CN112614539B

CN112614539B - Motor imagery detection method based on TEO-MIC algorithm

Info

Publication number: CN112614539B
Application number: CN202011642735.6A
Authority: CN
Inventors: 李亚兵; 陈墨; 王红玉; 李红叶
Original assignee: Xian University of Posts and Telecommunications
Current assignee: Xian University of Posts and Telecommunications
Priority date: 2020-12-31
Filing date: 2020-12-31
Publication date: 2023-04-11
Anticipated expiration: 2040-12-31
Also published as: CN112614539A

Abstract

The present invention proposes a motor imagery state detection method based on the TEO-MIC algorithm combining the Teager energy operator and the microstate. (2) establish a discrete time series based on the Teager model; (3) determine the corresponding whole brain field strength according to the discrete time series; (4) cluster the whole brain field strength and calculate the corresponding microstate; (5) Calculate the microstate parameters in each state according to the existing microstates, so as to determine the motor imagery state. The invention can effectively analyze the functional state of the brain and realize the detection of the motor imagery state through the method of combining the Teager energy operator and the microstate.

Description

Motion imagery detection method based on TEO-MIC algorithm

Technical Field

The invention relates to the crossing field of signal and information processing and neurobiology, in particular to a method for combining electroencephalogram (EEG) and micro-state algorithms based on Teager energy operators. It proposes and designs an algorithm that combines Teager with the micro-state algorithm (TEO-MIC). The EEG signal is easily interfered by Gaussian noise during acquisition, so an algorithm with better robustness is needed, and the Teager energy operator is a theory with better suppression characteristic on the Gaussian noise.

Background

EEG signals, which are non-stationary and complex signals, are generally thought to be produced by a combination of concussive activities in different brain regions. Compared with the traditional characteristic analysis method, the complex network analysis method has more visual and effective effect on analysis of non-stationary signals such as EEG. The microstate, a parameter that describes the information about the change in global brain functional state, reflects the complexity of the activity in different areas of the brain in different cognitive states. The change of the activity state of the brain is indicated by the mutual conversion between different micro states.

Therefore, the EEG signals are analyzed by using a complex network analysis method, the micro-states of the EEG signals under different cognitive states are calculated, the change of different motor imagery states can be detected, and the motor imagery states are further evaluated. However, the EEG signal adopted by the method is susceptible to gaussian noise interference, and robustness for motor imagery state detection is not high.

The Teager Energy Operator (Teager Energy Operator) is a theory with good inhibition characteristic on Gaussian noise, so that the method is suitable for carrying out Teager Operator modeling on an EEG signal to further calculate the micro state of the EEG signal.

Based on Teager energy operator and micro-state, the invention provides a motion imagery state detection method based on a TEO-MIC algorithm

Disclosure of Invention

The invention provides a method for detecting a motor imagery state by combining a Teager energy operator and a micro-state TEO-MIC algorithm, which adopts the Teager energy operator to perform Teager modeling on an original EEG signal to construct a new discrete time sequence, then performs calculation of micro-state parameters on the discrete time characteristic sequence, and realizes the detection of the motor imagery state by using the micro-state parameters, wherein the basic scheme is as follows:

1. preprocessing an original EEG, segmenting an EEG signal according to a specific experimental paradigm, and preparing for Teager modeling;

2. processing the segmented data, and establishing a discrete time sequence based on a Teager model;

3. extracting characteristics by using a Teager model-based discrete time sequence, and solving corresponding micro-state parameters:

the following method for extracting the micro-state features by utilizing the discrete time sequence is established:

(1) Calculating the whole brain Field intensity (GFP) of the electroencephalogram data by using the discrete time sequence based on the Teager model;

(2) Performing TAAHC (hierarchical admixture and aggregation hierarchical clustering) clustering calculation on the time sequence of the GFP local maximum value of the whole brain field intensity to obtain four basic microstate;

4. calculating the micro-state parameters in each state by using the variation trend of the four micro-states based on the time sequence: number of occurrences per second (occupancy), duration (Duration), microstate fraction (Coverage), spatial correlation coefficient (Mspatcorr);

5. and detecting the motor imagery state by using the obtained micro-state parameters.

The method has the advantages that in the process of finally detecting the state of the EEG signal, the Teager modeling method is adopted to reduce the interference of Gaussian noise, so that the robustness of the calculation of the micro state is further greatly improved; by comparing various algorithms with a method based on the combination of TEO energy operators and micro-states, the method obtains high state detection accuracy.

