Electroencephalogram signal feature extraction method and device based on frequency domain analysis

Abstract

The invention provides an electroencephalogram signal feature extraction method and device based on frequency domain analysis. According to the method, firstly, signals are adjusted through dynamic filtering according to the characteristics of electroencephalogram signals and noise environment, then time-varying characteristics of the signals are accurately captured through energy distribution extraction on a frequency-time plane, and functional connection between brain areas is predicted by applying a frequency domain coherence analysis technology. And frequency domain nonlinear feature extraction is carried out on the signal in combination with a nonlinear dynamics theory. Finally, the extracted frequency domain features are screened through a feature importance scoring method, a target feature set is obtained, and electroencephalogram signals are recognized and classified based on the target feature set. According to the method, by introducing dynamic filtering, frequency domain coherence analysis and a nonlinear dynamics theory, the problems that in the prior art, non-stationarity processing on electroencephalogram signals is insufficient, time-varying characteristics are ignored, and feature extraction accuracy and calculation efficiency are low are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram (EEG) signal processing, and in particular to an EEG signal feature extraction method based on frequency domain analysis, an EEG signal feature extraction device based on frequency domain analysis, a computer-readable storage medium, and an EEG wave analysis system. Background Art

[0002] In the field of EEG signal processing, frequency domain feature extraction methods have always been a hot topic of research. Early EEG signal feature extraction mainly relied on time domain analysis, such as autoregressive models, Higuchi algorithms, etc. These methods can reflect the characteristics of EEG signals to a certain extent. However, with the continuous development of signal processing technology, frequency domain analysis has gradually become the mainstream of research due to its unique advantages. The application of Fourier transform, power spectral density, band power and other technologies enables researchers to understand the inherent laws of EEG signals more deeply from the perspective of frequency. The development of these technologies provides strong support for the feature extraction, classification and interpretation of EEG signals.

[0003] However, the existing frequency domain analysis technology still has certain shortcomings in extracting EEG signal features. The traditional Fourier transform is suitable for processing stationary signals, but the non-stationarity of EEG signals means that the Fourier transform may cause the loss of important signal information during the processing. Existing frequency domain analysis methods often ignore the time-varying characteristics of the signal at different frequencies, resulting in the extracted features may not fully reflect the dynamic changes of EEG signals. In addition, the current frequency domain feature extraction method has a high computational cost and its feature identification ability still needs to be improved, making it difficult to meet the dual requirements of efficiency and accuracy in practical applications. Summary of the invention

[0004] The main purpose of this application is to provide a method for extracting EEG signal features based on frequency domain analysis, an apparatus for extracting EEG signal features based on frequency domain analysis, a computer-readable storage medium and an EEG analysis system, so as to at least solve the problems in the prior art of insufficient processing of non-stationarity of EEG signals, neglect of time-varying characteristics, and low feature extraction accuracy and computational efficiency.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for analyzing EEG signals based on frequency domain analysis is provided, including: obtaining an EEG signal to obtain a first target signal, dynamically filtering the first target signal based on the signal characteristics and noise environment of the first target signal to obtain a second target signal; extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain first distribution data; using coherence spectrum analysis technology, performing frequency domain coherence analysis on different second target signals based on the first distribution data to predict the functional connections between different brain regions to obtain a first analysis result; performing frequency domain nonlinear feature extraction on the second target signal based on the first distribution data combined with nonlinear dynamics theory to obtain a second analysis result; converting the first analysis result and the second analysis result into a frequency domain feature set, screening the frequency domain feature set by a feature importance scoring method to obtain a target feature set; and identifying and classifying the EEG signal based on the target feature set.

[0006] Optionally, an EEG signal is acquired to obtain a first target signal, and dynamic filtering is performed based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal, including: collecting historical data of the EEG signals of different target objects through electrodes to obtain a third target signal; collecting physiological parameter data of the target object to obtain first target data, collecting ambient sound through a microphone to obtain ambient noise data, and collecting vibration data through an accelerometer to obtain second target data; performing cluster analysis based on the third target signal, the first target data, the ambient noise data and the second target data to obtain different categories of the EEG signals and obtain multiple EEG signal groups; establishing a feature model for each of the EEG signal groups, and initializing filtering parameters corresponding to different categories based on the feature model; performing cluster analysis on the first target signal to obtain a target category corresponding to the first target signal, determining corresponding target filtering parameters based on the target category, and filtering the first target signal based on the target filtering parameters to obtain the second target signal.

[0007] Optionally, extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain first distribution data includes: determining a standard deviation of the second target signal based on the second target signal, and determining a scale parameter of the multi-scale analysis window based on the standard deviation: Wherein, λ is the scale parameter, σ(t) is the standard deviation, and ∈ is a value used to avoid division by 0; the second target signal is windowed based on the scale parameter to obtain a third target signal; an adaptive recursive wavelet transform is performed based on the third target signal to extract the instantaneous frequency and amplitude of the third target signal to obtain the first distribution data: E(f, t, λ) = |ARWT(x w (k),λ,t)| 2 ; Wherein, E(f, t, λ) is the first distribution data, which is used to characterize the three-dimensional distribution of the third target signal in energy-frequency-time, wherein ARWT(x w (k),λ,t) satisfies: Among them, x w (k) is the third target signal, is a parameter used to limit the range of wavelet basis functions, ψ λ (t) is the adaptive phase function, and j is the imaginary unit.

[0008] Optionally, using a coherence spectrum analysis technique, frequency domain coherence analysis is performed on different second target signals based on the first distribution data to predict functional connections between different brain regions to obtain a first analysis result, including: performing coherence calculation based on the second target signals of different channels and the first distribution data corresponding to the second target signals to obtain target coherence: Among them, C ij (f, t) is the target coherence, S ij (f, t) is the cross spectral density between the first distribution data corresponding to the second target signal of channels i and j, where S ij (f,t) satisfies: in, For X i(f,t,λ) The conjugate of X i(f,t,λ) is the data intercepted by the time window in the first distribution data, and T is the length of the time window; a weighted coherence index is calculated based on the target coherence to characterize the degree of functional connectivity between different brain regions: Among them, Γ ij (f) is the weighted coherence index, η(f) is the frequency energy weighting factor, which is used to enhance the coherence of a specific frequency band; the weighted coherence index and the target coherence are determined as the first analysis result.

[0009] Optionally, performing nonlinear feature extraction in the frequency domain on the second target signal based on the first distribution data in combination with nonlinear dynamics theory to obtain a second analysis result includes: determining the number of boxes based on the first distribution data, and determining a fractal dimension based on the number of boxes: Wherein, D is the fractal dimension, N(∈) is the number of boxes determined according to the signal energy distribution in the first distribution data; the complexity index of the second target signal is determined based on the fractal dimension and the time series entropy of the second target signal: ′ =D·H(X); where Ξ ′ is the complexity index, H(X) is the time series entropy; and the complexity index is determined as the second analysis result.

[0010] Optionally, converting the first analysis result and the second analysis result into a frequency domain feature set includes: converting the first analysis result and the second analysis result into a feature vector set to obtain the frequency domain feature set.

