Visual training evaluation method and device based on cloud computing, equipment and storage medium

Abstract

The invention relates to the technical field of visual training, and discloses a visual training evaluation method and device based on cloud computing, equipment and a storage medium, and the method comprises the steps: obtaining initial eye movement data; performing feature extraction according to the initial eye movement data, and constructing an initial eye movement matrix; inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; actual eye movement data are obtained, feature extraction is carried out, and an actual eye movement vector is constructed; performing coincidence degree calculation according to the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; and evaluating the visual training effect according to the training effect error. The method has the following effect that the accuracy of visual training evaluation can be improved.

Description

Technical Field

[0001] The present invention relates to the field of vision training technology, and in particular to a vision training evaluation method, device, equipment and storage medium based on cloud computing. Background Art

[0002] Currently, the research on brain visual perception intelligence is a cutting-edge field that combines knowledge from multiple disciplines such as neuroscience, psychology, and computer science, and aims to activate specific visual cortex layers of the brain through personalized visual training programs. This technology takes advantage of the plasticity of the nervous system, that is, the ability of the brain to repair or improve its function after appropriate stimulation. For patients with visual impairment due to disease, injury or other reasons, targeted training can be carried out through carefully designed visual perception learning models to enhance and restore impaired visual function. Personalized visual training programs can generate different categories of training tasks based on the user's binocular vision function level, ensuring that each user can get the training experience that best suits them.

[0003] In the prior art, in order to achieve the above goals, the system will first evaluate the user's binocular vision function level, and based on this information, provide the user with a series of specially designed visual training tasks. During the training process, the system will not only monitor and record the activity information and operation information of all elements, but also generate detailed training results after completing the visual training. These elements cover the various materials that make up the training tasks. Subsequently, the system will simulate and replay the entire visual training process based on the collected data, and evaluate the training based on the training results. This process allows users and their instructors to review the details of various operations in the training, help identify existing problems, and adjust future training plans accordingly to improve the training effect. This method can not only deeply explore the data indicators in the user's visual training, but also provide detailed and effective feedback guidance, thereby optimizing the user's visual training experience.

[0004] However, current technology still has some shortcomings. First, existing evaluation methods cannot fully and accurately capture the unique visual needs of each user, which results in the generated training tasks being imprecise. At the same time, although the simulation playback function helps to detect anomalies in training, it relies on pre-set parameters and rules and has limitations for certain complex or atypical visual impairments. In addition, since the effectiveness of visual training is affected by many factors, including but not limited to the user's motivation, health status, and environmental conditions, post-training simulation playback and data analysis alone are not sufficient to fully explain the reasons for poor training results. Therefore, there is a problem of low accuracy in evaluating the effectiveness of users' visual training. Summary of the invention

[0005] The present invention provides a visual training evaluation method, device, equipment and storage medium based on cloud computing to improve the accuracy of visual training effect evaluation.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a visual training evaluation method based on cloud computing, comprising: Obtain initial eye movement data; Perform feature extraction based on the initial eye movement data to construct an initial eye movement matrix; Inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; Obtain actual eye movement data, perform feature extraction, and construct actual eye movement vectors; Calculate the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; The visual training effect is evaluated according to the training effect error.

[0007] In an optional implementation, the obtaining of initial eye movement data includes: The initial eye movement data includes the first fixation point coordinates, the second fixation point coordinates, the sampling point coordinates during the tracking error test, the target test time during the saccade test, the saccade start time during the saccade test, the target start movement time during the smooth pursuit test, and the start tracking time during the smooth pursuit test.

[0008] In an optional implementation, extracting features based on the initial eye movement data to construct an initial eye movement matrix includes: The tracking error is calculated using the following formula:

[0009] in, is the tracking error, is the total number of sampling points, is the sampling point number, is the target horizontal coordinate of sampling point i, is the focus horizontal coordinate of sampling point i, is the target ordinate of sampling point i, is the focus ordinate of sampling point i; The amplitude of eye movement is calculated by the following formula:

[0010] in, is the saccade amplitude, is the latitude of the first gaze point, represents the latitude of the second gaze point, is the longitude of the first fixation point, is the longitude of the second gaze point; The saccadic reaction time was calculated using the following formula:

[0011] in, is the saccadic reaction time, is the target appearance time, is the saccade start time; The smooth pursuit time is calculated by the following formula:

[0012] in, To smoothly track time, Start moving time for the target, To start tracking time; An initial eye movement matrix is ​​constructed using the tracking error, the saccade amplitude, the saccade reaction time and the smooth tracking time corresponding to each training time point.

