EEG signal processing method and brain-computer interaction system
By preprocessing EEG signals and screening effective signal segments, combined with the grey constrained dynamic programming algorithm and sliding window features, the problems of time-varying and nonlinear EEG signals are solved, achieving more accurate EEG signal evaluation and stable brain-computer interaction.
Patent Information
- Application Number
- CN202511072619.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-01
- Publication Date
- 2025-09-30
- Estimated Expiration
- 2045-08-01
AI Technical Summary
The time-varying and nonlinear nature of EEG signals leads to random errors after measurement, affecting the reliability and stability of brain-computer interaction. Existing analysis methods make it difficult to effectively select personalized EEG signal segments for analysis.
By collecting EEG signals of the target object in different states, preprocessing and filtering and denoising are performed, and the judgment threshold is determined by using the grey constrained dynamic programming algorithm and sliding window features, the effective signal segments are screened, and the feature vectors are extracted for cluster analysis to evaluate the EEG data status.
The accuracy of EEG signal evaluation is improved, individual and scene interference is reduced, and a more reliable and stable brain-computer interaction system is achieved.
Smart Images
Figure CN120549514B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of bioelectric signals and electroencephalogram (EEG) signals, and in particular to an EEG signal processing method and a brain-computer interaction system. Background Art
[0002] With the development of brain-computer interaction technology, EEG signals collected from the brain that reflect neural activity will be transmitted to a decoder for translation, making action outputs such as controlling a computer cursor, controlling wheelchair movement, or controlling a robotic arm a reality.
[0003] Since the human brain is an open, time-varying and nonlinear system, the EEG signals that carry information about the characteristics or state of the human brain are also time-varying, nonlinear and relatively weak. Due to interference, EEG data signals will produce random errors after measurement, affecting the reliability and stability of brain-computer interaction. Therefore, how to obtain effective EEG data signals is the key to the application of brain-computer interaction technology.
[0004] Existing EEG signal analysis methods primarily include time-domain analysis, frequency-domain analysis, time-frequency analysis, multi-scale analysis, spatiotemporal analysis, and nonlinear dynamics analysis. These methods often use timestamps to capture EEG segments for analysis. However, due to individual differences and random external interference during acquisition, the effectiveness of these segments is significantly affected. Therefore, intelligently selecting valid EEG signal segments for analysis is crucial to the reliability and stability of brain-computer interaction systems. Summary of the Invention
[0005] In view of this, in order to at least partially solve at least one of the above-mentioned technical problems, the present invention provides an EEG signal processing method and a brain-computer interaction system, the technical solutions of which are as follows:
[0006] According to an embodiment of one aspect of the present invention, there is provided a method for processing EEG signals, comprising: collecting EEG signals of a target object in different states to obtain initial EEG data; selecting an EEG data segment from the initial EEG data and preprocessing it to obtain a preprocessed EEG data segment; screening based on the preprocessed EEG data segment to obtain a valid signal segment; intercepting the selected EEG data segment based on the valid signal segment to obtain a valid EEG information segment; extracting eigenvalues of the valid signal segment and the valid EEG information segment to form a eigenvector; and performing cluster analysis on the EEG data based on the eigenvector to evaluate the EEG signal and obtain an EEG data state classification result, so that an external device can be controlled to perform a specific task action based on the EEG data state classification result.
[0007] According to an embodiment of the present invention, collecting EEG signals of the target object in different states to obtain initial EEG data includes collecting EEG signals in a resting state to obtain EEG data in a resting state and collecting EEG signals in a task execution state to obtain EEG data in a task state; the resting state is the basic state when no task is executed; the task execution state is the state when the target is executing a specific task.
[0008] According to an embodiment of the present invention, selecting an EEG data segment from the initial EEG data and preprocessing it to obtain a preprocessed EEG data segment includes: denoising the EEG data segment, filtering it to remove interference, and performing dimensionless processing.
[0009] According to an embodiment of the present invention, screening the EEG data segments based on the preprocessed EEG data segments to obtain valid signal segments includes: determining sliding window characteristics; determining a judgment threshold; and screening the EEG data segments based on a gray constrained dynamic programming algorithm and the determined sliding window characteristics and judgment threshold to obtain valid signal segments.
[0010] According to an embodiment of the present invention, determining the sliding window characteristics includes: determining the sliding window according to the EEG data sampling frequency and the characteristics of the task.
[0011] According to an embodiment of the present invention, determining the judgment threshold includes: establishing a training set, and using the training set to obtain the judgment threshold of the valid signal segment.