[0011] Optionally, the frequency domain feature set is screened by a feature importance scoring method to obtain a target feature set, including: using part of the data in the frequency domain feature set as the initial state of the target feature set; calculating the target score based on the feature importance scoring method for the feature vectors in the frequency domain feature set based on a scoring function: R(f i )=α·Corr(f i ,y)-β·Redun(f i ,S); where R(f i ) is the target score, Corr(f i ,y) is the feature vector f i Correlation with the target variable y, Redun(f i , S) is the feature vector f i The redundancy with the current state of the target feature set S, α is a weight factor for balancing correlation, β is a weight factor for balancing redundancy; sorting is performed based on the target score, and the target feature set is updated with a previously preset number of feature vectors; the target score is recalculated based on the updated target feature set, and the target feature combination is updated according to the recalculated target score, until the difference in the target score between two adjacent iterations of the same feature vector is less than a second threshold or the maximum number of iterations is reached, and the target feature set is output.

[0012] According to another aspect of the present application, there is provided an apparatus for extracting features of an electroencephalogram signal based on frequency domain analysis, the apparatus comprising: a first acquisition unit, for acquiring an electroencephalogram signal to obtain a first target signal, and performing dynamic filtering based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal; a first calculation unit, for extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain first distribution data; a second calculation unit, for performing frequency domain coherence analysis on different second target signals based on the first distribution data by means of coherence spectrum analysis technology to predict the functional connections between different brain regions to obtain a first analysis result; a third calculation unit, for performing frequency domain nonlinear feature extraction on the second target signal based on the first distribution data in combination with nonlinear dynamics theory to obtain a second analysis result; a fourth calculation unit, for converting the first analysis result and the second analysis result into a frequency domain feature set, and screening the frequency domain feature set by means of a feature importance scoring method to obtain a target feature set; and a classification unit, for identifying and classifying the electroencephalogram signal based on the target feature set.

[0013] 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.

[0014] According to another aspect of the present application, a brain wave analysis 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.

[0015] Applying the technical solution of the present application, the method first adjusts the signal according to the characteristics of the EEG signal and the noise environment through dynamic filtering, then extracts the energy distribution on the frequency-time plane, accurately captures the time-varying characteristics of the signal, and applies the frequency domain coherence analysis technology to predict the functional connection between brain regions, and combines the nonlinear dynamics theory to extract the frequency domain nonlinear features of the signal. Finally, the extracted frequency domain features are screened by the feature importance scoring method to obtain the target feature set and identify and classify the EEG signal based on this. The present application introduces dynamic filtering, frequency domain coherence analysis and nonlinear dynamics theory. Dynamic filtering overcomes the defect that the traditional method may lose key information when processing non-stationary signals. Frequency domain coherence analysis and nonlinear feature extraction deeply explore the time-varying characteristics and nonlinear laws of the signal, and the feature importance scoring rule improves the efficiency and discrimination ability of feature screening, thereby comprehensively improving the accuracy and efficiency of EEG signal analysis. It solves the problems of insufficient processing of non-stationary EEG signals, ignoring time-varying characteristics, and low accuracy and computational efficiency of feature extraction in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A hardware structure block diagram of a mobile terminal for extracting EEG signal features based on frequency domain analysis provided in an embodiment of the present application is shown;

[0017] Figure 2 A schematic diagram of a process for extracting features of an electroencephalogram signal based on frequency domain analysis according to an embodiment of the present application is shown;

[0018] Figure 3 A structural block diagram of an electroencephalogram signal feature extraction device based on frequency domain analysis provided according to an embodiment of the present application is shown.

[0019] The above drawings include the following reference numerals:

[0020] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0021] 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.

[0022] 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.

[0023] 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.

[0024] As introduced in the background technology, the prior art has deficiencies in EEG signal analysis, which are mainly manifested in the inability to effectively process non-stationary signals, ignoring the time-varying characteristics of signals, and the low identification ability and computational efficiency of feature extraction, making it difficult to meet the needs of high precision and high efficiency. In order to address the problems of insufficient processing of EEG signal non-stationarity, ignoring time-varying characteristics, and low feature extraction accuracy and computational efficiency in the prior art, the embodiments of the present application provide an EEG signal feature extraction method based on frequency domain analysis, an EEG signal feature extraction device based on frequency domain analysis, a computer-readable storage medium, and an EEG analysis system.

[0025] 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.

[0026] 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 an EEG signal feature extraction method based on frequency domain analysis according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1Only 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 are shown.

[0027] 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.

[0028] In this embodiment, a method for extracting features of electroencephalogram signals based on frequency domain analysis that runs 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 can be executed in an order different from that shown here.

[0029] Figure 2 FIG. 1 is a flow chart of a method for extracting EEG signal features based on frequency domain analysis according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0030] Step S201, acquiring an EEG signal to obtain a first target signal, and performing dynamic filtering based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal;

[0031] Specifically, the EEG signal data is collected, and the complexity of signal processing and analysis caused by the differences in EEG signal characteristics and noise environment of different individuals is solved by real-time learning and adjustment of filtering parameters. Dynamic filtering is performed according to the EEG signal characteristics and noise environment of different individuals to obtain the second target signal. The introduction of dynamic filtering effectively improves the quality of EEG signals and reduces the interference caused by noise. The filtering strategy can adjust the parameters according to the characteristics of the signal, so that the second target signal finally obtained can more accurately reflect the EEG activity.

[0032] Step S202, extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain first distribution data;

[0033] Specifically, the dynamic change characteristics of EEG signals in the frequency-time plane are accurately captured to enable effective analysis and accurately evaluate the functional connections between different brain regions, providing key information for subsequent functional connection analysis and nonlinear feature extraction.

[0034] Step S203, performing frequency domain coherence analysis on different second target signals based on the first distribution data by using coherence spectrum analysis technology to predict functional connections between different brain regions and obtain a first analysis result;

[0035] Specifically, the frequency-time plane data of the second target signal is subjected to frequency domain coherence analysis through coherence spectrum analysis technology. This analysis can reveal the functional connectivity between different brain regions and reveal the coordinated activities of different brain regions in specific frequency bands through coherence spectrum.

[0036] Step S204, extracting nonlinear features of the second target signal in the frequency domain based on the first distribution data in combination with nonlinear dynamics theory to obtain a second analysis result;

[0037] Specifically, based on the frequency domain distribution data in step S202, combined with nonlinear dynamics theory, nonlinear feature extraction is performed on the second target signal. Nonlinear analysis methods can reveal complex dynamic features hidden in the signal and capture valuable information that traditional linear methods cannot discover.

[0038] Step S205, converting the first analysis result and the second analysis result into a frequency domain feature set, and screening the frequency domain feature set by a feature importance scoring method to obtain a target feature set;

[0039] Specifically, the analysis results obtained in steps S203 and S204 are converted into a frequency domain feature set, and then the feature importance scoring method is applied to screen these features to extract a target feature set with higher classification efficiency. This step aims to reduce redundant information and improve the accuracy of subsequent classification tasks through scientific feature selection methods. The screened target feature set contains the most influential information for the classification task, which significantly improves the classification accuracy and efficiency.

[0040] Step S206, identifying and classifying the EEG signal based on the target feature set.

[0041] Specifically, after obtaining the target feature set, these features are used to identify and classify EEG signals, which can efficiently identify different types of EEG signals. This step will provide classification results for subsequent EEG signal analysis and support practical applications such as clinical diagnosis and neurological function monitoring.

