Human factors intelligent EEG signal processing method, device, edge computing equipment and medium

Abstract

The present application discloses a human-factor intelligent EEG signal processing method, device, edge computing equipment and medium, which belongs to the field of signal processing technology. The method includes: obtaining EEG signals of different objects to construct a training set; using the training set to train a preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; inputting the target EEG signal of the target object into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application realizes the automated processing of EEG signals, can efficiently denoise EEG signals with low latency, and greatly improves the accuracy and efficiency of EEG signal processing.

Description

Technical Field

[0001] The present application belongs to the field of signal processing technology, and in particular, relates to a human factor intelligent EEG signal processing method, device, edge computing equipment and medium. Background Art

[0002] Electroencephalography (EEG) is an important tool for studying brain activity and cognitive processes. However, EEG signals are often interfered by physiological processes such as eye movements, electrocardiogram, and electromyography, which affects the accuracy and reliability of EEG signals.

[0003] Among the related technologies, the technology for denoising EEG signals mainly relies on offline processing. Specifically, after the EEG signal collection is completed, researchers need to carefully check and analyze the collected EEG signals to identify interference signals caused by physiological factors such as eye movements, electrocardiograms, and electromyography. Although this offline processing method can improve the signal-to-noise ratio of EEG signals, it often requires a lot of manual intervention and is inefficient. Summary of the invention

[0004] The present application proposes a human factors intelligent EEG signal processing method, device, edge computing device and medium.

[0005] In a first aspect, the present application provides a human factor intelligent EEG signal processing method, comprising:

[0006] Obtain EEG signals from different subjects to construct a training set;

[0007] The training set is used to train a preset wavelet filter to update target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy;

[0008] The target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

[0009] According to the human factors intelligent EEG signal processing method of the present application, a training set is constructed by acquiring EEG signals of different objects; the training set is used to train a preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; the target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application constructs a training set by acquiring EEG signals of different objects, and uses these data to train a preset wavelet filter to update parameters such as wavelet basis selection, decomposition layer number and threshold selection strategy of the wavelet filter, so that the trained wavelet filter can more accurately identify and remove noise signals in EEG signals, realize the automated processing of EEG signals, reduce the need for manual intervention, and can efficiently reduce the noise of EEG signals under low latency, greatly improving the accuracy and efficiency of EEG signal processing.

[0010] According to one embodiment of the present application, acquiring EEG signals of different subjects to construct a training set includes:

[0011] Obtaining EEG signals from different subjects;

[0012] Removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain reference EEG signals after the EOG signals are removed;

[0013] The training set is constructed by using the EEG signals of different subjects as training data and the reference EEG signals corresponding to the EEG signals of different subjects as labels of the EEG signals of different subjects.

[0014] In this embodiment, by collecting EEG signals of different subjects and constructing a training set, the basic data for training is diverse and extensive, and the original EEG signals are paired with reference EEG signals after removing electrooculogram signals as training data and labels, which not only improves the efficiency of training, but also enhances the ability of the wavelet filter to identify and remove electrooculogram artifacts through supervised learning.

[0015] According to an embodiment of the present application, removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different objects to obtain the reference EEG signals after the EOG signals are removed includes:

[0016] Decomposing the EEG signal into multiple statistically independent signal components by blind source separation;

[0017] Inputting a plurality of the signal components into a preset neural network model to obtain the category of each signal component output by the neural network model;

[0018] The signal components classified as valid EEG signals are combined to obtain a reference EEG signal after removing the electrooculogram signal.

[0019] In this embodiment, blind source separation is used to decompose the collected EEG signal into multiple statistically independent signal components, so that the noise signal can be separated from the signal by statistical inference. The decomposed signal components are input into a preset neural network model, and the signal is classified by a deep learning algorithm to distinguish between valid EEG signals and noise signals. This makes full use of blind source separation and machine learning algorithms to improve the accuracy of removing eye contact signals from EEG signals.

[0020] According to one embodiment of the present application, decomposing the EEG signal into multiple statistically independent signal components by blind source separation includes:

[0021] Preprocessing the EEG signal, wherein the preprocessing includes at least one of denoising, filtering, centering, and whitening;

[0022] Independent component analysis is performed on the preprocessed EEG signals to obtain multiple statistically independent signal components.

[0023] In this embodiment, the EEG signal is preprocessed by blind source separation, including denoising, filtering, centering, whitening and other steps, which can improve the signal-to-noise ratio of the EEG signal. The preprocessed EEG signal is subjected to independent component analysis to obtain multiple statistically independent signal components, which can effectively separate the mixed EEG signal into neural activity signals from different sources. These signals include brain activity signals and various noise sources. Further analysis of these signals can identify effective EEG signals and noise, thereby improving the EEG signal noise reduction effect.

[0024] According to one embodiment of the present application, decomposing the EEG signal into multiple statistically independent signal components by blind source separation includes:

[0025] Determining the number of channels of the EEG signal;

[0026] The number of the statistically independent signal components is determined according to the number of channels; wherein the number of the signal components is less than or equal to the number of channels.

[0027] In this embodiment, by determining the number of channels of the EEG signal, a reference can be provided for the blind source separation algorithm, thereby reducing the overfitting problem and enabling the algorithm to restore independent source signals as accurately as possible under limited information.

[0028] According to an embodiment of the present application, the step of using the training set to train a preset wavelet filter to update target parameters of the wavelet filter to obtain a trained wavelet filter includes:

[0029] Repeat the following steps until the loss function converges or reaches the preset stop condition to obtain the trained wavelet filter:

[0030] Initializing target parameters of the wavelet filter, inputting the training set into the wavelet filter, and obtaining an electroencephalogram signal filtered by the wavelet filter output;

[0031] The loss function is used to calculate the difference between the filtered EEG signal and the corresponding reference EEG signal;

[0032] Based on the difference, an optimization algorithm is used to adjust the target parameters of the wavelet filter.

[0033] In this embodiment, the target parameters of the wavelet filter are initialized and the training set is input into the wavelet filter to obtain preliminary filtering results. The loss function is used to measure the difference between the filtered EEG signal and the reference EEG signal. Based on these differences, an optimization algorithm is used to adjust the target parameters of the wavelet filter, such as wavelet basis selection, number of decomposition layers and threshold selection strategy, in order to reduce the value of the loss function. The training process is iterated until the loss function converges or reaches the preset stopping condition, so that the wavelet filter can achieve the expected performance and improve the accuracy of automated processing of EEG signals.

