Image aesthetics electroencephalogram evaluation model construction method and image aesthetics electroencephalogram evaluation method

Abstract

The invention discloses an image aesthetics electroencephalogram evaluation model construction method and an image aesthetics electroencephalogram evaluation method. The evaluation method comprises the following steps: stimulating a target object based on a preset aesthetic degree image of a preset type to obtain an original electroencephalogram signal; performing first preprocessing and second preprocessing on the original electroencephalogram signal to obtain a first type of electroencephalogram signal and a second type of electroencephalogram signal; processing the first type of electroencephalogram signals and the second type of electroencephalogram signals respectively based on a first preset strategy and a second preset strategy to obtain target ERP features and target time-frequency features, and then fusing the two features to obtain aesthetic fusion electroencephalogram features; based on the aesthetic fusion electroencephalogram features, the picture aesthetic preset classification information and a preset subjective aesthetic evaluation database, constructing an image aesthetic electroencephalogram evaluation model; and calling the model in response to an evaluation request, and analyzing the to-be-analyzed image based on the model to obtain aesthetic evaluation information. The method can realize comprehensive, accurate and reliable evaluation.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular to a method for constructing an image aesthetics electroencephalogram (EEG) evaluation model and an image aesthetics electroencephalogram (EEG) evaluation method. Background Art

[0002] Current studies on visual aesthetic evaluation mostly use machine learning or deep learning methods to extract features and classify images, but fail to fully consider the impact of human subjective cognition on image aesthetics. Due to the differences and subjectivity of individual visual evaluation, the collection and analysis of EEG signals provides a new method for the aesthetic evaluation of visual images. This method can make up for the shortcomings of traditional research and use the neural activity information in EEG signals to make more in-depth aesthetic judgments. However, in the existing aesthetic evaluation research based on EEG signals, it is not clear which EEG signals can effectively correspond to aesthetic recognition, and there is a lack of clear solutions for the precise separation of related features. For example, although ERP can reflect the brain's rapid response to specific events, it does not always correspond accurately to the subtle changes in aesthetic experience. For example, physical stimulation affects the EEG signal analysis of aesthetic evaluation, such as the brightness, contrast, color saturation, etc. of the image, which will cause direct reactions in the visual system. These reactions may be mixed with the EEG signals caused by aesthetic experience, thereby interfering with the accurate interpretation of aesthetic reactions.

[0003] In terms of EEG signal analysis methods, current research often focuses on the exploration of a single feature, lacking a comprehensive analysis of aesthetic features in different analysis methods. For example, ERP can accurately locate neural activities closely related to specific events, which is manifested in the consistency of latency and waveform. However, this feature of ERP analysis also limits its ability to capture non-phase-locked signals, resulting in the neglect of more complex dynamic changes in EEG signals, that is, ERP cannot fully capture the full complexity of EEG signals. Although time-frequency analysis can provide richer dynamic information, it lacks the accuracy of ERP in time positioning.

[0004] In addition, existing models built based on EEG signals have deficiencies in recognition accuracy and generalization, which limits their effectiveness in practical applications. The acquisition of EEG signals is often accompanied by high costs and complex operating procedures, which together lead to the limited size of available data sets. In the case of small samples, it is difficult for the model to learn enough features to ensure its universality and robustness in diverse scenarios. Therefore, how to improve the practical application value of the model, expand the EEG signal dataset, and increase the diversity and universality of samples is an urgent problem to be solved in current research.

[0005] YY+242232P Summary of the invention

[0006] In view of this, the embodiments of the present disclosure provide a method for constructing an image aesthetics EEG evaluation model and an image aesthetics EEG evaluation method, which can solve the problems existing in the prior art such as unclear image aesthetics-specific EEG indicators, small image research data sets, poor recognition accuracy, and poor generalization.

[0007] In a first aspect, the present disclosure provides a method for constructing an image aesthetics EEG evaluation model, which specifically includes the following solutions:

[0008] Stimulating the target object based on a preset type of preset aesthetic image to obtain a raw EEG signal;

[0009] Performing a first preprocessing and a second preprocessing on the original EEG signal to obtain a first type of EEG signal and a second type of EEG signal;

[0010] Processing the first type of EEG signal based on a first preset strategy to obtain a target ERP feature;

[0011] Processing the second type of EEG signal based on a second preset strategy to obtain target time-frequency features;

[0012] Based on the target ERP features and the target time-frequency features, obtaining aesthetic fusion EEG features;

[0013] Based on the aesthetic fusion EEG features, picture aesthetic preset classification information, and a preset subjective aesthetic evaluation database, an image aesthetic EEG evaluation model is constructed.

[0014] Optionally, the preset aesthetic images of the preset types include a first-class aesthetic image, a second-class aesthetic image and a third-class aesthetic image, and the aesthetic sensitivities of the first-class aesthetic image, the second-class aesthetic image and the third-class aesthetic image decrease in sequence.

[0015] Optionally, the method for obtaining the first type of EEG signal includes:

[0016] Determine whether there is abnormal lead data in the original EEG signal, if so, perform interpolation processing on the original EEG signal, and use the processed original EEG signal as the first type of target data; if not, use the original EEG signal as the first type of target data;

[0017] Determining a first signal segment in the first category of target data based on a first preset offline reference point and a first preset bandpass filter frequency;

[0018] Determine first electroencephalogram information corresponding to a first target time window based on the first signal segment; the first electroencephalogram information includes first time period information and second time period information;

[0019] Correcting the electroencephalogram data in the second time period information based on the average electroencephalogram amplitude in the first time period information as a baseline to obtain second electroencephalogram information;

[0020] The second electroencephalogram information includes first time period information and corrected second time period information;

[0021] The interference components in the second electroencephalogram information are eliminated to obtain the first type of electroencephalogram signal.

[0022] Optionally, the method for obtaining the second type of EEG signal includes:

[0023] Determine whether there is abnormal lead data in the original EEG signal. If so, interpolate the original EEG signal and use the processed original EEG signal as the second type of target data; if not, use the original YY+242232P

[0024] EEG signals are used as the second type of target data;

[0025] Determining a second signal segment in the second type of target data based on a second preset offline reference point, a second preset bandpass filter frequency, and a preset notch filter frequency;

[0026] Determine third electroencephalogram information corresponding to a second target time window based on the second signal segment; the third electroencephalogram information includes third time period information and fourth time period information;

[0027] Correcting the electroencephalogram data in the fourth time period information based on the average electroencephalogram amplitude in the third time period information as a baseline to obtain fourth electroencephalogram information;

[0028] The fourth electroencephalogram information includes the third time period information and the corrected fourth time period information;

[0029] The interference components in the fourth electroencephalogram information are eliminated to obtain the second type of electroencephalogram signal.

[0030] Optionally, the processing the first type of EEG signal based on a first preset strategy to obtain a target ERP feature includes:

[0031] The random fragmented graph is used as a control variable to analyze the first type of EEG signals and screen the initial EEG ERP features;

[0032] Acquire ERP component information and corresponding feature value information in the initial EEG ERP feature based on a plurality of first preset time windows;

[0033] Based on the ERP component information and the eigenvalue information, construct an ERP feature matrix;

[0034] The ERP feature matrix is ​​normalized to obtain the target ERP feature.

