Eye movement artifact noise reduction method and device based on multilayer decomposition and fractal analysis

Abstract

The invention discloses an eye movement artifact noise reduction method and device based on multilayer decomposition and fractal analysis, and belongs to the technical field of electroencephalogram signal processing.The method comprises the steps that empirical mode decomposition is conducted on original electroencephalogram signals to obtain a plurality of intrinsic mode functions, and the intrinsic mode functions are summed to obtain an eye movement artifact template; performing continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extracting a column vector of the time-frequency representation matrix as a feature vector; clustering the feature vectors based on frequency to obtain a plurality of different clusters, wherein each cluster comprises a plurality of column vectors under the same frequency classification; the fractal dimension of each cluster is calculated, the cluster with the minimum fractal dimension value is selected for time domain signal reduction, eye movement artifact signals are obtained and removed from the original electroencephalogram signals, and the electroencephalogram signals obtained after noise reduction are obtained. Under the condition that main components of the electroencephalogram signals are not affected, eye movement artifacts can be effectively removed, the accuracy and integrity of the electroencephalogram signals are guaranteed, and the method is suitable for various electroencephalogram signal processing and analysis tasks.

Description

Technical Field

[0001] The present invention belongs to the technical field of electroencephalogram signal processing, and in particular relates to a method and device for reducing eye movement artifact noise based on multi-layer decomposition and fractal analysis. Background Art

[0002] During the acquisition and analysis of electroencephalogram (EEG) signals, the interference of eye movement artifacts is one of the main factors affecting the signal quality. Eye movement artifacts are high-frequency noise caused by physiological activities such as blinking and eye movement, which usually appear as instantaneous fluctuations in the signal. The presence of these artifacts not only reduces the validity of EEG signals, but also greatly increases the complexity of subsequent signal processing and analysis, thereby affecting the accurate assessment of brain function status.

[0003] For a long time, researchers have been committed to developing various artifact removal technologies in order to eliminate the influence of eye movement artifacts from the source. Traditional artifact removal methods mostly rely on independent component analysis (ICA) and blind source separation (BSS) techniques. These methods decompose the signal to identify and remove artifact components. However, due to the similarity between artifacts and EEG signal components, these traditional methods may face multiple challenges such as algorithm convergence, model assumptions, and nonlinear relationships between signal components when processing complex signals, resulting in unsatisfactory denoising effects.

[0004] In order to overcome the limitations of traditional methods, some emerging signal processing technologies have been gradually introduced into the research of eye movement artifact removal in recent years. Among them, empirical mode decomposition (EMD), with its unique signal decomposition ability, can decompose complex EEG signals into a series of intrinsic mode functions (IMFs) with different frequency characteristics, providing a powerful tool for extracting potential signal features. The wavelet transform reveals the time-frequency characteristics of the signal at different scales through multi-resolution analysis, making it possible to accurately locate artifacts. In addition, with the development of big data and machine learning technology, methods such as cluster analysis have also been introduced into the detection and evaluation of artifacts, injecting new vitality into the research in this field.

[0005] However, most existing methods are still mainly based on the time domain characteristics of EEG signals for processing. Even if they are converted to the frequency domain for analysis through techniques such as wavelet transform, it is often difficult to fully and accurately capture all the key features of the signal. In this context, most of the eye movement artifact removal methods based on the above technologies still have certain limitations in practical applications. For example, some methods are not resistant to environmental noise and may be affected by external interference and affect detection accuracy; other methods require complex parameter settings and long calculation time, which limits their feasibility in real-time applications.

[0006] Therefore, the denoising effect of existing methods still needs to be further optimized and improved. There is an urgent need to develop a new eye movement artifact removal method that can improve processing efficiency and robustness while ensuring the denoising effect, so as to further provide strong support for the improvement of EEG signal quality and the in-depth understanding of brain functional status. Summary of the invention

[0007] In view of the above, the purpose of the present invention is to provide a method and device for reducing eye movement artifacts based on multi-layer decomposition and fractal analysis, which combines empirical mode decomposition, continuous wavelet transform, cluster analysis and fractal dimension technology to achieve denoising of EEG signals and improve the quality of EEG signals, thereby promoting a deeper understanding of the functional state of the brain.