[0013] In an optional implementation, the training process of the visual training model includes: Building a visual training model based on historical eye movement data, training the visual training model, setting the number of training times or determining that the training is completed after the loss function of the detection model meets the conditions, and obtaining the trained visual training model; Inputting the initial eye movement data into the trained visual training model to obtain a predicted eye movement matrix; Each row of the predicted eye movement matrix includes predicted training time, predicted tracking error, predicted saccade amplitude, predicted saccade reaction time and predicted smooth tracking time.

[0014] In an optional implementation, calculating the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error includes: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The overlap is calculated using the following formula:

[0015] in, To calculate the overlap between the predicted eye movement vector and the actual eye movement vector, To predict the eye movement vector, is the actual eye movement vector, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the j-th dimension weight, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The overlap degree is taken as the training effect error.

[0016] In an optional implementation, the evaluating the visual training effect according to the training effect error includes: When the training effect error is less than a first error threshold, it is determined that the training effect is excellent; When the training effect error is greater than the first error threshold, it is determined that there is a problem with the training effect.

[0017] In an optional implementation, after the training effect has problems, the method further includes: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The degree of overlap of each dimension of the predicted eye movement vector and the actual eye movement vector is calculated by the following formula:

[0018] in, is the dimension overlap of dimension j, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The visual training effect is evaluated in different dimensions according to the dimensional overlap and a preset dimensional threshold.

[0019] In a second aspect, the present invention provides a visual training and evaluation device based on cloud computing, comprising: A data acquisition module, used to acquire initial eye movement data; A feature extraction module, used to extract features according to the initial eye movement data and construct an initial eye movement matrix; An eye movement prediction module, used for inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; The actual acquisition module is used to acquire the actual eye movement data, perform feature extraction, and construct the actual eye movement vector; An error calculation module, used to calculate the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; The effect evaluation module is used to evaluate the visual training effect according to the training effect error.

[0020] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned cloud computing-based visual training evaluation methods.

[0021] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned cloud computing-based visual training evaluation methods.

[0022] Compared with the prior art, the present invention has the following beneficial effects: the present invention discloses a visual training evaluation method, device, equipment and storage medium based on cloud computing, the method comprising obtaining initial eye movement data; extracting features according to the initial eye movement data to construct an initial eye movement matrix; inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; obtaining actual eye movement data, extracting features and constructing an actual eye movement vector; calculating the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; and evaluating the visual training effect according to the training effect error. The accuracy of the user's visual training evaluation results can be improved.