[0035] Optionally, the processing the second type of EEG signal based on a second preset strategy to obtain target time-frequency features includes:

[0036] Based on the second type of EEG signals, obtaining a plurality of EEG signal images corresponding to each type of preset aesthetic image;

[0037] Averaging a plurality of the electroencephalogram signal images to obtain each type of event-related spectral perturbation;

[0038] Performing an average process on all types of event-related spectral perturbations to obtain initial EEG time-frequency characteristics;

[0039] Acquire frequency band information of the initial EEG time-frequency characteristics and the corresponding average neural oscillation power based on a plurality of second preset time windows;

[0040] Based on the frequency band information and the average neural oscillation power, constructing an EEG time-frequency feature matrix;

[0041] The EEG time-frequency feature matrix is ​​normalized to obtain the target time-frequency feature.

[0042] Optionally, obtaining an aesthetic fusion EEG matrix based on the target ERP feature and the target time-frequency feature includes:

[0043] Determining a first channel set based on the target ERP feature;

[0044] Determine a second channel set based on the target time-frequency feature;

[0045] Aligning the first channel set and the second channel set to obtain a target channel set;

[0046] Based on the target channel set, extracting the ERP feature vector and the EEG time-frequency feature vector on each channel;

[0047] Based on the ERP feature vector and the EEG time-frequency feature vector, respectively obtain an ERP feature matrix and an EEG time-frequency feature matrix;

[0048] Based on the ERP feature matrix and the EEG time-frequency feature matrix, a single-channel fusion matrix is ​​obtained; the single-channel fusion matrix includes the ERP component information and the frequency band information;

[0049] Combining the single-channel fusion matrix with the preset channel value to obtain a target matrix;

[0050] The target matrix is ​​subjected to a self-transposition product transformation to obtain the aesthetic fusion EEG feature.

[0051] Optionally, the step of constructing an image aesthetic EEG evaluation model based on the aesthetic fusion EEG features, the preset classification information of picture aesthetics, and the preset subjective aesthetic evaluation database includes:

[0052] Converting the aesthetic fusion EEG features into a one-dimensional concatenated vector;

[0053] Performing data enhancement on the one-dimensional concatenated vector based on a generative adversarial network to generate an EEG feature vector;

[0054] Rearranging the EEG feature vector and the one-dimensional splicing vector according to the original channel sequence to obtain a fusion feature matrix;

[0055] Determine the initial model;

[0056] The initial model is trained based on the fusion feature matrix, and the trained initial model is used as the image aesthetic EEG evaluation model.

[0057] In a second aspect, the present disclosure also provides an image aesthetics EEG evaluation method, including:

[0058] In response to the evaluation request, a target model is called; the target model is a model constructed according to any of the above-mentioned image aesthetic EEG evaluation model construction methods;

[0059] The image to be analyzed is input into the target model to obtain aesthetic evaluation information of the corresponding image.

[0060] In a third aspect, the present disclosure also provides a system for constructing an image aesthetic EEG evaluation model, including:

[0061] An acquisition module, used to stimulate the target object based on a preset type of preset aesthetic image to obtain an original EEG signal;

[0062] A preprocessing module, used for performing a first preprocessing and a second preprocessing on the original EEG signal to obtain a first type of EEG signal and a second type of EEG signal;

[0063] A first acquisition module, used for processing the first type of EEG signal based on a first preset strategy to obtain a target ERP feature;

[0064] A second acquisition module, used for processing the second type of EEG signal based on a second preset strategy to obtain target time-frequency features;

[0065] A fusion module is used to obtain an aesthetic fusion based on the target ERP feature and the target time-frequency feature.

[0066] Combined EEG features;

[0067] A construction module is used to construct an image aesthetic EEG evaluation model based on the aesthetic fusion EEG features, picture aesthetic preset classification information, and a preset subjective aesthetic evaluation database.

[0068] In a fourth aspect, the present disclosure also provides an image aesthetics evaluation system, including:

[0069] A calling module, used for calling a target model in response to an evaluation request; the target model is a model constructed by the image aesthetic EEG evaluation model construction method;

[0070] The analysis module is used to input the image to be analyzed into the target model to obtain aesthetic evaluation information of the corresponding image.

[0071] In a fifth aspect, the embodiments of the present disclosure further provide a computer device, which adopts the following technical solution:

[0072] The computer device comprises:

[0073] at least one processor; and,

[0074] a memory communicatively connected to the at least one processor; wherein,

[0075] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any of the above-mentioned image aesthetics EEG evaluation model construction methods or the image aesthetics EEG evaluation methods.

[0076] In a sixth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium storing computer instructions for enabling a computer to execute any of the above-mentioned methods for constructing an image aesthetics EEG evaluation model or the above-mentioned image aesthetics EEG evaluation method.

[0077] In the seventh aspect, the embodiments of the present disclosure also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for constructing an image aesthetic EEG evaluation model or the image aesthetic EEG evaluation method described in any of the above items.

[0078] The method for constructing an image aesthetic EEG evaluation model disclosed in the present application stimulates a target object through preset aesthetic images of preset types, thereby ensuring the diversity and comprehensive coverage of the original EEG signals obtained; different types of EEG signal information are obtained through different preprocessing of the original EEG signals, which is convenient for subsequent comprehensive analysis of multiple features in the EEG signals; different strategies are respectively adopted to extract different types of target ERP features and target time-frequency features from the corresponding preprocessed EEG signals, and more comprehensive and accurate EEG signal information can be obtained through multi-angle feature extraction, effectively reducing the error caused by a single feature; then the two features are fused to obtain aesthetic fusion EEG features, which can obtain features containing multiple information such as component information, feature information, frequency band information, etc., and improve the ability to refine the analysis of EEG information; finally, a comprehensive aesthetic evaluation system can be formed by combining aesthetic fusion EEG features, picture aesthetic preset classification information, and preset subjective aesthetic evaluation database, and the image aesthetic EEG evaluation model constructed based on these multi-dimensional information can take into account all-round information such as subjective and objective information at the same time, effectively improve the recognition accuracy and generalization ability of the model under different conditions, and improve the scientificity and accuracy of aesthetic evaluation through the model.

[0079] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the following preferred embodiments are specifically cited and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0081] Figure 1 A flowchart of a method for constructing an image aesthetics EEG evaluation model provided in an embodiment of the present disclosure.

[0082] Figure 2 A schematic flow chart of a method for obtaining a first type of EEG signal provided in an embodiment of the present disclosure.

[0083] Figure 3 A flowchart of a method for obtaining a second type of EEG signal provided in an embodiment of the present disclosure.

[0084] Figure 4 A flowchart of a method for obtaining target ERP features provided in an embodiment of the present disclosure.

[0085] Figure 5It is a schematic diagram of a specific embodiment of the ERP feature matrix in this application.

[0086] Figure 6 A flowchart of a method for obtaining target time-frequency features provided in an embodiment of the present disclosure.