[0034] According to an embodiment of the present application, the adopting the training set to train a preset wavelet filter to update a target parameter of the wavelet filter includes:

[0035] According to the characteristics of the EEG signals in the training set, at least one of the support length, vanishing moment, symmetry, regularity and similarity of the wavelet basis is used as a reference standard to select multiple wavelet bases;

[0036] During the training process, a wavelet basis is selected from the plurality of wavelet bases to determine the wavelet basis selection of the trained wavelet filter.

[0037] In this embodiment, multiple wavelet bases are selected based on the characteristics of the EEG signals in the training set by comprehensively considering multiple dimensions of the wavelet basis, such as support length, vanishing moment, symmetry, regularity and similarity. This method selects basis functions suitable for the characteristics of the EEG signals based on multiple attributes of the wavelet basis, thereby improving the denoising effect of the EEG signals.

[0038] In a second aspect, the present application provides a human factors intelligent EEG signal processing device, comprising:

[0039] An acquisition module, used to acquire EEG signals of different subjects to construct a training set;

[0040] A training module, used to train a preset wavelet filter using the training set to update target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy;

[0041] The input module is used to input the target EEG signal of the target object into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

[0042] According to the human factors intelligent EEG signal processing device of the present application, a training set is constructed by acquiring EEG signals of different objects; the training set is used to train a preset wavelet filter to update the target parameters of the wavelet filter and obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; the target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application constructs a training set by acquiring EEG signals of different objects, and uses these data to train a preset wavelet filter to update parameters such as wavelet basis selection, decomposition layer number and threshold selection strategy of the wavelet filter, so that the trained wavelet filter can more accurately identify and remove noise signals in EEG signals, realize the automated processing of EEG signals, reduce the need for manual intervention, and can efficiently reduce the noise of EEG signals under low latency, greatly improving the accuracy and efficiency of EEG signal processing.

[0043] In a third aspect, the present application provides an edge computing device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the human factors intelligent EEG signal processing method as described in the first aspect above is implemented.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the human factors intelligent EEG signal processing method as described in the first aspect above is implemented.

[0045] In a fifth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the human factors intelligent EEG signal processing method as described in the first aspect above.

[0046] In a sixth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the human factors intelligent EEG signal processing method as described in the first aspect above.

[0047] The above one or more technical solutions in the embodiments of the present application have at least one of the following technical effects:

[0048] According to the human factors intelligent EEG signal processing method of the present application, a training set is constructed by acquiring EEG signals of different objects; the training set is used to train a preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; the target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application constructs a training set by acquiring EEG signals of different objects, and uses these data to train a preset wavelet filter to update parameters such as wavelet basis selection, decomposition layer number and threshold selection strategy of the wavelet filter, so that the trained wavelet filter can more accurately identify and remove noise signals in EEG signals, realize the automated processing of EEG signals, reduce the need for manual intervention, and can efficiently reduce the noise of EEG signals under low latency, greatly improving the accuracy and efficiency of EEG signal processing.

[0049] Furthermore, in some embodiments, by collecting EEG signals of different subjects and constructing a training set, the basic data for training is diverse and extensive, and the original EEG signals are paired with reference EEG signals after removing electrooculogram signals as training data and labels, which not only improves the efficiency of training, but also enhances the ability of the wavelet filter to identify and remove electrooculogram artifacts through supervised learning.

[0050] Furthermore, in some embodiments, blind source separation is used to decompose the collected EEG signals into multiple statistically independent signal components, thereby separating the noise signal based on statistical inference. The decomposed signal components are input into a preset neural network model, and a deep learning algorithm is used to classify the signals to distinguish between valid EEG signals and noise signals. This fully utilizes blind source separation and machine learning algorithms to improve the accuracy of removing eye contact signals from EEG signals.

[0051] Furthermore, in some embodiments, the EEG signals are preprocessed through blind source separation, including denoising, filtering, centering, whitening and other steps, which can improve the signal-to-noise ratio of the EEG signals. Independent component analysis is performed on the preprocessed EEG signals to obtain multiple statistically independent signal components, which can effectively separate the mixed EEG signals into neural activity signals from different sources. These signals include brain activity signals and various noise sources. Further analysis of these signals can identify effective EEG signals and noise, thereby improving the EEG signal noise reduction effect.

[0052] Furthermore, in some embodiments, by determining the number of channels of the EEG signal, a reference can be provided for the blind source separation algorithm, thereby reducing the overfitting problem and enabling the algorithm to restore independent source signals as accurately as possible with limited information.

[0053] Furthermore, in some embodiments, the target parameters of the wavelet filter are initialized and the training set is input into the wavelet filter to obtain preliminary filtering results. The loss function is used to measure the difference between the filtered EEG signal and the reference EEG signal. Based on these differences, an optimization algorithm is used to adjust the target parameters of the wavelet filter, such as wavelet basis selection, number of decomposition layers, and threshold selection strategy, in order to reduce the value of the loss function. The training process is iterated until the loss function converges or reaches the preset stopping condition, so that the wavelet filter can achieve the expected performance and improve the accuracy of automated processing of EEG signals.

[0054] Furthermore, in some embodiments, multiple wavelet bases are selected based on the characteristics of the EEG signals in the training set by comprehensively considering multiple dimensions of the wavelet basis, such as support length, vanishing moment, symmetry, regularity and similarity. This method selects basis functions suitable for the characteristics of the EEG signals based on multiple attributes of the wavelet basis, thereby improving the denoising effect of the EEG signals.

[0055] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a flowchart of a human factor intelligent EEG signal processing method provided in an embodiment of the present application;

[0058] Figure 2 is a schematic diagram of signal components provided by an embodiment of the present application;

[0059] Figure 3 is a schematic diagram of denoising an EEG signal through a neural network model provided in an embodiment of the present application;

[0060] Figure 4 It is a structural schematic diagram of a human factor intelligent EEG signal processing device provided in an embodiment of the present application;

[0061] Figure 5 It is a structural diagram of the edge computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0063] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0064] EEG signals refer to electrophysiological signals recorded from the cerebral cortex, which reflect the electrical activity of neurons in the cerebral cortex. When a large number of cortical neurons emit potentials at the same time, detectable electric field signals are generated. By attaching electrodes to the scalp, these weak electric field signals, namely EEG signals, can be recorded. EEG signals provide real-time data on the functional state of the brain. By analyzing EEG signals, we can understand information such as human cognitive processes and emotional states, providing valuable data for a deep understanding of brain functions. In addition, EEG signals can also be used to realize non-invasive brain-computer interfaces, allowing humans to directly control external devices through electrical signals from the brain, which has broad application prospects in medical rehabilitation, entertainment interaction, neuroscience research and other fields.

