Method and device for removing artifacts of electroencephalogram signals

Abstract

The invention provides a method and a device for removing artifacts of electroencephalogram signals. The method comprises the following steps: recording and preprocessing electroencephalogram signals of participants by using brain wave acquisition equipment; marking artifacts by using an artifact identification model; performing quantitative analysis on influence degrees of different artifact types; and artifacts of the electroencephalogram signals are removed through a comprehensive PCA-LLE analysis method. According to the method, the electroencephalogram signal preprocessing and artifact recognition model and the PCA-LLE-based artifact removal method are introduced, so that the artifact signals are accurately classified and effectively removed, meanwhile, the artifact interference is quantitatively analyzed, the integrity and effectiveness of the electroencephalogram signals are reserved to the maximum extent, and the accuracy of the electroencephalogram signals is improved. Therefore, the problems that in the prior art, artifact recognition of the electroencephalogram signals is not accurate, the integrity of the electroencephalogram signals is reduced due to artifact processing, and the complex artifact processing capacity is limited are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram signal processing, and in particular to a method for removing electroencephalogram signal artifacts, an electroencephalogram signal artifact removal device, a computer-readable storage medium, and an electroencephalogram monitoring system. Background Art

[0002] EEG signals are weak nonlinear and non-stationary signals, which are very susceptible to artifacts, especially in the actual acquisition process, where a large number of artifacts are mixed in. In order to effectively apply EEG signals to scientific research, it is necessary to find ways to remove these artifacts.

[0003] In the early days of EEG signal research, researchers usually used experimental control methods to suppress the generation of artifacts. For example, by strictly restricting the behavior of the experimenter, such as avoiding eye movements, blinking, swallowing and other actions as much as possible, the interference of electrooculography and electromyography artifacts can be reduced. Although this method can reduce the impact of artifacts to a certain extent, it imposes an additional psychological burden on the experimental subjects, may activate unnecessary EEG activities, and have an adverse effect on the accuracy of the experimental results.

[0004] In addition, there is a method to remove artifacts by removing EEG signal bands that are highly similar to the artifact waveform characteristics. However, this method usually leads to the loss of some useful signals, further reducing the integrity of EEG signals and the reliability of analysis.

[0005] In the actual acquisition process, due to the interference of other physiological activities of the human body, such as eye blinking, muscle movement, heartbeat, etc., artifact signals will inevitably be generated. Blinking and eye movement artifacts are the most common sources of artifacts. The amplitude of the electrooculogram signal is usually hundreds or even thousands of times that of the EEG signal, which seriously interferes with the EEG signal. The electromyography signal is prone to appear when the subject is nervous, and its high-frequency characteristics make it difficult to filter directly. In addition, non-physiological artifacts such as power frequency interference, acquisition electrodes and amplifier noise also have a significant impact on the accurate recording of EEG signals.

[0006] Although the current methods have been able to partially alleviate the above problems, there are still many shortcomings. For example, although it is simple to reduce artifacts by restricting the behavior of the subjects, it may lead to limitations in the experimental environment and affect the natural state of the EEG signal; although the waveform removal method is effective, it cannot accurately distinguish between artifacts and useful signals, and there is a risk of losing important information. At the same time, the effectiveness of removing composite artifacts and real-time identification of artifact signals still needs to be further improved. These problems significantly limit the reliability and extensiveness of EEG signals in practical applications. Summary of the invention

[0007] The main purpose of the present application is to provide a method for removing artifacts from EEG signals, an apparatus for removing artifacts from EEG signals, a computer-readable storage medium and an EEG monitoring system, so as to at least solve the problems in the prior art of inaccurate artifact recognition of EEG signals, decreased integrity of EEG signals due to artifact processing, and limited complex artifact processing capabilities.

[0008] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for removing artifacts from an EEG signal is provided, comprising: acquiring signals from a target object through an EEG acquisition device to obtain a first target signal, preprocessing the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping and removing non-physiological artifact signals other than the first target signal; inputting the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; quantitatively analyzing the artifact signal to determine a first degree of influence of each artifact signal on the EEG signal; and removing the artifact signal in the second target signal through a PCA-LLE analysis method based on the first degree of influence to obtain a target EEG signal.

[0009] Optionally, the EEG acquisition device includes multiple electrode sheets, and the EEG acquisition device is used to acquire signals from the target object to obtain a first target signal, including: acquiring signals separately through the electrode sheets installed at preset positions to obtain multiple signals and integrating them to obtain the first target signal, the first target signal including EEG signals, EO signals, ECG signals, EMG signals and the artifact signals, the EO signals including blinking frequency and eye movement trajectory, the ECG signals including ECG data and heart rate variability data, the EMG signals including EMG data, EMG activity frequency and EMG activity amplitude, and the artifact signals including environmental noise data, equipment interference data and background variables.

[0010] Optionally, the first target signal is preprocessed to obtain a second target signal, including: performing bandpass filtering based on the first target signal to remove power frequency interference and low-frequency noise to obtain a third target signal; dividing the third target signal through a preset time window in a sliding window manner to obtain multiple fourth target signals; generating a spectrum diagram based on the fourth target signal, and calculating a power spectrum density based on the spectrum diagram; when the power spectrum density is greater than a first threshold, determining the fourth target signal as the artifact signal and screening it out; when the power spectrum density is less than or equal to the first threshold, determining the fourth target signal as the second target signal.

[0011] Optionally, the second target signal is input into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal, including: obtaining a reference signal, the reference signal including a reference electrooculogram signal, a reference electrocardiogram signal and a reference electromyography signal, calculating the recognition strength based on the reference signal to obtain the target recognition strength; calculating the classification score based on the reference signal by a strong classifier to obtain a first classification score, the strong classifier being obtained by combining a KNN classifier and a logistic regression classifier; determining a target discrimination threshold based on the first classification score and a first preset formula: T=T base +β·(C final -C mean ), where T is the target discrimination threshold, T base is the preset classification strength threshold, β is the adjustment coefficient of the first classification score to the target discrimination threshold, C mean is the mean value of the reference signal corresponding to the target discrimination threshold, C final is the target discrimination threshold corresponding to the reference signal; based on the target discrimination threshold, the second target signal is intensity updated to obtain a fifth target signal; based on the fifth target signal and the target discrimination threshold, classification is performed to obtain the artifact signal in the second target signal.

[0012] Optionally, updating the intensity of the second target signal based on the target discrimination threshold to obtain a fifth target signal, and classifying based on the fifth target signal and the target discrimination threshold to obtain the artifact signal in the second target signal, including: determining the influence of the artifact signal on the target EEG signal based on the intensity of the second target signal to obtain a second influence degree; updating the intensity of the second target signal based on the first influence degree and the target discrimination threshold to obtain the fifth target signal: Wherein, STR′ is the strength of the fifth target signal, I is the first influence degree, δ is the adjustment coefficient of the strength of the second target signal and the estimated strength of the artifact signal on the strength of the fifth target signal, STR total is the intensity of the second target signal, STR new is the estimated strength of the artifact signal; when the strength of the fifth target signal is greater than or equal to the second threshold, the fifth target signal is determined to be the electrooculogram signal; when the strength of the fifth target signal is greater than or equal to the third threshold and less than the second threshold, the fifth target signal is determined to be the electrocardiogram signal; when the strength of the fifth target signal is greater than or equal to the fourth threshold and less than the third threshold, the fifth target signal is determined to be the electromyography signal; the electrooculogram signal, the electrocardiogram signal and the electromyography signal are determined to be the artifact signals.

