normalize each first data point to obtain normalized data; calculate the information entropy based on each first data point, and multiply the information entropy by the corresponding normalized data, and use the result of the multiplication as the second data point; convert each second data point into a percentage system, and use it as the final detection result corresponding to each detection indicator of the pilot.

[0059] The consistency verification module is configured to form an index judgment matrix based on each final detection result, and then construct an interval intelligence matrix. It then uses a de-intelligentization method to transform the true, false, and uncertain intelligence intervals contained in the interval intelligence matrix into single calculated values, constructing a de-intelligentization judgment matrix. Here, the true, false, and uncertain intelligence intervals correspond to the membership degree T. A Hesitation level I A Membership degree F A Together they represent a set A in an interval;

[0060] Based on the decentralization judgment matrix, the consistency index CI is calculated; the average random consistency index RI corresponding to CI is found according to the interval central intelligence evaluation language scale, and the consistency ratio CR is obtained by dividing the RI and RI; it is determined that the CR is less than the set consistency ratio threshold. If not, the index judgment matrix is ​​modified.

[0061] The first weight value update module is configured to normalize each element in the decentralization judgment matrix to obtain normalized elements; and calculate the arithmetic mean of each row of the decentralization judgment matrix based on each normalized element, and update it as the final weight value of each detection module.

[0062] The interval-based intelligence evaluation language scale represents a pre-defined mapping relationship between the relative importance of two detection modules.

[0063] To more clearly illustrate the detection system for multi-dimensional cognitive abilities of civil aviation pilots according to the present invention, the following detailed description of each module in one embodiment of the system is provided in conjunction with the accompanying drawings.

[0064] First, the overall process of the system will be introduced, such as... Figure 3 As shown, the first step involves the test subject (i.e., the pilot whose cognitive ability is to be tested) logging into the system through the user login interface.

[0065] After entering the personal system, the person being tested creates a document for this test. Once created, they can proceed to the next stage. Different airlines have different selection criteria for different types and stages of pilots. Therefore, the order and frequency of the questions can be adjusted by the testing party and named as specific categories. The document creation process requires first selecting the specific test category and then creating the document.

[0066] After creating the document for this test, you can proceed to the question demonstration section. The first step is to test question demonstration 1, which will be displayed through the module (e.g., ...). Figure 1 As shown in the diagram, the test subject then attempts to perform the operation. After completing the attempt, the test subject can choose whether they understand the operation or not. If they understand the operation and have learned how to perform it, they can move on to the next operation. If they do not understand or cannot perform the operation, the system returns to the demonstration of operation 1 and demonstrates the operation again. This process continues until the test subject understands the content of all the operations and has learned how to perform them. Figure 3 As shown.

[0067] Once the test taker understands all the test questions and learns how to operate them, they can move on to the next stage and begin the formal cognitive ability test.

[0068] The specific content of the cognitive ability test is as follows: Figure 6 As shown, based on all the aspects and abilities a pilot needs to participate in during flight, the cognitive ability test mainly includes six modules: professional ability test module, memory ability test module, thinking ability test module, attention ability test module, spacetime rotation ability test module, and perceptual speed ability test module; as detailed below:

[0069] The memory testing module is used to test the memory of test subjects. The memory testing module includes sub-modules for textual memory, numerical memory, digital working memory, path memory, audio memory, episodic memory, character memory, and motor memory.

[0070] The text memory test is configured as follows: several (e.g., 10) Chinese words appear simultaneously for a certain period of time (e.g., 40 seconds). After the text disappears, the same number (e.g., 10) of different words appear one by one in sequence. The test asks whether these words have appeared before, in order to test the test taker's text memory ability.

[0071] The numerical memory test is configured as follows: Several (e.g., 10) multi-digit numbers (e.g., three-digit numbers) appear simultaneously for a certain period of time (e.g., 40 seconds). After the numbers disappear, the same number (e.g., 10) of different multi-digit numbers (e.g., three-digit numbers) appear one by one in sequence. The test asks whether these multi-digit numbers (e.g., three-digit numbers) have appeared before, in order to test the numerical memory ability of the test subjects.

[0072] The digital working memory test is configured as follows: a multi-square grid (e.g., a nine-square grid) with several (e.g., 2) squares lit up sequentially is displayed. A number is displayed in one of the lit squares. After multiple (e.g., four) displays, the sum of the historical numbers in a certain square is calculated to test the digital working memory ability of the test subject.

[0073] The path memory test is configured as follows: In a large square composed of n by n (n is a natural number greater than 1, such as 6 by 6) square grids, one square is randomly lit and then quickly turned off (for example, one square is randomly displayed as red for 0.75 seconds). This is repeated multiple times (for example, 3-4 times). The positions where the lit square (red square) appeared in all the times are selected to test the path memory ability of the test subjects.

[0074] The audio memory test is designed to assess the test taker's audio memory ability by asking them to answer questions related to a short audio recording (e.g., 30 words or less).