[0065] In the process of collecting EEG signals, the signals are often interfered by various noises. These noise sources include electrical signals generated by eye movements, electrocardiographic interference caused by heartbeats, myoelectric noise generated by muscle activity, and environmental noise such as power line interference. The presence of these noises not only reduces the quality of EEG signals, but may also mask or distort the real signals of brain activity, bringing challenges to data analysis and interpretation.

[0066] In order to improve the accuracy and availability of EEG signals, related technologies mainly use offline methods for denoising. In this process, researchers need to conduct detailed inspection and analysis of the original EEG signals to identify and mark interference signals caused by physiological factors such as eye movements, electrocardiograms, and electromyography. Manual inspection is a time-consuming process that requires a high degree of expertise. Researchers must carefully distinguish which signals are generated by brain activity and which are interference caused by noise. Although this offline processing method can improve the signal-to-noise ratio of EEG signals, it often requires a lot of manpower and is inefficient. Therefore, how to effectively remove noise from EEG signals, especially without sacrificing real-time performance, is an important challenge facing the current field of EEG signal processing.

[0067] It should be noted that in the optional embodiments of the present application, the object information and other related data involved, when the embodiments in the present application are applied to specific products or technologies, need to obtain the permission or consent of the object, and the collection, use and processing of the relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions. In other words, if the embodiments of the present application involve data related to the object, it needs to be obtained with the authorization and consent of the object, the authorization and consent of the relevant departments, and in compliance with the relevant laws, regulations and standards of the country and region. If personal information is involved in the embodiments, the acquisition of all personal information needs to obtain the consent of the individual. If sensitive information is involved, the separate consent of the information subject needs to be obtained. The embodiments also need to be implemented with the authorization and consent of the object.

[0068] In the following, in combination with the accompanying drawings, the human factors intelligent EEG signal processing method, device, edge computing device and medium provided in the embodiments of the present application are described in detail through specific embodiments and their application scenarios.

[0069] Among them, the human factors intelligent EEG signal processing method can be applied to a terminal, and can be specifically executed by hardware or software in the terminal.

[0070] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or tablet computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad). It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad).

[0071] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse and a joystick.

[0072] The execution subject of the human-factor intelligent EEG signal processing method provided in the embodiment of the present application may be an edge computing device or a functional module or functional entity in the edge computing device that can implement the human-factor intelligent EEG signal processing method. The edge computing device mentioned in the embodiment of the present application may be a portable device (such as a mobile phone, a tablet computer), a wearable device (such as a smart watch, a smart bracelet), an eye tracker, AR (Augmented Reality) / VR (Virtual Reality) / XR (Extended Reality) / MR (Mixed Reality) device, an edge server, an edge gateway, a vehicle-mounted device, a cloud computing device, etc. The human-factor intelligent EEG signal processing method provided in the embodiment of the present application is described below using the edge computing device as an example of the execution subject.

[0073] like Figure 1 As shown, the human factor intelligent EEG signal processing method includes: step 110, step 120 and step 130.

[0074] Step 110: Obtain EEG signals of different subjects to construct a training set.

[0075] In the embodiments of the present application, the object may be different objects according to different application scenarios. For example, in a simulated driving scenario, the object may be a driver, in a medical and health scenario, the object may be a patient, and in a test scenario, the object may be a person being tested.

[0076] In one example, in a simulated driving scenario, the driver's behavior can be effectively evaluated and monitored by analyzing the driver's EEG signals and eye movement signals. By using an eye tracker and EEG acquisition equipment, these two signals can be collected at the same time. Eye movement signals provide intuitive visual attention information, while EEG signals reveal the brain's deep cognitive processing. This data fusion enhances the richness of information and improves the accuracy of driver behavior analysis. Eye movement signals can be collected by an eye tracker to reflect the driver's visual attention, scanning patterns and other characteristics, so as to judge the driver's observation and reaction to road conditions. Eye movement signals help understand the driver's concentration, reaction speed and other key driving behaviors.

[0077] The collection of EEG signals can monitor the driver's fatigue, distraction or emotional state in real time and predict potential driving risks. However, the driver's EEG signals contain a large number of interference signals from physiological processes such as eye movement, electrocardiogram, and electromyography. Therefore, it is necessary to professionally filter and process the EEG signals to extract effective signals that reflect the cognitive state. After analyzing and processing the EEG signals, the driver's attention allocation, emotional changes, decision-making process and other information can be identified, thereby evaluating his driving status.

[0078] In the embodiments of the present application, EEG signals are electrical signals generated by brain neuron activity. EEG signals reflect the superposition of postsynaptic potentials generated by a large number of neurons and are a mapping of brain activity. Neurons in the brain are connected to each other through synapses to form a complex neural network. When neurons are activated, bioelectric phenomena are generated. These electrical signals can be captured by placing electrodes on the scalp or implanted directly into the brain.

[0079] In some embodiments, an EEG acquisition device may be used to acquire EEG signals of different subjects. Specifically, electrodes may be placed on the scalps of these subjects to acquire EEG signals of these subjects. The EEG acquisition device may be a head-mounted device, and after the subject wears the head-mounted device correctly, the EEG signals of the subject are acquired through the electrodes in the head-mounted device.

[0080] In an embodiment of the present application, a large number of EEG signals of different objects can be collected as a training set, and each EEG signal in the training set can correspond to a label. For example, the noise signal or artifact caused by eye movement, electrocardiogram, electromyography, etc. in each EEG signal in the training set can be removed manually or by other means to obtain a reference EEG signal, and the reference EEG signal corresponding to each EEG signal is used as a label. By setting a label for each EEG signal in the training set, it is convenient to compare the reference EEG signal with the EEG signal output by the wavelet filter during the training process to determine the training effect.

[0081] Step 120: Use the training set to train the preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy.

[0082] Wavelet filter is a signal processing tool based on wavelet transform. Wavelet filter analyzes the signal by decomposing it into wavelet components of different scales (frequencies) and positions, thereby removing the noise from the signal. Compared with the traditional Fourier transform, wavelet transform can provide both time and frequency information of the signal, which makes it have great advantages in processing non-stationary signals, such as EEG signals.

[0083] In the embodiment of the present application, the target parameters for updating the training wavelet filter may include at least one of wavelet basis selection, number of decomposition layers, and threshold selection strategy.

[0084] Wavelet basis is the basic function used to decompose the signal in wavelet transform. In wavelet transform, the signal is represented as a superposition of a series of wavelet functions, which are obtained by translation and scaling of a mother wavelet. In the embodiment of the present application, the update target of the wavelet basis is to determine a suitable wavelet basis from a plurality of different wavelet bases. Among them, the plurality of different wavelet bases may include Daubechies, Symlets, Coiflets, etc. The embodiment of the present application does not limit the wavelet basis used in the training process.