[0013] Optionally, performing quantitative analysis on the artifact signals to determine a first degree of influence of each artifact signal on the EEG signal includes: inputting the artifact signals into an artifact influence model to obtain the first degree of influence: ψ(W i ,No i )=μ1·(W i -μ2·No i ), wherein Q is the first influence degree of the artifact signal, n is the type of the artifact signal, I i is the second influence degree corresponding to the artifact signal in the ith i ,No i ) is the artifact signal W in the ith i and the artifact signal W i Noise component introduced during acquisition i The relationship between them, μ1 and μ2 are preset coefficients.

[0014] Optionally, based on the first influence degree, the artifact signal in the second target signal is removed by a PCA-LLE analysis method to obtain a target EEG signal, including: when the first influence degree is greater than a fifth threshold, removing the artifact signal in the second target signal by a PCA algorithm and an LLE algorithm in turn to obtain the target EEG signal; when the first influence degree is less than or equal to the fifth threshold, removing the artifact signal in the second target signal by an LLE algorithm and a PCA algorithm in turn to obtain the target EEG signal.

[0015] According to another aspect of the present application, a device for removing artifacts of an electroencephalogram signal is provided, and the device includes: a first acquisition unit, used to collect signals from a target object through an electroencephalogram acquisition device to obtain a first target signal, and preprocess the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping and removing non-physiological artifact signals other than the first target signal; a classification unit, used to input the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; a first calculation unit, used to quantitatively analyze the artifact signal to determine a first degree of influence of each artifact signal on the electroencephalogram signal; and a second calculation unit, used to remove the artifact signal in the second target signal through a PCA-LLE analysis method based on the first degree of influence to obtain a target electroencephalogram signal.

[0016] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0017] According to another aspect of the present application, a brain wave monitoring system is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the described methods.

[0018] Applying the technical solution of the present application, firstly, the target object is signal collected by an EEG acquisition device to obtain a first target signal, and the first target signal is preprocessed to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping and removing non-physiological artifact signals other than the first target signal; then the second target signal is input into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; then the artifact signal is quantitatively analyzed to determine the first degree of influence of each artifact signal on the EEG signal; finally, based on the first degree of influence, the artifact signal in the second target signal is removed by a PCA-LLE analysis method to obtain a target EEG signal. The present application achieves accurate classification and effective removal of artifact signals by introducing EEG signal preprocessing, an artifact recognition model and an artifact removal method based on PCA-LLE, and at the same time quantitatively analyzes artifact interference, thereby maximally retaining the integrity and effectiveness of EEG signals, thereby solving the problems of inaccurate artifact recognition of EEG signals, decreased integrity of EEG signals due to artifact processing and limited complex artifact processing capabilities in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A hardware structure block diagram of a mobile terminal for removing artifacts of an electroencephalogram signal provided in an embodiment of the present application is shown;

[0020] Figure 2 A schematic flow chart of a method for removing artifacts from an electroencephalogram signal provided according to an embodiment of the present application is shown;

[0021] Figure 3 A structural block diagram of a device for removing artifacts from an electroencephalogram signal provided according to an embodiment of the present application is shown.

[0022] The above drawings include the following reference numerals:

[0023] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0028] PCA-LLE (Principal Component Analysis-Locally Linear Embedding): PCA-LLE is a comprehensive processing technology that combines linear and nonlinear dimensionality reduction methods. It aims to use the advantages of both PCA and LLE to achieve efficient dimensionality reduction and feature extraction. The PCA part is used to perform preliminary dimensionality reduction on high-dimensional data, quickly remove noise and redundant features in the data, and retain the main linear structure of the data, thereby reducing computational complexity. The LLE part performs nonlinear dimensionality reduction on the data after PCA processing, further captures the local nonlinear relationship of the data, and retains the manifold structure and local geometric characteristics of the data. Through this staged processing method, PCA-LLE can take into account the linear and nonlinear characteristics of the data while reducing the computational cost, and is particularly suitable for the analysis and processing of complex high-dimensional data. This method is often used in scenarios that require precise feature extraction, such as pattern recognition, signal processing, and image analysis.

[0029] KNN (K-Nearest Neighbors): KNN is a simple and efficient classification and regression algorithm. Its basic idea is: for a sample to be classified, the classification result of the sample is determined according to the categories of the K nearest neighbors in the sample space. In classification tasks, voting is usually used, and the result is determined by the majority category among the neighbors; in regression tasks, the average value of the neighbors is taken as the predicted value. KNN does not require a training process, has a low model complexity, and is suitable for solving linear or nonlinear problems, but has low computational efficiency for large-scale data and is greatly affected by data distribution.

[0030] As introduced in the background technology, the prior art has deficiencies in terms of accuracy of artifact recognition, retention of signal integrity and complex artifact processing capabilities, and it is difficult to fully meet the needs of high-precision application of EEG signals. In order to solve the problems in the prior art of inaccurate artifact recognition of EEG signals, artifact processing leading to decreased integrity of EEG signals, and limited complex artifact processing capabilities, the embodiments of the present application provide a method for removing artifacts from EEG signals, an apparatus for removing artifacts from EEG signals, a computer-readable storage medium and an EEG monitoring system.

[0031] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0032] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a method for removing artifacts of an electroencephalogram signal according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0033] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the display method of device information in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific examples of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0034] In this embodiment, a method for removing artifacts from EEG signals running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0035] Figure 2 FIG. 1 is a flow chart of a method for removing artifacts of an EEG signal according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0036] Step S201, collecting signals from a target object through an EEG acquisition device to obtain a first target signal, preprocessing the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping, and removing non-physiological artifact signals other than the first target signal;

[0037] Specifically, the electrodes are placed according to the standard scalp electrode positions to record the participants' EEG signals at rest, while the electrooculogram signals, electrocardiogram signals, and electromyography signals are collected. During the collection process, the non-physiological signals of the equipment are recorded to extract artifact removal features. The preprocessing includes using bandpass filtering to remove power frequency interference and low-frequency noise signals, segmenting and cutting the signals, and removing obvious non-physiological artifact signals.

[0038] Step S202, inputting the second target signal into an artifact recognition model to classify the second target signal to obtain an artifact signal in the second target signal;

[0039] Specifically, the EEG signal obtained after preprocessing and after removing non-physiological artifact signals is input into a trained artifact recognition model. The artifact recognition model is based on deep learning or machine learning algorithms and can automatically classify the types of artifacts in the signal, such as eye movement artifacts, electromyographic artifacts, and electrocardiographic artifacts, and generate artifact annotations.