[0075] The contextual memory testing module is configured to display a scene image containing multiple (e.g., 4) special items and maintain the display for a certain period of time before disappearing (e.g., 4 seconds). After the scene image disappears, the test subject answers what item appears in one of the positions (e.g., the question mark position) to test their contextual memory ability.

[0076] The character memory test is configured as follows: a multi-digit number (e.g., 5 digits) is displayed for a certain period of time (e.g., 1.5s). After the multi-digit number disappears, a random number appears. The tester determines whether the randomly appearing number has appeared in the previously displayed multi-digit number (if it has, the space bar is pressed) in order to test the character memory ability of the test taker.

[0077] The motor memory test is configured as follows: four numbers (e.g., 1, 2, 3, 4) are displayed for a certain period of time (e.g., 3 seconds). The test subject must remember the direction keys corresponding to these four numbers. When the key instruction is issued (e.g., the instruction "PRESS" appears), the test subject must press the four direction keys in ascending order of the numbers to test the motor memory ability of the test subject.

[0078] The attention ability testing module is used to test the attention of the test subjects. The attention ability testing module includes: an attention allocation ability submodule and an attention concentration ability submodule;

[0079] Note that there are two configurations for assigning questions:

[0080] 1) The screen is divided into two parts, left and right. On the left screen, you need to use your hand to move the cursor to track the moving object (such as a ball) (i.e., the tracking task), while on the right side, the motion memory detection module mentioned above is used for detection (i.e., motion memory).

[0081] 2) The screen is divided into two parts, left and right. On the left screen, you need to move the cursor with your hand to track the moving ball (i.e., tracking task). At the same time, on the right, after a certain number (e.g., the number 0) appears, you silently count the set time (e.g., 5 seconds) and then press the specified key (e.g., the space bar) (i.e., time estimation).

[0082] The questions that require focus have the following two configurations:

[0083] 1) The screen is divided into two parts, left and right. On the left screen, you need to move the cursor with your hand to track the moving ball (i.e., tracking task). At the same time, when different colors appear on the right, you need to press the corresponding number (e.g., "green" is 1, "red" is 2, "yellow" is 3) (i.e. color reaction).

[0084] 2) The screen is divided into left and right parts. On the left screen, you need to move the cursor with your hand to track the moving ball (i.e., the tracking task). At the same time, you will hear three different frequencies of sound (e.g., 550Hz, 650Hz, and 750Hz). During the test, when the sound frequency you hear is one of these frequencies (e.g., 650Hz), press the designated button (e.g., press the designated button with your right hand) (i.e., the sound response).

[0085] The thinking ability assessment module includes: a sub-module for arithmetic operations, a sub-module for numerical reasoning, and a sub-module for image ability (i.e., image reasoning).

[0086] The arithmetic module is configured to perform arithmetic operations on multiple (e.g., three) numbers on the screen, with the final result being a single digit. After the question appears, the user inputs the result of the arithmetic operation to test the accuracy of the result and the input speed.

[0087] The numerical reasoning module is configured to deduce the correct answer that conforms to the pattern based on existing numbers, in order to test the numerical reasoning ability of the test subjects.

[0088] The image reasoning module is configured to deduce the correct answer that conforms to a pattern based on existing images, in order to test the image reasoning ability of the test subjects.

[0089] The spatiotemporal rotation capability detection module includes a spatiotemporal rotation capability submodule; the spatiotemporal rotation capability submodule is configured to select an image that can be obtained by rotating a specified 3D image within a certain time (e.g., 120s).

[0090] The perceptual speed ability detection module includes a digit recognition ability submodule, a digit reaction ability submodule, a color reaction ability submodule, and a sound reaction ability submodule;

[0091] The digit recognition ability submodule is configured to test the digit recognition ability of the test subject by having them press the corresponding number based on the displayed artistically processed image.

[0092] The digital reaction ability submodule is configured as follows: letters or numbers appear randomly on the screen, each appearing for a certain duration (e.g., 1 second). When a number is seen, the space bar is pressed immediately, and no action is taken when a letter is seen, in order to test the digital reaction ability of the test subject.

[0093] The color reaction ability submodule is configured to display different colors on the screen, and press the corresponding button according to the displayed color (e.g., "green" is 1, "red" is 2, "yellow" is 3) to test the color reaction ability of the person being tested.

[0094] The sound response ability submodule is configured such that when a subject hears a sound of different frequencies, they press the corresponding button according to the pitch they hear (e.g., "550Hz" is 1, "650Hz" is 2, and "750Hz" is 3) to test the subject's sound response ability.

[0095] The professional capability testing module includes sub-modules for dashboard recognition capability, scale estimation capability, special shape tracking capability, moving target tracking capability, scan comparison capability, position recognition capability, time estimation capability, and speed estimation capability.

[0096] The dashboard recognition capability submodule is configured as follows: Given a reference scale line (e.g., 0 degrees) and the degree of a certain scale line in the dial, determine the degree of another scale line.