[0087] Figure 7 It is a schematic diagram of a specific embodiment of the EEG time-frequency feature matrix in this application.

[0088] Figure 8 A flowchart of a method for obtaining aesthetically fused EEG features provided in an embodiment of the present disclosure.

[0089] Fig. 9 A schematic diagram of a specific embodiment of the target matrix in the present application.

[0090] Fig.10 This is a schematic diagram of a specific embodiment of the aesthetic fusion EEG features in the present application.

[0091] Fig.11 A schematic diagram of a process for constructing an image aesthetics EEG evaluation model provided in an embodiment of the present disclosure.

[0092] Fig.12 A schematic diagram of a specific embodiment of a one-dimensional splicing vector provided in an embodiment of the present disclosure.

[0093] Fig.13 A schematic diagram of a method for generating an EEG feature vector according to a specific embodiment of the present disclosure.

[0094] Fig.14 A schematic diagram of a specific embodiment of the fusion feature matrix provided in the embodiment of the present disclosure.

[0095] Fig.15 A schematic diagram of a convolutional neural network model constructed using a self-attention mechanism provided in an embodiment of the present disclosure.

[0096] Fig.16 A flowchart of EEG data collection provided in an embodiment of the present disclosure.

[0097] Fig.17 A schematic diagram of the flow chart of the method for image aesthetics EEG evaluation provided in an embodiment of the present disclosure.

[0098] Fig.18 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0099] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0100] It should be clear that the following embodiments of the present disclosure are described by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other in the absence of conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0101] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0102] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0103] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0104] Reference Figure 1 , the present application discloses a method for constructing an image aesthetic EEG evaluation model, comprising:

[0105] S100, stimulating a target object based on a preset type of preset aesthetic image to obtain an original EEG signal.

[0106] Among them, the preset types of preset aesthetic images include type I aesthetic images, type II aesthetic images and type III aesthetic images, and the aesthetics of type I aesthetic images, type II aesthetic images and type III aesthetic images decrease in sequence, that is, the images used as stimulus sources include all preset types.

[0107] In a specific embodiment, the first type of aesthetic image refers to visual images that are widely recognized as being labeled as beauty, the second type of aesthetic image refers to visual images that are widely recognized as being labeled as neutral, and the third type of aesthetic image refers to visual images that are widely recognized as being labeled as ugliness.

[0108] Specifically, different categories of visual picture stimuli (i.e., multi-source stimulation) are presented to the target object, and an electroencephalogram (EEG) device can be used to collect the EEG signals of the target object (i.e., the subject) when viewing each image; wherein the visual picture stimuli may include paper images, electronic images, and other forms.

[0109] The target objects can be individuals of all ages who have the ability of aesthetic judgment.

[0110] In this step, by selecting a variety of images, we ensure that various aesthetic types are covered, so that the obtained EEG signals can more comprehensively reflect the subjects' reactions to different types of images. At the same time, the original EEG signals can be obtained, laying the foundation for subsequent processing and analysis.

[0111] S200, performing a first preprocessing and a second preprocessing on the original EEG signal to obtain a first type of EEG signal and a second type of EEG signal.

[0112] By processing the original EEG signals through different preprocessing methods and improving the corresponding data quality, the EEG signals of the corresponding categories can be obtained.

[0113] S300, processing the first type of EEG signal based on the first preset strategy to obtain target ERP features; processing the second type of EEG signal based on the second preset strategy to obtain target EEG time-frequency features.

[0114] Among them, the target ERP (event-related potential) feature is the target event-related potential feature, which can provide information about the neural activity related to the EEG signal and a specific event, while the time-frequency feature can capture the changes in the signal in time and frequency.

[0115] The ERP features obtained through this step can provide cognitive response information related to visual stimulation, which helps to understand the subjects' immediate aesthetic evaluation of the image; the obtained EEG time-frequency features provide the changes of EEG signals at different time points and frequency ranges, which can reveal a more detailed aesthetic processing process.

[0116] S400, based on the target ERP features and target time-frequency features, obtain aesthetic fusion EEG features.

[0117] The obtained aesthetic fusion EEG features integrate different types of EEG features, making the final aesthetic features more comprehensive and rich, improving the ability to capture complex aesthetic information, and enhancing the accuracy of evaluation.

[0118] S500,constructs an image aesthetic EEG evaluation model based on aesthetic fusion EEG features, picture aesthetic preset classification information, and preset subjective aesthetic evaluation database.

[0119] The final aesthetic evaluation model can perform image aesthetic evaluation based on EEG signals, combining subjective and objective data to provide more reliable evaluation results. The model can be used in various application scenarios, such as image optimization, advertising design, artistic creation, etc., to help achieve designs that are more in line with aesthetic standards.

[0120] The method for constructing an image aesthetic EEG evaluation model disclosed in the present application stimulates a target object through preset aesthetic images of preset types, thereby ensuring the diversity and comprehensive coverage of the original EEG signals obtained; different types of EEG signal information are obtained through different preprocessing of the original EEG signals, which is convenient for subsequent comprehensive analysis of multiple features in the EEG signals; different strategies are respectively adopted to extract different types of target ERP features and target time-frequency features from the corresponding preprocessed EEG signals, and more comprehensive and accurate EEG signal information can be obtained through multi-angle feature extraction, effectively reducing the error caused by a single feature; then the two features are fused to obtain aesthetic fusion EEG features, which can obtain features containing multiple information such as component information, feature information, frequency band information, etc., and improve the ability to refine the analysis of EEG information; finally, a comprehensive aesthetic evaluation system can be formed by combining aesthetic fusion EEG features, picture aesthetic preset classification information, and preset subjective aesthetic evaluation database, and the image aesthetic EEG evaluation model constructed based on these multi-dimensional information can take into account all-round information such as subjective and objective information at the same time, effectively improve the recognition accuracy and generalization ability of the model under different conditions, and improve the scientificity and accuracy of aesthetic evaluation through the model.

[0121] Therefore, the image aesthetics EEG evaluation model construction method disclosed in the present application, through multi-level and multi-dimensional data analysis and fusion processing to construct the image aesthetics EEG evaluation model, can provide more accurate, personalized and comprehensive image aesthetics evaluation results.

[0122] Reference Figure 2 , the first type of EEG signal acquisition method includes:

[0123] A100, judging whether there is abnormal lead data in the original EEG signal, if so, performing interpolation processing on the original EEG signal and taking the processed original EEG signal as the first category target data; if not, taking the original EEG signal as the first category target data.

[0124] Specifically, if there is abnormal lead data in the original EEG signal, interpolation processing of the abnormal lead data in this embodiment includes: 1) determining the position of the abnormal lead data (i.e., the position of the abnormal data point); 2) obtaining normal data points around (i.e., adjacent to) the position of the abnormal lead data; 3) calculating the average value of the surrounding normal data points; 4) replacing the abnormal lead data with the average value, i.e., replacing the value of the abnormal data point with the average value. Through the judgment processing of this step, it is guaranteed to obtain smooth and consistent data information.