[0040] Step S203, performing quantitative analysis on the artifact signals to determine the first influence degree of each artifact signal on the EEG signal;

[0041] Specifically, an artifact influence model is constructed to quantitatively analyze the influence of different artifact types and to quantitatively analyze artifact signals, mainly including: statistically analyzing the average amplitude and standard deviation of artifact signals; analyzing the energy distribution of artifact signals in specific frequency bands; and evaluating the trend of artifact signals changing over time. Through these analyses, the interference degree of artifact signals on EEG signals is quantified, and the first influence degree of each artifact signal is obtained.

[0042] Step S204: based on the first influence degree, the artifact signal in the second target signal is removed by using a PCA-LLE analysis method to obtain a target EEG signal.

[0043] Specifically, based on the results of quantitative analysis, the PCA-LLE (principal component analysis combined with local linear embedding) algorithm is used to reduce the dimension of the second target signal and remove artifact-related components in the signal. PCA (principal component analysis) initially extracts the principal components of the signal and separates the artifact components; LLE (local linear embedding) further optimizes signal embedding, retains important features in the EEG signal, and removes artifact signals. Finally, a high-quality target EEG signal is obtained.

[0044] Through this embodiment, the present application provides a method for removing artifacts from EEG signals, using an EEG acquisition device to record the EEG signals of the participants and preprocess them, using an artifact recognition model to mark the artifacts, constructing an artifact influence model to quantitatively analyze the degree of influence of different artifact types, and removing artifacts from EEG signals through a comprehensive PCA-LLE analysis method. The present application aims to solve the technical problem that EEG signal acquisition is susceptible to interference from electrooculogram, electrocardiogram, electromyography artifacts and environmental noise, and by constructing an artifact recognition model and an artifact influence model, quantitative analysis and effective removal of artifact signals are achieved, thereby improving the reliability and analysis accuracy of EEG signal data. The method adopts a comprehensive PCA-LLE analysis method, which not only reduces the data dimension but also retains the local geometric structure of the signal, so as to achieve the purpose of improving the artifact removal effect, improve the quality of EEG signals, and solve the problems of inaccurate artifact recognition of EEG signals, decreased integrity of EEG signals due to artifact processing, and limited ability to process complex artifacts in the prior art.

[0045] As a possible implementation, the brain wave acquisition device includes a plurality of electrodes, and the brain wave acquisition device is used to acquire signals from a target object to obtain a first target signal, including:

[0046] Signals are collected separately by electrodes installed at preset positions to obtain multiple signals and integrate them to obtain a first target signal. The first target signal includes electroencephalogram signals, electrooculogram signals, electrocardiogram signals, electromyogram signals and artifact signals. The electrooculogram signals include blinking frequency and eye movement trajectory, the electrocardiogram signals include electrocardiogram data and heart rate variability data, the electromyogram signals include electromyogram data, electromyogram activity frequency and electromyogram activity amplitude, and the artifact signals include environmental noise data, equipment interference data and background variables.

[0047] Specifically, the EEG acquisition device collects the EEG signals of the target object through multiple electrodes. EEG signals are time series data collected by EEG equipment, which can reflect the energy distribution of different frequency bands. By recording the potential difference between different electrodes, a multi-channel voltage data matrix is ​​formed. These electrodes are installed on the scalp of the target object or other preset positions, and the electrophysiological signals of each area are collected through the electrodes respectively, and finally the first target signal is integrated. Blinking frequency is to record the number of blinks of the participants during the experiment, and the eye movement trajectory is to monitor the angle change of eye movement; electromyography data is to collect the potential changes, frequency and amplitude of myoelectric activity during muscle contraction and relaxation; equipment interference data is to record the equipment status changes or other non-physiological signals that may affect the signal quality during the experiment, and background variables are to monitor the potential impact of variables related to the experimental environment, such as temperature, humidity, etc. on signal quality. Among them, EEG signals are the main focus, while other signals (electrooculogram signals, electrocardiogram signals, electromyogram signals and artifact signals) may affect the accuracy of EEG signals. Therefore, in the subsequent processing process, these signals need to be removed and corrected to obtain cleaner and more reliable EEG signal data.

[0048] As a possible implementation manner, preprocessing the first target signal to obtain the second target signal includes:

[0049] Step S301, performing bandpass filtering based on the first target signal to remove power frequency interference and low-frequency noise to obtain a third target signal;

[0050] Specifically, the acquired EEG signal is filtered with a bandpass filter to remove 50 / 60 Hz power frequency interference and low frequency noise, and the effective frequency band of the EEG signal is retained to obtain a third target signal, which has removed power frequency interference and low frequency noise but still contains possible artifact signals.

[0051] Step S302, dividing the third target signal by a preset time window in a sliding window manner to obtain a plurality of fourth target signals;

[0052] Specifically, the signal is segmented, a 2-second time window is set, and a sliding window method is used. After the sliding window segmentation, different EEG signal segments S are obtained. i , is the fourth target signal, S i Represents the signal data of the i-th time period;

[0053] Step S303, generating a frequency spectrum diagram based on the fourth target signal, and calculating a power spectrum density based on the frequency spectrum diagram;

[0054] Specifically, the segmented signal data is used to make a data spectrum diagram and the segmented power spectrum density is calculated: Among them, PSD band (Si ) represents the i-th signal segment S i The power spectral density calculated in the specified frequency band, [f1, f2] is the physiological signal frequency band, [f3, f4] is the non-physiological artifact signal frequency band, f is the power of the signal at each frequency point, c is the important fixed value of the weighted artifact related frequency band, X noise (f) is the Fourier transform result of the environmental noise in the signal;

[0055] Step S304: when the power spectrum density is greater than the first threshold, the fourth target signal is determined as an artifact signal and is screened out;

[0056] Step S305: when the power spectrum density is less than or equal to the first threshold, determine the fourth target signal as the second target signal.

[0057] Specifically, the first artifact signal screening is performed to screen out obvious non-physiological artifact signals in the spectrum:

[0058] P i =w1·PSD band (S i )+w2·EMG feature (S i )+w3·EOG feature (S i );

[0059] Among them, P i represents the comprehensive artifact judgment standard of the i-th signal segment, which is the power spectrum density. w1, w2, and w3 represent the importance weights of various features, respectively. EMG feature (S i ) represents the EMG activity frequency of the i-th signal segment, which is used to determine whether the current frequency band is interfered by EMG artifacts. feature (S i ) represents the blinking frequency of the ith signal segment, which is used to determine whether the current frequency segment is interfered by electrooculographic artifacts;

[0060] If P th is the first threshold, when P i >P th When S i To filter out non-physiological artifact signals, when P i ≤P th Otherwise it is marked as a valid physiological signal.

[0061] Through the above steps, the first target signal is effectively preprocessed and the artifact signal is filtered out to obtain the second target signal. This method uses technical means such as bandpass filtering, sliding window segmentation, and spectrum analysis to ensure that the final second target signal has higher accuracy and reliability, providing cleaner data input for subsequent EEG signal analysis.