[0097] The scale estimation capability submodule is configured as follows: a bar scale appears in the upper half of the screen, with two scale values ​​already marked, and another marker (such as a triangle marker) points to a certain scale line. The scale value pointed to by the marker (triangle marker) is estimated.

[0098] The special shape tracking capability submodule is configured as follows: a movable object (such as a small airplane) and a special curved trajectory appear on the screen. The movable object (small airplane) will move towards the cursor position at a fixed speed. The test subject needs to control the movable object (small airplane) to move along the special curved trajectory in the figure, so that the movable object (small airplane) can circle the special curved trajectory once.

[0099] The moving target tracking capability submodule is configured as follows: a triangle and a circle appear on the screen. The triangle moves according to a built-in function, and the person being tested needs to manually control the movement of the circle and try to keep the triangle in the center of the circle.

[0100] The scanning comparison capability submodule is configured as follows: Two rows of paired graphics appear on the screen. The test subject needs to compare them and identify a pair of graphics with exactly the same shape (without needing to rotate), and then press the column number of the pair of graphics with the same shape.

[0101] The location recognition capability submodule is configured as follows: the person being tested observes the displayed coordinate system, reads the (x, y) coordinates of a specified cell (e.g., the red cell), and inputs the x and y values ​​using the keyboard.

[0102] The time estimation capability submodule is configured as follows: a certain number (e.g., the number 0) appears on the screen accompanied by interference sounds (e.g., "beep" interference). After the number (0) appears, the test subject needs to silently count to a set duration (e.g., 5 seconds) and then press a designated key (e.g., the space bar).

[0103] The speed estimation capability submodule is configured as follows: A uniform linear motion object (e.g., a ball) and a target object (e.g., a wall) appear on the screen. The motion speed is random. After the ball has traveled a certain distance (1 / 2 or 2 / 3), it will disappear, but in reality, it will continue to move at the original speed. The test subject needs to estimate the time when the ball reaches the target wall and press a specified key (e.g., the space bar) at that time.

[0104] The performance metrics for all the above questions are divided into three types: for action-related tasks, the mean squared error of distance (MSE); for questions that examine correctness, the accuracy rate; and for all other tasks except action-related tasks, reaction time is recorded.

[0105] To improve the credibility and reliability of the results, the system employs a two-tiered module approach for weighting (i.e., Figure 1 The visual feature dual-layer weighting module (in the text) has two layers. The core theory of the first layer is the interval intelligence evaluation method, and the core theory of the second layer is the entropy weight method. The function of the first layer module is to realize the weighting of visual features such as... Figure 5 The module weight allocation in the criterion layer primarily relies on the fuzzy scoring results of the expert team, with module weights obtained after defuzzification. Since new experts are added to the scoring and weighting process, the expert scoring system iteratively updates the weight values ​​of the criterion layer modules based on the scores provided by these new experts. The second-layer module's function is to implement the weight allocation for the indicator layer modules; changes in the system parameters correspond to updates to the indicator layer weights. The final weight of the second-layer module in the global context is shown below. The weights of the first-layer modules (w) i Multiply by the weight of the second-level module under the first-level module (w) i-j To characterize, that is It is important to note that the system assigns weights (w) to the first-layer modules. i The second-layer module corresponds to the weights (w) in the first-layer model. i-j The weighting process is relatively independent. The system adjusts the weights of the second-layer modules based on the visual feature analysis results. The adjustment targets the weights of the second-layer modules under the corresponding first-layer modules, i.e., w. i-j .

[0106] The second-layer module (i.e., the test result acquisition module; in addition to the modules mentioned above, the system also includes a consistency verification module and a first weight value update module) is configured to acquire the test score data (e.g., the pilot to be tested for cognitive ability) obtained after passing each test indicator test in each test module of the cognitive ability testing unit. Figure 2 As shown in the figure, the first data is used as the first data; the first data is normalized to obtain normalized data; based on the first data, the information entropy is calculated, and the information entropy is multiplied by the corresponding normalized data, and the result of the multiplication is used as the second data; the second data is converted into a percentage system and used as the final detection result corresponding to each detection indicator of the pilot; the specific content is as follows:

[0107] Based on the scores obtained by pilots after completing the tests in the system, the calculation process for assigning weights to each module in the indicator layer is as follows:

[0108] For the test index j, the transformation formula for the i-th record value (i.e., the test score data) is:

[0109]

[0110] Where, x ij This represents the test score data of pilot i corresponding to test index j for the cognitive ability test, x min,j and x max,j Let x be the minimum and maximum values ​​corresponding to the detection index j. i ′ j Represents the normalized x ij (That is, the test results data are transformed according to the distribution so that the value range is within the range of [0,1].)

[0111] The formula for calculating the information entropy of detection index j is as follows:

[0112]

[0113] in, This indicates that for detection index j, the i-th record value appears as x. ij The probability, k = 1 / ln(n) > 0, is given by the normalized result of introducing information entropy as x″. ij =x″ ij *e.