[0042] Through this embodiment, a method for extracting EEG signal features based on frequency domain analysis is provided, which collects signal data, optimizes filtering according to the EEG signal characteristics and noise environment of different individuals through real-time learning and adjustment of filtering parameters; extracts the energy distribution of the signal on the frequency-time plane to reveal the dynamic change characteristics of the signal, uses coherence spectrum analysis technology to identify the frequency domain coherence between different EEG signal channels, evaluates the functional connection between brain regions, and extracts the nonlinear characteristics of frequency domain signals in combination with nonlinear dynamics theory, and analyzes the complexity and chaotic nature of EEG signals; optimizes and selects the extracted frequency domain features to identify the most contributing feature subsets; outputs the optimized frequency domain feature set for classification, recognition and other advanced analysis of EEG signals. This application optimizes the frequency domain analysis algorithm, introduces dynamic filtering, frequency domain coherence analysis and nonlinear dynamics theory. Dynamic filtering overcomes the defect that traditional methods may lose key information when processing non-stationary signals. Frequency domain coherence analysis and nonlinear feature extraction deeply explore the time-varying characteristics and nonlinear laws of signals, and the feature importance scoring rule improves the efficiency and discrimination ability of feature screening, thereby comprehensively improving the accuracy and efficiency of EEG signal analysis, thereby providing more reliable technical support for the research and application of EEG signals. Compared with the existing technology, it may have significant advantages in processing non-stationary signals, extracting time-varying features and reducing computing costs. It solves the problems of insufficient processing of non-stationary EEG signals, ignoring time-varying characteristics, and low accuracy and computing efficiency of feature extraction in the existing technology.

[0043] As a possible implementation method, acquiring an EEG signal to obtain a first target signal, and dynamically filtering the first target signal based on the signal characteristics and noise environment of the first target signal to obtain a second target signal includes:

[0044] Step S301, collecting historical data of EEG signals of different target objects through electrodes to obtain a third target signal;

[0045] Step S302, collecting physiological parameter data of the target object to obtain first target data, collecting environmental sound through a microphone to obtain environmental noise data, and collecting vibration data through an accelerometer to obtain second target data;

[0046] Specifically, the electrode patch can accurately acquire the electrical activity of the brain. These signals can provide information on neural activity in various areas of the brain as the basis for subsequent analysis. The electrode patch collects voltage time series, physiological parameter data (such as heart rate, body temperature, etc.) recorded by scalp electrodes on different target objects, and environmental sounds (such as background noise) and vibration data collected by microphones and accelerometers. These data, as auxiliary information, can provide additional background information about the target object and its surrounding environment, making the analysis of EEG signals more accurate and reliable, especially in noisy environments.

[0047] Step S303, performing cluster analysis based on the third target signal, the first target data, the environmental noise data, and the second target data to obtain different categories of EEG signals and obtain multiple EEG signal groups;

[0048] Specifically, a feature model is established for each target object through preliminary EEG signal analysis. Cluster analysis is used to comprehensively analyze the third target signal (historical EEG signal data), the first target data (physiological parameter data), the environmental noise data, and the second target data (vibration data) to obtain EEG signals of different categories. Especially when the target object is in different physiological states or environmental conditions, cluster analysis helps to clearly divide the different categories of EEG signals, thereby providing an accurate signal classification basis for subsequent filtering processing.

[0049] Step S304, establishing a feature model for each EEG signal group, and initializing filter parameters corresponding to different categories based on the feature model;

[0050] Specifically, the feature model can help describe and summarize the typical characteristics of each category of signals, and the initialization of the filter parameters is adjusted according to these feature models. Establishing a feature model for each EEG signal group (each category) can accurately capture the unique properties of each type of EEG signal, thereby optimizing the parameter settings of the filter. Through reasonable initialization of filter parameters, it can ensure targeted processing of each signal category, minimize noise interference, and improve filtering effect.

[0051] Let Θ i is the filter parameter set for the i-th target object, initialized to: Θ i ={θ i1,θ i2 ,…,θ in}; where θ ij is the jth filter parameter for the i-th type of target object.

[0052] Monitor the characteristic changes of EEG signals in real time, and dynamically adjust the filtering parameters according to the real-time tracking characteristics to adapt to the current state of the target object.

[0053] Step S305, performing cluster analysis on the first target signal to obtain a target category corresponding to the first target signal, determining a corresponding target filtering parameter based on the target category, filtering the first target signal based on the target filtering parameter, and obtaining a second target signal.

[0054] Specifically, cluster analysis is performed on the first target signal to obtain the target category to which it belongs. Based on the category, the corresponding target filter parameters are selected, and the first target signal is filtered to obtain the second target signal. The filtered signal will remove unnecessary noise and retain the EEG signals that are valuable for analysis. By associating EEG signals with specific categories and applying corresponding filter parameters, more accurate noise suppression and signal enhancement can be achieved.

[0055] As a possible implementation manner, extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain the first distribution data includes:

[0056] Step S401, based on the second target signal, determining the standard deviation of the second target signal, and determining the scale parameter of the multi-scale analysis window based on the standard deviation: Where λ is the scale parameter, σ(t) is the standard deviation, and ∈ is a value used to avoid division by 0;

[0057] Step S402, windowing the second target signal based on the scale parameter to obtain a third target signal;

[0058] Specifically, extracting the energy distribution of the signal on the frequency-time plane includes automatically selecting a suitable multi-scale analysis window according to the waveform characteristics of the EEG signal to capture frequency components of different scales. By calculating the standard deviation, the volatility of the signal at different time points can be accurately reflected. Using the scale parameter λ to control the window size helps to adaptively adjust the width of the analysis window in frequency and time, thereby improving the time-frequency resolution of the signal. Based on the scale parameter λ calculated in step S401, the second target signal is further windowed to obtain the third target signal. The windowing process divides the signal into multiple time windows for analysis, and the size of each window is adaptively determined by the scale parameter λ. The windowing process can refine the analysis of the signal in time and frequency, and improve the ability to analyze the local characteristics of the signal.

[0059] It can be seen that the scale-adaptive windowing method can ensure more accurate local analysis of signals in different frequency and time ranges, thereby optimizing subsequent frequency domain analysis and feature extraction.

[0060] Step S403, performing adaptive recursive wavelet transform based on the third target signal to extract the instantaneous frequency and amplitude of the third target signal to obtain the first distribution data: E(f, t, λ) = |ARWT(x w (k),λ,t)| 2 ;

[0061] Wherein, E(f, t, λ) is the first distribution data, which is used to characterize the three-dimensional distribution of the third target signal in energy-frequency-time. w (k),λ,t) satisfies:

[0062]

[0063] Among them, x w (k) is the third target signal, is a parameter used to limit the range of wavelet basis functions, ψ λ (t) is the adaptive phase function, and j is the imaginary unit.