[0125] A200, determining a first signal segment in a first category of target data based on a first preset offline reference point and a first preset bandpass filter frequency.

[0126] In a specific embodiment, bilateral mastoid processes (M1, M2) can be used as the first preset offline reference points, and the first preset bandpass filter frequency includes 0.1 to 30 Hz, that is, by bandpass filtering the EEG data, the signal in the frequency range of 0.1 to 30 Hz is retained. This can remove noise above or below this frequency range and retain the EEG signal we are concerned about.

[0127] A300, determining first electroencephalogram information corresponding to a first target time window based on the first signal segment; the first electroencephalogram information includes first time period information and second time period information.

[0128] In a preferred embodiment, the moment when the aesthetic image appears is selected as the starting point of zero milliseconds, and the first target time window refers to 200 milliseconds before and 800 milliseconds after the zero millisecond starting point, which is recorded as [-200ms 800ms]. The first time period information includes the EEG data information of the time window from 200 milliseconds before the zero millisecond starting point to the zero millisecond starting point, that is, the corresponding time window is [-200ms 0]; the second time period information includes the EEG data information of the time window from the zero millisecond starting point to 800 milliseconds after the zero millisecond starting point, that is, the corresponding time window is [0800ms].

[0129] A400, correcting the electroencephalogram data in the second time period information based on the average electroencephalogram amplitude in the first time period information to obtain second electroencephalogram information; the second electroencephalogram information includes the first time period information and the corrected second time period information.

[0130] In this preferred embodiment, the average EEG amplitude in the first time period information refers to the average EEG amplitude in the time window corresponding to [-200ms 0], that is, the EEG data within 200 milliseconds before the event ([-200ms 0]) is used as the baseline, and the data after the event is corrected, that is, the data within [0800ms] is corrected to eliminate the influence of background noise and make event-related changes more obvious.

[0131] A500, removes the interference components in the second EEG information and obtains the first type of EEG signal.

[0132] Among them, the interference components include interference information caused by electrooculography (blinking, etc.) and electromyography (muscle activity).

[0133] Specifically, an independent component analysis (ICA) algorithm can be used to remove more than 90% of interference components that may be electrooculography and electromyography, so as to improve the clarity and accuracy of the EEG data.

[0134] In this step, the main task of EOG is to remove abnormal trials with EOG electrode amplitude exceeding ±100 μV to avoid extreme data affecting the analysis results, that is, to exclude data generated by abnormal activities or artifacts (such as poor electrode contact).

[0135] Reference Figure 3 , the method for obtaining the second type of EEG signal includes:

[0136] B100, determine whether there is abnormal lead data in the original EEG signal. If so, perform interpolation processing on the original EEG signal and use the processed original EEG signal as the second type of target data; if not, use the original EEG signal as the second type of target data.

[0137] Specifically, if there is abnormal lead data in the original EEG signal, interpolation processing of the abnormal lead data in this embodiment includes: 1) determining the position of the abnormal lead data (i.e., the position of the abnormal data point); 2) obtaining normal data points around (i.e., adjacent to) the position of the abnormal lead data; 3) calculating the average value of the surrounding normal data points; 4) replacing the abnormal lead data with the average value, i.e., replacing the value of the abnormal data point with the average value. Through the abnormal judgment processing of this step, it is guaranteed to obtain smooth and consistent data information.

[0138] B200, determining a second signal segment in the second type of target data based on a second preset offline reference point, a second preset bandpass filter frequency, and a preset notch filter frequency.

[0139] In a specific embodiment, bilateral mastoid processes (M1, M2) may be used as the second preset offline reference point, which is consistent with the first preset offline reference point; the second preset bandpass filter frequency includes 0.1-150 Hz, and the preset notch filter frequency includes 48-52 Hz.

[0140] In this step, two filtering processes are performed. First, the second type of target data is band-pass filtered through a band-pass filter to retain the signal in the frequency range of 0.1 to 150 Hz, that is, only the signal in the frequency range of interest is retained; then the second type of target data after one processing is filtered through a notch filter to remove the frequency component close to 50 Hz and obtain the second signal segment. The notch filter can remove the interference of specific frequencies to ensure that the signal is purer. It should be noted that the order of these two processes can be swapped as long as the processing requirements are met and the acquisition of the second signal segment is not affected.

[0141] B300, determining third electroencephalogram information corresponding to the second target time window based on the second signal segment; the third electroencephalogram information includes third time period information and fourth time period information.

[0142] In a preferred embodiment, the moment when the aesthetic image appears is selected as the zero millisecond starting point, and the second target time window refers to 500 milliseconds before and 2000 milliseconds after the zero millisecond starting point, which is recorded as [-500ms 2000ms], wherein the third time period information includes the electroencephalogram data information of the time window from 500 milliseconds before the zero millisecond starting point (i.e. 500ms before stimulation) to the zero millisecond starting point (i.e. the stimulation starting point), that is, the corresponding time window is [-500ms 0]; the fourth time period information includes the electroencephalogram data information of the time window from the zero millisecond starting point to 2000 milliseconds after the zero millisecond starting point, that is, the corresponding time window is [02000ms].

[0143] B400, correcting the EEG data in the fourth time period information based on the average EEG amplitude in the third time period information as a baseline to obtain fourth EEG information; the fourth EEG information includes the third time period information and the corrected fourth time period information.

[0144] In this preferred embodiment, the average EEG amplitude in the third time period information refers to the average EEG amplitude in the time window corresponding to [-500ms 0], that is, the EEG data within 500 milliseconds before the event ([-500ms 0]) is used as the baseline, and the data after the event is corrected, that is, the data within [02000ms] is corrected to eliminate the influence of background noise and make event-related changes more obvious.

[0145] B500, remove the interference components in the fourth EEG information and obtain the second type of EEG signal.

[0146] Specifically, the ICA algorithm is applied to eliminate electrooculography and electromyography interference, and manual screening is performed to resolve residual artifacts; the specific processing method can refer to the acquisition of the first type of EEG signal.

[0147] Reference Figure 4 , the method for obtaining the target ERP features includes:

[0148] C100, using random fragmentation graph as control variable to analyze the first type of EEG signal and screen the initial EEG ERP features.

[0149] Among them, the initial EEG ERP features refer to the features related to aesthetics that can distinguish aesthetic types.

[0150] Specifically, random fragment images were used as control variables to control the physical stimulation effects of the three types of aesthetic pictures: beautiful, neutral, and ugly. A variance analysis of 2 (image type: interface vs fragments) × 3 (aesthetic level: beautiful vs neutral vs ugly) was conducted to obtain some significant difference information, and the EEG components corresponding to the significant differences less than the preset threshold were used as the initial EEG ERP features. Through variance analysis, the EEG activities under different conditions can be compared to evaluate which event-related potential components can effectively distinguish the three types of aesthetic pictures and their corresponding random fragment images, which means finding those ERP features that show significant differences under different aesthetic levels but are not affected by basic visual features (color, brightness, contrast, etc.).