[0008] In order to achieve the above-mentioned invention object, the technical solution provided by the present invention is as follows: In a first aspect, an embodiment of the present invention provides a method for reducing eye movement artifact noise based on multi-layer decomposition and fractal analysis, comprising the following steps: Performing empirical mode decomposition on the collected and preprocessed original EEG signals to obtain a number of intrinsic mode functions, and obtaining an eye movement artifact template by summing up the intrinsic mode functions; Perform continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extract the column vector of the time-frequency representation matrix as a feature vector; The feature vectors are clustered based on frequency to be divided into multiple different clusters, each cluster contains several column vectors under the same frequency classification; The fractal dimension of each cluster is calculated, and the cluster with the smallest fractal dimension value is selected to restore the time domain signal to obtain the eye movement artifact signal; The restored eye movement artifact signal is removed from the original EEG signal to obtain the denoised EEG signal.

[0009] Preferably, the step of performing empirical mode decomposition on the collected and preprocessed original EEG signals to obtain a number of intrinsic mode functions includes: Performing preprocessing operations on the collected EEG signals, including removing the noise at the beginning and end caused by wearing or removing the device, to obtain the original EEG signals; The original EEG signal is subjected to empirical mode decomposition to obtain several intrinsic mode functions, and the calculation formula is expressed as: , in, represents the original EEG signal, Indicates time, represents the remainder, Represents the decomposition of The intrinsic mode functions, Indicates the maximum number of intrinsic mode functions.

[0010] Preferably, the step of obtaining the eye movement artifact template by summing up the intrinsic mode functions comprises: According to the several intrinsic mode functions obtained by decomposition, the third and all subsequent intrinsic mode functions are selected and summed up to obtain the eye movement artifact template.

[0011] Preferably, performing continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix includes: The eye movement artifact template is convolved with wavelet functions of different scales to generate a time-frequency representation matrix. The calculation formula is expressed as: , in, represents the time-frequency representation matrix, represents the eye movement artifact template to be analyzed, Represents the wavelet function The complex conjugate of Indicates time, represents the scale factor that controls the compression or expansion of the wavelet, Represents the translation factor that controls the position of the wavelet on the time axis.

[0012] Preferably, each column vector in the time-frequency representation matrix represents the time evolution characteristics of the signal at a certain frequency, that is, describes the change of the signal over time at the frequency.

[0013] Preferably, the K-Means clustering algorithm is used to cluster the feature vectors.

[0014] Preferably, the fractal dimension of each cluster is calculated using a box counting method, comprising: For each cluster, a box scale is selected as the starting scale to cover the signal data of the feature vector in the cluster, and the number of boxes containing non-zero values ​​of signal data at each scale is recorded. The scale of the box is gradually reduced during the iteration process until the error is less than the set error factor, then the iteration is stopped and the fractal dimension at this time is calculated. The formula is expressed as: , in, represents the fractal dimension, represents the number of boxes after which the iteration stops. Indicates the size of the box after stopping iteration, Indicates derivative.

[0015] In a second aspect, an embodiment of the present invention further provides an eye movement artifact denoising device based on multi-layer decomposition and fractal analysis, which is implemented using the above-mentioned eye movement artifact denoising method based on multi-layer decomposition and fractal analysis, and includes: an empirical mode decomposition module, a continuous wavelet transform module, a feature vector clustering module, a fractal dimension calculation module and an EEG signal denoising module; The empirical mode decomposition module is used to perform empirical mode decomposition on the collected and preprocessed original EEG signal to obtain a number of intrinsic mode functions, and obtain an eye movement artifact template based on the summation of the intrinsic mode functions; The continuous wavelet transform module is used to perform continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extract the column vector of the time-frequency representation matrix as a feature vector; The feature vector clustering module is used to cluster the feature vectors based on frequency to divide them into multiple different clusters, each cluster containing a number of column vectors under the same frequency classification; The fractal dimension calculation module is used to calculate the fractal dimension of each cluster, and select the cluster with the smallest fractal dimension value to perform time domain signal restoration to obtain the eye movement artifact signal; The EEG signal denoising module is used to remove the restored eye movement artifact signal from the original EEG signal to obtain a denoised EEG signal.

[0016] In a third aspect, an embodiment of the present invention further provides an electronic device comprising a memory and one or more processors, wherein the memory is used to store a computer program, and the processor is used to implement the above-mentioned eye movement artifact denoising method based on multi-layer decomposition and fractal analysis when executing the computer program.