[0085] The number of decomposition layers refers to the number of different scales (or frequency bands) into which the signal is decomposed in the wavelet transform. The higher the number of decomposition layers, the more obvious the different characteristics of noise and signal are, and the more conducive to the separation of noise and signal. On the other hand, the larger the number of decomposition layers, the greater the distortion of the reconstructed signal, which affects the final denoising effect to a certain extent. For EEG signals, the number of decomposition layers can be set to 3-5 layers to cover the important frequency bands in the EEG signals. Specifically, the appropriate number of decomposition layers can be determined according to the training process.

[0086] In wavelet denoising, the threshold is a critical value used to determine which wavelet coefficients (representing frequency components of different scales in the signal) are noise and should be removed. It can be used to distinguish between signals and noise, thereby achieving the purpose of denoising. Common threshold selection strategies can include fixed threshold estimation, extreme value threshold estimation, unbiased likelihood estimation, and heuristic estimation. These threshold selection strategies can estimate the threshold directly from the signal. Among them, the extreme value threshold estimation and unbiased likelihood estimation methods are relatively conservative. When the noise is less distributed in the high frequency band of the signal, these two threshold estimation methods can extract weak signals with better results. The fixed threshold estimation and heuristic threshold estimation denoising are more thorough and more effective in denoising, but they are also easy to mistake useful signals for noise and remove them. According to the characteristics of EEG signals, it is recommended to use extreme value estimation or unbiased likelihood estimation methods. Specifically, the appropriate threshold selection strategy can be determined according to the training process.

[0087] During the training process, the wavelet filter can filter and denoise the EEG signals in the training set. The process includes wavelet decomposition, threshold quantization and wavelet reconstruction of the signal. First, a wavelet basis will be selected and a wavelet decomposition level N will be determined during the training process. Then, the EEG signal will be decomposed into a series of wavelet coefficients by performing N-layer wavelet decomposition calculations. These wavelet coefficients correspond to the performance of the EEG signal at different time and frequency scales. Through wavelet decomposition, the local characteristics of the EEG signal can be captured, laying the foundation for subsequent noise identification and removal.

[0088] Since the wavelet coefficients of effective EEG signals are usually large, while the wavelet coefficients of noise are small, a threshold can be set, and the wavelet coefficients smaller than the threshold are regarded as noise and processed. Specifically, during the training process, a threshold selection strategy can be determined to determine a threshold, and the wavelet coefficients smaller than the threshold can be regarded as noise. These wavelet coefficients are set to zero or shrunk, thereby effectively suppressing the noise. Afterwards, the wavelet coefficients larger than the threshold can be recombined to restore the original form of the EEG signal, and the noise in the recombined EEG signal has been greatly reduced.

[0089] In this embodiment, each time the wavelet filter is trained, the filtered EEG signal is output, and then the difference between the filtered EEG signal and the corresponding reference EEG signal is calculated, and based on these differences, the target parameters of the wavelet filter are adjusted. The target parameters are continuously iterated through training to obtain a trained wavelet filter.

[0090] Step 130: input the target EEG signal of the target subject into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

[0091] In an embodiment of the present application, the target EEG signal of the target object can be input as input data into a trained wavelet filter. The wavelet filter will decompose the target EEG signal into wavelet coefficients of different scales through wavelet decomposition, and perform threshold processing on these wavelet coefficients. The wavelet coefficients after threshold processing will be reconstructed by wavelet to obtain the target EEG signal after noise reduction.

[0092] According to the human factors intelligent EEG signal processing method of the present application, a training set is constructed by acquiring EEG signals of different objects; the training set is used to train a preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; the target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application constructs a training set by acquiring EEG signals of different objects, and uses these data to train a preset wavelet filter to update parameters such as wavelet basis selection, decomposition layer number and threshold selection strategy of the wavelet filter, so that the trained wavelet filter can more accurately identify and remove noise signals in EEG signals, realize the automated processing of EEG signals, reduce the need for manual intervention, and can efficiently reduce the noise of EEG signals under low latency, greatly improving the accuracy and efficiency of EEG signal processing.

[0093] In some embodiments, obtaining EEG signals of different subjects to construct a training set includes:

[0094] Obtaining EEG signals from different subjects;

[0095] Removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain reference EEG signals after the EOG signals are removed;

[0096] The EEG signals of different objects are used as training data, and the reference EEG signals corresponding to the EEG signals of different objects are used as labels of the EEG signals of different objects to construct a training set.

[0097] In this embodiment, multiple EEG signals may be collected. For example, EEG signals of multiple different subjects may be obtained, or EEG signals of the same subject under different conditions may be obtained.

[0098] In EEG signals, the electrooculogram (EOG) signals generated by eye movements are a common source of interference, which may mask or distort the characteristics of EEG signals. Therefore, these EOG signals need to be removed.

[0099] In this embodiment, blind source separation (BSS), such as independent component analysis (ICA) and methods based on empirical mode decomposition (EMD), can be used to remove the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different objects. After the EOG signal removal process, relatively pure EEG signals are obtained, which can be used as reference EEG signals.

[0100] In this embodiment, the collected EEG signals of different subjects can be used as training data, and the corresponding reference EEG signals can be used as labels to construct a training set. Each pair of data (EEG signal and reference EEG signal) in the training set will be used to train the wavelet filter, so that the wavelet filter can learn how to extract effective EEG signals from noisy EEG signals.

[0101] In this embodiment, by collecting EEG signals of different subjects and constructing a training set, the basic data for training is diverse and extensive, and the original EEG signals are paired with reference EEG signals after removing electrooculogram signals as training data and labels, which not only improves the efficiency of training, but also enhances the ability of the wavelet filter to identify and remove electrooculogram artifacts through supervised learning.

[0102] In some embodiments, removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain the reference EEG signals after the EOG signals are removed includes:

[0103] Decompose the EEG signal into multiple statistically independent signal components through blind source separation;

[0104] Inputting multiple signal components into a preset neural network model to obtain the category of each signal component output by the neural network model;

[0105] The signal components classified as valid EEG signals are combined to obtain a reference EEG signal after removing the electrooculogram signal.

[0106] In this embodiment, blind source separation is a signal processing method that aims to recover the original independent signal sources from a mixed signal without any prior knowledge about the source signals or the mixing process. BSS is based on the assumption that the observed mixed signal is a linear combination of multiple independent source signals. These source signals are statistically independent, and nothing is known about the source signals except independence. The goal of BSS is to find an unmixing matrix so that the signals transformed by this matrix are as independent as possible.

[0107] In this embodiment, BSS can be used to separate the signal generated by brain activity and the noise signal caused by other physiological or external factors. BSS can decompose the mixed EEG signal into multiple statistically independent signal components, some of which may be signals generated by brain activity, i.e., signal components of valid EEG signals, while other signal components may be noise signals or artifacts caused by processes such as eye movement, electrocardiogram, and electromyography, i.e., invalid signal components.