[0023] Specifically, by introducing the calculation of tracking error (TE), saccade amplitude (SA), saccade reaction time (SRT) and smooth pursuit time (SPRT), the present invention provides a method for quantitative visual training evaluation. These indicators can objectively reflect the user's visual motion performance, including tracking accuracy, saccade characteristics and response speed to moving targets. The tracking error formula is used to measure the position deviation when the eyes follow the target. It comprehensively considers the differences in the horizontal and vertical coordinates and obtains an overall error value by taking the root mean square of the sum of squares. This method effectively captures the degree of position deviation in two-dimensional space and provides an intuitive measure of the user's tracking ability. The calculation of saccade amplitude is based on spherical trigonometry, and the latitude and longitude difference between two fixation points is used to determine the angle of the saccade. This method can accurately represent the length of the eye movement path from one fixation point to another, which is helpful to understand the user's rapid eye movement characteristics. The saccade reaction time and smooth pursuit time respectively reflect the speed at which the user initiates the corresponding eye movement behavior after the visual stimulus appears. These two time parameters are crucial for evaluating the responsiveness of the visual system because they reveal the time required for the nervous system to process visual information and convert it into specific actions. The process of constructing the initial eye movement matrix involves organizing the data of the above four key performance indicators into a multidimensional data set that represents the characteristics of the user's eye movement behavior at a specific training stage. When this data is input into a pre-trained visual training model, the eye movement pattern of the subsequent training stage can be predicted, that is, the predicted eye movement matrix. The acquisition and feature extraction steps of actual eye movement data ensure that the evaluation process is based on the latest user performance, rather than relying solely on historical or simulated data. By comparing the overlap between the predicted eye movement matrix and the actual eye movement vector, the training effect error can be calculated, which directly reflects the actual effectiveness of visual training. Evaluating the visual training effect based on the training effect error makes the evaluation results more accurate and reliable, and the multi-dimensional ability test makes the evaluation results of the training effect more scientific. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart of a visual training evaluation method based on cloud computing provided by the first embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a visual training and evaluation device based on cloud computing provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Currently, the research on brain visual perception intelligence is a cutting-edge field that combines knowledge from multiple disciplines such as neuroscience, psychology, and computer science, and aims to activate specific visual cortex layers of the brain through personalized visual training programs. This technology takes advantage of the plasticity of the nervous system, that is, the ability of the brain to repair or improve its function after appropriate stimulation. For patients with visual impairment due to disease, injury or other reasons, targeted training can be carried out through carefully designed visual perception learning models to enhance and restore impaired visual function. Personalized visual training programs can generate different categories of training tasks based on the user's binocular vision function level, ensuring that each user can get the training experience that best suits them.

[0027] In the prior art, in order to achieve the above goals, the system will first evaluate the user's binocular vision function level, and based on this information, provide the user with a series of specially designed visual training tasks. During the training process, the system will not only monitor and record the activity information and operation information of all elements, but also generate detailed training results after completing the visual training. These elements cover the various materials that make up the training tasks. Subsequently, the system will simulate and replay the entire visual training process based on the collected data, and evaluate the training based on the training results. This process allows users and their instructors to review the details of various operations in the training, help identify existing problems, and adjust future training plans accordingly to improve the training effect. This method can not only deeply explore the data indicators in the user's visual training, but also provide detailed and effective feedback guidance, thereby optimizing the user's visual training experience.

[0028] However, current technology still has some shortcomings. First, existing evaluation methods cannot fully and accurately capture the unique visual needs of each user, which results in the generated training tasks being imprecise. At the same time, although the simulation playback function helps to detect anomalies in training, it relies on pre-set parameters and rules and has limitations for certain complex or atypical visual impairments. In addition, since the effectiveness of visual training is affected by many factors, including but not limited to the user's motivation, health status, and environmental conditions, post-training simulation playback and data analysis alone are not sufficient to fully explain the reasons for poor training results. Therefore, there is a problem of low accuracy in evaluating the effectiveness of users' visual training.

[0029] To solve the above problems, refer to Figure 1 The first embodiment of the present invention provides a visual training evaluation method based on cloud computing, comprising the following steps: S11, obtaining initial eye movement data; S12, performing feature extraction according to the initial eye movement data to construct an initial eye movement matrix; S13, inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; S14, obtaining actual eye movement data, performing feature extraction, and constructing an actual eye movement vector; S15, calculating the overlap degree according to the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; S16, evaluating the visual training effect according to the training effect error.

[0030] In step S11, initial eye movement data is acquired.

[0031] In one embodiment, the initial eye movement data includes the first fixation point coordinates, the second fixation point coordinates, the sampling point coordinates during the tracking error test, the target test time during the saccade test, the saccade start time during the saccade test, the target start movement time during the smooth pursuit test, and the start tracking time during the smooth pursuit test.

[0032] It is worth noting that the initial eye movement data should be obtained before the user begins formal vision training. This means recording their eye movement patterns when the user is still unfamiliar with the upcoming training task. This ensures that the data obtained accurately represents the user's untrained state. It also includes the user's state before receiving the vision training evaluation. The more detailed the data, the more accurate the prediction of the training effect.