[0017] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a computer, the above-mentioned eye movement artifact denoising method based on multi-layer decomposition and fractal analysis is implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention combines empirical mode decomposition with continuous wavelet transform, which can effectively separate the eye movement artifact template in the original EEG signal through empirical mode decomposition, and convert the time domain signal into a frequency domain signal through continuous wavelet transform to further extract the time-frequency representation, which helps to achieve high-precision denoising processing. Compared with traditional methods such as independent component analysis, the present invention is more accurate in retaining the useful information of the original signal, and reduces the damage to the effective EEG components during the denoising process, and the denoising accuracy is significantly improved.

[0019] (2) Based on empirical mode decomposition and continuous wavelet transform, the present invention further introduces cluster analysis and fractal dimension feature extraction methods to process the time-frequency feature of the feature vector. This can better adapt to signal changes in different scenarios and effectively deal with complex artifact situations, making the identification of eye movement artifacts more intelligent and accurate, thereby enhancing the robustness of the algorithm.

[0020] (3) The overall solution provided by the present invention reduces the reliance on manual adjustment of a large number of parameters and can realize automatic recognition and adaptive processing of different signals. This not only improves the application convenience of the algorithm, but also greatly reduces the operational complexity. By simplifying the operation and parameter settings, the algorithm can be deployed more quickly in various application environments.

[0021] (4) The overall solution provided by the present invention can effectively cope with the interference of external noise, especially in complex experimental environments, and can still maintain a high eye movement artifact removal effect. Compared with some traditional methods that are not effective under environmental noise, the anti-noise performance of the present invention is significantly improved, ensuring the reliability of EEG signal processing.

[0022] (5) The overall solution provided by the present invention can be widely applied to various EEG signal acquisition and analysis scenarios, covering clinical diagnosis, scientific research experiments, cognitive research and other fields, especially in application scenarios requiring high-precision artifact removal. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 It is a flow chart of a method for reducing eye movement artifact noise based on multi-layer decomposition and fractal analysis provided by an embodiment of the present invention; Figure 2 is a schematic diagram of a clean EEG signal generated by an embodiment of the present invention; Figure 3 is a schematic diagram of an eye movement artifact signal generated by an embodiment of the present invention; Figure 4 is a schematic diagram of an electroencephalogram signal contaminated by an eye movement artifact signal constructed according to an embodiment of the present invention; Figure 5 is a schematic diagram comparing the original EEG signals before and after denoising provided by an embodiment of the present invention; Figure 6It is a structural schematic diagram of an eye movement artifact denoising device based on multi-layer decomposition and fractal analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0026] The inventive concept of the present invention is: in view of the problem of insufficient accuracy and robustness of the eye movement artifact removal method in the EEG signal in the prior art, the embodiment of the present invention provides an eye movement artifact denoising method and device based on multi-layer decomposition and fractal analysis. First, the original EEG signal is subjected to empirical mode decomposition to extract the intrinsic mode function, and the high-order intrinsic mode function is regarded as the potential template of the eye movement artifact; then, the continuous wavelet transform is used to obtain the time-frequency representation of the eye movement signal, and the K-Means clustering method is used to classify the signal; then, by calculating the fractal dimension of the signal in each cluster, the signal component with the smallest fractal dimension value is selected as the eye movement artifact signal and removed from the original EEG signal to achieve denoising. The method can effectively remove the eye movement artifact without affecting the main component of the EEG signal, ensure the accuracy and integrity of the EEG signal, and is suitable for various EEG signal processing and analysis tasks.

[0027] Figure 1 FIG. 1 is a flow chart of a method for reducing eye movement artifact noise based on multi-layer decomposition and fractal analysis provided by an embodiment of the present invention. Figure 1 As shown, the embodiment provides an eye movement artifact denoising method based on multi-layer decomposition and fractal analysis, comprising the following steps: S1, performing empirical mode decomposition on the collected and preprocessed original EEG signal to obtain a number of intrinsic mode functions, and obtaining an eye movement artifact template based on the summation of the intrinsic mode functions.

[0028] S1.1, place the EEG acquisition device on the subject's scalp surface for a certain period of time, keep the subject's head relatively still, and collect stable EEG signals.