[0114] The final score for a test item with one detection indicator is its normalized value converted to a percentage. For test items with two or more detection indicators, the final score is calculated by summing the normalized values, dividing by the sum of the information entropies corresponding to the indicators, and then converting to a percentage, as shown in the following formula:

[0115] Where x″ ij S represents the normalized test score of pilot i corresponding to test indicator j, i.e., the second data corresponding to the j-th test indicator of pilot i, j = 1, 2, ..., m, where m is the number of test indicators in the test questions. test The test score is the final test result corresponding to two or more test indicators.

[0116] Similarly, the overall detection results are calculated in each dimension using the following formula:

[0117] The first-layer module (i.e., the consistency verification module and the first weight value update module) is configured to form an index judgment matrix based on each final detection result, and then construct an interval intelligence matrix. It then uses a de-intelligentization method to transform the true, false, and uncertain intelligence intervals contained in the interval intelligence matrix into single calculated values, thus constructing a de-intelligentization judgment matrix. Here, the true, false, and uncertain intelligence intervals correspond to the membership degree T. A Hesitation level I A Membership degree F A , together represent a set of intelligences in an interval A. Each set of intelligences in an interval contains true, false, and uncertain intelligence intervals in a fixed order. De-interference is a method of defuzzification that transforms the three intelligence intervals into a single value; as shown in the following formula (3).

[0118] Based on the de-intermediation judgment matrix, the consistency index CI is calculated; the average random consistency index RI corresponding to CI is found according to the interval intermediation evaluation language scale, and the consistency ratio CR is obtained by dividing the RI and RI; it is determined that the CR is less than the set consistency ratio threshold. If not, the index judgment matrix is ​​modified (specifically, the modification method is to reorganize the expert scoring, explain to the experts the current situation of unqualified consistency test, and provide a detailed and accurate introduction to the concepts of each module. Based on their experience, cognition and understanding of the current situation, the experts re-score according to the language scale standard. After the scoring results are collected, the defuzzification process is repeated to obtain the modified index judgment matrix. Then, the consistency test is re-performed until the test passes, that is, the de-intermediation judgment matrix is ​​reconstructed and the consistency ratio is calculated). The first weight update module is configured to normalize each element in the decentralization judgment matrix to obtain normalized elements; calculate the arithmetic mean of each row of the decentralization judgment matrix based on each normalized element, and update it as the final weight value of each detection module; the interval-based intelligence evaluation language scale is a preset mapping relationship of relative importance between two detection modules; the specific content is as follows:

[0119] The core of the interval intelligence evaluation method is the assumption that A is an interval intelligence set in the universe of discourse E, determined by the membership function T. A Hesitation function I A Non-membership function F A They jointly stated:

[0120] A = { <T A (x), I A (x), F A (x)>|x∈E} (3)

[0121] Among them, T A (x), I A (x) and FA (x) represents the membership degree, hesitation degree, and non-membership degree, respectively, and is a subset of [0, 1], where 0 ≤ sup[T]. A (x)]+sup[I A (x)]+sup[F A [x]≤3, T A The lower limit of (x) is The upper limit value is If T A The closer (x) is to 1, the more it indicates that x belongs to T. A The higher the degree.

[0122] The expert scoring system scores each module of the criteria layer based on the existing test results data in the current system database. The Interval Central Intelligence Evaluation Language Scale is the reference standard for the expert system to score the importance of each cognitive dimension. The scale is based on the relative importance between two test elements and is represented by an interval central intelligence scale. The scale contains 11 importance levels, including equal importance, absolutely weak importance, very weak importance, and relatively weak importance. The values ​​belong to a range, such as <[0.50, 0.60], [0.35, 0.45], [0.40, 0.50]> in absolutely weak importance. Equal importance is represented by [0.50, 0.50], which means that its value is 0.50. The specific content is shown in Table 1 (Interval Central Intelligence Evaluation Language Scale). Each importance level corresponds to a specific interval central intelligence scale.

[0123] Table 1

[0124]

[0125]

[0126] Based on the specific importance scores from the expert system, the importance rating results are matrixed to form an indicator judgment matrix (which can be adjusted according to actual circumstances):

[0127]

[0128] In the indicator judgment matrix, the element in the i-th row and j-th column represents the detection module C. i Compared to detection module C j Importance level.

[0129] Next, each dimension of the judgment matrix corresponds to an importance level. Referring to the interval-based intelligence evaluation language scale, the importance levels in the judgment matrix are replaced with their corresponding interval-based intelligence sets, thus forming an interval-based intelligence matrix. Then, using the de-intelligence formula, the true, false, and uncertain intelligence intervals contained in each matrix element are transformed into single calculated values, forming the de-intelligence judgment matrix, along with each element U in the matrix.

[0130]

[0131] in, Indicate I A The corresponding upper limit value, F represents A The lower limit value, F represents A The upper limit.

