Eye movement data monitoring and analysis method for flight crew spatial orientation training

Abstract

The present invention discloses a method for monitoring and analyzing eye movement data of flight personnel's spatial orientation training, which belongs to the technical field of eye movement data analysis. The method comprises the following steps: S1: obtaining historical eye movement data of spatial orientation training and constructing a data distribution model; S2: dividing a number of functional areas and constructing a historical transfer probability matrix; S3: obtaining current eye movement data and projecting it into the data distribution model to construct a current transfer probability matrix; S4: calculating the difference matrix between the current transfer probability matrix and the historical transfer probability matrix to determine the deviation degree of functional area transfer selection. The present invention overcomes the problems that the prior art cannot carry out targeted regional division and analysis of eye movement data. By constructing a data distribution model, constructing a historical transfer probability matrix and a current transfer probability matrix, dividing the functional areas and conducting targeted analysis, a targeted optimization direction is provided for the flight personnel's spatial orientation training, and the training effect and spatial orientation ability are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of eye movement data analysis, and in particular to a method for monitoring and analyzing eye movement data for flight personnel in spatial orientation training. Background Art

[0002] In the field of flight training, the assessment of the spatial orientation ability of flight personnel is of vital importance. Among them, analyzing the information acquisition ability of flight personnel when facing complex instrument data is a key content. In this process, the analysis of eye movement data can provide a lot of valuable information, which can help understand the visual focus and information processing mode of flight personnel.

[0003] However, existing eye movement data analysis technologies have significant limitations. When processing eye movement data, current technologies usually compare the data points of eye movement data one by one, or use deep learning models for fuzzy processing. However, existing deep learning models are rarely used in analyzing eye movement data, and the analysis results are unstable. Existing technologies generally lack effective means for regional division and analysis of eye movement data. This means that when facing a flight cockpit environment containing many instruments, it is impossible to conduct targeted regional analysis of eye movement data based on the location of instrument data. When calculating the distance relationship between data points, traditional methods often ignore the key factor of the gaze duration of the data points. In actual flight training, the gaze duration of different data points can reflect the degree of attention and processing time of the flight personnel to the information in the area. If the gaze duration is not considered when calculating the distance, it is impossible to accurately measure the weight of different data points for the overall analysis, thereby affecting the judgment of the flight personnel's behavior pattern. Inappropriate classification and analysis of the sight transfer path cannot accurately evaluate the data viewing speed of the flight personnel, which is not conducive to optimizing the training content and methods.

[0004] These defects are particularly prominent in the spatial orientation training of flight personnel. The eye movement patterns of flight personnel may change during different stages of training. If the eye movement data cannot be analyzed by region, it will be difficult to accurately judge the performance of flight personnel in finding data in different functional areas, and it will be impossible to clearly understand their ability to find data in each functional area. This has created a major obstacle to comprehensively evaluating the information processing capabilities of flight personnel in complex flight environments, and then formulating personalized training plans to improve their spatial orientation capabilities. Therefore, it is urgent to develop an analysis method that can effectively divide eye movement data into regions to meet the needs of flight personnel spatial orientation training evaluation. Summary of the invention

[0005] The present invention overcomes the problem that the prior art cannot perform targeted regional division and analysis of eye movement data. A method for monitoring and analyzing eye movement data for flight personnel spatial orientation training is provided. By constructing a data distribution model, processing and associating historical eye movement data, and constructing a historical transfer probability matrix and a current transfer probability matrix, the attention changes of flight personnel in functional area transfer can be clearly reflected. By reasonably dividing functional areas and conducting targeted analysis, targeted optimization directions can be provided for flight personnel's spatial orientation training, thereby effectively improving training effects and spatial orientation capabilities. Through accurate eye movement data monitoring and analysis and training optimization, it is ensured that flight personnel can quickly and accurately obtain various instrument data in actual flight, make timely and correct decisions, and enhance flight safety assurance capabilities.