[0029] S1.2, preprocessing the collected EEG signals, including deleting the noise at the beginning and end caused by wearing or removing the device, to obtain the original EEG signals.

[0030] S1.3, perform empirical mode decomposition (EMD) on the original EEG signal to obtain several intrinsic mode functions (IMFs). EMD can decompose the different frequency components of the EEG signal. It is an adaptive signal decomposition method used to process nonlinear and non-stationary signal analysis. Its calculation formula is: , in, represents the original EEG signal, Indicates time, represents the remainder, Represents the decomposition of The intrinsic mode functions, Indicates the maximum number of intrinsic mode functions.

[0031] Through EMD decomposition, multiple IMF functions will be obtained. In the embodiment, the maximum number of IMFs (such as 50-100) is selected to ensure sufficient detail capture. IMF is the basic component obtained by EMD decomposition, which meets the following two conditions: (1) In the entire data set, the number of zero crossing points and extreme points must be equal or differ by no more than one; (2) At any point in the signal, the average value of the local extreme envelope curve is zero. The biggest advantage of EMD is its adaptability. It does not rely on preset basis functions, but decomposes according to the characteristics of the signal itself. EMD can naturally adapt to the nonlinearity and non-stationarity in the signal, so it performs well when processing complex physiological signals (such as EEG signals, ECG signals, etc.).

[0032] S1.4, according to the decomposed intrinsic mode functions, select the third and all subsequent intrinsic mode functions, calculate the sum of these IMFs, and obtain the eye movement artifact template. At the same time, draw a graph of the eye movement artifact template changing over time to facilitate further analysis and ensure that its morphology conforms to the characteristics of the eye movement artifact.

[0033] S2, performing continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extracting the column vector of the time-frequency representation matrix as a feature vector.

[0034] S2.1, perform continuous wavelet transform (CWT) on the eye movement artifact template, use complex Morlet wavelet, and select the appropriate scale range, convolve the eye movement artifact template with wavelet functions of different scales to generate a two-dimensional time-frequency representation, that is, the time-frequency representation matrix, and the calculation formula is expressed as: , in, represents the time-frequency representation matrix, represents the eye movement artifact template to be analyzed, Represents the wavelet function The complex conjugate of Indicates time, Represents the scale factor that controls the compression or expansion of the wavelet, similar to the change in frequency, Represents the translation factor that controls the position of the wavelet on the time axis.

[0035] Compared with Fourier transform and empirical mode decomposition, wavelet transform has the advantages of multi-resolution analysis, time-frequency localization and flexible selection of wavelet basis functions. It can effectively process non-stationary signals, capture transient characteristics, and maximize the retention of useful signals while removing noise.

[0036] Furthermore, by performing three-dimensional visualization on the time-frequency representation matrix, the three-dimensional coordinates with the horizontal and vertical coordinates of the bottom being time and frequency respectively and the height being signal intensity are obtained, which can be used to further analyze the time-frequency characteristics of the eye movement artifact signal.

[0037] S2.2, extract the column vectors of the time-frequency representation matrix as feature vectors. Each column vector represents the time evolution characteristics of the signal at a specific frequency, that is, it describes the change of the signal over time at this frequency.

[0038] S3, clustering the feature vectors based on frequency to divide them into multiple different clusters, each cluster contains several column vectors under the same frequency classification.

[0039] In the embodiment, the K-Means clustering algorithm is used to cluster the feature vectors, and the clustering results are converted into different signal components to obtain the time domain signal corresponding to each type of signal, that is, each cluster contains several column vectors under the same frequency classification.

[0040] K-Means clustering is used to divide a data set into multiple different clusters, so that the data points in the same cluster have greater similarity, while the data points between different clusters are more different, and the clustering purpose is achieved by minimizing the square distance from each data point in the cluster to the cluster center. Each cluster center (called the centroid) is the average value of all data points in the cluster. In an embodiment, K-Means clustering is used to separate the signal components related to eye movement artifacts from the time-frequency characteristics of the signal, thereby achieving the removal of artifacts.

[0041] The specific steps of the algorithm are as follows: (1) Initialize the centroid: Randomly select K points as the initial centroids. The initial centroids are selected from random points in the data set (i.e., the set of feature vectors).

[0042] (2) Assigning data points: For each data point, calculate its Euclidean distance to each centroid, and then assign the data point to the cluster to which the centroid with the closest distance belongs.