[0012] According to an embodiment of the present invention, a training data set is established based on preprocessed EEG data segments, resting EEG data, and task state EEG data, and a judgment threshold is determined based on the grey correlation coefficient between the preprocessed EEG data segments and the resting EEG data, and the grey correlation coefficient between the preprocessed EEG data segments and the task state EEG data, and the valid signal segment is determined through the judgment threshold.
[0013] According to an embodiment of the present invention, the judgment threshold is updated in real time based on the effective signal segment and the EEG data in the resting state.
[0014] According to an embodiment of the present invention, intercepting the selected EEG data segment according to the valid signal segment to obtain the valid EEG information segment includes: aligning the valid signal segment in the EEG data segment to obtain the valid EEG information segment.
[0015] According to another embodiment of the present invention, there is provided a brain-computer interaction system that applies the above-mentioned EEG signal processing method, comprising an acquisition module, a preprocessing module, a screening module, an interception module, a feature extraction module, and an evaluation module. Among them, the acquisition module is used to acquire EEG signals of different states of the target object to obtain initial EEG data; the preprocessing module is used to select an EEG data segment from the initial EEG data and perform preprocessing to obtain a preprocessed EEG data segment; the screening module is used to perform screening based on the preprocessed EEG data segment to obtain a valid signal segment; the interception module is used to intercept the selected EEG data segment based on the valid signal segment to obtain a valid EEG information segment; the feature extraction module is used to extract the eigenvalues of the valid signal segment and the valid EEG information segment to form a eigenvector; the evaluation module is used to perform cluster analysis on the EEG data based on the eigenvector to realize the evaluation of the EEG signal. The brain-computer interaction system is used to output the EEG data state classification result based on the evaluation result of the target EEG signal, and control the external device to perform a specific task action based on the EEG data state classification result. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:
[0017] Figure 1 Flowchart of an electrical signal processing method according to an embodiment of the present invention.
[0018] Figure 2 Schematic diagram of a brain-computer interaction system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0019] The present invention provides an EEG signal processing method and a brain-computer interaction system, which obtains effective signal segments by reducing the personalization and external interference of EEG data and combining the signal timing analysis method. Then, the effective signal segments and the eigenvalues in the corresponding EEG data are extracted and screened, and a multi-dimensional feature vector is constructed to enhance the homogeneity of the eigenvectors of the effective signals, thereby reducing the interference caused by cross-individual and cross-scenario interference, more accurately evaluating EEG signals and controlling peripherals, and ultimately laying a solid foundation for achieving reliable and stable brain-computer interaction.
[0020] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.
[0021] In the embodiment of the present invention, Figure 1 As shown, a method for processing an electroencephalogram signal is provided, including operations S1-S6:
[0022] S1: Collect EEG signals of the target subject in different states to obtain initial EEG data;
[0023] S2: selecting an EEG data segment from the initial EEG data and performing preprocessing to obtain a preprocessed EEG data segment;
[0024] S3: Filtering the pre-processed EEG data segments to obtain valid signal segments;
[0025] S4: intercepting the selected EEG data segment according to the valid signal segment to obtain a valid EEG information segment;
[0026] S5: extracting the eigenvalues of the valid signal segment and the valid EEG information segment to form a eigenvector; and
[0027] S6: Based on the feature vector, cluster analysis is performed on the EEG data to evaluate the EEG signal and obtain EEG data state classification results, so that the external device can be controlled to perform specific task actions based on the EEG data state classification results.
[0028] According to an embodiment of the present invention, operation S1 includes collecting EEG signals in a resting state to obtain EEG data in a resting state and collecting EEG signals in a task execution state to obtain EEG data in a task execution state; the resting state is the basic state when no task is executed.
[0029] The resting state can be, for example, a lying down and relaxing state; the task execution state is the state when the target performs a specific task, such as executing limb movements, including controlling hand grasping, moving objects, and controlling legs to move forward, backward, squat, stand, etc.
[0030] According to an embodiment of the present invention, operation S2 includes performing denoising, filtering and other processing on the selected initial EEG data segment to remove interference signals, and then further performing dimensionless processing to obtain a preprocessed EEG data segment.
[0031] According to an embodiment of the present invention, in operation S3, screening is performed using a sliding window method, including: determining sliding window characteristics; determining a judgment threshold; and finally screening the entire preprocessed EEG data segment based on a gray constrained dynamic programming algorithm and the determined sliding window characteristics and judgment threshold to obtain a valid signal segment.