[0006] In order to achieve the above object, the present invention adopts the following scheme:

[0007] The method for monitoring and analyzing eye movement data of flight personnel spatial orientation training is characterized by comprising the following steps:

[0008] S1: Acquire historical eye movement data of spatial orientation training, wherein the historical eye movement data includes fixation point, fixation duration and sight transfer path, generate historical data points with the fixation point of the historical eye movement data as coordinates and the fixation duration as distribution value, and construct a data distribution model with the historical data points;

[0009] S2: In the data distribution model, a number of functional areas are divided according to the position of the aircraft instrument data, and the number of gaze transfers between all functional areas is counted according to the gaze transfer paths in the historical eye movement data to construct a historical transfer probability matrix;

[0010] S3: obtaining the current eye movement data of the flight crew performing the spatial orientation training task through an eye tracker, wherein the current eye movement data includes the gaze point, gaze duration and gaze transfer path, wherein the gaze transfer path includes a starting point and a landing point, projecting the current eye movement data onto the data distribution model to obtain the current data point, and counting the number of gaze transfers between all functional areas according to the gaze transfer path in the current eye movement data to construct a current transfer probability matrix;

[0011] S4: Calculate the difference matrix between the current transfer probability matrix and the historical transfer probability matrix, and determine the deviation degree of the function area transfer selection for the spatial training task according to the difference matrix;

[0012] In step S2, the historical transition probability matrix is ​​constructed as follows:

[0013]

[0014] Where n is the number of functional areas, in the historical transition probability matrix P, element prs is the probability of transferring from the rth functional area to the sth functional area, , B is the total number of gaze transfer paths across functional areas in the historical eye movement data, B rs is the number of sight transfer paths from the rth functional area to the sth functional area.

[0015] Preferably, the method further comprises the following steps:

[0016] S5: selecting a historical data point in each functional area as a gathering point, and classifying the gaze transfer path in the current eye movement data into a cross-area path and an intra-area path, wherein the starting point and the landing point of the cross-area path are located in different functional areas, and the starting point and the landing point of the intra-area path are located in the same functional area;

[0017] For each cross-region path, the distance between its landing point and the gathering point of the functional area is calculated and recorded as the transfer deviation;

[0018] For each path within the area, the distance between its starting point and the landing point is calculated and recorded as the search deviation;

[0019] The data viewing speed of the flight personnel is determined based on the transfer deviation and the search deviation.

[0020] Preferably, in step S5, a method of selecting a historical data point in the function area as a gathering point is:

[0021] Select any historical data point in the functional area as the initial gathering point of the functional area, take the gathering point and several historical data points closest to it as candidate points, calculate the distance between each candidate point and other historical data points in the functional area to which it belongs, weight the distance by the gaze duration of the historical data point and then average it to obtain the weighted average distance between each candidate point and other historical data points in the functional area to which it belongs, update the gathering point to the candidate point with the smallest weighted average distance, and iteratively select candidate points, calculate the weighted average distance and update the gathering point in the same way until the gathering point no longer changes.

[0022] As a preferred method, the specific method of selecting the candidate point is:

[0023] Establish a rectangular coordinate system with the current gathering point as the origin, and diverge from the origin at a continuously increasing distance until at least one historical data point falls within the divergence range in each of the four quadrants of the rectangular coordinate system. Select at most K historical data points that fall within the divergence range in each quadrant as candidate points;

[0024] The method for calculating the weighted average distance of the candidate points is:

[0025]

[0026]

[0027] Among them, d j,i For the alternative point (x j ,y j ) to the i-th historical data point (x i ,y i ), , For the alternative point (x j ,y j ) to other historical data points in the corresponding functional area, M is the weighted average distance between the candidate point (x j ,y j ) belongs to the functional area, t i is the fixation duration of the i-th historical data point, t avg It is the average fixation duration of the historical data points in the functional area to which the candidate point belongs.