[0043] (3) Update the centroid: According to the distribution of data points, the new centroid of each cluster is calculated, that is, the average value of all data points in the cluster, and the new centroid will replace the original centroid.

[0044] (4) Iteration: Repeat steps (2) and (3) until the position of the centroid no longer changes or the preset maximum number of iterations is reached and the algorithm converges.

[0045] Through K-Means clustering, complex time-frequency features can be divided into multiple different clusters, which helps to effectively separate the artifact components in the subsequent steps and further improve the accuracy of eye movement artifact removal. Compared with the traditional filtering method based on frequency threshold, the present invention uses K-Means clustering to fully utilize the multi-dimensional characteristics of the signal and has higher flexibility and adaptability.

[0046] S4, calculate the fractal dimension for each cluster respectively, select the cluster with the smallest fractal dimension value to perform time domain signal restoration to obtain the eye movement artifact signal.

[0047] In the embodiment, the fractal characteristics of the signal are analyzed by the box counting method, and the fractal dimension of each signal is obtained by linear fitting, and the signal component with the smallest fractal dimension is selected as the estimation of the eye movement artifact. The box counting method is based on the idea of ​​segmenting the signal space, and the purpose is to estimate the fractal dimension of the signal by measuring the number of "boxes" covering the signal at different scales. The basic idea is to divide the time series of the signal into multiple boxes of different scales, and then count the number of boxes occupied by the signal at each scale, and finally calculate the fractal dimension through the relationship between the number of these boxes and the scale.

[0048] For the signal data in each cluster, the specific steps of the algorithm are as follows: (1) Scale initialization: The signal data of each cluster of feature vectors is embedded into an appropriate space, a box size is selected as the starting size, and a reduction factor is set to gradually reduce the box size during the iteration process, and an error factor is set as the judgment condition for stopping the iteration.

[0049] (2) Spatial division: At each selected box scale, the entire range of the signal is divided into a number of boxes of equal size (scale size of length by width), which are used to subsequently cover the data points in the signal.

[0050] (3) Box coverage and counting: For each box at the current scale, count how many boxes are needed to completely cover all data points in the signal. In other words, count the number of boxes that contain non-zero values ​​of the signal. For a one-dimensional signal, this is equivalent to counting the number of intervals spanned by the signal at a given scale.

[0051] (4) Iteration and recording: For different box scales, repeat step (3) of box coverage and counting. In each iteration, record the number of covered boxes at the current scale. During the iteration, set the minimum and maximum values ​​of the box size, and update the box size according to a certain rule (geometric ratio, arithmetic progression, etc.). If the box size is greater than the maximum value, stop the iteration.

[0052] (5) Calculate the logarithmic relationship: After stopping the iteration, the logarithmic relationship between the number of boxes obtained last time and the scale is linearly fitted to obtain the slope of the fitting line, according to the calculation formula of the box counting method: , in, represents the fractal dimension determined by the slope of the fitted straight line, represents the number of boxes after which the iteration stops. Indicates the size of the box after stopping iteration, Indicates derivative.

[0053] Through this process, the algorithm can accurately calculate the fractal dimension of each cluster of signal data, thereby providing an effective measure of signal complexity and irregularity. Then, the cluster with the smallest fractal dimension value is selected to restore the time domain signal to obtain the eye movement artifact signal.

[0054] S5, removing the restored eye movement artifact signal from the original EEG signal to obtain a denoised EEG signal.

[0055] In the embodiment, for the restored eye movement artifact signal, the specific position of the eye movement artifact signal is marked in the original EEG signal, and then the eye movement artifact signal is subtracted from the original EEG signal to obtain a noise-reduced EEG signal.

[0056] In summary, based on the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis provided by an embodiment of the present invention (hereinafter referred to as the denoising method), an experiment to evaluate the denoising performance of an EEG signal was conducted in an embodiment of the present invention. The specific steps are as follows: (1) Generate a reference signal: Figure 2 As shown, firstly, a segment of EEG signal is generated as a clean initial signal. This signal should have typical EEG characteristics without any noise interference.

[0057] (2) Adding random noise: Based on the clean initial signal, a certain proportion of Gaussian noise is randomly added to form a noisy EEG signal. This step is used to simulate the noise environment that may be encountered in the actual acquisition process.