[0032] The sliding window can be determined by the data sampling frequency of the EEG signal or EEG data and the type or characteristics of the task. For example, the window length and step size of the sliding window can be determined. The duration and frequency range of the EEG signal corresponding to different tasks will be different, so the characteristics of the sliding window will be adjusted accordingly. The sliding window is determined by the EEG data sampling frequency and the characteristics of the task, so that the sliding window can cover at least one cycle of the task execution to avoid being too long or too short.
[0033] When determining the judgment threshold, a training set is established and the judgment threshold of the valid signal segment is obtained using the training set. For example, the grey correlation coefficient is used as the judgment threshold. A training data set is established based on the pre-processed EEG data segment, the resting state EEG data, and the task state EEG data, and the judgment threshold is determined based on the grey correlation coefficient between the pre-processed EEG data segment and the resting state EEG data, and the grey correlation coefficient between the pre-processed EEG data segment and the task state EEG data, and the valid signal segment is determined by the judgment threshold. Furthermore, the judgment threshold can be adaptively updated in real time based on the valid signal segment and different resting state EEG data.
[0034] According to an embodiment of the present invention, operation S5 includes: aligning the valid signal segments within the EEG data segments (the EEG data from the selected initial EEG data segments that have been denoised and filtered to remove interference signals but have not been dimensionlessly processed) to obtain valid EEG information segments for use in extracting energy distribution eigenvalues and Lempel-Ziv complexity eigenvalues. The eigenvalues of the valid signal segments and the valid EEG information segments are extracted and combined with the grey correlation coefficient to form an eigenvector matrix, thereby performing cluster analysis on the EEG data to evaluate the EEG signals and obtain EEG data state classification results. Based on the EEG data state classification results, an external device can be controlled to perform specific task actions.
[0035] Example 1
[0036] First, the EEG data collected during resting state training Use ICA (Independent Component Analysis) and other algorithms to filter out interference signals such as electrooculography and electromyography , and use the mean normalization dimensionless processing to obtain the dimensionless clean EEG signal , determined by the type of task;
[0037] ;
[0038] ;
[0039] ;
[0040] in, is n sampling points in the EEG data, is the i-th sampling point, n is the number of EEG data points within a specific time length, is the average value of each sampling point, is the standard deviation of EEG data in the resting state, then:
[0041]
[0042]
[0043] In real-time tasks, resting state EEG data can be Updates to reduce interference caused by different testers, collection devices, or wearing errors are as follows:
[0044] ;
[0045] in, is the k-th resting state EEG data, The currently collected resting EEG signal, the updated EEG data is ; For the forgetting factor, generally
[0046] In the training data, the EEG data segment collected during the task execution state As the task status master signal, the preprocessing process is similar, using ICA and other algorithms to filter out interference signals such as electrooculography and electromyography. , and use the mean normalization dimensionless processing to obtain dimensionless clean preprocessed EEG data segments ; In real-time tasks, the pre-processed EEG data segments under task status Updates to reduce interference caused by different testers, collection devices, or wearing errors are as follows:
[0047] ;
[0048] in, is the pre-processed EEG data segment updated for the k+1th time, For the forgetting factor, generally , is the current pre-processed EEG data segment, Valid data after alignment for constraint dynamic programming;
[0049] If the tester needs to perform the task state, such as the kth task state, collect EEG data segments For the initial EEG data, use ICA and other algorithms to filter out interference signals such as electrooculogram and electromyography , and use the mean normalization dimensionless processing to obtain dimensionless clean preprocessed EEG data ; After the task tag is started, the signal segment is obtained by using the constrained dynamic programming algorithm is the effective signal segment, and the sliding window is It is determined according to the sampling rate of the EEG signal or EEG data and the characteristics of different tasks. For example, the length of the master signal of the collected EEG data is generally set to 0.1s.
[0050] Set constraints on the allowed path length for dynamic programming ,but ; Task state preprocessing EEG data segment The length is m, and the EEG data segment selected for the kth task state The data length is n; construct the distance matrix of two EEG data segments , , for matrix;
[0051] ;
[0052] ;
[0053] ;
[0054] Among them, i and j represent the row and column numbers of the distance matrix H respectively. , Represents the distance matrix H in The elements at represents the i-th EEG data point in the pre-processed EEG data segment under the task state, Represents the EEG data point of the kth task state; represents the preprocessed EEG data segment, represents the data points of the preprocessed EEG data segment, Represents the pre-processed EEG data segment under task state, Represents a data point of a preprocessed EEG data segment under task state.