[0064] Specifically, an adaptive complex modulated wavelet transform is applied to each windowed signal segment to extract the instantaneous frequency and amplitude of the signal, and obtain the energy-frequency-time three-dimensional distribution E(f, t, λ). The adaptive recursive wavelet transform can provide high time-frequency resolution and accurately capture the changes in the instantaneous frequency and amplitude of the signal. The ARWT method uses an adaptive phase function and a wavelet basis function to recursively analyze the signal in time and frequency, thereby obtaining the energy distribution of the signal in frequency, time and scale, thereby describing the dynamic characteristics of the signal in more detail. This method is particularly suitable for analyzing non-stationary signals, such as EEG signals, and can efficiently extract the dynamic change information of the signal, providing important features for subsequent analysis and classification.

[0065] As a possible implementation, frequency domain coherence analysis is performed on different second target signals based on the first distribution data by using coherence spectrum analysis technology to predict functional connections between different brain regions, and obtain a first analysis result, including:

[0066] Step S501, based on the second target signal of different channels and the first distribution data corresponding to the second target signal, a coherence calculation is performed to obtain the target coherence:

[0067] Among them, C ij (f, t) is the target coherence, Sij (f, t) is the cross spectral density between the first distribution data corresponding to the second target signal of channels i and j,

[0068] Among them, S ij (f,t) satisfies: in, For X i(f,t,λ) The conjugate of X i(f,t,λ) is the data intercepted by the time window in the first distribution data, and T is the length of the time window;

[0069] Specifically, coherence calculation is performed based on the second target signal of different channels and its corresponding first distribution data. Coherence measures the linear relationship between signals. Through coherence calculation, the degree of functional connectivity between different brain regions can be effectively evaluated, revealing the collaborative working mode of different regions in EEG signals. Using frequency domain coherence analysis can provide more accurate functional connectivity information than time domain analysis, helping to study the dynamic changes of brain networks.

[0070] Step S502, calculating a weighted coherence index based on the target coherence to characterize the degree of functional connectivity between different brain regions:

[0071] Among them, Γ ij (f) is the weighted coherence index, η(f) is the frequency energy weighting factor, which is used to enhance the coherence of a specific frequency band;

[0072] Specifically, a weighted coherence index is calculated based on the target coherence to further quantify the strength of functional connections between different brain regions. η(f) helps to emphasize the coherence of specific frequency bands, especially in some specific frequency bands, such as alpha waves, beta waves or theta waves, which are particularly important for the functional connection of brain activity. The weighted coherence index can focus on specific frequency bands of the brain by frequency weighting, further improving the ability to detect important EEG activity areas. This method can not only quantify the functional connections between brain regions, but also reveal the contribution of specific frequency bands to functional connections, which helps to deeply understand the impact of different frequency bands on the working mode of the brain.

[0073] Step S503: determining the weighted coherence index and the target coherence as the first analysis result.

[0074] Specifically, by combining the target coherence and weighted coherence index, a richer and more accurate analysis of brain functional connectivity can be provided. This comprehensive analysis not only focuses on the overall coherence, but also emphasizes the role of frequency, thus providing deep insights into the interaction and synergy between different brain regions.

[0075] As a possible implementation, based on the first distribution data combined with nonlinear dynamics theory, the second target signal is subjected to nonlinear feature extraction in the frequency domain to obtain a second analysis result, including:

[0076] Step S601, determining the number of boxes based on the first distribution data, and determining the fractal dimension based on the number of boxes: Wherein, D is the fractal dimension, N(∈) is the number of boxes determined according to the signal energy distribution in the first distribution data;

[0077] Specifically, the extraction of fractal dimension can reveal the self-similarity of EEG signals at different scales and help analyze the complex structure of brain signals. In neuroscience, fractal dimension is closely related to the dynamic activity pattern of the brain and can effectively reflect the complexity of brain activity. The fractal dimension D is calculated by the box counting method. By changing the scale of the box, the self-similarity of the signal at different scales is analyzed, and N(∈) is calculated by selectively covering the points in the high-energy frequency band. By calculating the fractal dimension, the complexity and nonlinear characteristics of the signal can be quantified.

[0078] Step S602: determining the complexity index of the second target signal based on the fractal dimension and the time series entropy of the second target signal: ′ =D·H(X);

[0079] Among them, ′ is the complexity index, H(X) is the time series entropy; it indicates the uncertainty or disorder of the signal. The higher the entropy, the more irregular the signal is, the more difficult it is to predict, and it reflects its more complex dynamic characteristics.

[0080] Specifically, the complexity index combines the fractal dimension and time series entropy of the signal to measure the overall complexity of the signal. By introducing the combination of time series entropy and fractal dimension, the complexity index can more comprehensively evaluate the nonlinear characteristics and dynamic changes of EEG signals. The higher the complexity index, the more complex the dynamic structure of the EEG signal, reflecting a more complex interaction pattern between brain regions.

[0081] Step S603: determine the complexity index as the second analysis result.

[0082] Specifically, the complexity index, as a second analysis result, can provide high-dimensional features for further analysis of brain signals and help to deeply understand the nonlinear dynamic characteristics of signals. This analysis result can not only improve the accuracy of signal classification, but also reveal the functional complexity of different brain regions.

[0083] As a possible implementation, converting the first analysis result and the second analysis result into a frequency domain feature set includes:

[0084] The first analysis result and the second analysis result are converted into a feature vector set to obtain a frequency domain feature set.

[0085] Specifically, the first analysis result (weighted coherence index) obtained in step S501 (coherence spectrum analysis) and the second analysis result (complexity index) obtained in step S602 (nonlinear dynamic feature extraction) are first converted into a set of feature vectors. The weighted coherence index reflects the degree of functional connectivity between different brain regions. Its value can be regarded as a dimension of a multidimensional feature vector, indicating the interaction between brain regions in different frequency bands. The complexity index reflects the dynamic complexity of brain signals, and a feature vector can also be constructed through its value. In the form of feature vectors, these two analysis results are converted into a unified frequency domain feature set for subsequent signal recognition and classification.

[0086] As a possible implementation method, the extracted frequency domain features are optimized and selected to identify the most contributing feature subset. The frequency domain feature set is screened by the feature importance scoring method to obtain the target feature set, including:

[0087] Step S701, using part of the data in the frequency domain feature set as the initial state of the target feature set;

[0088] Specifically, the extracted frequency domain features are represented as a set of feature vectors, F = {f1, f2, ..., f n}; where f i is the i-th eigenvector, and the set contains multiple eigenvectors extracted from the frequency domain analysis, which represent the multi-dimensional information of the EEG signal.

[0089] Step S702, based on the feature importance scoring method, the feature vectors in the frequency domain feature set are calculated based on the scoring function to calculate the target score: R(f i )=α·Corr(f i ,y)-β·Redun(f i ,S);

[0090] In this case, a target feature subset S is initialized, the feature importance scoring method is used, and the scoring function is defined. R(f i ) is the target score, Corr(f i ,y) is the feature vector f i The correlation with the target variable y reflects the impact of this feature on the final recognition result. i ,S) is the feature vector f iThe redundancy with the current state of the target feature set S is aimed at removing redundant features with high similarity and reducing the complexity of the feature set. α is the weight factor for balancing relevance, and β is the weight factor for balancing redundancy.

[0091] Step S703, sorting based on the target score, and updating the target feature set with a preset number of feature vectors;

[0092] Specifically, the feature vectors are sorted based on the calculated target scores, and the feature vector with the highest score is selected to update the target feature set.

