CN115826743A - SSVEP brain-computer interface-oriented multi-channel electroencephalogram signal modeling method - Google Patents

Abstract

This application discloses a multi-channel EEG signal modeling method for SSVEP brain-computer interface. The steps include: determining the SSVEP signal frequency range and signal number of the subject; based on the signal number, establishing a basic neuron group model for Generate a variety of rhythmic narrowband signals in the frequency range; based on the frequency range with a variety of rhythmic narrowband signals, determine the multi-dynamic neuron group model, which is used to modulate a single-channel SSVEP frequency signal suitable for the subject; based on the multi-dynamic neuron group The model and the single-channel SSVEP frequency signal construct a multi-channel multi-dynamic neuron group model, which is used to set the coupling coefficient matrix, and then by adjusting the coupling coefficient between different basic neuron groups, a multi-channel EEG signal is constructed, and it reflects the inter-tester variability. This application generates multi-channel EEG signals for SSVEP brain-computer interface for application in brain-controlled intelligent devices from the perspective of models, and solves the problem of high cost in the prior art.

Description

A multi-channel EEG signal modeling method for SSVEP brain-computer interface

technical field

The present application relates to the field of signal processing, in particular to a multi-channel EEG signal modeling method for SSVEP brain-computer interface.

Background technique

Brain Computer Interface (Brain Computer Interface, BCI) provides a direct real-time information exchange and control channel for users and external physical devices, which can directly decode the user's brain activity from neurophysiological signals into control instructions for external devices. BCI technology completes the acquisition of EEG signals, data preprocessing, feature extraction, classification and command quantification. For the acquisition of EEG signals, the data collected by EEG is only the activity of some neurons, but there are infinitely many specific activities in the brain. In order to solve this problem, researchers mainly use invasive and non-invasive Two methods are used to collect EEG signals.

Invasive methods are somewhat traumatic, while non-invasive methods are relatively cheap compared to invasive methods. In current research, non-invasive methods are widely used and are easily accepted by the public. In the non-invasive method, the most widely used is the electroencephalogram (Electroencephalograph, EEG). EEG is a comprehensive expression of the post-synaptic currents of a large number of neuron groups in brain tissue, which can be decomposed into specific frequency ranges (δ: 1–4Hz, θ: 4–8Hz, α:8–12Hz, β:12–30Hz, gamma: 30–70Hz). The monitoring method is specifically that electrodes are placed along the scalp, and then through multiple electrodes placed on the scalp, the brain's spontaneous electrical activity is recorded over a period of time. Despite its limited spatial resolution and high signal artifacts, the EEG remains an invaluable tool for research and diagnosis.

Steady-State Visual Evoked Potentials (SSVEP) refers to the EEG EEG signals that the brain will induce for visual stimuli of a specific frequency. When the retina receives visual stimuli from 3.5Hz to 75Hz, the brain will produce the same Electrical activity at a frequency or multiples of frequency. BCI technology based on SSVEP-EEG has been widely used in the development of various brain-controlled intelligent devices, such as brain-controlled cursors, brain-controlled virtual keyboards, brain-controlled web browsing, brain-controlled prosthetics, brain-controlled wheelchairs, brain-controlled vehicles, and brain-controlled robots. and other systems. At present, in the research of the above-mentioned brain-controlled intelligent devices, it is still necessary to recruit subjects, and it is necessary to wear an acquisition device or even apply conductive paste, wash hair, etc., which consumes a certain amount of manpower and material resources. It is very important and meaningful to generate correct, similar and applicable EEG signals.

A variety of neural models have been proposed, which can generally be divided into two categories. One type is the detail-oriented model, that is, the modeling of neurons. Obviously, information transmission is completed by the cooperation of many neurons, and it is not enough to study the complex behavior of the brain only by studying the discharge of a single neuron at the microscopic level. Therefore, some researchers directly study the neural network composed of multiple neurons, and study the discharge behavior of the neural network. However, there are many types of neurons, and it is difficult to determine the parameters of each neuron model. At the same time, the connections between various neurons are very complex. It is complex and has a huge amount of calculation, so it is quite difficult to simulate the actual neural network at the neuron level. Another type of neural model is the neural group model, which does not need to model individual cells in the network structure, but models the overall characteristics of neuron groups composed of specific types of cells.