[0055] Create the cumulative cost matrix , for Matrix, initialization starting point ;other is infinite; for every possible path length , and each location ,calculate:
[0056] ;
[0057] Among them, i and j represent the row and column numbers of the distance matrix H respectively. , represents the path length, L represents the maximum path length, , I∈[max(m,n), L], Indicates the creation of the specified three-dimensional element of the cumulative cost matrix, Represents the distance from template i to data group j in EEG data segment. When the path length , but Does not exist.
[0058] In all paths that meet the minimum path length and do not exceed the maximum path length In the path, find the value with the minimum cumulative cost, backtrack the cumulative cost matrix, and find the optimal alignment path.
[0059] According to the time coordinate corresponding to the selected optimal alignment path, the dimensionless clean pre-processed EEG data segment Filter out the signal segment data that needs to be compared , ;beg and the kth resting state EEG data Grey correlation coefficient :
[0060] ;
[0061] ;
[0062] in, represents the absolute difference sequence, is the minimum value, , , .
[0063] To compare signal segment data and task state EEG data The grey relational coefficient of , then it is considered necessary to compare the signal segment data is the effective signal segment. At the same time, the EEG data segment with the interference signal removed is The time coordinate selection corresponds to the valid EEG information segment .
[0064] ;
[0065] when hour, .
[0066] like , it is believed that this segment needs to be compared with the signal segment data It is interference data and should be discarded. for and Grey correlation coefficient of . Solve the effective EEG information segment The energy distribution eigenvalue, Lempel-Ziv complexity eigenvalue, and grey correlation coefficient Forming a eigenvector (eigenvector matrix), the eigenvector matrix is weighted or clustered according to the application system to realize the evaluation of the EEG signal, or outputting the EEG data state classification result according to the evaluation result of the target EEG signal, and controlling the external device to perform a specific task action based on the EEG data state classification result.
[0067] Another embodiment of the present invention further provides a brain-computer interaction system using the above EEG signal processing method, which is used to output EEG data state classification results based on the results of evaluating the target EEG signal, or to control an external device to perform a specific task action. Figure 2 As shown, the brain-computer interaction system includes:
[0068] An acquisition module is used to acquire EEG signals of the target object in different states to obtain initial EEG data;
[0069] A preprocessing module, configured to select an EEG data segment from the initial EEG data and perform preprocessing to obtain a preprocessed EEG data segment;
[0070] A screening module, used for screening and obtaining effective signal segments based on preprocessed EEG data segments;
[0071] An interception module is used to intercept the selected EEG data segment according to the valid signal segment to obtain a valid EEG information segment;
[0072] A feature extraction module is used to extract the eigenvalues of the effective signal segment and the effective EEG information segment to form a feature vector; and
[0073] The evaluation module performs cluster analysis on EEG data based on feature vectors to evaluate EEG signals.
[0074] The brain-computer interaction system is used to output an EEG data state classification result based on the evaluation result of the target's EEG signal, and control an external device to perform a specific task action based on the EEG data state classification result. The external device connects to the target through a wired or wireless connection to achieve signal connection.
[0075] In order to verify the accuracy of the EEG signal evaluation by the EEG signal processing method and the brain-computer interaction system of the present invention, 5 subjects were selected, and 20 segments of EEG data including resting state and different task states were collected from each subject. The sampling rate was 256 Hz and the sampling accuracy was 24 bits. The EEG signal processing method and the brain-computer interaction system of the present invention (referred to as the present invention in Table 1 below) were used to compare the evaluation results of the EEG signals with different EEG signal analysis methods in the prior art. The recognition rate and false recognition rate of different task states in the evaluation results are shown in Table 1 below. It can be seen that the evaluation results of the EEG signals by the EEG signal processing method and the brain-computer interaction system of the present invention are better than the evaluation results of the spatiotemporal domain analysis method, the time-frequency analysis method and the nonlinear dynamics analysis method in the prior art. The recognition rates of the task movements of hand grasping, stepping and squatting are all higher than 85%, and the false recognition rates are no more than 3%.
[0076] Table 1
[0077]
[0078] The beneficial effects of the EEG signal processing method and brain-computer interaction system provided by the present invention include:
[0079] The EEG signal processing method of the present invention is used to process and extract features of EEG signals, screen effective signal segments in EEG data, reduce personalization and external interference, and then enhance the homogeneity of the characteristic vectors of effective signals, thereby reducing interference due to cross-individual and cross-scenario interference, and more accurately realize the evaluation of EEG signals and control peripherals, ultimately providing a solid foundation for realizing a reliable and stable brain-computer interaction system.