[0093] Step S704, recalculate the target score based on the updated target feature set, and update the target feature combination according to the recalculated target score, until the difference in target scores between two adjacent iterations of the same feature vector is less than a second threshold or reaches the maximum number of iterations, and output the target feature set.

[0094] Specifically, the target score is recalculated based on the updated target feature set, and the target feature set is adjusted in the next iteration. Through multiple iterations, the algorithm continuously adjusts the feature subset to ensure the gradual improvement of the feature subset until the optimal or approximately optimal feature subset is found. Furthermore, through the iterative optimization process, the feature subset with the highest fitness is selected. In each iteration, the result of the previous iteration is used to guide the search direction of the next iteration to ensure the correlation between the formulas. This process continues until the change in the target score is less than the set second threshold, or the maximum number of iterations is reached, and the optimal target feature set is finally output. The final optimized frequency domain features are composed of the most contributing feature subsets selected for classification, recognition, and other advanced analysis of EEG signals.

[0095] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for extracting EEG signal features based on frequency domain analysis of the present application will be described in detail below in combination with specific embodiments.

[0096] This embodiment relates to a specific method for extracting features of EEG signals based on frequency domain analysis. In order to verify the beneficial effects of this application, scientific demonstration is carried out through experiments.

[0097] Twenty healthy adults aged between 25 and 35 were selected as subjects. Before the experiment, all subjects signed informed consent and rested in a quiet laboratory to adapt to the environment. The subjects wore a customized high-density EEG acquisition device, which contained 64 channels and recorded EEG activity at a sampling rate of 2,048 Hz. At the same time, synchronous equipment was used to collect physiological data such as heart rate and breathing, as well as environmental noise and vibration information. The experimental data are shown in Table 1.

[0098] Table 1

[0099]

[0100] Under the condition of the same signal-to-noise ratio (20dB), the feature extraction time (1.2s) of the present application is shorter than that of the prior art A (1.5s) and B (1.8s), showing higher processing efficiency. At the same time, the classification accuracy rate (95%) of the present application method is higher than that of the prior art A (90%) and B (85%), indicating that it has higher accuracy in EEG signal classification.

[0101] In terms of feature optimization rate, the method of the present application reaches 85%, while the prior art A and B are only 75% and 70% respectively. This shows that the method of the present application is more effective in optimizing feature subsets and can better screen out features that contribute to classification.

[0102] When the signal-to-noise ratio is reduced to 15 dB, the method of the present application still maintains a high classification accuracy (92%) and feature optimization rate (80%), while the performance of the prior art decreases. This further proves the robustness and superiority of the method of the present application in a noisy environment.

[0103] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0104] The embodiment of the present application also provides an EEG signal feature extraction device based on frequency domain analysis. It should be noted that the EEG signal feature extraction device based on frequency domain analysis in the embodiment of the present application can be used to execute the EEG signal feature extraction method based on frequency domain analysis 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 conceived.

[0105] The following is an introduction to the EEG signal feature extraction device based on frequency domain analysis provided in an embodiment of the present application.

[0106] Figure 3 is a structural block diagram of an EEG signal feature extraction device based on frequency domain analysis according to an embodiment of the present application. Figure 3 As shown, the device includes: a first acquisition unit 10, a first calculation unit 20, a second calculation unit 30, a third calculation unit 40, a fourth calculation unit 50 and a classification unit 60.

[0107] A first acquisition unit 10 is used to acquire an EEG signal to obtain a first target signal, and to perform dynamic filtering based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal;

[0108] Specifically, the EEG signal data is collected, and the complexity of signal processing and analysis caused by the differences in EEG signal characteristics and noise environment of different individuals is solved by real-time learning and adjustment of filtering parameters. Dynamic filtering is performed according to the EEG signal characteristics and noise environment of different individuals to obtain the second target signal. The introduction of dynamic filtering effectively improves the quality of EEG signals and reduces the interference caused by noise. The filtering strategy can adjust the parameters according to the characteristics of the signal, so that the second target signal finally obtained can more accurately reflect the EEG activity.

[0109] The first calculation unit 20 is used to extract the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal and obtain first distribution data;

[0110] Specifically, the dynamic change characteristics of EEG signals in the frequency-time plane are accurately captured to enable effective analysis and accurately evaluate the functional connections between different brain regions, providing key information for subsequent functional connection analysis and nonlinear feature extraction.

[0111] The second calculation unit 30 is used to perform frequency domain coherence analysis on different second target signals based on the first distribution data by using a coherence spectrum analysis technique to predict the functional connections between different brain regions and obtain a first analysis result;

[0112] Specifically, the frequency-time plane data of the second target signal is subjected to frequency domain coherence analysis through coherence spectrum analysis technology. This analysis can reveal the functional connectivity between different brain regions and reveal the coordinated activities of different brain regions in specific frequency bands through coherence spectrum.

[0113] A third calculation unit 40 is used to extract nonlinear features in the frequency domain of the second target signal based on the first distribution data combined with nonlinear dynamics theory to obtain a second analysis result;

[0114] Specifically, based on the frequency domain distribution data and combined with nonlinear dynamics theory, the nonlinear features of the second target signal are extracted. Nonlinear analysis methods can reveal the complex dynamic features hidden in the signal and capture valuable information that traditional linear methods cannot discover.

[0115] The fourth calculation unit 50 is used to convert the first analysis result and the second analysis result into a frequency domain feature set, and screen the frequency domain feature set by a feature importance scoring method to obtain a target feature set.

[0116] Specifically, the analysis results are converted into a frequency domain feature set, and then the feature importance scoring method is applied to screen these features to extract a target feature set with high classification efficiency. This step aims to reduce redundant information and improve the accuracy of subsequent classification tasks through scientific feature selection methods. The screened target feature set contains the most influential information for the classification task, which significantly improves the classification accuracy and efficiency.

[0117] The classification unit 60 is used to identify and classify the EEG signal based on the target feature set.

[0118] Through this embodiment, a device for extracting features of an electroencephalogram signal based on frequency domain analysis is provided, and the device includes: a first acquisition unit, a first calculation unit, a second calculation unit, a third calculation unit, a fourth calculation unit, and a classification unit. The first calculation unit is used to extract the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal and obtain the first distribution data; the first calculation unit is used to extract the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal and obtain the first distribution data; the second calculation unit is used to perform frequency domain coherence analysis on different second target signals based on the first distribution data by coherence spectrum analysis technology to predict the functional connection between different brain regions and obtain the first analysis result; the third calculation unit is used to extract nonlinear features in the frequency domain of the second target signal based on the first distribution data combined with nonlinear dynamics theory to obtain the second analysis result; the fourth calculation unit is used to convert the first analysis result and the second analysis result into a frequency domain feature set, and screen the frequency domain feature set by a feature importance scoring method to obtain a target feature set; the classification unit is used to identify and classify the electroencephalogram signal based on the target feature set. This application solves the problems of insufficient processing of non-stationarity of EEG signals, neglect of time-varying characteristics, and low feature extraction accuracy and computational efficiency in the prior art by optimizing the frequency domain analysis algorithm, introducing dynamic filtering, frequency domain coherence analysis and nonlinear dynamics theory.

[0119] As a possible implementation, the first acquisition unit includes: a first acquisition module, a second acquisition module, a first analysis module, a building module and a second analysis module.