[0062] As a possible implementation manner, inputting the second target signal into an artifact recognition model to classify the second target signal to obtain an artifact signal in the second target signal includes:

[0063] Step S401, obtaining a reference signal, the reference signal including a reference electrooculogram signal, a reference electrocardiogram signal and a reference electromyography signal, and calculating the recognition strength based on the reference signal to obtain the target recognition strength;

[0064] Specifically, reference signals are obtained, including reference electrooculogram signals, reference electrocardiogram signals, and reference electromyography signals. These reference signals are usually physiological signals of the participants under normal conditions, which are used to compare with the target signals. The recognition strength can represent the obviousness of the artifact signal, which is calculated by comparing the difference between the target signal and the reference signal.

[0065] The EEG signal obtained after preprocessing and after removing non-physiological artifact signals is input into the trained artifact recognition model, and the historical data of the participants are continuously input to calculate the recognition strength STR of the artifact:

[0066]

[0067] Among them, STR is the artifact signal strength calculated dynamically, N is the number of samples of participants, and STR 0,i represents the initial estimate of the signal strength of the ith artifact, F i is the specific characteristic parameter of the i-th artifact signal, F mean represents the average characteristic of the historical data of the participants, α is the adjustment factor to control the influence of characteristic deviation, and i is the variable index;

[0068] Step S402, calculating a classification score through a strong classifier based on the reference signal to obtain a first classification score, where the strong classifier is obtained by combining a KNN classifier and a logistic regression classifier;

[0069] Specifically, an integrated learning strategy is adopted to combine the KNN classifier and the logistic regression classifier into a strong classifier, and the output of each classifier is defined as C j , comprehensive classification score C final for:

[0070] Among them, C finalIt represents the final classification result after all the basic classifiers output, M is the number of basic classifiers, C j is the output result of the jth basic classifier's judgment on the current situation, j is the variable index, STR max Indicates the maximum intensity of all artifact signals;

[0071] Step S403, determining a target discrimination threshold based on the first classification score and the first preset formula:

[0072] T=T base +β·(C final -C mean );

[0073] Among them, T is the target discrimination threshold, T base is the preset classification strength threshold, β is the adjustment coefficient of the first classification score to the target discrimination threshold, C mean is the mean value of the target discrimination threshold corresponding to the reference signal, C final is the target discrimination threshold corresponding to the reference signal;

[0074] Specifically, this formula automatically adjusts the target discrimination threshold according to the changes in the classification score, ensuring that the classification process is more flexible and accurate.

[0075] Step S404: update the intensity of the second target signal based on the target discrimination threshold to obtain a fifth target signal, and classify the fifth target signal based on the target discrimination threshold to obtain an artifact signal in the second target signal.

[0076] Specifically, the second target signal is updated in intensity based on the target discrimination threshold to obtain a fifth target signal. By adjusting the signal intensity, the distinction between the artifact and the valid EEG signal is enhanced. According to the updated signal intensity and the target discrimination threshold, the fifth target signal is classified to identify the artifact signal. Through this process, the artifact signal in the second target signal can be distinguished from the valid signal.

[0077] As a possible implementation manner, updating the intensity of the second target signal based on the target discrimination threshold to obtain a fifth target signal, and classifying based on the fifth target signal and the target discrimination threshold to obtain an artifact signal in the second target signal, including:

[0078] Step S501, based on the strength of the second target signal, determining the influence degree of the artifact signal on the target EEG signal to obtain a second influence degree;

[0079] Specifically, the artifact influence I is calculated by comprehensively considering the strength of the artifact signal and its influence on the EEG signal:

[0080] I=ξ·STR total +γ·Jcov ;

[0081] Among them, I is the influence of the artifact signal on the EEG signal, ξ is the adjustment of STR total The weight coefficient in the influence calculation, STR total is the strength of the comprehensive artifact signal, which represents the combined effect of the sum of the artifact signal strengths of all artifacts; γ is the adjustment J cov The weight coefficient of influence, J cov It indicates the degree of coverage of the artifact signal on the EEG signal and characterizes the interference caused by the artifact.

[0082] Step S502: Update the strength of the second target signal based on the first impact degree and the target discrimination threshold to obtain a fifth target signal:

[0083]

[0084] Wherein, STR′ is the strength of the fifth target signal, I is the first influence degree, δ is the adjustment coefficient of the strength of the second target signal and the estimated strength of the artifact signal on the strength of the fifth target signal, STR total is the strength of the second target signal, STR new is the estimated strength of the artifact signal;

[0085] Specifically, the signal strength is updated according to the influence degree and the target discrimination threshold, providing refined strength data for artifact signal classification.

[0086] Step S503, when the intensity of the fifth target signal is greater than or equal to the second threshold, determining that the fifth target signal is an electrooculogram signal;

[0087] Step S504, when the strength of the fifth target signal is greater than or equal to the third threshold and less than the second threshold, determining that the fifth target signal is an electrocardiogram signal;

[0088] Step S505, when the strength of the fifth target signal is greater than or equal to the fourth threshold and less than the third threshold, determining that the fifth target signal is an electromyographic signal;

[0089] Specifically, the updated artifact intensity value is combined with the discrimination threshold set according to the comprehensive classification score to determine the category of the current artifact signal. By setting multiple thresholds, the signal is classified according to different intensity thresholds to determine whether the fifth target signal belongs to an electrooculogram signal, an electrocardiogram signal or an electromyography signal, so as to classify and identify the artifact signal.

[0090] Step S506: determine the electrooculogram signal, the electrocardiogram signal and the electromyography signal as artifact signals.

[0091] Specifically, the electrooculogram (EOG) signal, the electrocardiogram (ECG) signal and the electromyography (EMG) signal are finally identified as artifact signals, and these artifact signals are removed during EEG signal processing.

[0092] This embodiment updates the signal strength in a refined manner by comprehensively considering the strength of the artifact signal and its impact on the EEG signal, and classifies the signal according to different strength thresholds. Through these steps, the artifact signal can be effectively identified and marked, and a clear basis can be provided for subsequent artifact removal processing.

[0093] As a possible implementation method, the artifact signals are quantitatively analyzed to determine the first degree of influence of each artifact signal on the EEG signal, including:

[0094] The artifact signal is input into the artifact influence model to obtain the first influence degree:

[0095]

[0096] ψ(W i ,No i )=μ1·(W i -μ2·No i );

[0097] Wherein, Q is the first influence degree of the artifact signal, n is the type of the artifact signal, and I i is the second influence degree corresponding to the artifact signal in the ith i ,No i ) is the artifact signal W in the ith i and the artifact signal W i Noise component introduced during acquisition i The relationship between them, μ1 and μ2 are preset coefficients.

[0098] Specifically, an artifact influence model is constructed to quantitatively analyze the influence of different artifact types, and the application order of PCA (principal component analysis) and LLE (local linear embedding) algorithms is determined according to the influence degree, so as to remove artifact signals and obtain target EEG signals.