[0108] In some embodiments, the BSS algorithm used may include independent component analysis (ICA), a method based on empirical mode decomposition (EMD), etc. Among them, ICA assumes that the source signals are statistically independent and at least some sources are non-Gaussian distributed. ICA finds independent components of the signal source by maximizing non-Gaussianity or minimizing mutual information.

[0109] In this embodiment, the EEG signal can be input into the BSS algorithm, and the BSS algorithm can find a set of unmixing matrices so that the signal components obtained after the EEG signal is transformed by the unmixing matrix are as independent as possible, that is, multiple statistically independent signal components are obtained. Statistical independence means that there is no correlation or almost no correlation between two or more random variables, that is, the occurrence of one variable will not affect the probability distribution of another variable. In statistics, if the joint probability distribution of two random variables X and Y is equal to the product of their respective probability distributions, then X and Y are said to be statistically independent.

[0110] In this embodiment, the categories of signal components may include valid EEG signals and invalid EEG signals. Valid EEG signals include signals generated by brain activities, and invalid EEG signals include noise signals caused by eye movements, electrocardiograms, myoelectricity, and other processes.

[0111] A neural network model can be pre-trained so that the neural network model has learned how to identify and distinguish different EEG signals, so that it can output the categories of each signal component. For example, a deep learning model such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) can be selected to design the network structure of the neural network model, and a large number of labeled EEG signals can be used as training samples for end-to-end learning. Through multiple iterative learning, the neural network model can better extract features from signal components and perform accurate classification.

[0112] In this embodiment, after the neural network model is trained, the application can be deployed to input multiple signal components obtained from blind source separation into the neural network model for inference. Each signal component will obtain a category label in the output layer of the neural network model, and the category label is the classification result of the neural network model for the signal component.

[0113] In this embodiment, the signal components classified as valid EEG signals may be combined, and other signal components may be removed to obtain a de-noised EEG signal. For example, the signal components classified as valid EEG signals may be linearly superimposed to reconstruct the EEG signal.

[0114] In this embodiment, blind source separation is used to decompose the collected EEG signal into multiple statistically independent signal components, so that the noise signal can be separated from the signal by statistical inference. The decomposed signal components are input into a preset neural network model, and the signal is classified by a deep learning algorithm to distinguish between valid EEG signals and noise signals. This makes full use of blind source separation and machine learning algorithms to improve the accuracy of removing eye contact signals from EEG signals.

[0115] In some embodiments, before decomposing the EEG signal into a plurality of statistically independent signal components by blind source separation, the method comprises:

[0116] The collected EEG signal is segmented to obtain multiple EEG signals with a preset length.

[0117] In this embodiment, an EEG acquisition device can be used to collect the EEG signal of the target object. Since the continuous EEG signal is long, in order to realize the online processing of the EEG signal and reduce the computational complexity, the continuous EEG signal can be divided into multiple shorter time periods, and the length of each EEG signal is a preset duration. Among them, the preset duration can be predetermined, for example, it can be 5s, 10s, 15s, etc., and the embodiment of the present application does not limit the preset duration.

[0118] In this embodiment, for application scenarios such as brain-computer interfaces that require real-time processing, too long signal segments will increase the overall processing delay, while the use of short segment analysis with a preset duration can significantly reduce the delay in EEG signal processing and provide support for real-time feedback and control. In addition, short-duration signal segment processing can also improve the parallelization capability of the algorithm, thereby further improving the overall EEG signal processing efficiency.

[0119] In some embodiments, the EEG signal is decomposed into a plurality of statistically independent signal components by blind source separation, including:

[0120] Preprocessing the EEG signal, wherein the preprocessing includes at least one of denoising, filtering, centering, and whitening;

[0121] Independent component analysis is performed on the preprocessed EEG signals to obtain multiple statistically independent signal components.

[0122] In this embodiment, before decomposing the EEG signal, the EEG signal can be preprocessed, including denoising, filtering, centering, whitening and other operations. Among them, denoising can reduce the noise in the EEG signal, such as electromyographic interference, power line interference, etc., to improve the signal quality; the bandpass filter can reduce the unnecessary frequency components in the EEG signal, such as removing low-frequency drift and high-frequency noise; centering is to adjust the mean of the observed data to zero; whitening is to convert the EEG signal into the form of a unit matrix with a covariance matrix.

[0123] In this embodiment, the preprocessed EEG signal may be decomposed using an independent component analysis algorithm, such as FastICA, Infomax, etc., to obtain a plurality of statistically independent signal components.

[0124] In this embodiment, the EEG signal is preprocessed by blind source separation, including denoising, filtering, centering, whitening and other steps, which can improve the signal-to-noise ratio of the EEG signal. The preprocessed EEG signal is subjected to independent component analysis to obtain multiple statistically independent signal components, which can effectively separate the mixed EEG signal into neural activity signals from different sources. These signals include brain activity signals and various noise sources. Further analysis of these signals can identify effective EEG signals and noise, thereby improving the EEG signal noise reduction effect.

[0125] In some embodiments, the EEG signal is decomposed into a plurality of statistically independent signal components by blind source separation, including:

[0126] Determine the number of channels of EEG signals;

[0127] The number of statistically independent signal components is determined based on the number of channels; wherein the number of signal components is less than or equal to the number of channels.

[0128] In this embodiment, the EEG signal is a multi-channel signal obtained by placing electrodes on the scalp or in the skull, and the number of channels depends on the number and position of the electrodes used. Multiple electrodes, such as 32 electrodes, 64 electrodes, 128 electrodes, etc., are used to collect EEG signals. Each electrode records an EEG signal of one channel. The number of channels is the number of these electrodes, and the number of channels determines the dimension of the EEG signal that can be recorded.

[0129] In blind source separation, the number of signal components obtained by decomposing the EEG signal is less than the number of channels of the EEG signal. Figure 2 As shown, for an EEG signal with 32 channels, at most 32 signal components can be obtained after the EEG signal is processed by ICA.

[0130] In this embodiment, by determining the number of channels of the EEG signal, a reference can be provided for the blind source separation algorithm, thereby reducing the overfitting problem and enabling the algorithm to restore independent source signals as accurately as possible under limited information.

[0131] In some embodiments, the training process of the neural network model includes:

[0132] Obtaining EEG signals of different subjects as training data;

[0133] Decompose the training data into multiple statistically independent data components through independent component analysis;

[0134] Label the categories of each data component to obtain a training set;

[0135] The training set is used to train the neural network model to obtain a trained neural network model.

