Neural network evaluation method for interface readability

Abstract

The embodiment of the invention provides an interface readability-oriented neural network evaluation method, which is used for objectively evaluating the readability of a display interface based on eye movement behavior data to obtain the readability level of the display interface, and comprises the following steps: collecting the eye movement behavior data of each tested object for executing a specified task; collecting readability questionnaire data of each tested object after executing the specified task to obtain subjective evaluation data of each tested object on the readability level of the interface; preprocessing the collected eye movement behavior data and readable questionnaire data; neural network hyper-parameters are set, and neural network model construction and code implementation are carried out based on the set neural network hyper-parameters; carrying out neural network model training; testing and optimizing the neural network model; and obtaining a neural network evaluation model for interface readability. Objective evaluation on the readability of the display interface is realized, and an evaluation means is provided for the readability level of an information system interface and optimization of man-machine interaction design.

Description

Technical Field

[0001] The present invention relates to the technical field of interface reading and evaluation, and in particular to a neural network evaluation method for interface readability. Background Art

[0002] The neural network evaluation method for the display interface readability of equipment software systems mainly provides a method and approach for evaluating the readability level of the interface.

[0003] Currently, for the objective evaluation of interface readability, different display interfaces can only be subjectively evaluated through questionnaires, lacking an objective evaluation method, so the level of human-computer interface design cannot be known. Summary of the Invention

[0004] In view of the above problems existing in the prior art, an embodiment of the present invention provides a neural network evaluation method for interface readability. The method is used to objectively evaluate the readability of a display interface based on eye movement behavior data to obtain its readability level. The method includes the following steps:

[0005] Collect the eye movement behavior data of each subject performing a specified task;

[0006] Collect the readability questionnaire data of each subject after performing the specified task to obtain the subjective evaluation data of each subject on the interface readability level;

[0007] Preprocess the collected eye movement behavior data and readability questionnaire data;

[0008] Extract eye movement indicators such as pupil diameter as feature vectors according to the eye movement data that can reflect the visual attention distribution and cognitive load of users when processing information;

[0009] Normalize the feature vectors as the input of the neural network; set the neural network hyperparameters, and construct and implement the neural network model based on the set neural network hyperparameters;

[0010] Carry out neural network model training;

[0011] Perform the test and optimization of the neural network model;

[0012] Obtain a neural network evaluation model for interface readability.

[0013] In some embodiments of the present invention, the collecting the eye movement behavior data of each subject performing a specified task includes:

[0014] According to the task to be evaluated, select an experimental platform and subjects. Among them, the subjects are selected from a suitable group of subjects according to the characteristics of the task to be evaluated. The subjects, as samples in the experiment, include all corresponding characteristics of the population; and the subjects are randomly selected from the group of subjects.

[0015] Complete the ergonomics experiment. Under the condition of controlling relevant variables, each subject wears an eye movement tracking system to complete the specified task and collect relevant data.

[0016] Record the eye movement behavior data of the subjects during the execution of the evaluation task in the ergonomics experiment. The eye movement behavior data includes the fixation data, blink data, and saccade data of all subjects.

[0017] In some embodiments of the present invention, collect the readability questionnaire data of each subject after completing the specified task, including:

[0018] Design a readability questionnaire before the experiment;

[0019] After each subject completes the specified task, fill out the readability questionnaire respectively.

[0020] In some embodiments of the present invention, preprocess the collected eye movement behavior data and readability questionnaire data, including:

[0021] Eliminate invalid and abnormal eye movement data from the eye movement behavior data;

[0022] Calculate the subjective evaluation level of readability according to the readability questionnaire of each subject;

[0023] Research and screen the eye movement behavior data related to the interface readability level according to the information of at least the literature.

[0024] In some embodiments of the present invention, set the neural network hyperparameters, including:

[0025] Set the neural network hyperparameters including at least the structure, layer, and number of nodes of the neural network.

[0026] In some embodiments of the present invention, carry out the training of the neural network model, specifically including:

[0027] Eliminate the missing items in the original data. Adopt the 8:2 classification and allocation principle, select 80% as the training set and 20 as the test set;

[0028] The number of samples in each batch of iterations is 100;

[0029] The number of times to traverse the samples is 1000.

[0030] In some embodiments of the present invention, the testing and optimization of the neural network model specifically include:

[0031] Set the learning rate to 0.01;

[0032] Prevent overfitting by randomly ignoring a set proportion of neurons.

[0033] In some embodiments of the present invention, obtaining the neural network evaluation model for interface readability includes:

[0034] According to the testing and optimization of the neural network model, obtain a neural network model with an accuracy of not less than 0.7, a level of 4, and 64 nodes.