[0132] Based on the attributes of the judgment matrix, the consistency index (CI) is calculated using the following formula:

[0133]

[0134] Where, λ max The largest eigenvalue is the eigenvalue of a matrix with the largest absolute value. The largest eigenvalue can be calculated using the following steps: Given an n-order matrix A, represented as A = [a_{ij}], where 1 <= i, j <= n; use the eigenvalue solving method to calculate the eigenvalues ​​λ of matrix A. The eigenvalues ​​satisfy the following equation: A*v = λ*v, where v is a non-zero vector. Solving the eigenvalue equation yields all eigenvalues ​​λ_1, λ_2, ..., λ_n of matrix A; by comparing the absolute values ​​of the eigenvalues, find the eigenvalue with the largest absolute value, i.e., the largest eigenvalue, where n is the matrix order. Based on the interval-based intelligence evaluation language scale, find the average random consistency index RI corresponding to the calculated result CI, and divide by RI to obtain the consistency ratio CR. If the CR meets the set consistency ratio threshold (preferably 0.1 in this invention), the consistency of the judgment matrix can be considered acceptable; otherwise, the index judgment matrix needs to be modified, and expert review and scoring need to be reorganized until the consistency test passes.

[0135] If the consistency check passes, then according to the normalization calculation formula, each element in the decentralization judgment matrix is ​​processed to obtain the normalized element N. ij :

[0136]

[0137] Calculate the arithmetic mean of each row of the matrix:

[0138]

[0139] Among them, W i U represents the final weight value of the i-th detection module. ij This represents the element in the i-th row and j-th column of the decentralization judgment matrix.

[0140] Since each row of the matrix corresponds to a cognitive ability index, W at this point... i This is the final weight value of the i-th detection module.

[0141] The system for detecting the multi-dimensional cognitive abilities of civil aviation pilots also includes a second weight update module;

[0142] The second weight update module is configured to acquire the visual features of the detection links corresponding to the sub-modules in each detection module and the visual features of each sub-link in the simulated flight scenario. The eye movement indicators corresponding to the visual features include fixation duration percentage, average fixation duration, saccade amplitude, left eye pupil diameter, and right eye pupil diameter. Based on the visual features of each sub-link in the simulated flight scenario, the importance parameters of the detection links corresponding to the sub-modules in each detection module are predicted through a pre-constructed multiple linear regression analysis model. Based on the importance parameters, a weight increase list and a weight not adjusted list corresponding to the sub-modules of each detection module are acquired, and then the weights corresponding to the sub-modules of each detection module are adjusted (specifically: the adjustment strategy is: all sub-modules of the detection module are adjusted simultaneously, the weight adjustment step size is set to α, and the number of people in a unit batch is X (preferably set to 100 in this application), that is, the default number of newly added detection people in each batch is 100. If the number is less than or more than this number, the step size is α* The weight is defined as (actual number of participants / 100), and the weight adjustment direction is increased based on its importance value. An update iteration is performed according to the step size. The weight iteration step size α is set to 0.05. For example, if the initial weight of the motor memory submodule under the memory ability module is 0.27, and if motor memory is listed in the significant cognitive ability module list after the eye movement structure analysis results of 100 test subjects (unit batch) are published, then the weight value of the motor memory submodule under the memory ability module is adjusted upwards by 0.05*(100 / 100) = 0.05, and the adjusted weight is 0.32. This operation is repeated as the number of test subjects increases. The average of the updated weight values ​​of the submodules of each detection module after the (N-1)th iteration, the unupdated weights of the submodules of each detection module after the Nth iteration, and the adjusted weights of the submodules of each detection module after the Nth iteration is used as the updated weight value for the Nth iteration, and then updated (i.e....). Figure 1 (Parameter update module in the middle); where N represents the current iteration. Specifically as follows:

[0143] When updating the weights of each module (i.e., sub-module) in the indicator layer, visual feature information is introduced to assist in weight adjustment decisions. Since cognitive ability detection is designed to ensure flight safety in actual flight scenarios, the weights of each cognitive ability dimension should be considered to be significantly (or important) related to performance during flight. Therefore, two scenario paradigms are created: one is the cognitive ability detection scenario, and the other is the simulated flight scenario. The simulated flight scenario is implemented using a flight simulator (the simulator model depends on the actual application scenario). Initial visual signals are collected in both scenarios using an eye-tracking device (eye tracker).

[0144] The cognitive ability testing scenario is divided into multiple sub-sections according to various question types. Among them, four question types—"tracking task + motor memory," "tracking task + time memory," "tracking task + color reaction," and "tracking task + sound reaction"—specifically test the pilot's attention allocation ability under "dual-task" conditions. Therefore, these were merged into a single sub-section during sub-section segmentation. This resulted in 22 cognitive testing scenario sub-sections, numbered 1-22. The simulated flight scenario includes the two most demanding operational procedures in flight: takeoff and approach / landing. Takeoff is divided into three sub-stages based on pilot operation characteristics: taxiing to takeoff, initial climb from 0-200ft, and climb above 200ft. Approach and landing are divided into five sub-stages: capture glide to 200ft above the runway, final approach from 200-100ft, pre-touch-off from 100-20ft, touch-off from 20-0ft, and post-touch-off. These eight sub-stages are numbered sequentially as sub-stages 23-30, for a total of 30 sub-stages, as shown in Table 2 (all sub-stages and their corresponding numbers).