[0028] Preferably, the method further comprises the following steps:

[0029] S6: For each current data point projected into the data distribution model, the average fixation duration of several historical data points adjacent to it is calculated, and the ratio of the fixation duration of the current data point to the corresponding average fixation duration is calculated. If this ratio of a current data point exceeds a preset ratio range, the current data point is marked as having an abnormal fixation duration.

[0030] Preferably, in step S6, N historical data points closest to the current data point are selected, and the ratio d of the gaze duration of the current data point to the average gaze duration of the N historical data points is calculated. The lower limit and upper limit of the preset ratio range are set to e and f respectively, and e<1<f. The method for determining whether to mark the current data point as abnormal gaze duration is as follows:

[0031] If e≤d≤f, the gaze duration of the corresponding current data point is judged to be normal;

[0032] If d<e, it is judged that the fixation time of the corresponding current data point is too short;

[0033] If d>f, it is judged that the gaze duration of the corresponding current data point is too long.

[0034] Preferably, in step S1, abnormal data processing is performed on historical data points when constructing the data distribution model, and the abnormal data processing method is:

[0035] Obtaining the gaze transfer rate and gaze transfer acceleration corresponding to the gaze transfer path of the historical eye movement data, if the gaze transfer rate of a gaze transfer path exceeds a preset rate threshold or the gaze transfer acceleration exceeds a preset acceleration threshold, determining that the gaze transfer path is an invalid path and deleting the corresponding historical data point;

[0036] If a historical data point is outside the data area of ​​all instruments or the coordinate distance from other historical data points is greater than the preset distance threshold, the historical data point is determined as abnormal data and deleted.

[0037] Preferably, in step S3, the current transition probability matrix is ​​constructed in the following manner:

[0038]

[0039] In the current transition probability matrix G, element g rs is the probability of transferring from the rth functional area to the sth functional area, , C is the total number of gaze transfer paths across functional areas in the current eye movement data, C rs is the number of sight transfer paths from the rth functional area to the sth functional area;

[0040] The difference matrix between the current transition probability matrix and the historical transition probability matrix is: T d =GP, difference matrix T d The absolute value of the elements in is the transfer selection deviation between the two functional areas.

[0041] The present invention includes at least the following beneficial effects: (1) by constructing a data distribution model, processing and associating historical eye movement data, constructing a historical transfer probability matrix and a current transfer probability matrix and calculating the difference, it is possible to clearly reflect the changes in the attention of flight personnel in transferring data information in functional areas; (2) by selecting representative gathering points in the functional area through weighted calculation, based on the analysis results of the functional area transfer deviation, it is possible to provide targeted optimization directions for the spatial orientation training of flight personnel, such as strengthening training in key areas, correcting visual transfer habits, improving information search skills, etc., thereby effectively improving the training effect and spatial orientation ability. Through accurate eye movement data monitoring and analysis and training optimization, it is possible to ensure that flight personnel can quickly and accurately obtain various instrument data in actual flight, make timely and correct decisions, and enhance flight safety assurance capabilities; (3) by shifting the line of sight to the right and left, the flight personnel can obtain the data of various instruments quickly and accurately, make timely and correct decisions, and enhance the flight safety assurance capabilities. The migration paths are classified into cross-zone paths and intra-zone paths, and the calculation of transfer deviation and search deviation provides a quantitative basis for accurately judging the data viewing speed of flight personnel, which helps to gain a deeper understanding of the efficiency of flight personnel in obtaining information in different functional areas; (4) By calculating the ratio of the average gaze duration of the current data point to the adjacent historical data points, the abnormal gaze duration can be accurately detected, fully considering the individual differences of flight personnel. The preset ratio range can also be adjusted according to their own situation and training progress to achieve personalized evaluation and adaptive monitoring; (5) Abnormal data processing is performed on historical data points. By deleting abnormal gaze transfer path data points caused by emergencies or external interference, as well as historical data points located outside the instrument data area or with abnormal coordinate distances, it is ensured that the data based on which the data distribution model is constructed truly reflects the eye movement rules under normal training conditions, thereby improving the reliability and effectiveness of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A step flow chart of the method for monitoring and analyzing eye movement data for flight personnel spatial orientation training provided by the present invention. DETAILED DESCRIPTION