[0151] YY+242232P

[0152] In a specific embodiment, the results of variance analysis showed that the EEG components N1 ([80-120]ms), P2 ([140-180]ms, N2 ([200-280]ms and LPP ([400-600]ms) had significant differences in the interaction effects of different image types and aesthetic levels (i.e., P<0.05), and the above components were selected as the initial EEG ERP features related to aesthetics, and the initial EEG ERP features included corresponding component features and channel values.

[0153] C200, obtaining ERP component information and corresponding feature value information in the initial EEG ERP features based on a plurality of first preset time windows.

[0154] Specifically, ERP component information refers to N1 ([80-120]ms), P2 ([140-180]ms, N2 ([200-280]ms, LPP ([400-600]ms), and feature value information refers to average amplitude, latency, and peak.

[0155] In this preferred embodiment, the characteristic value information is preferably these three types. It should be noted that those skilled in the art can also select other information according to their needs, all of which are within the protection scope of this application.

[0156] C300,constructs the ERP feature matrix based on ERP component information and eigenvalue information.

[0157] Specific reference Figure 5 In this preferred embodiment, based on the ERP component information (i.e., N1, P2, N2, LPP) and the eigenvalue information (Amplitude, Latency, Peak), the constructed ERP feature matrix is ​​a 4×3 feature matrix.

[0158] C400, normalize the ERP feature matrix to obtain the target ERP features.

[0159] Specifically, normalization processing can adjust the feature values ​​to a standard range and eliminate the dimensional differences between different features; wherein the shape of the target ERP feature is (4,3).

[0160] Reference Figure 6 , the method for obtaining the target time-frequency features includes:

[0161] D100, based on the second type of EEG signal, obtains several EEG signal images corresponding to each type of preset aesthetic image.

[0162] Through this step, the EEG signal images based on the same type of stimulation are screened out.

[0163] D200, several copies of the EEG signal images were averaged to obtain each type of event-related spectral perturbation.

[0164] Specifically, the Morlet wavelet transform was used to calculate the time-frequency representation of the single trial data (i.e., each EEG signal graph), where the wavelet width was set to 3 for low frequency (1-30 Hz) and 7 for high frequency (30-100 Hz), and the power change was calculated with a frequency step of 1 Hz and a time step of 10 ms. Multiple EEG signal graphs of the same type of stimulation were then averaged to reduce random noise and improve the signal-to-noise ratio of the signal, thereby reducing noise and instability factors in a single measurement, thereby obtaining a more stable event-related spectral perturbation (ERSP).

[0165] D300, all types of event-related spectral perturbations are averaged to obtain the initial EEG time-frequency characteristics.

[0166] In this step, the ERSPs of the different types of aesthetic images obtained previously are averaged to obtain a unified time-frequency feature; by averaging different types of ERSPs, the commonalities and differences in the brain's perception of different aesthetic levels can be highlighted, providing a basis for establishing a more consistent model.

[0167] D400, obtaining frequency band information of the initial EEG time-frequency characteristics and the corresponding average neural oscillation power based on multiple second preset time windows.

[0168] Specifically, a point-to-point sample t-test was performed on all electrodes in all frequency bands (1-100 Hz) and time windows (-100 ms to 2000 ms) to compare the power activations of the three types of interface images (alpha = 0.05), and the false discovery rate (FDR) correction method was used for processing; based on the corrected test results, statistically significant frequencies, time windows, and brain regions were selected to define the region of interest (ROI).

[0169] Then, taking the average neural oscillation power within the ROI as the dependent variable, we further tested the main effects and interactions of image type and aesthetic level through 2 (image type: interface vs fragments) × 3 (aesthetic level: beautiful vs neutral vs uglier) repeated measures analysis of variance to identify those EEG signal features that showed significant differences at different aesthetic levels.

[0170] In a specific embodiment, the results of variance analysis show that the average neural oscillation power of alpha ([200270]ms), beta ([200270]ms) and theta ([200300]ms) has significant differences in the interaction effect of different image types and aesthetic levels (P<0.05), and the above-mentioned frequency bands are selected as time-frequency features related to aesthetics, which preferably include frequency band information and its corresponding average neural oscillation power. It should be noted that other parameters can also be selected, all of which are within the scope of protection of this application.

[0171] By analyzing frequency band information within a specific time window, we can focus on neural oscillation activities that are closely related to specific cognitive processes or psychological states.

[0172] D500, constructs the EEG time-frequency feature matrix based on frequency band information and average neural oscillation power.

[0173] Also refer to Figure 7 The EEG time-frequency feature matrix constructed based on frequency band information and average neural oscillation power is a 3×1 feature matrix. The EEG time-frequency feature matrix can present complex time and frequency information in a structured manner, which is convenient for machine learning and statistical analysis.

[0174] D600, normalize the EEG time-frequency feature matrix to obtain the target time-frequency features.

[0175] Specifically, normalization processing can adjust the feature value to a standard range and eliminate the dimensional differences between different features; wherein the shape of the target time-frequency feature is (3,1).

[0176] Reference Figure 8 ,The method for obtaining aesthetic fusion EEG features includes:

[0177] S410, determining a first channel set based on the target ERP feature;

[0178] A second channel set is determined based on the target time-frequency feature.

[0179] In a preferred embodiment, the first channel set determined based on the target ERP features includes FP1, FPz, FP2, O3, Oz, O4, F3, Fz, F4, C3, Cz, C4, P3, Pz, and P4 channels of ERP components related to aesthetics.

[0180] The second channel set determined based on the target time-frequency features includes PO4, PO6, PO7, PO8, P7, P8, O2, and F7 channels of time-frequency EEG features related to aesthetics.

[0181] By determining the first channel set based on the target ERP feature, the precise channel of the EEG activity related to the target event can be determined, ensuring that the selected channel has a high signal-to-noise ratio, which helps to capture the ERP feature more accurately.

[0182] By determining the second channel set based on the target time-frequency features, more fine-grained EEG activities can be captured, revealing important information in the frequency band, so that both time domain response and frequency domain information are considered in subsequent fusion.

[0183] S420, aligning the first channel set and the second channel set to obtain a target channel set.

[0184] In this step, alignment is to adjust the dimension and format of the feature vector on each channel to be consistent, and the obtained target channel set includes 23 channels, that is, the number of channels in the target channel set is the sum of the number of channels in the first channel set and the number of channels in the second channel set.

[0185] S430, based on the target channel set, extracting the ERP feature vector and the EEG time-frequency feature vector on each channel; based on the ERP feature vector and the EEG time-frequency feature vector, respectively obtaining the ERP feature matrix and the EEG time-frequency feature matrix.

[0186] Among them, the shape of the ERP feature matrix obtained based on the extracted ERP feature vector on each channel is (4,3), and the shape of the EEG time-frequency feature matrix obtained based on the extracted EEG time-frequency feature vector on each channel is (3,1).

[0187] S440, obtaining a single channel fusion matrix based on the ERP feature matrix and the EEG time-frequency feature matrix.

[0188] The single-channel fusion matrix includes ERP component information and frequency band information on the channels of the target channel set.