[0033] In one embodiment, an eye tracking device (such as an eye tracker) is used for measurement. Such devices track the position of the eyes through infrared or other optical technology and calculate the coordinates of the gaze point on the screen or in the real world. The eye tracker can record the user's eye movements at a high frequency, thereby accurately determining the user's gaze position and collecting the first gaze point coordinates and the second gaze point coordinates. Let the user follow the moving target, and the eye tracker will continue to record the movement of the user's eyes with the target to obtain the tracking error sampling points. The target test time refers to the time point when the target presented to the user appears; the eye saccade start time refers to the time when the user's eyes move quickly from one gaze point to another. These times can be recorded by software that is programmed to control the experimental process, and the eye tracker synchronously records the actual moment when the eye saccade occurs. The smooth pursuit test requires the user to smoothly track a moving target. The target start movement time and the user start tracking time are also recorded by the experimental control system, and the eye tracker is responsible for capturing how the user's eyes respond to the moving target.

[0034] In step S12, feature extraction is performed based on the initial eye movement data to construct an initial eye movement matrix.

[0035] In one implementation, the tracking error is calculated using the following formula:

[0036] in, is the tracking error, is the total number of sampling points, is the sampling point number, is the target horizontal coordinate of sampling point i, is the focus horizontal coordinate of sampling point i, is the target ordinate of sampling point i, is the focus ordinate of sampling point i; The amplitude of eye movement is calculated by the following formula:

[0037] in, is the saccade amplitude, is the latitude of the first gaze point, represents the latitude of the second gaze point, is the longitude of the first fixation point, is the longitude of the second gaze point; The saccadic reaction time was calculated using the following formula:

[0038] in, is the saccadic reaction time, is the target appearance time, is the saccade start time; The smooth pursuit time is calculated by the following formula:

[0039] in, To smoothly track time, Start moving time for the target, To start tracking time; An initial eye movement matrix is ​​constructed using the tracking error, the saccade amplitude, the saccade reaction time and the smooth tracking time corresponding to each training time point.

[0040] It is worth mentioning that tracking error is a measure of the accuracy of the eyes when following a moving target. It is achieved by calculating the average distance between the eye gaze position and the actual position of the target over a series of sample points. Lower tracking error indicates that the user is able to follow the target more accurately, reflecting good dynamic vision and hand-eye coordination. Saccade amplitude refers to the angular change from one gaze point to another quickly. This formula uses spherical trigonometry to calculate the angular distance between two gaze points. Larger saccade amplitude does not necessarily mean better performance; it mainly reflects the range of transitions between gaze points. Frequent large saccades mean that the user has difficulty maintaining a stable gaze or is not efficient in searching for information. Saccade reaction time is the time difference from the appearance of the target to the start of the saccade. This directly reflects the speed at which the brain processes visual information and directs the eye muscles to respond.

[0041] Among them, a shorter saccade reaction time indicates that the user has a faster visual processing speed and a higher level of attention. Smooth pursuit time is the time interval from the start of the target to the start of the user's smooth tracking of the target. It reflects the user's ability to initiate smooth tracking. A shorter smooth tracking time indicates that the user can enter the tracking mode faster, indicating that he has better predictive ability and motion control. For example: A user's tracking error is significantly lower than the average level, which means that he performs well in following moving targets and has excellent dynamic vision and stability; for a reader, if his smooth tracking time is shorter, then he is more efficient when reading because he can follow the movement of the text line faster.

[0042] In step S13, the initial eye movement matrix is ​​input into the trained visual training model to obtain a predicted eye movement matrix.

[0043] In one embodiment, the training process of the visual training model includes: constructing a visual training model based on historical eye movement data, training the visual training model, setting the number of training times or determining that the training is completed after the loss function of the detection model meets the conditions, and obtaining a trained visual training model; inputting the initial eye movement data into the trained visual training model to obtain a predicted eye movement matrix; wherein each row of the predicted eye movement matrix includes predicted training time, predicted tracking error, predicted eye movement amplitude, predicted eye movement reaction time and predicted smooth tracking time.