[0136] In this embodiment, EEG signals from multiple different subjects can be collected as training data for the neural network model. These training data need to cover as wide a range of EEG signals as possible when collected to improve the diversity and representativeness of EEG signals so that the neural network model can learn the characteristics of EEG signals in different situations.

[0137] In this embodiment, the continuous EEG signal can be divided into multiple shorter time periods, and the length of each EEG signal is a preset duration. The preset duration can be predetermined, for example, 5s, 10s, 15s, etc., and the embodiment of the present application does not limit the preset duration.

[0138] The segmented EEG signals can be decomposed using the ICA algorithm to obtain multiple statistically independent signal components, which are then labeled with categories. For example, the category information labeled by developers for these signal components can be obtained. After the labeling is completed, these signal components and category labels constitute a training set.

[0139] In this embodiment, a suitable neural network architecture, such as CNN, can be selected, and the structure of the neural network can be designed according to the task requirements, including selecting the appropriate number of layers, the number of neurons in each layer, and the activation function, etc. The training set is input into the neural network model for training, and the difference between the model prediction and the actual label is measured by the loss function. In each iteration, the input data is forward propagated through the neural network, the loss is calculated, and then the gradient is calculated and the weight of the model is updated by the back propagation algorithm. During the training process, the neural network model will automatically learn the rules of feature extraction and classification, and finally obtain a trained model, that is, the ICA component judgment classifier.

[0140] In this embodiment, before the neural network model is deployed and applied, the neural network model needs to be tested. First, the EEG signal of any object needs to be obtained as test data. The continuous EEG signal can be divided into multiple shorter time periods, and the length of each EEG signal is a preset time. The segmented EEG signal can be decomposed using the ICA algorithm to obtain multiple statistically independent signal components, and these signal components are input into the ICA component judgment classifier. The ICA component judgment classifier can output the category label of each signal component. When the category label is determined to be a valid signal, the signal component can be used for the combination of EEG signals. When the category label is determined to be an invalid signal, the signal component can be removed to finally obtain the denoised EEG signal.

[0141] In this embodiment, by collecting EEG signals of different objects as training data, the diversity of model training is increased, so that the model can be generalized to EEG signals of different individuals. The independent component analysis algorithm is used to decompose the training data into multiple statistically independent components, and the training set is constructed by labeling the categories of these components. The decomposition ability of the independent component analysis algorithm is used to extract useful EEG activities and various noise sources from the mixed EEG signals. These labeled data are used to train the neural network model, so that the model can learn how to identify and distinguish different EEG signal components, thereby improving the accuracy and efficiency of EEG signal processing.

[0142] In some embodiments, multiple signal components are input into a preset neural network model to obtain the categories of each signal component output by the neural network model, including:

[0143] In the case where the neural network model includes a binary classifier, the categories of each signal component output by the neural network model include signal components that are valid EEG signals and signal components that are invalid EEG signals;

[0144] In the case where the neural network model includes multiple classifiers, the categories of each signal component output by the neural network model include at least one of the signal components effective electroencephalogram signal, electrocardiogram signal, electrooculogram signal, electromyography signal, and unknown signal.

[0145] In this embodiment, according to the requirements of the classification task, the neural network model can adopt a simple binary classification neural network; for multi-classification, a deeper or more complex network structure can be adopted.

[0146] During the training process, when the neural network model includes a binary classifier, the category labels for the signal components can be valid EEG signals and invalid EEG signals, so that the outputs of the neural network model are also valid EEG signals and invalid EEG signals.

[0147] In one example, if Figure 3 As shown, the collected EEG signals can first be segmented, for example, 10 seconds per segment, to obtain EEG signal segments, and then the EEG signal segments are decomposed by ICA to obtain multiple statistically independent signal components, and these signal components are input into the neural network model to obtain the classification results of each signal component. For example, the neural network model is a binary classifier, and the output of the neural network model can include the category label of the signal component. For example, the category label "1" indicates that the category of the signal component is a valid EEG signal, and the category label "0" indicates that the category of the signal classification is an invalid EEG signal.

[0148] After obtaining the classification results of each signal component, the post-processing step is entered. Specifically, the post-processing step may include: removing the signal components classified as invalid EEG signals, combining the signal components classified as valid EEG signals, and obtaining a de-noised EEG signal.

[0149] When the neural network model is a multi-classifier, the category labeling of the signal component can be effective EEG signal, ECG signal, EO signal, EMG signal, unknown signal, etc., so that the neural network model can learn the mapping relationship between the characteristics of the signal component and the category, and output effective EEG signal, ECG signal, EO signal, EMG signal, unknown signal, etc. Figure 2 As shown, Figure 2The signal components ICA003, ICA010, and ICA011 are eye movement signals.

[0150] In this embodiment, the learning ability of the neural network model is used to achieve automatic classification of signal components. In the case of a binary classifier, the model can distinguish between valid EEG signals and invalid EEG signals, and in the case of a multi-classifier, the model further refines the classification and identifies multiple signals such as valid EEG signals, ECG signals, EOG signals, EMG signals, and unknown signals, greatly improving the accuracy and efficiency of EEG signal processing.

[0151] In some embodiments, a preset wavelet filter is trained using a training set to update target parameters of the wavelet filter to obtain a trained wavelet filter, including:

[0152] Repeat the following steps until the loss function converges or reaches the preset stop condition to obtain the trained wavelet filter:

[0153] Initialize the target parameters of the wavelet filter, input the training set into the wavelet filter, and obtain the EEG signal after the wavelet filter output is filtered;

[0154] The loss function is used to calculate the difference between the filtered EEG signal and the corresponding reference EEG signal;

[0155] Based on the differences, an optimization algorithm is used to adjust the target parameters of the wavelet filter.

[0156] In this embodiment, before starting to train the wavelet filter, it is necessary to first construct an initial wavelet filter. For example, an initial wavelet basis, decomposition level and threshold strategy can be selected, and a wavelet analysis toolkit, such as the wavelet toolbox in MATLAB, is used to construct the initial wavelet filter using the selected wavelet basis, decomposition level and threshold strategy.

[0157] At the beginning of each training iteration, the target parameters of the wavelet filter can be initialized, and the training set can be output to the wavelet filter. The wavelet filter processes the EEG signals in the training set according to the current target parameters to obtain filtered EEG signals.

[0158] Since the EEG signals in the training set correspond to reference EEG signals after noise removal, the filtered EEG signals can be compared with the corresponding reference EEG signals. If the difference is small, it means that the denoising effect of the target parameters used in this training is better. If the difference is large, it means that the denoising effect of the target parameters used in this training is poor. The difference between the filtered EEG signals and the reference EEG signals can be measured by a loss function. For example, the loss function can measure the difference between the filtered EEG signals and the reference EEG signals by calculating the minimum mean square error between the filtered EEG signals and the reference EEG signals.