[0120] The first acquisition module is used to collect historical data of EEG signals of different target objects through electrodes to obtain a third target signal;

[0121] A second acquisition module is used to acquire physiological parameter data of the target object to obtain first target data, acquire environmental sound through a microphone to obtain environmental noise data, and acquire vibration data through an accelerometer to obtain second target data;

[0122] Specifically, the electrode patch can accurately acquire the electrical activity of the brain. These signals can provide information on neural activity in various areas of the brain as the basis for subsequent analysis. The electrode patch collects voltage time series, physiological parameter data (such as heart rate, body temperature, etc.) recorded by scalp electrodes on different target objects, and environmental sounds (such as background noise) and vibration data collected by microphones and accelerometers. These data, as auxiliary information, can provide additional background information about the target object and its surrounding environment, making the analysis of EEG signals more accurate and reliable, especially in noisy environments.

[0123] A first analysis module is used to perform cluster analysis based on the third target signal, the first target data, the environmental noise data, and the second target data to obtain different categories of EEG signals and obtain multiple EEG signal groups;

[0124] Specifically, a feature model is established for each target object through preliminary EEG signal analysis. Cluster analysis is used to comprehensively analyze the third target signal (historical EEG signal data), the first target data (physiological parameter data), the environmental noise data, and the second target data (vibration data) to obtain EEG signals of different categories. Especially when the target object is in different physiological states or environmental conditions, cluster analysis helps to clearly divide the different categories of EEG signals, thereby providing an accurate signal classification basis for subsequent filtering processing.

[0125] Establishing a module for establishing a feature model for each EEG signal group and initializing filtering parameters corresponding to different categories based on the feature model;

[0126] Specifically, the feature model can help describe and summarize the typical characteristics of each category of signals, and the initialization of the filter parameters is adjusted according to these feature models. Establishing a feature model for each EEG signal group (each category) can accurately capture the unique properties of each type of EEG signal, thereby optimizing the parameter settings of the filter. Through reasonable initialization of filter parameters, it can ensure targeted processing of each signal category, minimize noise interference, and improve filtering effect.

[0127] Let Θ i is the filter parameter set for the i-th target object, initialized to: Θ i ={θ i1 ,θ i2 ,…,θ in}; where θ ij is the jth filter parameter for the i-th type of target object.

[0128] Monitor the characteristic changes of EEG signals in real time, and dynamically adjust the filtering parameters according to the real-time tracking characteristics to adapt to the current state of the target object.

[0129] The second analysis module is used to perform cluster analysis on the first target signal to obtain a target category corresponding to the first target signal, determine a corresponding target filtering parameter based on the target category, and filter the first target signal based on the target filtering parameter to obtain a second target signal.

[0130] Specifically, cluster analysis is performed on the first target signal to obtain the target category to which it belongs. Based on the category, the corresponding target filter parameters are selected, and the first target signal is filtered to obtain the second target signal. The filtered signal will remove unnecessary noise and retain the EEG signals that are valuable for analysis. By associating EEG signals with specific categories and applying corresponding filter parameters, more accurate noise suppression and signal enhancement can be achieved.

[0131] As a possible implementation, the first calculation unit includes: a standard deviation module, a windowing module and a transformation module.

[0132] A standard deviation module is used to determine the standard deviation of the second target signal based on the second target signal, and determine the scale parameter of the multi-scale analysis window based on the standard deviation: Where λ is the scale parameter, σ(t) is the standard deviation, and ∈ is a value used to avoid division by 0;

[0133] A windowing module, used for performing windowing on the second target signal based on the scale parameter to obtain a third target signal;

[0134] Specifically, extracting the energy distribution of the signal on the frequency-time plane includes automatically selecting a suitable multi-scale analysis window based on the waveform characteristics of the EEG signal to capture frequency components of different scales. It can be seen that the scale-adaptive windowing method can ensure more accurate local analysis of the signal within different frequency and time ranges, thereby optimizing subsequent frequency domain analysis and feature extraction.

[0135] A transformation module is used to perform an adaptive recursive wavelet transform based on the third target signal to extract the instantaneous frequency and amplitude of the third target signal to obtain the first distribution data: E(f, t, λ) = |ARWT(x w (k),λ,t)| 2 ;

[0136] Wherein, E(f, t, λ) is the first distribution data, which is used to characterize the three-dimensional distribution of the third target signal in energy-frequency-time. w (k),λ,t) satisfies:

[0137]

[0138] Among them, x w (k) is the third target signal, is a parameter used to limit the range of wavelet basis functions, ψ λ (t) is the adaptive phase function, and j is the imaginary unit.

[0139] Specifically, adaptive complex modulation wavelet transform is applied to each windowed signal segment to extract the instantaneous frequency and amplitude of the signal and obtain the energy-frequency-time three-dimensional distribution E(f, t, λ). Adaptive recursive wavelet transform can provide high time-frequency resolution and accurately capture the changes in the instantaneous frequency and amplitude of the signal.

[0140] As a possible implementation, the second calculation unit includes: a first calculation module, a second calculation module and a first result module.

[0141] The first calculation module is used to perform coherence calculation based on the second target signal of different channels and the first distribution data corresponding to the second target signal to obtain the target coherence:

[0142] Among them, C ij (f, t) is the target coherence, S ij (f, t) is the cross spectral density between the first distribution data corresponding to the second target signal of channels i and j,

[0143] Among them, S ij (f,t) satisfies: in, For X i(f,t,λ) The conjugate of X i(f,t,λ) is the data intercepted by the time window in the first distribution data, and T is the length of the time window;

[0144] Specifically, coherence calculation is performed based on the second target signal of different channels and its corresponding first distribution data. Coherence measures the linear relationship between signals. Through coherence calculation, the degree of functional connectivity between different brain regions can be effectively evaluated, revealing the collaborative working mode of different regions in EEG signals. Using frequency domain coherence analysis can provide more accurate functional connectivity information than time domain analysis, helping to study the dynamic changes of brain networks.

[0145] The second calculation module is used to calculate the weighted coherence index based on the target coherence to characterize the degree of functional connectivity between different brain regions:

[0146] Among them, Γ ij (f) is the weighted coherence index, η(f) is the frequency energy weighting factor, which is used to enhance the coherence of a specific frequency band;

[0147] Specifically, a weighted coherence index is calculated based on the target coherence to further quantify the strength of functional connections between different brain regions. η(f) helps to emphasize the coherence of specific frequency bands, especially in some specific frequency bands, such as alpha waves, beta waves or theta waves, which are particularly important for the functional connection of brain activity. The weighted coherence index can focus on specific frequency bands of the brain by frequency weighting, further improving the ability to detect important EEG activity areas. This method can not only quantify the functional connections between brain regions, but also reveal the contribution of specific frequency bands to functional connections, which helps to deeply understand the impact of different frequency bands on the working mode of the brain.

[0148] The first result module is used to determine the weighted coherence index and the target coherence as the first analysis result.

[0149] Specifically, by combining the target coherence and weighted coherence index, a richer and more accurate analysis of brain functional connectivity can be provided. This comprehensive analysis not only focuses on the overall coherence, but also emphasizes the role of frequency, thus providing deep insights into the interaction and synergy between different brain regions.