[0044] It is worth noting that the main function of the visual training model is to predict the visual-motor ability changes of the user at different time points after receiving visual training by inputting the visual-motor ability of the user at a specific time point (i.e., the initial eye movement data). This prediction result is presented in the form of a matrix, called the predicted eye movement matrix. The historical eye movement data is constructed from the values ​​of the visual-motor parameters of users who have received visual training in the past at different time points.

[0045] In one implementation, a visual training model is constructed based on a long short-term memory neural network. Prepare features such as historical user training time, tracking error, saccade amplitude, saccade reaction time, and smooth tracking time. According to task requirements, it is necessary to construct multi-step input time series data, and use data from N past time points to predict performance at M future time points. The input layer accepts processed feature vectors as input. The mean square error is selected for the loss function, and the Adam optimizer is used. The data is divided into a training set and a validation set to monitor overfitting, and the training parameters are set for training.

[0046] In step S14, actual eye movement data is obtained, and feature extraction is performed to construct an actual eye movement vector.

[0047] In one embodiment, an eye tracker is used to collect the user's actual eye movement data. The experimental eye movement data mentioned here is the eye movement data after visual training. The actual eye movement data is feature extracted by using the same formula in step S12. Since there is only one set of data, the generated data is in vector form.

[0048] In step S15, the overlap degree is calculated based on the predicted eye movement matrix and the actual eye movement vector to obtain the training effect error.

[0049] In one implementation, a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time is found and recorded as a predicted eye movement vector; The overlap is calculated using the following formula:

[0050] in, To calculate the overlap between the predicted eye movement vector and the actual eye movement vector, To predict the eye movement vector, is the actual eye movement vector, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the j-th dimension weight, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The overlap degree is taken as the training effect error.

[0051] It is worth noting that the above formula is used to calculate the overlap between the predicted eye movement vector and the actual eye movement vector, which reflects the similarity between the two in terms of time and feature dimensions. This overlap is ultimately used as a measure of the training effect error. The time difference part uses a negative exponent to weaken the difference in time, but since the closest row is taken in the step, there will be some differences, which is taken into consideration. The value of the time coefficient is 1. For the selection and setting of the weight, set it according to the time proportion of different training projects. The more time spent on the training project, the greater the proportion, and the sum is 1.

[0052] In step S16, the visual training effect is evaluated according to the training effect error.

[0053] In one implementation, when the training effect error is less than a first error threshold, it is determined that the training effect is excellent; when the training effect error is greater than the first error threshold, it is determined that there is a problem with the training effect.

[0054] It is worth noting that in order to improve the evaluation efficiency, the overall training effect is first evaluated. When the calculated training effect error is less than the first error threshold, the training effect is determined to be excellent. This means that the predicted eye movement vector is highly consistent with the actual eye movement vector, indicating that the user's visual movement ability has been effectively improved or remained stable during the training process. If the training effect error is greater than the first error threshold, it is considered that there is a problem with the training effect, and it is necessary to further analyze the cause and adjust the training plan.

[0055] In one implementation, a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time is found and recorded as a predicted eye movement vector; The degree of overlap of each dimension of the predicted eye movement vector and the actual eye movement vector is calculated by the following formula:

[0056] in, is the dimension overlap of dimension j, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The visual training effect is evaluated in different dimensions according to the dimensional overlap and a preset dimensional threshold.

[0057] It is worth noting that the formula adjusts the impact of time differences through an exponential decay function and multiplies it by the difference in the feature dimension to finally obtain the overlap of each dimension. This value reflects the similarity between the predicted value and the true value in a specific dimension. For example, after a user has received a period of visual training, he finds that the overall training effect has a large error, which exceeds the first error threshold. Next, follow the above steps: use the formula to calculate the overlap of the four dimensions of tracking error, saccade amplitude, saccade reaction time, and smooth tracking time. Compare the overlap of each dimension with the preset dimension threshold to identify which dimensions perform well and which need improvement.