[0035] Compared with the prior art, the beneficial effect of the neural network evaluation method for interface readability provided by the embodiments of the present invention is that it realizes the objective evaluation of the readability of the display interface and provides an evaluation means for the readability level of the information system interface and the optimization of human-computer interaction design. Description of the Drawings

[0036] Figure 1 It is a flowchart of the neural network evaluation method for interface readability provided by the embodiments of the present invention. Detailed Embodiments

[0037] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be described in detail below with reference to the drawings and specific embodiments.

[0038] Reference is made herein to the various aspects and features of the present application with reference to the drawings.

[0039] These and other features of the present application will become apparent from the following description of the preferred forms of the embodiments given as non-limiting examples with reference to the drawings.

[0040] It should also be understood that although the present application has been described with reference to some specific examples, those skilled in the art can surely implement many other equivalent forms of the present application, which have the features as described in the claims and thus are all within the protection scope defined thereby.

[0041] When combined with the drawings, the above and other aspects, features, and advantages of the present application will become more apparent in view of the following detailed description.

[0042] Specific embodiments of the present application will be described hereinafter with reference to the accompanying drawings. However, it should be understood that the claimed embodiments are merely examples of the present application, which can be implemented in various ways. Well-known and / or repetitive functions and structures are not described in detail to clarify the true intention based on the user's historical operations and avoid unnecessary or redundant details from obscuring the present application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but are merely used as a basis for the claims and a representative basis for teaching those skilled in the art to use the present application in a substantially appropriate detailed structure in various ways.

[0043] This specification may use the phrase "in one embodiment", "in another embodiment", "in yet another embodiment", or "in other embodiments", which may each refer to one or more of the same or different embodiments according to the present application.

[0044] An embodiment of the present invention provides a neural network evaluation method for interface readability, which is used to objectively evaluate the readability of a display interface based on eye movement behavior data and obtain its readability level. As Figure 1 shown, the method includes the following steps:

[0045] Collect eye movement behavior data of each subject performing a specified task;

[0046] Collect readability questionnaire data of each subject after performing the specified task to obtain subjective evaluation data of each subject on the interface readability level;

[0047] Preprocess the collected eye movement behavior data and readability questionnaire data;

[0048] Extract eye movement indicators such as pupil diameter as feature vectors according to the eye movement data that can reflect the visual attention distribution and cognitive load of the user when processing information;

[0049] Normalize the feature vectors as the input of the neural network;

[0050] Set the hyperparameters of the neural network and construct and implement the neural network model based on the set hyperparameters of the neural network;

[0051] Carry out neural network model training;

[0052] Test and optimize the neural network model;

[0053] Obtain a neural network evaluation model for interface readability.

[0054] In this embodiment, the collecting eye movement behavior data of each subject performing a specified task includes:

[0055] According to the task to be evaluated, select the experimental platform and subjects. Among them, the experimental subjects are selected from a suitable subject group according to the characteristics of the task to be evaluated. As samples in the experiment, the representativeness of the subjects is a key factor. Representativeness means that the selected subject samples should include all corresponding characteristics of the population. Moreover, in order to obtain a representative sample population, sample individuals should be randomly selected from the group. Randomness means that each individual in the group has an equal chance of being selected into the sample. At the same time, factors such as randomness and the size of the sample space should also be considered when selecting subjects;

[0056] Conduct an ergonomics experiment, that is, under the condition of controlling relevant variables, each subject wears an eye movement tracking system to complete the specified task and collect relevant data;

[0057] Record the eye movement behavior data of the subjects during the execution of the evaluation task in the ergonomics experiment; the eye movement behavior data includes the fixation data, blink data, and saccade data of all subjects;

[0058] Collect the readability questionnaire data of each subject after completing the specified task, including:

[0059] Before the experiment, design a readability questionnaire;

[0060] After each subject completes the specified task, fill out the readability questionnaire respectively.

[0061] Preprocess the collected eye movement behavior data and readability questionnaire data, including:

[0062] According to the readability questionnaire of each subject, detect and eliminate invalid questionnaires;

[0063] Calculate the subjective evaluation level of readability and complete the preprocessing of the readability questionnaire data;

[0064] Mark the data according to the experimental conditions, experimental tasks, and corresponding subjective evaluations of readability;

[0065] Export the selected and marked eye movement data from the data processing software of the eye tracker to the TXT format;

[0066] Use the threshold method to detect and delete outliers in the data, and use eye movement event detection and elimination based on speed to detect and eliminate eye movement events such as fixations and saccades to complete the preprocessing of eye movement behavior data;

[0067] Extract eye movement indicators such as pupil diameter as feature vectors according to the eye movement data that can reflect the visual attention distribution and cognitive load of users when processing information, including:

[0068] Extract the fixation data, blink data, and saccade data from the preprocessed eye movement behavior data respectively;

[0069] Extract data such as pupil diameter and fixation time from the fixation data and mark them;

[0070] Extract blink time data from the eye blink data and mark them;

[0071] Extract data such as saccade time and saccade rate from the saccade data and mark them to complete the extraction of feature vectors;

[0072] Perform normalization processing on the feature vectors as the input of the neural network, including:

[0073] Select the Z-score method to complete the normalization processing of the extracted feature vectors;

[0074] Mark the mean and standard deviation of the eye movement feature vectors to complete the normalization processing;

[0075] The setting of the neural network hyperparameters includes:

[0076] Set neural network hyperparameters that at least include the structure, layers, and number of nodes of the neural network.