[0150] As a possible implementation, the third calculation unit includes: a fractal dimension module, a complexity module and a second result module.

[0151] A fractal dimension module, used to determine the number of boxes based on the first distribution data, and to determine the fractal dimension based on the number of boxes: Wherein, D is the fractal dimension, N(∈) is the number of boxes determined according to the signal energy distribution in the first distribution data;

[0152] Specifically, the box counting method is used to calculate the fractal dimension D. By changing the scale of the box, the self-similarity of the signal at different scales is analyzed, and N(∈) is calculated by selectively covering the points in the high-energy frequency band. By calculating the fractal dimension, the complexity and nonlinear characteristics of the signal can be quantified.

[0153] A complexity module, for determining a complexity index of the second target signal based on the fractal dimension and the time series entropy of the second target signal: Ξ ′ =D·H(X);

[0154] Among them, ′ is the complexity index, H(X) is the time series entropy; it indicates the uncertainty or disorder of the signal. The higher the entropy, the more irregular the signal is, the more difficult it is to predict, and it reflects its more complex dynamic characteristics.

[0155] Specifically, by introducing a combination of time series entropy and fractal dimension, the complexity index can more comprehensively evaluate the nonlinear characteristics and dynamic changes of EEG signals. The higher the complexity index, the more complex the dynamic structure of the EEG signal, reflecting a more complex interaction pattern between brain regions.

[0156] The second result module is used to determine the complexity index as the second analysis result.

[0157] Specifically, the complexity index, as a second analysis result, can provide high-dimensional features for further analysis of brain signals and help to deeply understand the nonlinear dynamic characteristics of signals. This analysis result can not only improve the accuracy of signal classification, but also reveal the functional complexity of different brain regions.

[0158] As a possible implementation, the fourth calculation unit includes: a conversion module.

[0159] The conversion module is used to convert the first analysis result and the second analysis result into a feature vector set to obtain a frequency domain feature set.

[0160] Specifically, the weighted coherence index reflects the degree of functional connectivity between different brain regions. Its value can be regarded as a dimension of a multidimensional feature vector, indicating the interaction between brain regions in different frequency bands. The complexity index reflects the dynamic complexity of brain signals, and a feature vector can also be constructed through its value. In the form of feature vectors, these two analysis results are converted into a unified frequency domain feature set for subsequent signal recognition and classification.

[0161] As a possible implementation, the fourth calculation unit further includes: an initial module, a scoring module, a sorting module and an iteration module.

[0162] The initial module is used to use part of the data in the frequency domain feature set as the initial state of the target feature set;

[0163] Specifically, the extracted frequency domain features are represented as a set of feature vectors, F = {f1, f2, ..., f n}; where f i is the i-th eigenvector, and the set contains multiple eigenvectors extracted from the frequency domain analysis, which represent the multi-dimensional information of the EEG signal.

[0164] The scoring module is used to calculate the target score based on the scoring function of the feature vector in the frequency domain feature set based on the feature importance scoring method: R(f i )=α·Corr(f i ,y)-β·Redun(f i ,S);

[0165] In this case, a target feature subset S is initialized, the feature importance scoring method is used, and the scoring function is defined. R(f i ) is the target score, Corr(f i ,y) is the feature vector f i The correlation with the target variable y reflects the impact of this feature on the final recognition result. i ,S) is the feature vector f i The redundancy with the current state of the target feature set S is aimed at removing redundant features with high similarity and reducing the complexity of the feature set. α is the weight factor for balancing relevance, and β is the weight factor for balancing redundancy.

[0166] A sorting module, used to sort based on the target score, and a preset number of feature vectors are used to update the target feature set;

[0167] Specifically, the feature vectors are sorted based on the calculated target scores, and the feature vector with the highest score is selected to update the target feature set.

[0168] The iterative module is used to recalculate the target score based on the updated target feature set, and update the target feature combination according to the recalculated target score, until the difference in target scores between two consecutive iterations of the same feature vector is less than a second threshold or the maximum number of iterations is reached, and output the target feature set.

[0169] Specifically, the target score is recalculated based on the updated target feature set, and the target feature set is adjusted in the next iteration. Through multiple iterations, the algorithm continuously adjusts the feature subset, ensuring the gradual improvement of the feature subset until the optimal or nearly optimal feature subset is found.

[0170] The above-mentioned device for extracting features of EEG signals based on frequency domain analysis includes a processor and a memory. The above-mentioned first acquisition unit, first calculation unit, second calculation unit, third calculation unit, fourth calculation unit and classification unit 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 the form of any combination.

[0171] 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.

[0172] 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.

[0173] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the EEG signal feature extraction method based on frequency domain analysis.

[0174] The present application also provides an EEG analysis 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 feature extraction method based on frequency domain analysis.

[0175] 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.

[0176] 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.

[0177] 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.

[0178] 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.

[0179] 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.

[0180] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0181] 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.

[0182] 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.

[0183] 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.

[0184] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0185] 1) The EEG signal feature extraction method based on frequency domain analysis of the present application collects signal data, optimizes filtering according to the EEG signal characteristics and noise environment of different individuals through real-time learning and adjustment of filtering parameters; extracts the energy distribution of the signal on the frequency-time plane, reveals the dynamic change characteristics of the signal, uses coherence spectrum analysis technology to identify the frequency domain coherence between different EEG signal channels, evaluates the functional connection between brain regions, combines nonlinear dynamics theory to extract the nonlinear characteristics of frequency domain signals, and analyzes the complexity and chaotic nature of EEG signals; optimizes the extracted frequency domain features and identifies the most contributing feature subsets; outputs the optimized frequency domain feature set for classification, recognition and other advanced analysis of EEG signals. This application optimizes the frequency domain analysis algorithm, introduces dynamic filtering, frequency domain coherence analysis and nonlinear dynamics theory. Dynamic filtering overcomes the defect that traditional methods may lose key information when processing non-stationary signals. Frequency domain coherence analysis and nonlinear feature extraction deeply explore the time-varying characteristics and nonlinear laws of signals, and the feature importance scoring rule improves the efficiency and discrimination ability of feature screening, thereby comprehensively improving the accuracy and efficiency of EEG signal analysis, thereby providing more reliable technical support for the research and application of EEG signals. Compared with the existing technology, it may have significant advantages in processing non-stationary signals, extracting time-varying features and reducing computing costs. It solves the problems of insufficient processing of non-stationary EEG signals, ignoring time-varying characteristics, and low accuracy and computing efficiency of feature extraction in the existing technology.

[0186] 2) The device for extracting EEG signal features based on frequency domain analysis of the present application. The device includes: a first acquisition unit, a first calculation unit, a second calculation unit, a third calculation unit, a fourth calculation unit and a classification unit. The first calculation unit is used to extract the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal and obtain the first distribution data; the first calculation unit is used to extract the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal and obtain the first distribution data; the second calculation unit is used to perform frequency domain coherence analysis on different second target signals based on the first distribution data by coherence spectrum analysis technology to predict the functional connection between different brain regions and obtain the first analysis result; the third calculation unit is used to extract nonlinear features in the frequency domain of the second target signal based on the first distribution data combined with nonlinear dynamics theory to obtain the second analysis result; the fourth calculation unit is used to convert the first analysis result and the second analysis result into a frequency domain feature set, and screen the frequency domain feature set by feature importance scoring method to obtain a target feature set; the classification unit is used to identify and classify the EEG signal based on the target feature set. This application solves the problems of insufficient processing of non-stationarity of EEG signals, neglect of time-varying characteristics, and low feature extraction accuracy and computational efficiency in the prior art by optimizing the frequency domain analysis algorithm, introducing dynamic filtering, frequency domain coherence analysis and nonlinear dynamics theory.