[0099] As a possible implementation method, the application order of the PCA and LLE analysis methods is determined according to the first influence degree of the artifact signal, so as to optimize the artifact removal process. Based on the first influence degree, the artifact signal in the second target signal is removed by the PCA-LLE analysis method to obtain the target EEG signal, including:

[0100] Step S601, when the first influence degree is greater than a fifth threshold, removing artifact signals in the second target signal by using a PCA algorithm and an LLE algorithm in sequence to obtain a target EEG signal;

[0101] Specifically, when the influence of the artifact signal is large, PCA is used to remove the artifact signal in order to effectively reduce the global noise, reduce the data dimension, and clarify the principal components, so that the subsequent LLE can work on a clear structure and reflect the main characteristics of the signal.

[0102] Furthermore, when PCA is selected to process the signal first, a feature-preserving nonlinear transformation is performed:

[0103]

[0104] Among them, D PCA is the output data matrix after PCA processing, D is the input data matrix of multi-channel EEG signal data, E PCA is the PCA principal component matrix, which is used to project the data into the new feature space; η is the adjustment parameter that controls the weight of the original signal and the nonlinear mapping result, ranging from [0,1]; n is the number of signal channels, j is the variable index, and D j Signal data of the jth channel, σ is the activation function;

[0105] LLE dynamic weighting adjustment is used to preserve local geometric structure: h(D PCA,k )=R(D PCA )-ρ(D PCA,k );

[0106] Among them, D LLE is the output data matrix after LLE processing, K represents the number of neighbors used in LLE reconstruction, ω k Represents the importance weight of the kth neighbor in the reconstruction, h(D PCA,k ) is the local reconstruction function, which calculates the adjustment associated with the kth neighbor; R(D PCA ) represents the signal reconstructed from the global feature, ρ(D PCA,k ) is a nonlinear function based on the local characteristics of the PCA output signal;

[0107] The artifact removal output is:

[0108] Among them, D cleaned is the final output signal matrix after artifact removal, π(Q) is the adjustment function of the artifact removal strength with Q, Ω(Q) represents the artifact component that needs to be removed, ε1 is the basic adjustment parameter for setting the minimum artifact removal strength, ε2 is the adjustment parameter for the proportion of the artifact removal strength, and λ is the attenuation parameter for adjusting the influence of Q on the artifact removal strength.

[0109] Step S602: When the first influence degree is less than or equal to the fifth threshold, the artifact signal in the second target signal is removed by the LLE algorithm and the PCA algorithm in sequence to obtain a target EEG signal.

[0110] Specifically, when the influence of the artifact signal is small and the local signal structure is complex, LLE priority processing is selected, which is suitable for processing signals with complex local structures, maintaining the local geometric characteristics of the data, and may better extract effective information from complex artifacts, but may retain global artifacts, and then use PCA to extract the main feature information and remove residual noise. PCA and D LLE The calculated order positions are permuted.

[0111] By quantitatively analyzing the influence of artifact signals on EEG signals, this embodiment determines the order of using PCA and LLE algorithms. PCA is used first when the influence is higher, and LLE is used first when the influence is lower. This flexible analysis method can optimize the artifact removal process according to the influence of different artifact signals, thereby improving the quality of EEG signals.

[0112] The embodiment of the present application also provides an artifact removal device for EEG signals. It should be noted that the artifact removal device for EEG signals in the embodiment of the present application can be used to execute the artifact removal method for EEG signals provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0113] The following is an introduction to the device for removing artifacts of EEG signals provided in an embodiment of the present application.

[0114] Figure 3 is a structural block diagram of an artifact removal device for an electroencephalogram signal according to an embodiment of the present application. Figure 3 As shown, the device includes: a first acquisition unit 10, a classification unit 20, a first calculation unit 30 and a second calculation unit 40.

[0115] The first acquisition unit 10 is used to collect signals from the target object through the EEG acquisition device to obtain a first target signal, and preprocess the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping, and removing non-physiological artifact signals other than the first target signal;

[0116] Specifically, the electrodes are placed according to the standard scalp electrode positions to record the participants' EEG signals at rest, while the electrooculogram signals, electrocardiogram signals, and electromyography signals are collected. During the collection process, the non-physiological signals of the equipment are recorded to extract artifact removal features. The preprocessing includes using bandpass filtering to remove power frequency interference and low-frequency noise signals, segmenting and cutting the signals, and removing obvious non-physiological artifact signals.

[0117] The classification unit 20 is used to input the second target signal into the artifact recognition model to classify the second target signal and obtain the artifact signal in the second target signal;

[0118] Specifically, the EEG signal obtained after preprocessing and after removing non-physiological artifact signals is input into a trained artifact recognition model. The artifact recognition model is based on deep learning or machine learning algorithms and can automatically classify the types of artifacts in the signal, such as eye movement artifacts, electromyographic artifacts, and electrocardiographic artifacts, and generate artifact annotations.

[0119] A first calculation unit 30, used for performing quantitative analysis on the artifact signals to determine a first influence degree of each artifact signal on the EEG signal;

[0120] Specifically, an artifact influence model is constructed to quantitatively analyze the influence of different artifact types and to quantitatively analyze artifact signals, mainly including: statistically analyzing the average amplitude and standard deviation of artifact signals; analyzing the energy distribution of artifact signals in specific frequency bands; and evaluating the trend of artifact signals changing over time. Through these analyses, the interference degree of artifact signals on EEG signals is quantified, and the first influence degree of each artifact signal is obtained.

[0121] The second calculation unit 40 is used to remove the artifact signal in the second target signal by using a PCA-LLE analysis method based on the first influence degree to obtain a target EEG signal.

[0122] Specifically, based on the results of quantitative analysis, the PCA-LLE (principal component analysis combined with local linear embedding) algorithm is used to reduce the dimension of the second target signal and remove artifact-related components in the signal. PCA (principal component analysis) initially extracts the principal components of the signal and separates the artifact components; LLE (local linear embedding) further optimizes signal embedding, retains important features in the EEG signal, and removes artifact signals. Finally, a high-quality target EEG signal is obtained.

[0123] The embodiment of the present application provides an artifact removal device for EEG signals, which includes: a first acquisition unit, a classification unit, a first calculation unit, and a second calculation unit. The first acquisition unit is used to collect signals from a target object through an EEG acquisition device to obtain a first target signal, and preprocess the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping, and removing non-physiological artifact signals in non-first target signals; the classification unit is used to input the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; the first calculation unit is used to quantitatively analyze the artifact signal to determine the first degree of influence of each artifact signal on the EEG signal; the second calculation unit is used to remove the artifact signal in the second target signal based on the first degree of influence by a PCA-LLE analysis method to obtain a target EEG signal. The device uses an EEG acquisition device to record the EEG signals of the participants and preprocess them, uses an artifact recognition model to mark the artifacts, constructs an artifact influence model to quantitatively analyze the degree of influence of different artifact types, and removes artifacts from the EEG signals by a comprehensive PCA-LLE analysis method. The present invention solves the problems in the prior art of inaccurate artifact recognition of EEG signals, reduced integrity of EEG signals due to artifact processing, and limited ability to process complex artifacts.