[0189] Specifically, the ERP feature matrix and the EEG time-frequency feature matrix are multiplied to generate a new single-channel fusion matrix, and the shape of the obtained single-channel fusion matrix is ​​(4,1).

[0190] In terms of feature processing, time-frequency analysis technology is used, which can not only capture the instantaneous changes of signals in the time and frequency domains, but also ensure the effectiveness of the fused features through matrix product fusion with ERP features, while reducing redundancy. This fusion strategy optimizes the expressiveness of the features and lays a good foundation for subsequent model training and evaluation.

[0191] S450, combining the single-channel fusion matrix with the preset channel value to obtain a target matrix.

[0192] The preset channel value is the number of channels in the target channel set.

[0193] Also refer to Fig. 9 , combining the single-channel fusion matrix with the preset channel value, and obtaining the shape of the target matrix is ​​(4, 23). Through this step, the data can be reduced in dimension while retaining the original data information, thereby reducing the data complexity.

[0194] S460, performing a self-transposition product transformation on the target matrix to obtain aesthetic fusion EEG features.

[0195] Also refer to Fig.10 In a preferred embodiment, the target matrix of the (4, 23) shape is transformed into a matrix of the (23, 23) shape by self-transposition product transformation as the aesthetic fusion EEG feature (i.e., fusion feature matrix). The self-transposition product transformation can extract high-order features and relationships in the data and enhance the expressive power of the features.

[0196] Reference Fig.11 , the construction method of the image aesthetic EEG evaluation model includes:

[0197] S510, converting the aesthetic fusion EEG features into a one-dimensional splicing vector.

[0198] Specifically, the aesthetic fusion EEG matrix is ​​spliced ​​by a horizontal splicing method to obtain a converted one-dimensional splicing vector. In a preferred embodiment, the specific one-dimensional splicing vector is as follows: Fig.12 , the high-dimensional EEG data is converted into a one-dimensional vector for subsequent processing and analysis.

[0199] S520, performing data enhancement on the one-dimensional concatenated vector based on a generative adversarial network to generate an EEG feature vector.

[0200] Specifically, refer to Fig.13 , construct a generative adversarial network (GAN), including a generator and a discriminator. The generator accepts random noise input and generates an enhanced EEG feature vector, while the discriminator evaluates whether the generated feature vector is real, further optimizing the performance of the generator; use GAN to perform data enhancement on the original one-dimensional concatenated vector to generate multiple different EEG feature vectors, increasing the diversity and quantity of data.

[0201] In addition, the application of data augmentation (DADA) increases data diversity, which not only improves the applicability of the EEG signal evaluation model when the sample size is small, but also reduces the reliance on expert knowledge through automation, reducing the workload of manually designing data augmentation strategies. At the same time, DADA's differentiable automatic data augmentation strategy search method significantly improves search efficiency, reduces computational costs, and speeds up the process of model training and tuning.

[0202] The generated EEG feature vector is similar to the original EEG data (i.e., the one-dimensional concatenated vector) but slightly changed. This process helps to increase the diversity of training samples and thus improve the generalization ability of the model.

[0203] Data augmentation in this step can generate more samples, which helps improve the robustness and generalization ability of the model. If the original data samples are insufficient, GAN can help generate synthetic data to alleviate the problem of insufficient data.

[0204] S530, rearrange the EEG feature vector and the one-dimensional concatenated vector according to the original channel sequence to obtain a fusion feature matrix.

[0205] The original channel sequence may be the channel sequence of the target channel set.

[0206] Also refer to Fig.14 , the obtained fusion feature matrix has the same dimensional structure as the original feature matrix (i.e., aesthetic fusion EEG features).

[0207] S540, determining an initial model.

[0208] Specifically, select the appropriate model architecture according to the needs, and configure the model parameters and training settings, including optimization algorithm, learning rate, etc. Establishing the initial model provides the basis for subsequent training, helps to form a systematic evaluation framework, and ensures that the model can effectively process the fusion feature matrix by reasonably configuring the model parameters.

[0209] S550, training the initial model based on the fused feature matrix, and using the trained initial model as an image aesthetic EEG evaluation model.

[0210] The initial model is trained using the fused feature matrix, and the model weights and parameters are adjusted so that it can accurately predict the aesthetic score of the image. The performance of the trained model is evaluated through methods such as cross-validation to ensure that it has a high prediction accuracy.

[0211] The trained model obtained by the method can effectively evaluate the aesthetic value of the image and provide users with scientific and objective evaluation results. The trained model can be applied to multiple fields such as image aesthetic evaluation and design optimization.

[0212] Specifically, refer to Fig.15 In this embodiment, the initial model is preferably a convolutional neural network (Transformer-based CNN) with a self-attention mechanism.

[0213] The local features of EEG data are extracted through convolutional neural network (CNN). As a feature extractor, CNN can capture the pattern and structure information in the data and provide rich input features for the subsequent Transformer layer.

[0214] The Transformer layer is used to process the feature sequence extracted by CNN. Through the self-attention mechanism and position encoding, the Transformer layer can effectively capture the global dependencies between input features and enhance the model's ability to understand data. The parallel computing characteristics of self-attention are also suitable for enhanced large-scale data sets.

[0215] The categorical cross entropy loss function is then used to measure the difference between the model prediction and the actual label, and the model parameters are updated through the Adam optimizer. This combination ensures the efficiency and stability of the model during training.

[0216] The performance of the model is evaluated on the validation set, and the model structure and hyperparameters are adjusted according to the evaluation results to further improve the accuracy and generalization ability of the model. Finally, the trained initial model is used as the image aesthetic EEG evaluation model.

[0217] In this embodiment, the introduction of the self-attention mechanism effectively reduces the noise interference that may be introduced and improves the ability to capture aesthetic features. Its parallel computing characteristics make the application of the model on large-scale data sets after data enhancement more efficient. During the model training process, the combination of the Adam optimizer and the classification cross entropy loss function ensures the efficiency and stability of the model training process. The addition of the evaluation and optimization steps further improves the accuracy of the model.

[0218] Finally, the model combining CNN and Transformer can capture local features and global dependencies at the same time, providing richer information for aesthetic evaluation. The flexibility of this model structure enables it to adapt to different data volumes and classification tasks, and has good generalization and universality.

[0219] In summary, this scheme not only improves the accuracy of aesthetic assessment, but also reduces resource consumption and improves computational efficiency through automation and optimization techniques; its application on different data sets has demonstrated obvious superiority, providing an innovative, efficient and reliable technical solution for the field of aesthetic assessment.

[0220] Further, refer to Fig.16 Regarding the analysis methods of EEG signals, current research often focuses on the exploration of single features, and lacks a comprehensive analysis of aesthetic features in different analysis methods. Through the scheme disclosed in this application, three classification standards of beautiful, average, and ugly are adopted, and random fragmented images are introduced as control variables to effectively separate the influence of physical stimulation of pictures on EEG signals. Through this processing, we successfully extracted the neural signal features directly related to aesthetics, and constructed a feature database and recognition model, which provided a solid foundation for the evaluation of visual images. By taking into account multiple feature analyses, the comprehensiveness and accuracy of the data are guaranteed.