[0058] (3) Introducing eye movement noise: randomly generate eye movement noise, such as Figure 3The eye movement artifact signal shown in the figure is generated, and the generated eye movement noise is added to the EEG signal containing random noise to obtain the EEG signal contaminated by the eye movement noise (such as Figure 4 ). The position and intensity of the added eye movement noise are known to facilitate the subsequent evaluation of the detection ability of the algorithm.

[0059] (4) Running the noise reduction method: The EEG signal containing random noise and eye movement noise is input into the noise reduction method for processing to generate a noise-reduced EEG signal.

[0060] (5) Detection effect evaluation: The detection accuracy of eye movement noise is determined by comparing the position of eye movement noise detected by the noise reduction method with the actual set position.

[0061] (6) Signal information retention evaluation: Further, by comparing the signal correlation before and after noise reduction, the signal information retention level of the noise reduction method is evaluated. Figure 5 The figure shows the comparison of the original EEG signal before and after noise reduction. Figure 5 (a) is a schematic diagram of the noise-free signal (blue) and the noise-reduced signal (orange). Figure 5 (b) in the figure is a schematic diagram of the noisy signal (blue) and the denoised signal (orange). Figure 5 It can be seen from (a) in the figure that the denoised signal largely restores the frequency and amplitude of the original noise-free signal, indicating that the denoising method retains the useful information in the signal to a great extent. Figure 5 As can be seen from (b) in the figure, the denoised signal has basically removed the eye movement noise, the positioning is accurate, the removal is clean, and no missed detection or wrong detection occurs.

[0062] Through this effective experimental design, the performance of the denoising method in processing eye movement artifacts was systematically evaluated to ensure the accuracy of the denoising method provided by the present invention and its superiority in practical applications. Table 1 shows the denoising results of the denoising method under different signals and noises. In the first five rounds, 50-60 eye movement noises were randomly added to the EEG signal, and in the last five rounds, 90-100 eye movement noises were randomly added to the EEG signal. The higher the Pearson correlation coefficient, the higher the similarity between the denoised EEG signal and the clean EEG signal, indicating that effective information can be retained more accurately, and the damage to the effective EEG components in the denoising process is reduced, and the denoising accuracy is significantly improved.

[0063] Table 1 Denoising results of different denoising methods under different signals and noises

[0064] Based on the same inventive concept, Figure 6As shown, an embodiment of the present invention also provides an eye movement artifact denoising device 600 based on multi-layer decomposition and fractal analysis, comprising: an empirical mode decomposition module 610, a continuous wavelet transform module 620, a feature vector clustering module 630, a fractal dimension calculation module 640 and an EEG signal denoising module 650.

[0065] The empirical mode decomposition module 610 is used to perform empirical mode decomposition on the collected and preprocessed original EEG signal to obtain a number of intrinsic mode functions, and obtain an eye movement artifact template based on the summation of the intrinsic mode functions.

[0066] The continuous wavelet transform module 620 is used to perform continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extract the column vector of the time-frequency representation matrix as a feature vector.

[0067] The feature vector clustering module 630 is used to cluster the feature vectors based on frequency to divide them into multiple different clusters, each cluster containing a number of column vectors under the same frequency classification.

[0068] The fractal dimension calculation module 640 is used to calculate the fractal dimension of each cluster, and select the cluster with the smallest fractal dimension value to perform time domain signal restoration to obtain the eye movement artifact signal.

[0069] The EEG signal denoising module 650 is used to remove the restored eye movement artifact signal from the original EEG signal to obtain a denoised EEG signal.

[0070] Based on the same inventive concept, an embodiment of the present invention also provides an electronic device, including a memory and one or more processors, the memory is used to store a computer program, and the processor is used to implement the above-mentioned eye movement artifact denoising method based on multi-layer decomposition and fractal analysis when executing the computer program.

[0071] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a computer, the above-mentioned eye movement artifact denoising method based on multi-layer decomposition and fractal analysis is implemented.

[0072] It should be noted that the eye movement artifact denoising device, electronic device and computer-readable storage medium based on multi-layer decomposition and fractal analysis provided in the above embodiments all belong to the same inventive concept as the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis. The specific implementation process is detailed in the embodiment of the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis, which will not be repeated here.