[0043] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0044] like Figure 1 As shown, the eye movement data monitoring and analysis method for flight personnel spatial orientation training provided by the present invention comprises the following steps:

[0045] S1: Acquire historical eye movement data of spatial orientation training, wherein the historical eye movement data includes fixation point, fixation duration and line of sight transfer path, generate historical data points with the fixation point of the historical eye movement data as coordinates and the fixation duration as distribution value, and construct a data distribution model with the historical data points.

[0046] In practical applications, the first thing to do is to collect historical eye movement data. The historical eye movement data comes from the database of flight training records. By installing high-precision eye tracking equipment on the flight simulation training equipment, the eye movements of the flight personnel during multiple spatial orientation training can be recorded. The collected data includes information such as the flight personnel's line of sight, the location of the gaze point, and the gaze duration of each gaze point. The collected historical eye movement data can mainly use the eye movement data of flight personnel with high flying proficiency during spatial orientation training, and the data collection covers the entire training task process to ensure the reference and integrity of the historical eye movement data.

[0047] S2: In the data distribution model, several functional areas are divided according to the positions of the aircraft instrument data, and the number of gaze transfers between all functional areas is counted according to the gaze transfer paths in the historical eye movement data to construct a historical transfer probability matrix.

[0048] According to the characteristics of the training scenario, the training area is divided into multiple functional areas, such as the navigation information area, the flight attitude area, the engine parameter area, etc. The collected historical eye movement data can be preprocessed to remove noise and erroneous data. These data are then projected into the data distribution model, and each historical eye movement data point is associated with the corresponding functional area. For example, if a gaze point is within a specific coordinate range of the navigation information area, it is marked as belonging to the navigation information functional area. This way of dividing functional areas fits the layout and usage logic of instruments in actual flight operations. It can focus on the eye movement correlation of different functional-related areas, accurately grasp the habits and probabilities of flight personnel's visual attention transfer between different functional areas, and is critical to understanding the visual attention allocation pattern under conventional training conditions.

[0049] S3: The current eye movement data of the flight personnel performing this spatial orientation training task is obtained through an eye tracker, the current eye movement data includes the gaze point, the gaze duration and the sight transfer path, the sight transfer path includes the starting point and the landing point, the current eye movement data is projected into the data distribution model to obtain the current data point, and the number of sight transfers between all functional areas is counted according to the sight transfer path in the current eye movement data to construct a current transfer probability matrix. When the current eye movement data is projected into the data distribution model, the obtained current data point is annotated to distinguish it from the historical data point.

[0050] S4: Calculate the difference matrix between the current transfer probability matrix and the historical transfer probability matrix, and determine the deviation of the functional area transfer selection for this spatial training task based on the difference matrix. By calculating the difference matrix between the current transfer probability matrix and the historical transfer probability matrix, the deviation of the functional area transfer selection for this spatial training task is determined. Based on this deviation, the training organizer or the flight personnel themselves can clearly know whether there is a large deviation in the eye movement attention transfer in this training compared with previous training, so as to take targeted training adjustment measures in time, such as intensive training on key focus areas, correcting unreasonable visual transfer habits, etc., which will help improve the effect of spatial orientation training and ensure the spatial orientation ability and operation accuracy of flight personnel in actual flight.

[0051] In another technical solution, Figure 1 As shown, the following steps are also included:

[0052] S5: selecting a historical data point in each functional area as a gathering point, and classifying the gaze transfer path in the current eye movement data into a cross-area path and an intra-area path, wherein the starting point and the landing point of the cross-area path are located in different functional areas, and the starting point and the landing point of the intra-area path are located in the same functional area;

[0053] For each cross-region path, the distance between its landing point and the gathering point of the functional area is calculated and recorded as the transfer deviation;

[0054] For each path within the area, the distance between its starting point and the landing point is calculated and recorded as the search deviation;

[0055] The data viewing speed of the flight personnel is determined based on the transfer deviation and the search deviation.