[0145] Table 2

[0146]

[0147] Each sub-stage in the cognitive ability test and simulated flight scenario contains a segment of raw visual signal information, presented in video format, which is difficult to quantify. Therefore, four eye-tracking metrics are selected: fixation duration percentage, average fixation duration, saccade amplitude, and pupil diameter, to describe the pilot's visual behavior characteristics in each sub-stage. Thus, a vector containing five values ​​(where pupil diameter is divided into left and right eye values) can be used to characterize the visual features of each sub-stage, as shown in the following equation:

[0148]

[0149] Among them, a i a represents the visual feature of the detection stage corresponding to the i-th detection indicator (i.e., sub-module or sub-stage), where a i1 To a i5These represent the values ​​of the eye movement indices corresponding to the visual features.

[0150] Subsequently, by performing multiple linear regression analysis on the vectors of each sub-stage in the two scenarios, the importance of the cognitive ability dimension in each sub-stage of the flight scenario was calculated.

[0151] In the cognitive ability testing scenario, the sub-stage vectors corresponding to the 22 cognitive ability dimensions (i.e., the visual features of the testing stages corresponding to the sub-modules) are used as independent variables in the regression analysis. When a sub-stage vector in a certain simulated flight scenario is selected as the dependent variable, the following multiple linear regression model is constructed.

[0152] a i = w1a1 + w2a2 + w3a3 + ... + w 22 a 22 +ε,i=23,24,…,30

[0153] Among them, a i The vector representing the i-th detection stage, i.e., the visual features of each sub-stage in the simulated flight scenario, w i ε represents the importance parameter of the i-th vector, which is the importance parameter of the detection link corresponding to the submodule in each detection module, and ε represents the set error value.

[0154] Multiple linear regression can be used to predict the importance coefficients and rankings of each indicator layer module (i.e., sub-module). If, in a certain flight process sub-stage, the number of sub-module vectors with non-zero importance coefficients exceeds 5 (M is preferably set to 5), then the top five sub-modules are selected as significant cognitive ability modules. If there are fewer than 5, then all sub-modules with non-zero importance coefficients are selected as significant cognitive ability modules. Sub-modules with an importance coefficient of 0 are determined to be insignificant cognitive ability modules. Therefore, significant and insignificant cognitive ability modules in each flight process sub-stage can be identified. For significant cognitive ability modules appearing in more than 3 sub-stages, they are determined to be significant cognitive ability modules throughout the entire process and included in the weight-increased list; for insignificant cognitive ability modules appearing in more than 3 sub-stages, they are determined to be significant cognitive ability modules throughout the entire process and included in the weight-unadjusted list.

[0155] Based on the list of weight increases and decreases for cognitive ability modules, the weights of each module are adjusted sequentially to obtain the relatively optimal weighting result. The adjustment method is as follows: all cognitive ability modules are adjusted simultaneously, with a weight adjustment step size of 0.02 (the default number of new experimental samples in each batch is 100; if the number is less or more than this, the step size is set to 0.02 * (actual number of people / 100)). The direction of weight adjustment (increase or decrease) is determined according to its importance value, and an update iteration is performed according to the step size.

[0156] As the number of pilots undergoing cognitive ability testing increases, their physiological signal data is continuously collected and added to the analysis dataset, enabling iterative updates to the weights of the indicator-level modules. After the Nth batch of pilots receiving their cognitive ability test scores, the following weights are generated: the cognitive module weights after the (N-1)th iteration (i.e., the updated and corrected weights of each sub-module in the (N-1)th batch), the cognitive module weights calculated for the Nth batch (i.e., the weights of each sub-module before being updated and corrected based on eye-tracking signal analysis), and the cognitive module weights for the Nth batch after correction based on eye-tracking signal analysis (i.e., the adjusted weights of the aforementioned sub-modules based on eye-tracking signal correction). The average of these three weights yields the final weight value for the Nth batch. This process is repeated, and eventually, as the number of pilots undergoing cognitive ability testing and the number of batches increase, the weight changes will tend to stabilize. For example, after three batches of experimental data updates and iterations, the weight of the motor memory submodule under the memory ability module is 0.27. After the fourth batch of experimental participants obtained their scores, the weight was updated to 0.29 based on the test results. Based on the eye movement data analysis results of the fourth batch of experimental participants, the weight was further adjusted to 0.28. Therefore, after the fourth batch of tests, the final weight of the motor memory submodule under the memory ability module is (0.27+0.29+0.28) / 3=0.28.

[0157] After updating the system's dimensional weighting system based on visual information and test results, the next pilot awaiting cognitive ability testing will receive a cognitive ability test score based on the current system weighting system. After receiving the score, the pilot can choose whether to accept it. If satisfied, the score will be recorded in the system database. As new scores from pilots awaiting cognitive ability testing are collected, the system continuously updates the weights of the current indicator layer modules. This process repeats and updates as the number of pilots awaiting cognitive ability testing increases.