[0058] In summary, the present invention discloses a visual training evaluation method, device, equipment and storage medium based on cloud computing, which aims to achieve a more accurate evaluation of the visual training effect through a series of steps. The method first involves obtaining initial eye movement data, which not only includes the coordinates of the first gaze point and the second gaze point of the user in a specific task, but also covers the sampling point coordinates during the tracking error test, the target test time and the saccade start time during the saccade test, and the target start movement time and start tracking time during the smooth pursuit test. Multi-dimensional information. In order to ensure the comprehensiveness and accuracy of the data, the initial eye movement data should be collected before the user formally receives visual training. This step is crucial to capturing the user's eye movement pattern in an untrained state. After obtaining the initial eye movement data, the method further uses feature extraction technology to construct an initial eye movement matrix.

[0059] Specifically, this process is completed by calculating tracking error, saccade amplitude, saccade reaction time, and smooth pursuit time. Tracking error reflects the position deviation of the eyes when following a moving target, while saccade amplitude measures the angular change of a rapid movement from one fixation point to another. Saccade reaction time is the time difference from the appearance of the target to the start of the saccade, which reflects the speed at which the brain processes visual information and directs the eye muscles to respond. Smooth pursuit time is the time interval from the start of the target to the user's smooth tracking of the target, reflecting the user's ability to initiate smooth pursuit. By organizing the data of the above four key performance indicators, a multidimensional data set is formed, which represents the characteristics of the user's eye movement behavior at a specific training stage. This process lays a solid foundation for subsequent prediction and evaluation.

[0060] Next, the constructed initial eye movement matrix is ​​input into the pre-trained visual training model to obtain the predicted eye movement matrix. This visual training model is constructed based on historical eye movement data and is trained multiple times until the loss function meets the preset conditions or the specified number of training times is reached. The trained model can predict the changes in the user's visual movement ability at different time points after visual training based on the user's visual movement ability at a specific time point (i.e., the initial eye movement data). Each row of the predicted eye movement matrix contains the predicted training time, predicted tracking error, predicted eye movement amplitude, predicted eye movement reaction time, and predicted smooth tracking time, which provides theoretical expected results for evaluation.

[0061] Subsequently, the method also involves the acquisition and feature extraction of actual eye movement data to construct the actual eye movement vector. This part of the work also relies on professional equipment such as eye trackers to record the user's eye movements after visual training. By using the same formula as that used to construct the initial eye movement matrix, the consistency of the previous and subsequent data can be guaranteed, making the comparison between the predicted value and the actual value more scientific and reasonable. The introduction of actual eye movement data ensures that the evaluation process is based on the latest user performance, rather than relying solely on historical or simulated data, which enhances the timeliness and reliability of the evaluation results.

[0062] Finally, the training effect error can be obtained based on the overlap calculation between the predicted eye movement matrix and the actual eye movement vector. The overlap calculation takes into account the time difference and the importance of each dimensional feature. The influence of the time difference is adjusted by the exponential decay function and multiplied by the difference in the feature dimension to finally obtain a comprehensive error index. When the training effect error is less than the preset first error threshold, the training effect is considered to be excellent; otherwise, it is considered that there is a problem and the cause needs to be further analyzed and the training plan adjusted. In addition, for situations where there are errors, further dimensional evaluation can be performed to identify which aspects perform well and which aspects need to be improved, so as to optimize future visual training plans in a more targeted manner.

[0063] In summary, the present invention provides a complete set of cloud computing-based visual training evaluation solutions, which objectively reflects the user's visual motion performance, including tracking accuracy, eye movement characteristics, and response speed to moving targets, by quantitatively evaluating visual training methods, such as introducing the calculation of tracking error (TE), saccade amplitude (SA), saccade reaction time (SRT), and smooth pursuit time (SPRT). This method not only overcomes the problem in the prior art that it is impossible to fully and accurately capture the unique visual needs of each user, but also avoids the limitation of relying solely on post-training simulation playback and data analysis to explain the reasons for poor training results. Therefore, the evaluation method provided by the present invention can more accurately and reliably reflect the actual effectiveness of visual training, and provide strong support for personalized customization of visual training programs.

[0064] Reference Figure 2 The second embodiment of the present invention provides a visual training evaluation device based on cloud computing, comprising: A data acquisition module, used to acquire initial eye movement data; A feature extraction module, used to extract features according to the initial eye movement data and construct an initial eye movement matrix; An eye movement prediction module, used for inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; The actual acquisition module is used to acquire the actual eye movement data, perform feature extraction, and construct the actual eye movement vector; An error calculation module, used to calculate the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; The effect evaluation module is used to evaluate the visual training effect according to the training effect error.