[0080] Although the present invention is described above with reference to exemplary embodiments, it will be understood by those skilled in the art that various modifications in form and detail may be made to the present invention without departing from the spirit and scope of the present invention as defined by the appended claims. The preferred embodiments should be considered merely illustrative, not restrictive. Therefore, the detailed description of the present invention does not limit the scope of the present invention, which should be defined by the appended claims, and all distinguishing technical features within the scope of the present invention should be understood to be included in the present invention.
Claims
1. A method for processing electroencephalogram signals, characterized in that: include: Collecting EEG signals of the target object in different states to obtain initial EEG data; Selecting an EEG data segment from the initial EEG data and preprocessing it to obtain a preprocessed EEG data segment; Filtering the pre-processed EEG data segments to obtain valid signal segments; intercepting the selected EEG data segment according to the valid signal segment to obtain a valid EEG information segment; Extract the eigenvalues of the effective signal segment and the effective EEG information segment, Composition feature vector; as well as Based on the feature vector, cluster analysis is performed on the EEG data to evaluate the EEG signal and obtain EEG data state classification results, so that external devices can be controlled to perform specific task actions based on the EEG data state classification results.
2. The method for processing EEG signals according to claim 1, wherein: The collecting of EEG signals of the target object in different states to obtain initial EEG data includes collecting EEG signals in a resting state to obtain EEG data in a resting state and collecting EEG signals in a task-performing state to obtain EEG data in a task-performing state; The resting state is the basic state when no task is performed; The task execution state is the state of the target when executing a specific task.
3. The method for processing EEG signals according to claim 1, wherein: The step of selecting an EEG data segment from the initial EEG data and preprocessing the segment to obtain a preprocessed EEG data segment includes: The EEG data segments are subjected to denoising and filtering to remove interference, and are dimensionless.
4. The method for processing EEG signals according to claim 2, wherein: The screening based on the pre-processed EEG data segments to obtain valid signal segments includes: Determine sliding window characteristics; Determine judgment thresholds; and Based on the grey constrained dynamic programming algorithm, the determined sliding window characteristics and the judgment threshold, the EEG data segments are screened to obtain valid signal segments.
5. The method for processing EEG signals according to claim 4, wherein: Determining the sliding window characteristics includes: The sliding window is determined by the EEG data sampling frequency and the characteristics of the task.
6. The method for processing EEG signals according to claim 4, wherein: Determining the judgment threshold includes: A training set is established and the judgment threshold of the effective signal segment is obtained using the training set.
7. The method for processing EEG signals according to claim 6, wherein: A training data set is established based on the preprocessed EEG data segments, the resting EEG data, and the task state EEG data, and a judgment threshold is determined based on the grey correlation coefficient between the preprocessed EEG data segments and the resting state EEG data, and the grey correlation coefficient between the preprocessed EEG data segments and the task state EEG data, and the valid signal segment is determined by the judgment threshold.
8. The method for processing EEG signals according to claim 7, wherein: The judgment threshold is updated in real time according to the effective signal segment and the EEG data in the resting state.
9. The method for processing EEG signals according to claim 1, wherein: The step of intercepting the selected EEG data segment according to the valid signal segment to obtain a valid EEG information segment includes: The valid signal segments are aligned in the EEG data segments to obtain valid EEG information segments.
10. A brain-computer interaction system using the electroencephalogram signal processing method according to any one of claims 1 to 9, characterized in that: include: An acquisition module is used to acquire EEG signals of the target object in different states to obtain initial EEG data; A preprocessing module, configured to select an EEG data segment from the initial EEG data and perform preprocessing to obtain a preprocessed EEG data segment; A screening module, used for screening and obtaining effective signal segments based on pre-processed EEG data segments; An interception module is used to intercept the selected EEG data segment according to the valid signal segment to obtain a valid EEG information segment; Feature extraction module, used to extract the feature values of valid signal segments and valid EEG information segments, Composition feature vector; as well as An evaluation module, used for performing cluster analysis on EEG data based on feature vectors to achieve evaluation of EEG signals; The brain-computer interaction system is used to output an EEG data state classification result according to an evaluation result of a target's EEG signal, and to control an external device to perform a specific task action based on the EEG data state classification result.