[0159] In this embodiment, according to the calculation result of the loss function, an optimization algorithm such as gradient descent, Adam, etc. can be used to adjust the target parameters, and the target parameters are initialized to the adjusted target parameters before starting the next training. Through multiple trainings, the target parameters are continuously iteratively updated, so that the loss function is continuously reduced, thereby improving the filtering effect of the wavelet filter. Of course, the optimization algorithm can also be selected by enumeration. Wavelet basis, decomposition layer number and threshold selection strategy, so as to find the best parameter combination.

[0160] In this embodiment, the stopping condition of the training process may be the convergence of the loss function, that is, the value of the loss function no longer decreases significantly, or a preset number of iterations is executed, etc. After stopping the training, the target parameters at this time may be saved, and the trained wavelet filter may be tested with test data. If the test passes, the trained wavelet filter may be deployed and applied for subsequent EEG signal denoising tasks.

[0161] In this embodiment, the target parameters of the wavelet filter are initialized and the training set is input into the wavelet filter to obtain preliminary filtering results. The loss function is used to measure the difference between the filtered EEG signal and the reference EEG signal. Based on these differences, an optimization algorithm is used to adjust the target parameters of the wavelet filter, such as wavelet basis selection, number of decomposition layers and threshold selection strategy, in order to reduce the value of the loss function. The training process is iterated until the loss function converges or reaches the preset stopping condition, so that the wavelet filter can achieve the expected performance and improve the accuracy of automated processing of EEG signals.

[0162] According to an embodiment of the present application, a preset wavelet filter is trained using a training set to update target parameters of the wavelet filter, including:

[0163] According to the characteristics of the EEG signals in the training set, multiple wavelet bases are selected by taking at least one of the support length, vanishing moment, symmetry, regularity and similarity of the wavelet base as a reference standard;

[0164] During the training process, a wavelet basis is selected from multiple wavelet bases to determine the wavelet basis selection of the trained wavelet filter.

[0165] In this embodiment, before training the small filter, multiple wavelet bases can be selected based on the characteristics of the EEG signal using the support length, vanishing moment, symmetry, regularity and similarity of the wavelet base as reference standards. Specifically, the support length represents the length of the wavelet filter. The shorter the length of the wavelet filter, the lower the computational complexity of the wavelet transform. Generally, the support length is not less than 2 times the vanishing moment minus 1, and a wavelet with a support length between 5 and 9 can be selected, because if the support length is too large, boundary problems will occur, and if the support length is too small, the vanishing moment is too low, which is not conducive to the concentration of signal energy.

[0166] The vanishing moment is the concentration of energy after wavelet transformation. The higher the vanishing moment, the smaller the wavelet coefficients of the high-frequency subband, and the more wavelet coefficients are close to 0. The size of the vanishing moment of the wavelet determines the ability of the wavelet to approximate a smooth signal. The larger the vanishing moment, the flatter the corresponding filter, and the stronger the oscillation of the wavelet function.

[0167] In this embodiment, the selected wavelet basis should be symmetrical, because a symmetrical wavelet can effectively avoid phase distortion, because the wavelet filter corresponding to the symmetrical wavelet has the characteristic of linear phase.

[0168] When quantizing or ingesting wavelet coefficients, in order to reduce the impact of reconstruction errors, the smoothness or continuous differentiability of the wavelet should be increased as much as possible. "Irregular" errors are more sensitive than "smooth" errors. Regularity adjustment needs to be imposed, that is, a wavelet with good regularity can achieve a better smoothing effect in signal reconstruction and reduce the impact of quantization or rounding errors.

[0169] In this embodiment, a wavelet basis similar to the waveform of the EEG signal may be selected, thereby improving the matching degree of the wavelet based on the waveform of the EEG signal.

[0170] After selecting multiple wavelet bases, the wavelet bases can be selected by enumeration during the training process to determine the wavelet base with the best denoising effect.

[0171] In this embodiment, multiple wavelet bases are selected based on the characteristics of the EEG signals in the training set by comprehensively considering multiple dimensions of the wavelet basis, such as support length, vanishing moment, symmetry, regularity and similarity. This method selects basis functions suitable for the characteristics of the EEG signals based on multiple attributes of the wavelet basis, thereby improving the denoising effect of the EEG signals.

[0172] The human factors intelligent EEG signal processing method provided in the embodiment of the present application can be executed by a human factors intelligent EEG signal processing device. In the embodiment of the present application, the human factors intelligent EEG signal processing device executing the human factors intelligent EEG signal processing method is taken as an example to illustrate the human factors intelligent EEG signal processing device provided in the embodiment of the present application.

[0173] The embodiment of the present application also provides a human factors intelligent EEG signal processing device.

[0174] like Figure 4 As shown, the human factor intelligent EEG signal processing device includes:

[0175] An acquisition module 410 is used to acquire EEG signals of different subjects to construct a training set;

[0176] The training module 420 is used to train the preset wavelet filter using the training set to update the target parameters of the wavelet filter and obtain the trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy;

[0177] The input module 430 is used to input the target EEG signal of the target object into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

[0178] According to the human factors intelligent EEG signal processing device of the present application, a training set is constructed by acquiring EEG signals of different objects; the training set is used to train a preset wavelet filter to update the target parameters of the wavelet filter to obtain a trained wavelet filter; the target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; the target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter. The embodiment of the present application constructs a training set by acquiring EEG signals of different objects, and uses these data to train a preset wavelet filter to update parameters such as wavelet basis selection, decomposition layer number and threshold selection strategy of the wavelet filter, so that the trained wavelet filter can more accurately identify and remove noise signals in EEG signals, realize the automated processing of EEG signals, reduce the need for manual intervention, and can efficiently reduce the noise of EEG signals under low latency, greatly improving the accuracy and efficiency of EEG signal processing.

[0179] In some embodiments, the acquisition module 410 is further configured to:

[0180] Obtaining EEG signals from different subjects;

[0181] Removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain reference EEG signals after the EOG signals are removed;

[0182] The EEG signals of different objects are used as training data, and the reference EEG signals corresponding to the EEG signals of different objects are used as labels of the EEG signals of different objects to construct a training set.

[0183] In some embodiments, the acquisition module 410 is further configured to:

[0184] Decompose the EEG signal into multiple statistically independent signal components through blind source separation;

[0185] Inputting multiple signal components into a preset neural network model to obtain the category of each signal component output by the neural network model;

[0186] The signal components classified as valid EEG signals are combined to obtain a reference EEG signal after removing the electrooculogram signal.