[0124] As a possible implementation manner, the first acquisition unit includes: a collection module.

[0125] The acquisition module is used to collect signals through electrodes installed at preset positions, obtain multiple signals and integrate them to obtain a first target signal. The first target signal includes electroencephalogram signals, electrooculogram signals, electrocardiogram signals, electromyogram signals and artifact signals. The electrooculogram signals include blinking frequency and eye movement trajectory, the electrocardiogram signals include electrocardiogram data and heart rate variability data, the electromyogram signals include electromyogram data, electromyogram activity frequency and electromyogram activity amplitude, and the artifact signals include environmental noise data, equipment interference data and background variables.

[0126] Specifically, EEG signals are the main focus, while other signals (e.g., electrooculogram, electrocardiogram, electromyography, and artifact signals) may affect the accuracy of EEG signals. Therefore, in the subsequent processing, these signals need to be removed and corrected to obtain cleaner and more reliable EEG signal data.

[0127] As a possible implementation manner, the first acquisition unit further includes: a filtering module, a segmentation module, a generation module, a screening module and a first determination module.

[0128] A filtering module, configured to perform bandpass filtering based on the first target signal to remove power frequency interference and low-frequency noise to obtain a third target signal;

[0129] Specifically, the acquired EEG signal is filtered with a bandpass filter to remove 50 / 60 Hz power frequency interference and low frequency noise, and the effective frequency band of the EEG signal is retained to obtain a third target signal, which has removed power frequency interference and low frequency noise but still contains possible artifact signals.

[0130] A segmentation module, used for segmenting the third target signal through a preset time window in a sliding window manner to obtain a plurality of fourth target signals;

[0131] Specifically, the signal is segmented, a 2-second time window is set, and a sliding window method is used. After the sliding window segmentation, different EEG signal segments S are obtained. i , is the fourth target signal, S i Represents the signal data of the i-th time period;

[0132] A generation module is used to generate a spectrum diagram based on the fourth target signal, and calculate the power spectrum density based on the spectrum diagram; specifically, the segmented signal data is used to make a data spectrum diagram and calculate the segmented power spectrum density: Among them, PSD band (S i ) represents the i-th signal segment S i The power spectral density calculated in the specified frequency band, [f1, f2] is the physiological signal frequency band, [f3, f4] is the non-physiological artifact signal frequency band, f is the power of the signal at each frequency point, c is the important fixed value of the weighted artifact related frequency band, X noise (f) is the Fourier transform result of the environmental noise in the signal;

[0133] A screening module, used for determining the fourth target signal as a pseudo signal and screening it out when the power spectrum density is greater than the first threshold;

[0134] The first determination module is used to determine the fourth target signal as the second target signal when the power spectrum density is less than or equal to the first threshold.

[0135] Specifically, the first artifact signal screening is performed to screen out obvious non-physiological artifact signals in the spectrum:

[0136] P i =w1·PSD band (S i )+w2·EMG feature (S i )+w3·EOG feature (S i );

[0137] Among them, P irepresents the comprehensive artifact judgment standard of the i-th signal segment, which is the power spectrum density. w1, w2, and w3 represent the importance weights of various features, respectively. EMG feature (S i ) represents the EMG activity frequency of the i-th signal segment, which is used to determine whether the current frequency band is interfered by EMG artifacts. feature (S i ) represents the blinking frequency of the ith signal segment, which is used to determine whether the current frequency segment is interfered by electrooculographic artifacts;

[0138] If P th is the first threshold, when P i >P th When S i To filter out non-physiological artifact signals, when P i ≤P th Otherwise it is marked as a valid physiological signal.

[0139] It can be seen that the first target signal is effectively preprocessed and the artifact signal is filtered out to obtain the second target signal. This method uses technical means such as bandpass filtering, sliding window segmentation, and spectrum analysis to ensure that the final second target signal has higher accuracy and reliability, providing cleaner data input for subsequent EEG signal analysis.

[0140] As a possible implementation manner, the classification unit includes: a strength calculation module, a score calculation module, a threshold determination module and an update module.

[0141] The strength calculation module is used to obtain a reference signal, which includes a reference electrooculogram signal, a reference electrocardiogram signal and a reference electromyography signal, and calculate the recognition strength based on the reference signal to obtain the target recognition strength;

[0142] Specifically, reference signals are obtained, including reference electrooculogram signals, reference electrocardiogram signals, and reference electromyography signals. These reference signals are usually physiological signals of the participants under normal conditions, which are used to compare with the target signals. The recognition strength can represent the obviousness of the artifact signal, which is calculated by comparing the difference between the target signal and the reference signal.

[0143] The EEG signal obtained after preprocessing and after removing non-physiological artifact signals is input into the trained artifact recognition model, and the historical data of the participants are continuously input to calculate the recognition strength STR of the artifact:

[0144]

[0145] Among them, STR is the artifact signal strength calculated dynamically, N is the number of samples of participants, and STR 0,i represents the initial estimate of the signal strength of the ith artifact, F iis the specific characteristic parameter of the i-th artifact signal, F mean represents the average characteristic of the historical data of the participants, α is the adjustment factor to control the influence of characteristic deviation, and i is the variable index;

[0146] A score calculation module, used to calculate a classification score through a strong classifier based on the reference signal to obtain a first classification score, where the strong classifier is obtained by combining a KNN classifier and a logistic regression classifier;

[0147] Specifically, an integrated learning strategy is adopted to combine the KNN classifier and the logistic regression classifier into a strong classifier, and the output of each classifier is defined as C j , comprehensive classification score C final for:

[0148]

[0149] Among them, C final It represents the final classification result after all the basic classifiers output, M is the number of basic classifiers, C j is the output result of the jth basic classifier's judgment on the current situation, j is the variable index, STR max Indicates the maximum intensity of all artifact signals;

[0150] A threshold determination module is used to determine a target discrimination threshold based on the first classification score and a first preset formula:

[0151] T=T base +β·(C final -C mean );

[0152] Among them, T is the target discrimination threshold, T base is the preset classification strength threshold, β is the adjustment coefficient of the first classification score to the target discrimination threshold, C mean is the mean value of the target discrimination threshold corresponding to the reference signal, C final is the target discrimination threshold corresponding to the reference signal;

[0153] Specifically, this formula automatically adjusts the target discrimination threshold according to the changes in the classification score, ensuring that the classification process is more flexible and accurate.

[0154] The updating module is used to update the intensity of the second target signal based on the target discrimination threshold to obtain a fifth target signal, and to classify the fifth target signal and the target discrimination threshold to obtain an artifact signal in the second target signal.