[0221] Specifically, Fig.16 In a, the aesthetic classification interface diagram and its corresponding random fragment diagram are used to analyze the EEG differences generated by these two types of pictures. The influence of visual physical stimulation (such as color and brightness) can be controlled, and the EEG ERP and time-frequency characteristic signals directly related to aesthetics can be separated more accurately. Fig.16 The b in it is the process of EEG signal collection. After 500ms of fixed gaze, the stimulus picture is presented for 4000ms, and the target object makes a subjective evaluation of the picture on a 7-point scale. After an interval of 800-1200ms, the next picture stimulation process is carried out. The EEG signals are recorded throughout the process, and complete EEG activity data can be obtained to ensure the comprehensiveness and meticulousness of the data, which is helpful to analyze the impact of different stimuli on EEG signals and identify features related to aesthetic perception.

[0222] In the method disclosed in the application, 1) the aesthetic signal can be accurately separated, and the EEG difference analysis of two types of images can be combined to more accurately separate and identify EEG features related to aesthetic perception; 2) the scientific nature of aesthetic evaluation can be improved: by combining objective analysis of EEG signals with subjective evaluation, a more scientific aesthetic evaluation method can be provided; 3) the effectiveness of experimental design can be enhanced: by designing appropriate stimulus presentation time and interval time, the effectiveness of the experiment and the reliability of data can be improved; 4) the understanding of aesthetics can be enriched: the impact of visual stimulation on EEG activity can be deeply understood, providing new perspectives and data support for aesthetic research. This EEG signal analysis method provides a scientific and objective evaluation method for the study of image aesthetics, and can play an important role in theoretical research and practical applications.

[0223] Reference Fig.17 The second aspect of the present application discloses an image aesthetic EEG evaluation method, which is an image aesthetic EEG evaluation method based on EEG ERP and time-frequency feature data, specifically comprising:

[0224] S10, calling the target model in response to the evaluation request.

[0225] Specifically, the called target model is a model constructed using the image aesthetic EEG evaluation model construction method disclosed in the first aspect of the present application.

[0226] Specifically, the system receives a user's image aesthetics evaluation request, which can be a user uploading a picture, or an evaluation request initiated through a system interface; then loads a target model, which is a model constructed using the image aesthetics EEG evaluation model construction method disclosed in the first aspect of the present application. After the model is loaded, the system performs necessary initialization, such as setting model parameters, loading pre-trained weights, etc., to ensure that the model can work normally in the subsequent evaluation.

[0227] In this step, we clarify user needs, ensure the pertinence and accuracy of the evaluation task, ensure that an optimized and professional model is used, and improve the reliability of the evaluation results. We also initialize the model to ensure the stability and performance of the model during the evaluation process and avoid affecting the evaluation results due to abnormal model status.

[0228] S20, inputting the image to be analyzed into the target model to obtain aesthetic evaluation information of the corresponding image.

[0229] Specifically, before the image to be analyzed is input into the model, necessary preprocessing can be performed on the image, such as scaling, denoising, adjusting brightness, etc., to meet the input requirements of the model; then the processed image data is input into the target model, and the model will analyze the image, extract features and evaluate it. Finally, the model outputs the aesthetic evaluation information of the image, which may include one or more of the following: aesthetic score, image quality evaluation, and recommended improvement suggestions.

[0230] Through this model, images can be automatically analyzed in depth and evaluated using its training and learning capabilities. It can provide detailed aesthetic evaluation information to help users understand the aesthetic quality of images and make improvements based on the evaluation results.

[0231] This method can automatically evaluate the aesthetic quality of images, reduce manual intervention, and improve evaluation efficiency. It uses advanced models and methods for evaluation and can provide more professional and accurate aesthetic evaluations. It uses fixed models and evaluation criteria and can provide consistent evaluation results between different images. The evaluation results include not only aesthetic scores but also specific improvement suggestions to help users optimize image quality.

[0232] This method can provide users with fast aesthetic evaluation services and improve the quality of users' decision-making in the process of image creation and selection. That is, through modeled image aesthetic evaluation, it can help users obtain accurate aesthetic feedback and improve overall work efficiency and evaluation quality through automated means.

[0233] In a third aspect, the present application discloses a system for constructing an image aesthetic EEG evaluation model, comprising:

[0234] An acquisition module, used to stimulate the target object based on a preset type of preset aesthetic image to obtain an original EEG signal;

[0235] A preprocessing module, used to perform a first preprocessing and a second preprocessing on the original EEG signal to obtain a first type of EEG signal and a second type of EEG signal;

[0236] A first acquisition module, used to process the first type of EEG signal based on a first preset strategy to obtain a target ERP feature;

[0237] The second acquisition module is used to process the second type of EEG signal based on the second preset strategy to obtain YY+242232P

[0238] Target time-frequency characteristics;

[0239] A fusion module is used to obtain an aesthetic fusion EEG matrix based on target ERP features and target time-frequency features;

[0240] The construction module is used to construct an image aesthetic EEG evaluation model based on the aesthetic fusion EEG matrix, picture aesthetic preset classification information, and preset subjective aesthetic evaluation database.

[0241] It should be noted that the scheme in the method for constructing an image aesthetic EEG evaluation model disclosed in the first aspect of the present application is applicable to the system for constructing an image aesthetic EEG evaluation model disclosed in the third aspect of the present application.

[0242] In a fourth aspect, the present application discloses an image aesthetics evaluation system, comprising:

[0243] A calling module is used to call a target model in response to an evaluation request; the target model is a model constructed by the image aesthetics EEG evaluation model construction method;

[0244] The analysis module is used to input the image to be analyzed into the target model to obtain aesthetic evaluation information of the corresponding image.

[0245] It should be noted that the scheme in the image aesthetics EEG evaluation method disclosed in the second aspect of the present application is applicable to the image aesthetics evaluation system disclosed in the fourth aspect of the present application.

[0246] The computer device according to the embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0247] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory, so that the computer device executes the aforementioned image aesthetics EEG evaluation model construction method of each embodiment of the present disclosure or all or part of the steps of the image aesthetics EEG evaluation method.

[0248] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0249] like Fig.18A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure is shown, which is a schematic diagram of the structure of a computer device suitable for implementing the embodiment of the present disclosure. Fig.18 The computer device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0250] like Fig.18 As shown, the computer device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer device are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0251] Typically, the following devices can be connected to the I / O interface: input devices such as sensors or visual information acquisition devices; output devices such as display screens; storage devices such as tapes, hard disks, etc.; and communication devices. The communication device can allow the computer device to communicate with other devices (such as edge computing devices) wirelessly or by wire to exchange data. Fig.18 A computer device having various devices is shown, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0252] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the image aesthetics EEG evaluation model construction method or the image aesthetics EEG evaluation method of the embodiment of the present disclosure are executed.

[0253] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0254] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the image aesthetics EEG evaluation model construction method or the image aesthetics EEG evaluation method of the aforementioned embodiments of the present disclosure are executed.