[0187] In some embodiments, the acquisition module 410 is further configured to:

[0188] Preprocessing the EEG signal, wherein the preprocessing includes at least one of denoising, filtering, centering, and whitening;

[0189] Independent component analysis is performed on the preprocessed EEG signals to obtain multiple statistically independent signal components.

[0190] In some embodiments, the acquisition module 410 is further configured to:

[0191] Determine the number of channels of EEG signals;

[0192] The number of statistically independent signal components is determined based on the number of channels; wherein the number of signal components is less than or equal to the number of channels.

[0193] In some embodiments, the training module 420 is further configured to:

[0194] Repeat the following steps until the loss function converges or reaches the preset stop condition to obtain the trained wavelet filter:

[0195] Initialize the target parameters of the wavelet filter, input the training set into the wavelet filter, and obtain the EEG signal after the wavelet filter output is filtered;

[0196] The loss function is used to calculate the difference between the filtered EEG signal and the corresponding reference EEG signal;

[0197] Based on the differences, an optimization algorithm is used to adjust the target parameters of the wavelet filter.

[0198] In some embodiments, the training module 420 is further configured to:

[0199] According to the characteristics of the EEG signals in the training set, multiple wavelet bases are selected by taking at least one of the support length, vanishing moment, symmetry, regularity and similarity of the wavelet base as a reference standard;

[0200] During the training process, a wavelet basis is selected from multiple wavelet bases to determine the wavelet basis selection of the trained wavelet filter.

[0201] The human factor intelligent EEG signal processing device in the embodiment of the present application can be an edge computing device, or a component in an edge computing device, such as an integrated circuit or a chip. The edge computing device can be a terminal, or it can be other devices other than a terminal. Exemplarily, the edge computing device can be a mobile phone, a tablet computer, a laptop computer, an eye tracker, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not specifically limited. In an optional example, an edge deployment solution in which the human factors intelligent EEG signal processing device is deployed as an edge computing device or component on the end side can migrate at least part of the computing tasks to the end side device, giving full play to the advantages of edge computing to improve data processing efficiency, reduce latency and network resource occupancy and dependence, and also facilitate data security and privacy protection.

[0202] The human factor intelligent EEG signal processing device in the embodiment of the present application may be a device having an operating system. The operating system may be a Microsoft (Windows) operating system, an Android (Android) operating system, an IOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0203] In some embodiments, Figure 5 As shown, the embodiment of the present application also provides an edge computing device 500, including a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the program is executed by the processor 501, each process of the above-mentioned human factors intelligent EEG signal processing method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0204] It should be noted that the edge computing devices in the embodiments of the present application include the above-mentioned mobile edge computing devices and non-mobile edge computing devices.

[0205] The embodiment of the present application also provides a computer-readable storage medium, on which programs or instructions are stored. When the programs or instructions are executed by the processor, the various processes of the above-mentioned human factors intelligent EEG signal processing method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, they will not be repeated here.

[0206] The processor is the processor in the edge computing device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0207] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above-mentioned human factors intelligent EEG signal processing method when executed by a processor.

[0208] The processor is the processor in the edge computing device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0209] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, which are coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-mentioned human factors intelligent EEG signal processing method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0210] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0211] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0212] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0213] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0214] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0215] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A human factor intelligent EEG signal processing method, characterized in that: include: Obtain EEG signals from different subjects to construct a training set; Using the training set to train a preset wavelet filter to update target parameters of the wavelet filter to obtain a trained wavelet filter; The target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; The target EEG signal of the target object is input into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

2. The method according to claim 1, characterized in that The step of obtaining EEG signals of different subjects to construct a training set includes: Obtaining EEG signals from different subjects; Removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain reference EEG signals after the EOG signals are removed; The training set is constructed by using the EEG signals of different subjects as training data and the reference EEG signals corresponding to the EEG signals of different subjects as labels of the EEG signals of different subjects.

3. The method according to claim 2, characterized in that The step of removing the electrooculogram (EOG) signals generated by eye movements from the EEG signals of different subjects to obtain the reference EEG signals after the EOG signals are removed includes: Decomposing the EEG signal into multiple statistically independent signal components by blind source separation; Inputting a plurality of the signal components into a preset neural network model to obtain the category of each signal component output by the neural network model; The signal components classified as valid EEG signals are combined to obtain a reference EEG signal after removing the electrooculogram signal.

4. The method according to claim 3, characterized in that The step of decomposing the EEG signal into a plurality of statistically independent signal components by blind source separation comprises: Preprocessing the EEG signal, wherein the preprocessing includes at least one of denoising, filtering, centering, and whitening; Independent component analysis is performed on the preprocessed EEG signals to obtain multiple statistically independent signal components.

5. The method according to claim 3, characterized in that: The step of decomposing the EEG signal into a plurality of statistically independent signal components by blind source separation comprises: Determining the number of channels of the EEG signal; The number of the statistically independent signal components is determined according to the number of channels; wherein the number of the signal components is less than or equal to the number of channels.

6. The method according to claim 1, characterized in that The step of training a preset wavelet filter using the training set to update target parameters of the wavelet filter and obtain a trained wavelet filter comprises: Repeat the following steps until the loss function converges or reaches the preset stop condition to obtain the trained wavelet filter: Initializing target parameters of the wavelet filter, inputting the training set into the wavelet filter, and obtaining an electroencephalogram signal filtered by the wavelet filter output; The loss function is used to calculate the difference between the filtered EEG signal and the corresponding reference EEG signal; Based on the difference, an optimization algorithm is used to adjust the target parameters of the wavelet filter.

7. The method according to claim 1, characterized in that The adopting the training set to train the preset wavelet filter to update the target parameters of the wavelet filter includes: According to the characteristics of the EEG signals in the training set, at least one of the support length, vanishing moment, symmetry, regularity and similarity of the wavelet basis is used as a reference standard to select multiple wavelet bases; During the training process, a wavelet basis is selected from the plurality of wavelet bases to determine the wavelet basis selection of the trained wavelet filter.

8. A human factor intelligent EEG signal processing device, characterized in that: include: An acquisition module, used to acquire EEG signals of different subjects to construct a training set; A training module, used for training a preset wavelet filter using the training set to update target parameters of the wavelet filter and obtain a trained wavelet filter; The target parameters include at least one of wavelet basis selection, decomposition layer number and threshold selection strategy; The input module is used to input the target EEG signal of the target object into the trained wavelet filter to obtain the denoised target EEG signal output by the trained wavelet filter.

9. An edge computing device, characterized in that: It includes a processor and a memory, the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the human factor intelligent EEG signal processing method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the human factors intelligent EEG signal processing method according to any one of claims 1 to 7 is implemented.

Images