[0065] Preferably, the data acquisition module is used to: Obtain initial eye movement data, including: The initial eye movement data includes but is not limited to the first fixation point coordinates, the second fixation point coordinates, the sampling point coordinates during the tracking error test, the target test time and the saccade start time during the saccade test, and the target start movement time and the start tracking time during the smooth pursuit test.

[0066] Preferably, the feature extraction module is used to: Extracting features according to the initial eye movement data and constructing an initial eye movement matrix include: The tracking error is calculated using the following formula:

[0067] in, is the tracking error, is the total number of sampling points, is the sampling point number, is the target horizontal coordinate of sampling point i, is the focus horizontal coordinate of sampling point i, is the target ordinate of sampling point i, is the focus ordinate of sampling point i; The amplitude of eye movement is calculated by the following formula:

[0068] in, is the saccade amplitude, is the latitude of the first gaze point, represents the latitude of the second gaze point, is the longitude of the first fixation point, is the longitude of the second gaze point; The saccadic reaction time was calculated using the following formula:

[0069] in, is the saccadic reaction time, is the target appearance time, is the saccade start time; The smooth pursuit time is calculated by the following formula:

[0070] in, To smoothly track time, Start moving time for the target, To start tracking time; An initial eye movement matrix is ​​constructed using the tracking error, the saccade amplitude, the saccade reaction time and the smooth tracking time corresponding to each training time point.

[0071] Preferably, the eye movement prediction module is used to input the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix.

[0072] Preferably, the training process of the visual training model includes: Building a visual training model based on historical eye movement data, training the visual training model, setting the number of training times or determining that the training is completed after the loss function of the detection model meets the conditions, and obtaining the trained visual training model; Inputting the initial eye movement data into the trained visual training model to obtain a predicted eye movement matrix; Each row of the predicted eye movement matrix includes predicted training time, predicted tracking error, predicted saccade amplitude, predicted saccade reaction time and predicted smooth tracking time.

[0073] Preferably, the actual acquisition module is used to acquire actual eye movement data, perform feature extraction, and construct an actual eye movement vector; Preferably, the error calculation module is used to calculate the overlap degree according to the predicted eye movement matrix and the actual eye movement vector to obtain the training effect error, including: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The overlap is calculated using the following formula:

[0074] in, To calculate the overlap between the predicted eye movement vector and the actual eye movement vector, To predict the eye movement vector, is the actual eye movement vector, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the j-th dimension weight, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The overlap degree is taken as the training effect error.

[0075] Preferably, the effect evaluation module is used to evaluate the visual training effect according to the training effect error, and includes: When the training effect error is less than a first error threshold, it is determined that the training effect is excellent; When the training effect error is greater than the first error threshold, it is determined that there is a problem with the training effect.

[0076] Preferably, after the training effect has problems, the method further comprises: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The degree of overlap of each dimension of the predicted eye movement vector and the actual eye movement vector is calculated by the following formula:

[0077] in, is the dimension overlap of dimension j, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The visual training effect is evaluated in different dimensions according to the dimensional overlap and a preset dimensional threshold.

[0078] It should be noted that the cloud computing-based visual training evaluation device provided in an embodiment of the present invention is used to execute all the process steps of the cloud computing-based visual training evaluation method in the above embodiment. The working principles and beneficial effects of the two correspond one to one, and thus will not be described in detail.

[0079] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, the steps in the above-mentioned cloud computing-based visual training and evaluation device embodiments are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data acquisition module.

[0080] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0081] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0082] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0083] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0084] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0085] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0086] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A visual training evaluation method based on cloud computing, characterized in that: include: Obtain initial eye movement data; Perform feature extraction based on the initial eye movement data to construct an initial eye movement matrix; Inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; Obtain actual eye movement data, perform feature extraction, and construct actual eye movement vectors; Calculate the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; The visual training effect is evaluated according to the training effect error.