[0073] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for reducing eye movement artifact noise based on multi-layer decomposition and fractal analysis, characterized in that: The following steps are involved: Performing empirical mode decomposition on the collected and preprocessed original EEG signals to obtain a number of intrinsic mode functions, and obtaining an eye movement artifact template by summing up the intrinsic mode functions; Perform continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extract the column vector of the time-frequency representation matrix as a feature vector; The feature vectors are clustered based on frequency to be divided into multiple different clusters, each cluster contains several column vectors under the same frequency classification; The fractal dimension of each cluster is calculated, and the cluster with the smallest fractal dimension value is selected to restore the time domain signal to obtain the eye movement artifact signal; The restored eye movement artifact signal is removed from the original EEG signal to obtain the denoised EEG signal.

2. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1, characterized in that: The empirical mode decomposition is performed on the collected and preprocessed original EEG signal to obtain several intrinsic mode functions, including: Performing preprocessing operations on the collected EEG signals, including removing the noise at the beginning and end caused by wearing or removing the device, to obtain the original EEG signals; The original EEG signal is subjected to empirical mode decomposition to obtain several intrinsic mode functions, and the calculation formula is expressed as: ， in, represents the original EEG signal, Indicates time, represents the remainder, Represents the decomposition of The intrinsic mode functions, Indicates the maximum number of intrinsic mode functions.

3. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1 or 2, characterized in that: The step of obtaining the eye movement artifact template by summing up the intrinsic mode function includes: According to the several intrinsic mode functions obtained by decomposition, the third and all subsequent intrinsic mode functions are selected and summed up to obtain the eye movement artifact template.

4. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1, characterized in that: The step of performing continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix includes: The eye movement artifact template is convolved with wavelet functions of different scales to generate a time-frequency representation matrix. The calculation formula is expressed as: ， in, represents the time-frequency representation matrix, represents the eye movement artifact template to be analyzed, Represents the wavelet function The complex conjugate of Indicates time, represents the scale factor that controls the compression or expansion of the wavelet, Represents the translation factor that controls the position of the wavelet on the time axis.

5. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1, characterized in that: Each column vector in the time-frequency representation matrix represents the time evolution characteristics of the signal at a specific frequency, that is, it describes the change of the signal over time at this frequency.

6. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1, characterized in that: The K-Means clustering algorithm is used to cluster the feature vectors.

7. The eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to claim 1, characterized in that: The fractal dimension of each cluster was calculated using the box counting method, including: For each cluster, a box scale is selected as the starting scale to cover the signal data of the feature vector in the cluster, and the number of boxes containing non-zero values ​​of signal data at each scale is recorded. The scale of the box is gradually reduced during the iteration process until the error is less than the set error factor, then the iteration is stopped and the fractal dimension at this time is calculated. The formula is expressed as: ， in, represents the fractal dimension, represents the number of boxes after which iteration stops. Indicates the size of the box after stopping iteration, Indicates derivative.

8. An eye movement artifact denoising device based on multi-layer decomposition and fractal analysis, implemented by the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis according to any one of claims 1 to 7, characterized in that: include: Empirical mode decomposition module, continuous wavelet transform module, eigenvector clustering module, fractal dimension calculation module and EEG signal denoising module; The empirical mode decomposition module is used to perform empirical mode decomposition on the collected and preprocessed original EEG signal to obtain a number of intrinsic mode functions, and obtain an eye movement artifact template based on the summation of the intrinsic mode functions; The continuous wavelet transform module is used to perform continuous wavelet transform on the eye movement artifact template to obtain a time-frequency representation matrix, and extract the column vector of the time-frequency representation matrix as a feature vector; The feature vector clustering module is used to cluster the feature vectors based on frequency to divide them into multiple different clusters, each cluster containing a number of column vectors under the same frequency classification; The fractal dimension calculation module is used to calculate the fractal dimension of each cluster, and select the cluster with the smallest fractal dimension value to perform time domain signal restoration to obtain the eye movement artifact signal; The EEG signal denoising module is used to remove the restored eye movement artifact signal from the original EEG signal to obtain a denoised EEG signal.

9. An electronic device comprising a memory and one or more processors, wherein the memory is used to store a computer program, characterized in that: The processor is used to implement the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a computer, the eye movement artifact denoising method based on multi-layer decomposition and fractal analysis as described in any one of claims 1 to 7 is implemented.

Images