The main idea of the neural group model is "average area approximation", that is, the aggregated state variables are used in the model to represent the average behavior of the entire cell population in the neural network. The model is simple and has physiological significance. "Construct the model of EEG signal from the point of view. The coupled neural group model can reflect the interconnection between neuron groups, and can simulate large-scale interactive neural networks at the macro level.

In 1995, researchers such as Jasen first proposed the generation of EEG signals and visual evoked potentials in the mathematical model of cerebral cortex column coupling. On this basis, it is possible to connect BCI with physiological models, especially the brain-computer interface based on the SSVEP paradigm. The key technology of the electrical signal model is how to make the model correctly simulate the multi-channel EEG signals for the SSVEP brain-computer interface. The model needs to be considered from several angles of signal frequency and the distinction of different subjects.

Contents of the invention

This application constructs a basic neuron group model, so that excitatory or inhibitory subgroups are weighted and controlled by several linear transfer functions with different dynamic characteristics in parallel, thereby generating signals with richer frequencies and wider frequency bands, which are suitable for The signal frequency required by the SSVEP paradigm is then adjusted to distinguish between different subjects by adjusting the coupling coefficient of the multi-channel multi-dynamic neuron group model.

In order to achieve the above purpose, the application discloses a multi-channel EEG signal modeling method for SSVEP brain-computer interface, the steps include:

Determine the SSVEP signal frequency range and signal number of the subject;

Based on the frequency range and the number of signals, a basic neuron group model is established for generating a variety of rhythmic narrowband signals in the frequency range;

Based on the frequency range with a variety of rhythmic narrowband signals, determine a multi-dynamic neuron group model for modulating a single-channel SSVEP frequency signal suitable for the subject;

Construct a multi-channel multi-dynamic neuron group model based on the multi-dynamic neuron group model and the single-channel SSVEP frequency signal, for setting the coupling coefficient matrix, and then by adjusting the coupling coefficients between different basic neuron groups , to reflect the differences among the subjects.

Preferably, the multi-dynamic neuron group model can limit the required EEG signal frequency through the setting of weight coefficients.

Preferably, the method for obtaining the frequency range and signal number of the SSVEP signal includes: obtaining the subject's EEG signal through SSVEP stimulation; selecting the SSVEP frequency required by the subject according to the EEG signal; modulating the The frequency corresponding to the SSVEP stimulation is determined to determine the frequency range and signal number of the SSVEP signal suitable for controlling the brain-controlled intelligent device.

Preferably, the method for generating the rhythmic narrowband signal includes: using the basic neuron group model to adjust the parameters of excitatory and inhibitory linear transformation functions, and generating the narrowband signal at different frequencies.

Preferably, the method for modulating the single-channel SSVEP frequency signal includes: according to the frequency selected by the subject in the frequency range, determining the required excitation and activation of the multi-dynamic neuron group model and its linear transfer function Inhibitory parameter combinations, and by adjusting the weights, a suitable single-channel SSVEP frequency signal is modulated.

Preferably, the method for modulating the single-channel SSVEP frequency signal includes: adjusting weights through the multi-dynamic neuron group model and setting excitatory and inhibitory parameters of several different linear transfer functions, adjusting neurons with different dynamic characteristics The relative proportion of cortical area signals, thereby generating the single-channel SSVEP frequency signal with richer frequency and wider frequency band.

Preferably, the method for distinguishing the differences includes: the multi-channel multi-dynamic neuron group model realizes the coupling of different intensities between EEG signals of different frequencies by setting the coupling coefficient matrix, and selects the SSVEP-oriented brain-computer interface channel, after the channel selection is completed, the coupling coefficient between the channels is adjusted through the multi-channel multi-dynamic neuron group model, and then the EEG signals of different subjects are modulated, and the difference is adjusted by adjusting the coupling coefficient coefficients are distinguished.

Preferably, the parameters of the multi-channel multi-dynamic neuron group model include: average excitatory synaptic gain, average inhibitory synaptic gain, excitatory membrane average time constant and dendritic average time delay, inhibitory membrane average time constant, tree Synaptic average time delay, average number of synaptic connections on excitatory feedback loop, average number of synaptic connections on inhibitory feedback loop, nonlinear S-function parameters, several pairs of weight coefficients and coupling coefficients to other channels.