[0255] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0256] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0257] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0258] In the present disclosure, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. The block diagrams of the devices, devices, equipment, and systems involved in the present disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0259] Additionally, as used herein, "or" used in a list of items beginning with "at least one" indicates a separate list, so that, for example, a list of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not mean that the example described is preferred or better than other examples.

[0260] It should also be noted that in the system and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0261] Various changes, substitutions, and modifications of the techniques described herein may be made without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of the present disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of events, means, methods, and actions described above. Currently existing or later to be developed processes, machines, manufactures, compositions of events, means, methods, or actions that perform substantially the same functions or achieve substantially the same results as the corresponding aspects described herein may be utilized. Thus, the appended claims include such processes, machines, manufactures, compositions of events, means, methods, or actions within their scope.

[0262] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0263] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for constructing an image aesthetic EEG evaluation model, characterized in that: include: Stimulating the target object based on a preset type of preset aesthetic image to obtain a raw EEG signal; Performing a first preprocessing and a second preprocessing on the original EEG signal to obtain a first type of EEG signal and a second type of EEG signal; Processing the first type of EEG signal based on a first preset strategy to obtain a target ERP feature; Processing the second type of EEG signal based on a second preset strategy to obtain target time-frequency features; Based on the target ERP features and the target time-frequency features, obtaining aesthetic fusion EEG features; Based on the aesthetic fusion EEG features, picture aesthetic preset classification information, and a preset subjective aesthetic evaluation database, an image aesthetic EEG evaluation model is constructed.

2. The method for constructing an image aesthetic EEG evaluation model according to claim 1, characterized in that: The preset aesthetic images of the preset types include a first-class aesthetic image, a second-class aesthetic image and a third-class aesthetic image, and the aesthetic sensitivities of the first-class aesthetic image, the second-class aesthetic image and the third-class aesthetic image decrease in sequence.

3. The method for constructing an image aesthetic EEG evaluation model according to claim 1, characterized in that: The method for obtaining the first type of EEG signal comprises: Determine whether there is abnormal lead data in the original EEG signal, if so, perform interpolation processing on the original EEG signal, and use the processed original EEG signal as the first type of target data; if not, use the original EEG signal as the first type of target data; Determining a first signal segment in the first category of target data based on a first preset offline reference point and a first preset bandpass filter frequency; Determine first electroencephalogram information corresponding to a first target time window based on the first signal segment; the first electroencephalogram information includes first time period information and second time period information; Correcting the electroencephalogram data in the second time period information based on the average electroencephalogram amplitude in the first time period information as a baseline to obtain second electroencephalogram information; The second electroencephalogram information includes first time period information and corrected second time period information; The interference components in the second electroencephalogram information are eliminated to obtain the first type of electroencephalogram signal.

4. The method for constructing an image aesthetic EEG evaluation model according to claim 1, characterized in that: The method for obtaining the second type of EEG signal includes: Determine whether there is abnormal lead data in the original EEG signal. If so, perform interpolation processing on the original EEG signal and use the processed original EEG signal as the second type of target data; if not, use the original EEG signal as the second type of target data; Determining a second signal segment in the second type of target data based on a second preset offline reference point, a second preset bandpass filter frequency, and a preset notch filter frequency; Determine third electroencephalogram information corresponding to a second target time window based on the second signal segment; the third electroencephalogram information includes third time period information and fourth time period information; Correcting the electroencephalogram data in the fourth time period information based on the average electroencephalogram amplitude in the third time period information as a baseline to obtain fourth electroencephalogram information; The fourth electroencephalogram information includes the third time period information and the corrected fourth time period information; The interference components in the fourth electroencephalogram information are eliminated to obtain the second type of electroencephalogram signal.

5. The method for constructing an image aesthetic EEG evaluation model according to claim 1, characterized in that: The step of processing the first type of EEG signal based on a first preset strategy to obtain a target ERP feature includes: The random fragmented graph is used as a control variable to analyze the first type of EEG signals and screen the initial EEG ERP features; Acquire ERP component information and corresponding feature value information in the initial EEG ERP feature based on a plurality of first preset time windows; Based on the ERP component information and the eigenvalue information, construct an ERP feature matrix; The ERP feature matrix is ​​normalized to obtain the target ERP feature.

6. The method for constructing an image aesthetic EEG evaluation model according to claim 5, characterized in that: The processing of the second type of EEG signal based on the second preset strategy to obtain target time-frequency features includes: Based on the second type of EEG signals, obtaining a plurality of EEG signal images corresponding to each type of preset aesthetic image; Averaging a plurality of the electroencephalogram signal images to obtain each type of event-related spectral perturbation; Performing an average process on all types of event-related spectral perturbations to obtain initial EEG time-frequency characteristics; Based on multiple second preset time windows, the frequency band information YY+242232P of the initial EEG time-frequency feature is obtained Information and the corresponding average neural oscillation power; Based on the frequency band information and the average neural oscillation power, constructing an EEG time-frequency feature matrix; The EEG time-frequency feature matrix is ​​normalized to obtain the target time-frequency feature.

7. The method for constructing an image aesthetic EEG evaluation model according to claim 6, characterized in that: The step of obtaining aesthetic fusion EEG features based on the target ERP features and the target time-frequency features includes: Determining a first channel set based on the target ERP feature; Determine a second channel set based on the target time-frequency feature; Aligning the first channel set and the second channel set to obtain a target channel set; Based on the target channel set, extracting the ERP feature vector and the EEG time-frequency feature vector on each channel; Based on the ERP feature vector and the EEG time-frequency feature vector, respectively obtain an ERP feature matrix and an EEG time-frequency feature matrix; Based on the ERP feature matrix and the EEG time-frequency feature matrix, a single-channel fusion matrix is ​​obtained; the single-channel fusion matrix includes the ERP component information and the frequency band information; Combining the single-channel fusion matrix with the preset channel value to obtain a target matrix; The target matrix is ​​subjected to a self-transposition product transformation to obtain the aesthetic fusion EEG feature.

8. The method for constructing an image aesthetic EEG evaluation model according to claim 1, characterized in that: The method of constructing an image aesthetic EEG evaluation model based on the aesthetic fusion EEG features, the preset classification information of the picture aesthetics, and the preset subjective aesthetic evaluation database includes: Converting the aesthetic fusion EEG features into a one-dimensional concatenated vector; Performing data enhancement on the one-dimensional concatenated vector based on a generative adversarial network to generate an EEG feature vector; Rearranging the EEG feature vector and the one-dimensional splicing vector according to the original channel sequence to obtain a fusion feature matrix; Determine the initial model; The initial model is trained based on the fusion feature matrix, and the trained initial model is used as the image aesthetic EEG evaluation model.

9. A method for evaluating image aesthetics using electroencephalogram, characterized in that: include: Invoking a target model in response to an evaluation request; The target model is a model constructed by the method for constructing an image aesthetic EEG evaluation model according to any one of claims 1 to 8; The image to be analyzed is input into the target model to obtain aesthetic evaluation information of the corresponding image.

10. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the image aesthetics EEG evaluation model construction method described in any one of claims 1-8 or the image aesthetics EEG evaluation method described in claim 9.

Images