[0155] Specifically, the second target signal is updated in intensity based on the target discrimination threshold to obtain a fifth target signal. By adjusting the signal intensity, the distinction between the artifact and the valid EEG signal is enhanced. According to the updated signal intensity and the target discrimination threshold, the fifth target signal is classified to identify the artifact signal. Through this process, the artifact signal in the second target signal can be distinguished from the valid signal.

[0156] As a possible implementation manner, the updating module includes: an influence degree submodule, a strength updating submodule, a second determining submodule, a third determining submodule, a fourth determining submodule and a determining submodule.

[0157] An influence degree submodule, used to determine the influence degree of the artifact signal on the target EEG signal based on the strength of the second target signal, and obtain a second influence degree;

[0158] Specifically, the artifact influence I is calculated by comprehensively considering the strength of the artifact signal and its influence on the EEG signal:

[0159] I=ξ·STR total +γ·J cov ;

[0160] Among them, I is the influence of the artifact signal on the EEG signal, ξ is the adjustment of STR total The weight coefficient in the influence calculation, STR total is the strength of the comprehensive artifact signal, which represents the combined effect of the sum of the artifact signal strengths of all artifacts; γ is the adjustment J cov The weight coefficient of influence, J cov It indicates the degree of coverage of the artifact signal on the EEG signal and characterizes the interference caused by the artifact.

[0161] The intensity updating submodule is used to update the intensity of the second target signal based on the first influence degree and the target discrimination threshold to obtain a fifth target signal:

[0162]

[0163] Wherein, STR′ is the strength of the fifth target signal, I is the first influence degree, δ is the adjustment coefficient of the strength of the second target signal and the estimated strength of the artifact signal on the strength of the fifth target signal, STR total is the strength of the second target signal, STR new is the estimated strength of the artifact signal;

[0164] Specifically, the signal strength is updated according to the influence degree and the target discrimination threshold, providing refined strength data for artifact signal classification.

[0165] A second determination submodule, configured to determine that the fifth target signal is an electrooculogram signal when the intensity of the fifth target signal is greater than or equal to a second threshold;

[0166] a third determination submodule, configured to determine that the fifth target signal is an electrocardiogram signal when the strength of the fifth target signal is greater than or equal to the third threshold and less than the second threshold;

[0167] a fourth determination submodule, configured to determine that the fifth target signal is an electromyographic signal when the intensity of the fifth target signal is greater than or equal to the fourth threshold and less than the third threshold;

[0168] Specifically, the updated artifact intensity value is combined with the discrimination threshold set according to the comprehensive classification score to determine the category of the current artifact signal. By setting multiple thresholds, the signal is classified according to different intensity thresholds to determine whether the fifth target signal belongs to an electrooculogram signal, an electrocardiogram signal or an electromyography signal, so as to classify and identify the artifact signal.

[0169] The determination submodule is used to determine the electrooculogram signal, the electrocardiogram signal and the electromyography signal as artifact signals.

[0170] Specifically, the electrooculogram (EOG) signal, the electrocardiogram (ECG) signal and the electromyography (EMG) signal are finally identified as artifact signals, and these artifact signals are removed during EEG signal processing.

[0171] This embodiment updates the signal strength in a refined manner by comprehensively considering the strength of the artifact signal and its impact on the EEG signal, and classifies the signal according to different strength thresholds. Through these steps, the artifact signal can be effectively identified and marked, and a clear basis can be provided for subsequent artifact removal processing.

[0172] As a possible implementation manner, the first calculation unit includes: an artifact influence module.

[0173] The artifact impact module is used to input the artifact signal into the artifact impact model to obtain a first impact degree:

[0174]

[0175] ψ(W i ,No i )=μ1·(W i -μ2·No i );

[0176] Wherein, Q is the first influence degree of the artifact signal, n is the type of the artifact signal, and I i is the second influence degree corresponding to the artifact signal in the ith i ,No i ) is the artifact signal W in the ith i and the artifact signal W iNoise component introduced during acquisition i The relationship between them, μ1 and μ2 are preset coefficients.

[0177] Specifically, an artifact influence model is constructed to quantitatively analyze the influence of different artifact types, and the application order of PCA (principal component analysis) and LLE (local linear embedding) algorithms is determined according to the influence degree, so as to remove artifact signals and obtain target EEG signals.

[0178] As a possible implementation manner, the second calculation unit includes: a first removal module and a second removal module.

[0179] A first removal module is used for removing the artifact signal in the second target signal by using the PCA algorithm and the LLE algorithm in sequence to obtain a target EEG signal when the first influence degree is greater than a fifth threshold;

[0180] Specifically, when the influence of the artifact signal is large, PCA is used to remove the artifact signal in order to effectively reduce the global noise, reduce the data dimension, and clarify the principal components, so that the subsequent LLE can work on a clear structure and reflect the main characteristics of the signal.

[0181] The second removal module is used to remove the artifact signal in the second target signal by LLE algorithm and PCA algorithm in sequence to obtain the target EEG signal when the first influence degree is less than or equal to the fifth threshold.

[0182] Specifically, when the influence of the artifact signal is small and the local signal structure is complex, LLE is selected for priority processing, which is suitable for processing signals with complex local structures and maintaining the local geometric characteristics of the data. It may better extract effective information from complex artifacts, but may retain global artifacts. PCA is then used to extract the main feature information and remove residual noise.

[0183] The above-mentioned artifact removal device for EEG signals includes a processor and a memory, and the above-mentioned first acquisition unit, classification unit, first calculation unit, second calculation unit, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0184] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the communication efficiency can be improved by adjusting the kernel parameters.

[0185] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0186] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute the method for removing artifacts from electroencephalogram signals.

[0187] An embodiment of the present invention provides an EEG monitoring system, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the above-mentioned EEG signal artifact removal method.

[0188] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0189] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0190] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0191] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0193] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0194] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0195] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0196] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0197] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0198] 1) The artifact removal method of the EEG signal of the present application uses an EEG acquisition device to record the EEG signals of the participants and pre-process them, uses an artifact recognition model to mark the artifacts, constructs an artifact influence model to quantitatively analyze the degree of influence of different artifact types, and removes EEG signal artifacts through a comprehensive PCA-LLE analysis method. The present application aims to solve the technical problem that the EEG signal acquisition process is susceptible to interference from electrooculogram, electrocardiogram, electromyography artifacts and environmental noise. By constructing an artifact recognition model and an artifact influence model, quantitative analysis and effective removal of artifact signals are achieved, and the reliability and analysis accuracy of EEG signal data are improved. The method adopts a comprehensive PCA-LLE analysis method, which not only reduces the data dimension but also retains the local geometric structure of the signal to achieve the purpose of improving the artifact removal effect, improve the quality of EEG signals, and solve the problems of inaccurate artifact recognition of EEG signals, decreased integrity of EEG signals due to artifact processing, and limited ability to process complex artifacts in the prior art.