[0056] In each functional area, a historical data point is selected as a gathering point. This gathering point is generally located in the center of multiple data-intensive areas to ensure that the gathering point can fully represent the location information of each historical data point in the entire functional area. When the flight personnel's sight falls on this gathering point, the required data can be quickly found at a relatively small regional search cost. The difference between the actual sight point and the gathering point is also used as an evaluation basis for the flight personnel's sight deviation. The sight transfer path is also the scanning path of the eyes, which represents the process of the flight personnel searching for the data location. When the sight is transferred from one functional area to another, the speed is relatively uniform. A sight transfer path across functional areas is used as a transfer record, and the transfer record has directionality.

[0057] For cross-zone paths, the distance between the landing point and the gathering point of the functional area is calculated as the transfer deviation. This indicator can intuitively reflect the accuracy of the flight personnel in reaching the target area when switching functional areas. The smaller the deviation, the faster and more accurately the flight personnel can locate key information when switching to a new functional area, which indirectly reflects their familiarity with the information of different functional areas and their ability to switch quickly, and thus indirectly reflects the faster data viewing speed. For intra-zone paths, by calculating the distance between the starting point and the landing point as the search deviation, the efficiency of flight personnel in searching for information in the same functional area can be effectively measured. A small search deviation means that flight personnel can quickly find the required information in the functional area, which also indicates that their data viewing speed is faster. This calculation method based on different path deviations provides a quantitative basis for accurately judging the data viewing speed of flight personnel.

[0058] In step S5, a method of selecting a historical data point in the function area as a gathering point is:

[0059] Select any historical data point in the functional area as the initial gathering point of the functional area, take the gathering point and several historical data points closest to it as candidate points, calculate the distance between each candidate point and other historical data points in the functional area to which it belongs, weight the distance by the gaze duration of the historical data point and then average it to obtain the weighted average distance between each candidate point and other historical data points in the functional area to which it belongs, update the gathering point to the candidate point with the smallest weighted average distance, and iteratively select candidate points, calculate the weighted average distance and update the gathering point in the same way until the gathering point no longer changes.

[0060] The clustering points are selected in an iterative manner, starting from any initial historical data point, and the clustering points are continuously updated to make them more representative of the overall distribution of historical data points in the functional area. This iterative mechanism can gradually exclude those points that are not relatively "central" or cannot reflect the data characteristics of the functional area well, so that the final clustering points are as close as possible to the concentrated area of ​​most historical data points in the functional area, thereby more scientifically reflecting the core characteristics of the eye movement data of the functional area. When calculating the distance between each candidate point and other historical data points in the functional area, the fixation time of the historical data points is introduced for weighting. The fixation time itself reflects the degree of visual attention paid by the flight crew to the corresponding fixation point. Giving it a weight means that when determining the clustering points, not only the spatial position relationship is considered, but also the difference in the degree of attention of different fixation points is taken into account. The clustering points selected in this way can better focus on the areas that the flight crew focuses on in the functional area, avoid the one-sidedness that may be caused by selection based solely on spatial distance, and make the clustering points more representative of the eye movement data in the functional area.

[0061] The specific method of selecting the candidate points is:

[0062] A rectangular coordinate system is established with the current gathering point as the origin, and diverges from the origin at a continuously increasing distance until at least one historical data point falls into the divergence range in each of the four quadrants of the rectangular coordinate system. In each quadrant, at most K historical data points that fall into the divergence range are selected as candidate points; the value of K is generally 2 or 3, ensuring that the number of selected candidate points is 6-10. This selection method ensures that the historical data points in the functional area are covered from different directions and different spatial positions, avoiding the omission of data points in certain areas, and ensuring the comprehensiveness of the source of candidate points. In addition, at most K historical data points are selected as candidate points in each quadrant, and a certain number of restrictions are imposed on a comprehensive basis, achieving relatively balanced selection among the quadrants, avoiding the unreasonable influence of too many or too few data points in a quadrant on the subsequent gathering point determination, making the selection of candidate points more reasonable and balanced.