Drawings

FIG. 1 computing flow based on TEO-MIC algorithm

FIG. 2 is a comparison graph of the ROC curves of the micro-state parameters of the micro-state and the classical algorithm calculated by the scheme

FIG. 3 is a comparison graph of AUC area of the micro-state parameter of the present solution and the micro-state parameter of the classical algorithm

FIG. 4 is a schematic diagram showing comparison between micro-state parameters calculated by the present scheme and recognition performance of a classical algorithm

Detailed Description

The following describes in detail embodiments of the present invention with reference to the drawings.

The flow of the TEO-MIC algorithm proposed by the present invention is shown in FIG. 1, and the following description will explain the embodiments of the present invention in detail with reference to the accompanying drawings.

1. Preprocessing raw EEG, including: segmenting the EEG data to facilitate Teager energy operator modeling of the data; according to the design of an experimental paradigm, task segmentation processing is carried out on an EEG signal, and 3s data of a motor imagery segment are selected for analysis;

2. teager energy operator modeling is carried out on the segmented EEG data, namely:

wherein x (n) represents a discrete EEG signal,

representing the Teager energy operator modeling output, where a, b, c, d are selected to be 0,1,2,3, respectively;

3. calculating the micro-state of the brain electrical signal:

1) Calculating the whole brain field strength GFP of the brain electrical signal, which can reflect the change of the brain to the motor imagery state:

in the formula (I), the compound is shown in the specification,

for the voltage value of the i-th lead>

The average voltage of all leads is shown, and N represents the number of leads.

2) Calculating the time point corresponding to the maximum value of the whole brain field intensity obtained in the step 1), and constructing a time sequence of the local maximum value by taking the electric field intensity corresponding to the time point of the local maximum value as a local field potential to obtain a brain map;

3) Performing TAAHC algorithm clustering on the time sequence of the local maximum value of the whole brain field intensity to obtain four micro states;

4) And sequentially calculating the micro-state parameters of each tested moment according to the four micro-states obtained in the step 3).

4. And detecting the motor imagery state by utilizing the relation between the brain nerve activity degree, the spatial information and the micro state.

The algorithm of the present invention is compared to the classical micro-state algorithm. Experimental results, as shown in fig. 2-4, the algorithm of the present invention has better recognition performance than the classical micro-state algorithm.

In the process of motor imagery, the robustness of the calculation of the micro state is insufficient, and the system identification performance can be greatly improved by combining the Teager operator with the micro state. In the examples, experimental data was derived from the BCI Competition IV 2008 database. The scheme proves the effectiveness of the scheme by analyzing the significance of the micro states of different motor imaginations and showing the experimental results in fig. 3 and 4. Meanwhile, the TEO-MIC algorithm and the calculation of the micro state are compared, and an experimental result is shown in figure 4, so that the result shows that the algorithm can obviously improve the identification performance.

Claims

1. A method for detecting motion-imagined states based on the TEO-MIC algorithm, comprising the following steps:

(1) Preprocess the original EEG to prepare for the Teager energy operator modeling, and then perform coordinate transformation to establish the Teager model of the discrete time series. The specific process includes the following steps:

Step 1: Preprocess the raw EEG data, including: segmenting the EEG data, performing task segmentation processing on the EEG signal according to the experimental paradigm design, and selecting the motion imagery task segment data for analysis.

Step 2: Perform coordinate transformation on the segmented EEG data to achieve Teager energy operator modeling, i.e.:

Where x(n) represents the discrete EEG signal,

This represents the output of the Teager energy operator modeling, where a, b, c, and d are selected as 0, 1, 2, and 3, respectively.

(2) Discrete time series constructed using the Teager model of discrete time series

Calculate discrete time series

The whole-brain field strength GFP was analyzed, and the time series of GFP local maxima were clustered using the TAAHC algorithm to obtain the corresponding microstate parameters. The specific process includes the following steps:

Step 1: Calculate the whole-brain field strength GFP of the EEG signal, which can reflect changes in the brain's response to motor imagery:

In the formula,

Let be the voltage value of the i-th lead.

The voltage is the average value across all leads, where N represents the number of leads.

Step 2: Based on the whole brain field strength obtained in Step 1, find the time point corresponding to its maximum value, and use the electric field strength corresponding to the local maximum time point as the local field potential to construct the time series of local maxima and obtain the brain topography map.

Step 3: Perform TAAHC algorithm clustering on the time series of local maxima of the whole brain field strength to obtain four microstates;

Step 4: Based on the four microstates obtained in Step 3, calculate the microstate parameters for each subject at each time step.

(3) Determine the motion imagination state based on the obtained micro-state parameters.