[0199] 2) The artifact removal device of the EEG signal of the present application includes: a first acquisition unit, a classification unit, a first calculation unit and a second calculation unit. The first acquisition unit is used to collect signals from the target object through an EEG acquisition device to obtain a first target signal, preprocess the first target signal to obtain a second target signal, and the preprocessing at least includes filtering, clipping and removing non-physiological artifact signals in the first target signal; the classification unit is used to input the second target signal into the artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; the first calculation unit is used to quantitatively analyze the artifact signal to determine the first influence degree of each artifact signal on the EEG signal; the second calculation unit is used to remove the artifact signal in the second target signal based on the first influence degree through the PCA-LLE analysis method to obtain the target EEG signal. The device uses an EEG acquisition device to record the EEG signal of the participant and preprocess it, uses an artifact recognition model to mark the artifact, constructs an artifact influence model to quantitatively analyze the influence degree of different artifact types, and removes artifacts from the EEG signal through a comprehensive PCA-LLE analysis method. The present invention solves the problems in the prior art of inaccurate artifact recognition of EEG signals, reduced integrity of EEG signals due to artifact processing, and limited ability to process complex artifacts.

[0200] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for removing artifacts from an electroencephalogram signal, characterized in that: include: Acquiring signals from a target object through an electroencephalogram acquisition device to obtain a first target signal, and preprocessing the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping, and removing non-physiological artifact signals other than the first target signal; Inputting the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; Performing quantitative analysis on the artifact signals to determine a first degree of influence of each artifact signal on the EEG signal; Based on the first influence degree, the artifact signal in the second target signal is removed by a PCA-LLE analysis method to obtain a target EEG signal.

2. The method according to claim 1, characterized in that The brain wave acquisition device includes a plurality of electrode sheets. The first target signal is obtained by collecting signals from the target object through the brain wave collection device, including: Signals are collected separately by the electrode sheets installed at preset positions to obtain multiple signals and integrate them to obtain the first target signal. The first target signal includes electroencephalogram signals, electrooculogram signals, electrocardiogram signals, electromyogram signals and the artifact signals. The electrooculogram signals include blinking frequency and eye movement trajectory, the electrocardiogram signals include electrocardiogram data and heart rate variability data, the electromyogram signals include electromyogram data, electromyogram activity frequency and electromyogram activity amplitude, and the artifact signals include environmental noise data, equipment interference data and background variables.

3. The method according to claim 2, characterized in that Preprocessing the first target signal to obtain a second target signal includes: Performing bandpass filtering on the first target signal to remove power frequency interference and low-frequency noise to obtain a third target signal; The third target signal is divided by a preset time window in a sliding window manner to obtain a plurality of fourth target signals; generating a frequency spectrogram based on the fourth target signal, and calculating a power spectral density based on the frequency spectrogram; When the power spectrum density is greater than a first threshold, determining the fourth target signal as the artifact signal and filtering it out; When the power spectral density is less than or equal to the first threshold, the fourth target signal is determined as the second target signal.

4. The method according to claim 2, characterized in that: Inputting the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal, comprising: Acquire a reference signal, wherein the reference signal includes a reference electrooculogram signal, a reference electrocardiogram signal, and a reference electromyography signal, and calculate the recognition strength based on the reference signal to obtain the target recognition strength; Calculating a classification score based on the reference signal by a strong classifier to obtain a first classification score, wherein the strong classifier is obtained by combining a KNN classifier and a logistic regression classifier; Based on the first classification score and based on the first preset formula, a target discrimination threshold is determined: T=T base +β·(C final -C mean ); Wherein, T is the target discrimination threshold, T base is the preset classification strength threshold, β is the adjustment coefficient of the first classification score to the target discrimination threshold, C mean is the mean value of the reference signal corresponding to the target discrimination threshold, C final is the target discrimination threshold corresponding to the reference signal; The second target signal is updated in intensity based on the target discrimination threshold to obtain a fifth target signal, and the artifact signal in the second target signal is obtained by performing classification based on the fifth target signal and the target discrimination threshold.

5. The method according to claim 4, characterized in that The method updates the intensity of the second target signal based on the target discrimination threshold to obtain a fifth target signal, and classifies the fifth target signal and the target discrimination threshold to obtain the artifact signal in the second target signal, including: Determining the influence degree of the artifact signal on the target EEG signal based on the strength of the second target signal to obtain a second influence degree; The second target signal is updated in intensity based on the first influence degree and the target discrimination threshold to obtain the fifth target signal: Wherein, STR′ is the strength of the fifth target signal, I is the first influence degree, δ is the adjustment coefficient of the strength of the second target signal and the estimated strength of the artifact signal on the strength of the fifth target signal, and STR total is the intensity of the second target signal, STR new is the estimated strength of the artifact signal; When the intensity of the fifth target signal is greater than or equal to the second threshold, determining that the fifth target signal is the electrooculogram signal; When the intensity of the fifth target signal is greater than or equal to the third threshold and less than the second threshold, determining that the fifth target signal is an electrocardiogram signal; When the intensity of the fifth target signal is greater than or equal to the fourth threshold and less than the third threshold, determining that the fifth target signal is an electromyographic signal; The electrooculogram signal, the electrocardiogram signal and the electromyography signal are determined as the artifact signals.

6. The method according to claim 5, characterized in that Quantitatively analyzing the artifact signals to determine a first degree of influence of each artifact signal on the EEG signal includes: The artifact signal is input into the artifact influence model to obtain the first influence degree: ψ(W i ,No i )=μ1·(W i -μ2·No i ); Wherein, Q is the first influence degree of the artifact signal, n is the type of the artifact signal, I i is the second influence degree corresponding to the artifact signal in the ith i ,No i ) is the artifact signal W in the ith i and the artifact signal W i Noise component introduced during acquisition i The relationship between them, μ1 and μ2 are preset coefficients.

7. The method according to claim 6, characterized in that Removing the artifact signal in the second target signal by a PCA-LLE analysis method based on the first influence degree to obtain a target EEG signal includes: When the first influence degree is greater than a fifth threshold, the artifact signal in the second target signal is removed by a PCA algorithm and an LLE algorithm in sequence to obtain the target EEG signal; When the first influence degree is less than or equal to the fifth threshold, the artifact signal in the second target signal is removed by LLE algorithm and PCA algorithm in sequence to obtain the target EEG signal.

8. An artifact removal device for electroencephalogram signals, characterized in that: The device comprises: A first acquisition unit is used to collect signals from a target object through an electroencephalogram acquisition device to obtain a first target signal, and preprocess the first target signal to obtain a second target signal, wherein the preprocessing at least includes filtering, clipping, and removing non-physiological artifact signals other than the first target signal; a classification unit, configured to input the second target signal into an artifact recognition model to classify the second target signal to obtain the artifact signal in the second target signal; A first calculation unit, used for performing quantitative analysis on the artifact signals to determine a first influence degree of each artifact signal on the EEG signal; The second calculation unit is used to remove the artifact signal in the second target signal through a PCA-LLE analysis method based on the first influence degree to obtain a target EEG signal.

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

10. A brain wave monitoring system, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of claims 1 to 7.

Images