[0187] The above description is only the preferred embodiment of the present application and is 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 analyzing EEG signals based on frequency domain analysis, characterized in that: include: Acquire an EEG signal to obtain a first target signal, and perform dynamic filtering based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal; Extracting energy distribution of the second target signal on a frequency-time plane to determine dynamic change characteristics of the second target signal and obtain first distribution data; Using a coherence spectrum analysis technique, based on the first distribution data, frequency domain coherence analysis is performed on different second target signals to predict functional connections between different brain regions, thereby obtaining a first analysis result; Based on the first distribution data and nonlinear dynamics theory, nonlinear feature extraction is performed on the second target signal in the frequency domain to obtain a second analysis result; Converting the first analysis result and the second analysis result into a frequency domain feature set, and screening the frequency domain feature set by a feature importance scoring method to obtain a target feature set; The EEG signal is identified and classified based on the target feature set.

2. The method according to claim 1, characterized in that Acquiring an EEG signal to obtain a first target signal, and performing dynamic filtering based on the signal characteristics of the first target signal and the noise environment to obtain a second target signal, including: Collecting historical data of EEG signals of different target objects through electrodes to obtain a third target signal; Collecting physiological parameter data of the target object to obtain first target data, collecting environmental sound through a microphone to obtain environmental noise data, and collecting vibration data through an accelerometer to obtain second target data; Performing cluster analysis based on the third target signal, the first target data, the environmental noise data, and the second target data to obtain different categories of the EEG signals and obtain multiple EEG signal groups; Establishing a feature model for each of the EEG signal groups, and initializing filtering parameters corresponding to different categories based on the feature model; Performing cluster analysis on the first target signal to obtain a target category corresponding to the first target signal, determining a corresponding target filtering parameter based on the target category, and filtering the first target signal based on the target filtering parameter to obtain the second target signal.

3. The method according to claim 2, characterized in that Extracting the energy distribution of the second target signal on the frequency-time plane to determine the dynamic change characteristics of the second target signal to obtain first distribution data, including: Based on the second target signal, a standard deviation of the second target signal is determined, and a scale parameter of the multi-scale analysis window is determined based on the standard deviation: Wherein, λ is the scale parameter, σ(t) is the standard deviation, and ∈ is a value used to avoid division by 0; Windowing the second target signal based on the scale parameter to obtain a third target signal; Performing an adaptive recursive wavelet transform based on the third target signal to extract the instantaneous frequency and amplitude of the third target signal to obtain the first distribution data: E(f,t,λ)=|ARWT(x w (k),λ,t)| 2 ; Wherein, E(f, t, λ) is the first distribution data, which is used to characterize the three-dimensional distribution of the third target signal in energy-frequency-time, wherein ARWT(x w (k),λ,t) satisfies: Among them, x w (k) is the third target signal, is a parameter used to limit the range of wavelet basis functions, ψ λ (t) is the adaptive phase function, and j is the imaginary unit.

4. The method according to claim 3, characterized in that: By using the coherence spectrum analysis technology, frequency domain coherence analysis is performed on different second target signals based on the first distribution data to predict the functional connections between different brain regions, and a first analysis result is obtained, including: Based on the second target signals of different channels and the first distribution data corresponding to the second target signals, coherence calculation is performed to obtain target coherence: Among them, C ij (f, t) is the target coherence, S ij (f, t) is the cross spectral density between the first distribution data corresponding to the second target signal of channels i and j, where S ij (f,t) satisfies: in, For X i(f,t,λ) The conjugate of X i(f,t,λ) is the data intercepted by the time window in the first distribution data, and T is the length of the time window; A weighted coherence index is calculated based on the target coherence to characterize the degree of functional connectivity between different brain regions: Among them, Γ ij (f) is the weighted coherence index, and η(f) is a frequency energy weighting factor used to enhance the coherence of a specific frequency band; The weighted coherence index and the target coherence are determined as the first analysis result.

5. The method according to claim 4, characterized in that Based on the first distribution data combined with nonlinear dynamics theory, nonlinear features of the second target signal are extracted in the frequency domain to obtain a second analysis result, including: Determine the number of boxes based on the first distribution data, and determine the fractal dimension based on the number of boxes: Wherein, D is the fractal dimension, and N(∈) is the number of boxes determined according to the signal energy distribution in the first distribution data; Determine a complexity index of the second target signal based on the fractal dimension and the time series entropy of the second target signal: X ′ =D·H(X); Among them, ′ is the complexity index, H(X) is the time series entropy; The complexity index is determined as the second analysis result.

6. The method according to claim 5, characterized in that Converting the first analysis result and the second analysis result into a frequency domain feature set includes: The first analysis result and the second analysis result are converted into a feature vector set to obtain the frequency domain feature set.

7. The method according to claim 6, characterized in that The frequency domain feature set is screened by a feature importance scoring method to obtain a target feature set, including: Using part of the data in the frequency domain feature set as the initial state of the target feature set; Based on the feature importance scoring method, the feature vector in the frequency domain feature set is based on the scoring function to calculate the target score: R(f i )=α·Corr(f i ,y)-β·Redun(f i ,S); Among them, R(f i ) is the target score, Corr(f i ,y) is the feature vector f i Correlation with the target variable y, Redun(f i , S) is the feature vector f i The redundancy with the current state of the target feature set S, α is a weight factor for balancing relevance, and β is a weight factor for balancing redundancy; Sorting based on the target score, a previously preset number of the feature vectors update the target feature set; The target score is recalculated based on the updated target feature set, and the target feature combination is updated according to the recalculated target score, until the difference in the target score between two adjacent iterations of the same feature vector is less than a second threshold or reaches a maximum number of iterations, and the target feature set is output.

8. A device for extracting features of electroencephalogram signals based on frequency domain analysis, characterized in that: The device comprises: A first acquisition unit is used to acquire an EEG signal to obtain a first target signal, and dynamically filter the first target signal based on the signal characteristics and noise environment of the first target signal to obtain a second target signal; a first calculation unit, configured to extract energy distribution of the second target signal on a frequency-time plane to determine a dynamic change feature of the second target signal and obtain first distribution data; a second computing unit, configured to perform frequency domain coherence analysis on different second target signals based on the first distribution data by using a coherence spectrum analysis technique, so as to predict the functional connections between different brain regions and obtain a first analysis result; A third calculation unit, configured to extract nonlinear features in the frequency domain of the second target signal based on the first distribution data in combination with nonlinear dynamics theory to obtain a second analysis result; a fourth computing unit, configured to convert the first analysis result and the second analysis result into a frequency domain feature set, and screen the frequency domain feature set by a feature importance scoring method to obtain a target feature set; A classification unit is used to identify and classify the EEG signal based on the target feature set.

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 analysis 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