[0077] The implementation of neural network model training specifically includes: removing the missing items in the original data, adopting the 8:2 classification and allocation principle, selecting 80% as the training set and 20 as the test set;

[0078] The number of samples in each batch of iterations is 100;

[0079] The number of times to traverse the samples is 1000.

[0080] The implementation of neural network model testing and optimization specifically includes: setting the learning rate to 0.01; preventing overfitting by randomly ignoring a set proportion of neurons (hidden nodes).

[0081] The obtaining of the neural network evaluation model for interface readability includes: according to the neural network model testing and optimization, obtaining a neural network model with an accuracy of not less than 0.7, 4 layers, and 64 nodes.

[0082] It can be seen from the above technical solutions that the neural network evaluation method for interface readability provided by the above embodiments of the present invention can objectively evaluate the readability of the display interface and provide an evaluation means for the readability level of the information system interface and the optimization of the human-computer interaction design.

[0083] The above embodiments are only exemplary embodiments of the present invention and are not used to limit the present invention. The protection scope of the present invention is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements within the essence and protection scope of the present invention, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present invention.

Claims

1. A neural network evaluation method for interface readability, characterized in that: The method is used to objectively evaluate the readability of a display interface based on eye movement behavior data to obtain its readability level, and the method comprises the following steps: Collect eye movement behavior data of each subject performing a specified task; Collect readability questionnaire data from each subject after they complete the specified task, and obtain each subject's subjective evaluation data on the readability level of the interface; Preprocess the collected eye movement behavior data and readability questionnaire data; Set the neural network hyperparameters, and build the neural network model and implement the code based on the set neural network hyperparameters; Conduct neural network model training; Test and optimize neural network models; A neural network evaluation model for interface readability is obtained.

2. The neural network evaluation method for interface readability according to claim 1, characterized in that: The collecting of eye movement behavior data of each subject performing a specified task includes: According to the task to be evaluated, the experimental platform and the subjects are selected, wherein the subjects are selected from a suitable subject group according to the characteristics of the task to be evaluated, and the subjects as samples in the experiment contain all the corresponding characteristics of the population; and the subjects are randomly selected from the subject group; Complete ergonomic experiments. Under the condition of controlling relevant variables, each subject wears an eye tracking system to complete the specified tasks and collect relevant data; The eye movement behavior data of the subjects in the ergonomic experiment during the performance of the evaluation task are recorded, and the eye movement behavior data include the gaze data, blink data and saccade data of all the subjects.

3. The neural network evaluation method for interface readability according to claim 2, characterized in that: The collecting of readability questionnaire data after each subject has completed the specified task includes: A readability questionnaire was designed before the experiment; After completing the assigned task, each subject filled out a readability questionnaire.

4. The neural network evaluation method for interface readability according to claim 3, characterized in that: The preprocessing of the collected eye movement behavior data and readability questionnaire data includes: Eliminate invalid and abnormal eye movement behavior data; Based on the readability questionnaire of each subject, the subjective evaluation level of readability was calculated; Eye movement behavior data that are relevant to the level of interface readability are selected based on data information that at least includes literature.

5. The neural network evaluation method for interface readability according to claim 4, characterized in that: The setting of neural network hyperparameters includes: Set the neural network hyperparameters including at least the structure, number of layers and nodes of the neural network.

6. The neural network evaluation method for interface readability according to claim 5, characterized in that: The carrying out of neural network model training specifically includes: After eliminating missing items in the original data, the 8:2 classification allocation principle was adopted, and 80% was selected as the training set and 20 as the test set; The number of samples in each batch of iterations is 100; The number of times the sample is traversed is 1000.

7. The neural network evaluation method for interface readability according to claim 6, characterized in that: The neural network model testing and optimization specifically includes: Set the learning rate to 0.01; Overfitting is prevented by randomly ignoring a set proportion of neurons.

8. The neural network evaluation method for interface readability according to claim 7, characterized in that: The neural network evaluation model for interface readability is obtained, including: According to the neural network model testing and optimization, a neural network model with an accuracy of ≮0.7, 4 levels, and 64 nodes was obtained.

Images