[0063] The method for calculating the weighted average distance of the candidate points is:

[0064]

[0065]

[0066] Among them, d j,i For the alternative point (x j ,y j ) to the i-th historical data point (x i ,y i ), , For the alternative point (x j ,y j ) to other historical data points in the corresponding functional area, M is the weighted average distance between the candidate point (x j ,y j ) belongs to the functional area, t i is the fixation duration of the i-th historical data point, t avg The average fixation time of the historical data points in the functional area to which the candidate point belongs. The weighted average distance of the candidate point is calculated through a clear formula, and the weighted distance and weighted average distance are accurately quantified. This quantitative calculation method provides an objective and accurate basis for comparing the degree of correlation between different candidate points and other historical data points in the functional area, and can clearly determine which candidate point is more suitable as the final gathering point, avoiding the ambiguity and uncertainty of subjective judgment.

[0067] In another technical solution, Figure 1 As shown, the method also includes the following steps:

[0068] S6: For each current data point projected into the data distribution model, the average fixation duration of several historical data points adjacent to it is calculated, and the ratio of the fixation duration of the current data point to the corresponding average fixation duration is calculated. If this ratio of a current data point exceeds a preset ratio range, the current data point is marked as having an abnormal fixation duration.

[0069] In step S6, N historical data points closest to the current data point are selected, and the ratio d of the gaze duration of the current data point to the average gaze duration of the N historical data points is calculated. The lower limit and upper limit of the preset ratio range are set to e and f respectively, and e<1<f. The method for determining whether to mark the current data point as abnormal gaze duration is as follows:

[0070] If e≤d≤f, the gaze duration of the corresponding current data point is judged to be normal;

[0071] If d<e, it is judged that the fixation time of the corresponding current data point is too short;

[0072] If d>f, it is judged that the gaze duration of the corresponding current data point is too long.

[0073] By calculating the ratio of the fixation time of the current data point to the average fixation time of the adjacent historical data points, the anomaly in fixation time can be accurately found. This helps to timely detect the possible inattention or excessive attention to a certain area of ​​the pilot during the spatial orientation training of the pilot. For example, when the ratio d < e, it indicates that the pilot's fixation time on the current area is too short, and there may not be enough time to accurately read and understand the information in the area (such as an important parameter display area of ​​the aircraft instrument). When d > f, the fixation time is too long, which may mean that the pilot has encountered cognitive difficulties or unnecessary pauses in this area. Taking the historical data points as a reference, the individual differences of each pilot are fully considered. Due to factors such as experience and habits, different pilots may have different normal fixation times when looking at various aircraft instrument areas. By comparing with their own historical data, personalized evaluation standards can be established for each pilot. For example, for experienced pilots, their normal fixation time may be relatively short when observing certain familiar instrument areas, while novice pilots may need longer to understand the information in the same area. Therefore, the lower limit and upper limit of the preset ratio range can be adjusted according to the flight crew's own situation and training progress to achieve an adaptive tolerance limit.

[0074] In another technical solution, in step S1, abnormal data processing is performed on historical data points when building the data distribution model. The abnormal data processing method is:

[0075] Obtaining the gaze transfer rate and gaze transfer acceleration corresponding to the gaze transfer path of the historical eye movement data, if the gaze transfer rate of a gaze transfer path exceeds a preset rate threshold or the gaze transfer acceleration exceeds a preset acceleration threshold, determining that the gaze transfer path is an invalid path and deleting the corresponding historical data point;

[0076] If a historical data point is outside the data area of ​​all instruments or the coordinate distance from other historical data points is greater than the preset distance threshold, the historical data point is determined as abnormal data and deleted.