[0158] The process involves simultaneously testing multiple pilots whose cognitive abilities are to be assessed, such as... Figure 4 As shown in the diagram, pilots 1 to N undergo cognitive ability testing on testing devices 1 to N respectively, generating 1 to N operation records. All operation records of the pilots undergoing cognitive ability testing are sent to the back-end server (i.e., the computer device) for scoring, generating N test scores. The test scores are ultimately entered into the database and displayed through the front-end interface, as shown in the diagram. Figure 4 As shown.

[0159] During the testing process, this system records the pilots' answers and processes for each cognitive ability test, based on the different questions. The main recorded information includes name, test paper name, question category, answer time, correctness, answer content, test time, and question number. The answer time is measured in milliseconds. It's important to note that for questions assessed by answer time, the answer time is recorded as 0, and the time is recorded in the answer content, such as the question "Tracking Task (Left) Sound Response (Right)". The order in which the questions appear is randomized during the testing process. The specific content of the answers varies depending on the question, recording the time required for each answer and the specific content of each answer for the pilots undergoing the cognitive ability test. The responses included three types: first, the "KBJ0" response in the listening data breadth question, where the pilot being tested answered "KBJ0" based on what they heard; second, a dual-task question, where the response contained a sequence of numbers in quotation marks, representing the standard deviation of the distance between the finger touch point and the aircraft pixel position measured every 0.5 seconds; and third, a situational awareness question, where the numerical part of the response represented the time the finger touched the screen, with "true" and "false" indicating the aircraft's position every 0.5 seconds—true for being in the white irregular corridor and false for being in the red restricted area. This section displayed the raw test data and... Figure 4 The pilot's operation record shown is equivalent to the one to be tested for cognitive ability. After obtaining the raw data, the raw data is imported into the database of the computing device for corresponding score calculation and processing to obtain the score.

[0160] This invention provides a multi-dimensional cognitive ability testing system for pilot selection, aiming to offer a more scientific, accurate, and fair testing method for the pilot selection process. First, the system employs multiple testing dimensions, such as professional skills, memory, thinking ability, attention, spatiotemporal rotation ability, and speed perception, to assess candidate pilots' performance in different cognitive aspects. Compared to traditional selection methods, this system can comprehensively and meticulously test pilots' cognitive abilities, thereby better determining their suitability for flight work. Second, the system utilizes intelligent and digital testing technology to improve the scientific rigor, accuracy, and objectivity of the testing. The system uses the results of eye-tracking signal comparison analysis during cognitive ability system testing and simulated flight as a reference, combined with visual eye-tracking characteristics for optimization and adjustment. Furthermore, it uses normalization and entropy weighting methods to score different testing indicators to better identify outstanding pilots. Finally, the system can automate and meet the selection requirements of multiple personnel, multiple companies, and multiple levels, thereby significantly reducing labor and time costs and improving selection efficiency. Simultaneously, this system can also be used in conjunction with other selection methods to improve selection effectiveness and reliability. Therefore, the multi-dimensional cognitive ability detection system of the present invention has broad application prospects and great promotion and application value in pilot selection and related fields.

[0161] It should be noted that the multi-dimensional cognitive ability detection system for civil aviation pilots provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the various modules or steps and are not considered as an improper limitation of the present invention.

[0162] Those skilled in the art will recognize that the module steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The program corresponding to the software module can be placed in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the invention.