2. The cloud computing-based visual training evaluation method according to claim 1, characterized in that: The obtaining of initial eye movement data comprises: The initial eye movement data includes the first fixation point coordinates, the second fixation point coordinates, the sampling point coordinates during the tracking error test, the target test time during the saccade test, the saccade start time during the saccade test, the target start movement time during the smooth pursuit test, and the start tracking time during the smooth pursuit test.

3. The cloud computing-based visual training evaluation method according to claim 1, characterized in that: The step of extracting features according to the initial eye movement data and constructing an initial eye movement matrix comprises: The tracking error is calculated using the following formula: in, is the tracking error, is the total number of sampling points, is the sampling point number, is the target horizontal coordinate of sampling point i, is the focus horizontal coordinate of sampling point i, is the target ordinate of sampling point i, is the focus ordinate of sampling point i; The amplitude of eye movement is calculated by the following formula: in, is the saccade amplitude, is the latitude of the first gaze point, represents the latitude of the second gaze point, is the longitude of the first fixation point, is the longitude of the second gaze point; The saccadic reaction time was calculated using the following formula: in, is the saccadic reaction time, is the target appearance time, is the saccade onset time; The smooth pursuit time is calculated by the following formula: in, To smoothly track time, Start moving time for the target, To start tracking time; An initial eye movement matrix is ​​constructed using the tracking error, the saccade amplitude, the saccade reaction time and the smooth tracking time corresponding to each training time point.

4. The cloud computing-based visual training evaluation method according to claim 1, characterized in that: The training process of the visual training model includes: Building a visual training model based on historical eye movement data, training the visual training model, setting the number of training times or determining that the training is completed after the loss function of the detection model meets the conditions, and obtaining the trained visual training model; Inputting the initial eye movement data into the trained visual training model to obtain a predicted eye movement matrix; Each row of the predicted eye movement matrix includes predicted training time, predicted tracking error, predicted saccade amplitude, predicted saccade reaction time and predicted smooth tracking time.

5. The cloud computing-based visual training evaluation method according to claim 1, characterized in that: The calculating the overlap degree according to the predicted eye movement matrix and the actual eye movement vector to obtain the training effect error includes: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The overlap is calculated using the following formula: in, To calculate the overlap between the predicted eye movement vector and the actual eye movement vector, To predict the eye movement vector, is the actual eye movement vector, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the j-th dimension weight, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The overlap degree is taken as the training effect error.

6. The cloud computing-based visual training evaluation method according to claim 1, characterized in that: The evaluating the visual training effect according to the training effect error includes: When the training effect error is less than a first error threshold, it is determined that the training effect is excellent; When the training effect error is greater than the first error threshold, it is determined that there is a problem with the training effect.

7. The cloud computing-based visual training evaluation method according to claim 6, characterized in that: After the training effect is problematic, it also includes: Find a row in the predicted eye movement matrix whose predicted training time is closest to the actual training time, and record it as a predicted eye movement vector; The degree of overlap of each dimension of the predicted eye movement vector and the actual eye movement vector is calculated by the following formula: in, is the dimension overlap of dimension j, is a natural constant, is the time coefficient, is the actual training time, To predict the training time, is the dimension number of the vector, is the difference between the predicted eye movement vector and the actual eye movement vector in dimension j; The visual training effect is evaluated in different dimensions according to the dimensional overlap and a preset dimensional threshold.

8. A visual training and evaluation device based on cloud computing, characterized in that: include: A data acquisition module, used to acquire initial eye movement data; A feature extraction module, used to extract features according to the initial eye movement data and construct an initial eye movement matrix; An eye movement prediction module, used for inputting the initial eye movement matrix into a trained visual training model to obtain a predicted eye movement matrix; The actual acquisition module is used to acquire the actual eye movement data, perform feature extraction, and construct the actual eye movement vector; An error calculation module, used to calculate the overlap between the predicted eye movement matrix and the actual eye movement vector to obtain a training effect error; The effect evaluation module is used to evaluate the visual training effect according to the training effect error.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the cloud computing-based visual training evaluation method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the cloud computing-based visual training evaluation method according to any one of claims 1 to 7.

Images