[0077] By deleting the historical data points corresponding to invalid paths where the speed of gaze transfer exceeds the rate threshold or the acceleration of gaze transfer exceeds the acceleration threshold, it is possible to remove abnormal eye movement data caused by possible emergencies of flight personnel (such as sudden head turning, external interference, etc.). For example, during training, if the flight personnel's gaze shifts rapidly and irregularly due to unexpected situations such as coughing or sudden body shaking, these data will interfere with the construction of the data distribution model. After deleting the data points of these invalid paths, the data on which the model is based is more in line with the eye movement rules under normal training conditions, so that the data distribution model can more accurately reflect the eye movement characteristics of flight personnel during normal spatial orientation training.

[0078] For historical data points that are outside the data area of ​​all instruments or whose coordinate distance from other historical data points is greater than the preset distance threshold, it is likely caused by the error of the eye tracker or the irrelevant actions of the flight crew (such as looking at irrelevant objects in the training environment). Deleting these abnormal data points can avoid mixing these irrelevant data into the model building process, ensuring that the data distribution model is built based on data closely related to spatial orientation training and aircraft instrument observation, thereby improving the reliability and effectiveness of the model.

[0079] In another technical solution, in step S2, the historical transition probability matrix is ​​constructed as follows:

[0080]

[0081] Where n is the number of functional areas, in the historical transition probability matrix P, element p rs is the probability of transferring from the rth functional area to the sth functional area, , B is the total number of gaze transfer paths across functional areas in the historical eye movement data, B rs is the number of sight transfer paths from the rth functional area to the sth functional area;

[0082] In step S3, the current transition probability matrix is ​​constructed as follows:

[0083]

[0084] In the current transition probability matrix G, element g rs is the probability of transferring from the rth functional area to the sth functional area, , C is the total number of gaze transfer paths across functional areas in the current eye movement data, C rs is the number of sight transfer paths from the rth functional area to the sth functional area;

[0085] The difference matrix between the current transition probability matrix and the historical transition probability matrix is: T d =GP, difference matrix T d The absolute value of the elements in is the transfer selection deviation between the two functional areas.

[0086] It should be noted that, although the above describes the various steps in a specific order, it does not mean that the various steps must be performed in the above specific order. In fact, some of these steps can be performed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention are obvious to those skilled in the art.

[0087] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for monitoring and analyzing eye movement data for flight personnel spatial orientation training, characterized in that: The following steps are involved: S1: Acquire historical eye movement data of spatial orientation training, wherein the historical eye movement data includes fixation point, fixation duration and sight transfer path, generate historical data points with the fixation point of the historical eye movement data as coordinates and the fixation duration as distribution value, and construct a data distribution model with the historical data points; S2: In the data distribution model, a number of functional areas are divided according to the position of the aircraft instrument data, and the number of gaze transfers between all functional areas is counted according to the gaze transfer paths in the historical eye movement data to construct a historical transfer probability matrix; S3: obtaining the current eye movement data of the flight crew performing the spatial orientation training task through an eye tracker, wherein the current eye movement data includes the gaze point, gaze duration and gaze transfer path, wherein the gaze transfer path includes a starting point and a landing point, projecting the current eye movement data onto the data distribution model to obtain the current data point, and counting the number of gaze transfers between all functional areas according to the gaze transfer path in the current eye movement data to construct a current transfer probability matrix; S4: Calculate the difference matrix between the current transfer probability matrix and the historical transfer probability matrix, and determine the deviation degree of the function area transfer selection for the spatial training task according to the difference matrix; In step S2, the method of constructing the historical transition probability matrix is: Where n is the number of functional areas, in the historical transition probability matrix P, element p rs is the probability of transferring from the rth functional area to the sth functional area, , B is the total number of gaze transfer paths across functional areas in the historical eye movement data, B rs is the number of sight transfer paths from the rth functional area to the sth functional area.

2. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 1, characterized in that: The following steps are also included: S5: selecting a historical data point in each functional area as a gathering point, and classifying the gaze transfer path in the current eye movement data into a cross-area path and an intra-area path, wherein the starting point and the landing point of the cross-area path are located in different functional areas, and the starting point and the landing point of the intra-area path are located in the same functional area; For each cross-region path, the distance between its landing point and the gathering point of the functional area is calculated and recorded as the transfer deviation; For each path within the area, the distance between its starting point and the landing point is calculated and recorded as the search deviation; The data viewing speed of the flight personnel is determined based on the transfer deviation and the search deviation.

3. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 2, characterized in that: In step S5, a method of selecting a historical data point in the function area as a gathering point is: Select any historical data point in the functional area as the initial gathering point of the functional area, take the gathering point and several historical data points closest to it as candidate points, calculate the distance between each candidate point and other historical data points in the functional area to which it belongs, weight the distance by the gaze duration of the historical data point and then average it to obtain the weighted average distance between each candidate point and other historical data points in the functional area to which it belongs, update the gathering point to the candidate point with the smallest weighted average distance, and iteratively select candidate points, calculate the weighted average distance and update the gathering point in the same way until the gathering point no longer changes.

4. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 3, characterized in that: The specific method of selecting the candidate points is: Establish a rectangular coordinate system with the current gathering point as the origin, and diverge from the origin at a continuously increasing distance until at least one historical data point falls within the divergence range in each of the four quadrants of the rectangular coordinate system. Select at most K historical data points that fall within the divergence range in each quadrant as candidate points; The method for calculating the weighted average distance of the candidate points is: Among them, d j,i For the alternative point (x j ,y j ) to the i-th historical data point (x i ,y i ), , For the alternative point (x j ,y j ) to other historical data points in the corresponding functional area, M is the weighted average distance between the candidate point (x j ,y j ) belongs to the functional area, t i is the fixation duration of the i-th historical data point, t avg It is the average fixation duration of the historical data points in the functional area to which the candidate point belongs.

5. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 1, characterized in that: The following steps are also included: S6: For each current data point projected into the data distribution model, the average fixation duration of several historical data points adjacent to it is calculated, and the ratio of the fixation duration of the current data point to the corresponding average fixation duration is calculated. If this ratio of a current data point exceeds a preset ratio range, the current data point is marked as having an abnormal fixation duration.

6. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 5, characterized in that: In step S6, N historical data points closest to the current data point are selected, and the ratio d of the gaze duration of the current data point to the average gaze duration of the N historical data points is calculated. The lower limit and upper limit of the preset ratio range are set to e and f respectively, and e<1<f. The method for determining whether to mark the current data point as abnormal gaze duration is as follows: If e≤d≤f, the gaze duration of the corresponding current data point is judged to be normal; If d<e, it is judged that the fixation time of the corresponding current data point is too short; If d>f, it is judged that the gaze duration of the corresponding current data point is too long.

7. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 1, characterized in that: In step S1, abnormal data processing is performed on historical data points when constructing the data distribution model. The abnormal data processing method is: Obtaining the gaze transfer rate and gaze transfer acceleration corresponding to the gaze transfer path of the historical eye movement data, if the gaze transfer rate of a gaze transfer path exceeds a preset rate threshold or the gaze transfer acceleration exceeds a preset acceleration threshold, determining that the gaze transfer path is an invalid path and deleting the corresponding historical data point; If a historical data point is outside the data area of ​​all instruments or the coordinate distance from other historical data points is greater than the preset distance threshold, the historical data point is determined as abnormal data and deleted.

8. The method for monitoring and analyzing eye movement data for flight personnel spatial orientation training according to claim 1, characterized in that: In step S3, the current transition probability matrix is ​​constructed as follows: In the current transition probability matrix G, element g rs is the probability of transferring from the rth functional area to the sth functional area, , C is the total number of gaze transfer paths across functional areas in the current eye movement data, C rs is the number of sight transfer paths from the rth functional area to the sth functional area; The difference matrix between the current transition probability matrix and the historical transition probability matrix is: T d =GP, difference matrix T d The absolute value of the elements in is the transfer selection deviation between the two functional areas.

Images