[0163] The terms “first,” “second,” “third,” etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0164] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A multi-dimensional cognitive ability detection system for civil aviation pilots, the system comprising a cognitive ability detection unit, said cognitive ability detection unit including a spatiotemporal rotation ability detection module, a perceptual speed ability detection module, a professional ability detection module, an attention ability detection module, a memory ability detection module, and a thinking ability detection module; characterized in that, The system also includes: The test result acquisition module is configured to acquire the test score data obtained by the pilot to be tested after passing each test index in each test module of the cognitive ability test unit, and use it as the first data. Normalize each first data point to obtain normalized data; calculate the information entropy based on each first data point, and multiply the information entropy by the corresponding normalized data, and use the result of the multiplication as the second data point; convert each second data point into a percentage system, and use it as the final detection result corresponding to each detection indicator of the pilot. The consistency verification module is configured to form an index judgment matrix based on each final detection result, and then construct an interval intelligence matrix. It then uses a de-intelligentization method to transform the true, false, and uncertain intelligence intervals contained in the interval intelligence matrix into single calculated values, constructing a de-intelligentization judgment matrix. Here, the true, false, and uncertain intelligence intervals correspond to the membership degrees. Hesitation level Membership degree Together they represent a set of intelligences within an interval. ; Based on the decentralization judgment matrix, the consistency index CI is calculated; the average random consistency index RI corresponding to CI is found according to the interval central intelligence evaluation language scale, and the consistency ratio CR is obtained by dividing the RI and RI; it is determined that the CR is less than the set consistency ratio threshold. If not, the index judgment matrix is ​​modified. The first weight value update module is configured to normalize each element in the decentralization judgment matrix to obtain normalized elements; and calculate the arithmetic mean of each row of the decentralization judgment matrix based on each normalized element, and update it as the final weight value of each detection module. The interval-based intelligence evaluation language scale represents a pre-defined mapping relationship between the relative importance of two detection modules. The second weight update module is configured to obtain the visual features of the detection links corresponding to the sub-modules in each detection module and the visual features of each sub-link in the simulated flight scenario. The eye movement indicators corresponding to the visual features include fixation duration ratio, average fixation duration, saccade amplitude, left eye pupil diameter, and right eye pupil diameter. Based on the visual features of each sub-stage in the simulated flight scenario, the importance parameters of the detection stages corresponding to the sub-modules in each detection module are predicted by a pre-constructed multiple linear regression analysis model. Based on the importance parameters, obtain the weight increase list and weight not adjusted list corresponding to the sub-modules of each detection module, and then adjust the weights corresponding to the sub-modules of each detection module. Calculate the average of the updated weight values ​​of the sub-modules of each detection module after the (N-1)th iteration, the unupdated weights of the sub-modules of each detection module after the Nth iteration, and the adjusted weights of the sub-modules of each detection module after the Nth iteration. Use this average as the updated weight value for the Nth iteration, and then update the module accordingly. Here, N represents the current iteration. The method for predicting importance parameters is as follows: ； in, Indicates the first The vectors of each detection stage, i.e., the visual features of each sub-stage in the simulated flight scenario. Indicates the first The importance parameters of each vector, that is, the importance parameters of the detection steps corresponding to the sub-modules in each detection module. This indicates the set error value; The weights corresponding to the sub-modules of each detection module are adjusted as follows: In any sub-stage of a simulated flight scenario, if the If the number of non-zero values ​​is greater than the set first threshold, the first M sub-modules are selected as significant cognitive ability modules; otherwise, all sub-modules with non-zero importance coefficients are selected as significant cognitive ability modules; sub-modules with zero importance coefficients are selected as non-significant cognitive ability modules. For significant cognitive ability modules that appear in sub-stages of more than a set number of simulated flight scenarios, add them to the weight-enhanced list; for insignificant cognitive ability modules that appear in sub-stages of more than a set number of simulated flight scenarios, add them to the weight-unadjusted list. Based on the weight increase list and the weight not adjusted list, the weights corresponding to the sub-modules of each detection module are adjusted. The adjustment strategy is as follows: all sub-modules of the detection module start adjusting simultaneously, and the weight adjustment step size is set to... The unit batch size is X people, meaning the default number of new testers added in each batch is X person-times. If the number is less than or more than this, the step size will be adjusted according to... (Actual number of people / X) is defined, and the weight is adjusted upwards based on its importance value. An update iteration is performed according to the step size.

2. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 1, characterized in that, The information entropy is calculated as follows: ； in, Indicates the detection indicators The corresponding information entropy, , indicating the detection index No. The record value appears The probability, Pilots awaiting cognitive ability testing Corresponding detection indicators The test results data, , where represents the coefficient, and n represents the number of pilots participating in the detection.

3. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 2, characterized in that, The final test results corresponding to each of the pilot's test indicators are as follows: ,in, Indicates the second data; The final test results for two or more test indicators are as follows: ； in, Pilots awaiting cognitive ability testing Corresponding detection indicators The normalized test score, i.e., the pilot's cognitive ability test No. The second data corresponding to each detection indicator , To determine the number of test indicators in the test question, The test score is the final test result corresponding to two or more test indicators.

4. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 2, characterized in that, The method for constructing the intelligence matrix in the interval is as follows: Based on the final test results, specific importance scores are obtained, and the importance scores are matrixed to form an indicator judgment matrix. Each dimension in the indicator judgment matrix is ​​determined to have a corresponding importance level. By referring to the pre-constructed interval intelligence set evaluation language scale, the importance level in the indicator judgment matrix is ​​replaced with the corresponding interval intelligence set, thereby forming an interval intelligence matrix.

5. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 4, characterized in that, The elements in the decentralization and intelligent judgment matrix are: ； in, This represents the elements in the de-intelligence judgment matrix. Indicates membership degree The lower limit value, Indicates membership degree The upper limit, Indicates the degree of hesitation The upper limit, Indicates non-membership degree The lower limit value, Indicates non-membership degree The upper limit.

6. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 5, characterized in that, The consistency index is obtained as follows: ； in, Indicators of consistency This represents the largest eigenvalue, which is the eigenvalue of the matrix with the largest absolute value. This indicates the order of the matrix.

7. The detection system for multi-dimensional cognitive abilities of civil aviation pilots according to claim 6, characterized in that, The arithmetic mean of each row of the decentralization judgment matrix is ​​calculated based on each normalized element, and used as the final weight value of each detection module. The method is as follows: ； ； in, Indicates the first i The final weight value of each detection module, Represents the normalized element. In the decentralization judgment matrix, the first... i Line number j The elements of the column.

Images