Compared with the prior art, the beneficial effects of the present application are as follows:

In this application, by adjusting the excitatory and inhibitory parameters in the transfer function, the basic neuron group model can generate a variety of rhythmic narrow-band signals, and adjust the relative proportion of cortical area signals with different dynamic characteristics, thereby generating more abundant frequencies and frequency bands. Wider signals, adjust the appropriate SSVEP frequency, adjust the coupling coefficient between channels through the model, and then distinguish different subjects. This application generates SSVEP-oriented brain-controlled intelligent devices from the perspective of the model. The multi-channel EEG signal of the brain-computer interface solves the problem of high cost of collecting real EEG signals.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings in the following description are only some of the present application. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without paying creative labor.

Fig. 1 is the schematic flow chart of the method in application embodiment one;

Fig. 2 is a schematic diagram of the function of the basic neuron group model of neural oscillations generated inside the channel in Example 1 of the application;

Fig. 3 is a schematic diagram of the function of the multi-channel multi-dynamic neuron group model of neural oscillations generated inside the channel in Embodiment 1 of the present application;

FIG. 4 is a schematic diagram of the composition principle of the multi-channel coupled neural group model in Embodiment 2 of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Embodiment one

As shown in Figure 1, it is a schematic flow chart of the method of Embodiment 1 of the present application. The steps include: determining the frequency range and signal number of the SSVEP signal of the subject; based on the frequency range and signal number, establishing a basic neuron group model for generating A variety of rhythmic narrowband signals in the frequency range; based on the frequency range with a variety of rhythmic narrowband signals, determine the multi-dynamic neuron group model, which is used to modulate a single-channel SSVEP frequency signal suitable for the subject; based on the multi-dynamic neuron group model Construct a multi-channel multi-dynamic neuron group model with the single-channel SSVEP frequency signal, which is used to set the coupling coefficient matrix, and then construct a multi-channel EEG signal by adjusting the coupling coefficient between different basic neuron groups, and reflect the measured difference between them.

Among them, this embodiment 1 is aimed at the practical application of brain-controlled intelligent devices, and develops a method for establishing an SSVEP EEG signal model. This method can be extended from the establishment of a basic neuron group model to a multi-dynamic neuron group model, and finally realizes multi-channel The construction of multi-dynamic neuron group model; the basic neuron group model is used to adjust the parameters of the excitatory and inhibitory linear transformation functions, and generate narrow-band signals of different frequencies; the multi-dynamic neuron group model adjusts the weight and sets several Different excitatory and inhibitory parameters of the linear transfer function can adjust the relative ratio of cortical area signals with different dynamic characteristics, thereby generating signals with richer frequencies and wider frequency bands; the multi-channel multi-dynamic neuron group model is used for By setting the coupling coefficient matrix, different intensities of coupling between EEG signals of different frequencies can be realized. It is not necessary to model individual cells in the network structure, but to model the overall characteristics of neuron groups composed of specific types of cells, which can realize the coupling between different regions of the brain.

In the first embodiment, the parameter information of the multi-channel EEG signal model includes the average excitatory synaptic gain, the average inhibitory synaptic gain, the average time constant of the excitatory membrane and the average time delay of the dendrites, the average time constant of the inhibitory membrane and the average time delay of the model. Dendritic mean time delay, mean number of synaptic connections in excitatory feedback loops, mean number of synaptic connections in inhibitory feedback loops, nonlinear S-function parameters, several pairs of weighting coefficients, and coupling coefficients to other channels.

The operation steps of the method of the present application will be described in detail below in conjunction with the first embodiment.

First, the subject controls the brain-controlled intelligent device by "staring" at the SSVEP stimulation-induced EEG signal, for example, staring at a square that flickers at a certain frequency, the frequency and its frequency can be identified in the EEG signal collected by the visual area. Harmonic, so that the control end knows that the subject has completed a selection, and the brain-computer interface selects the SSVEP frequency required by the subject; this step modulates the frequency corresponding to the corresponding stimulus, and determines that it is suitable for controlling the brain-controlled intelligent device SSVEP signal frequency range and number of signals;

Afterwards, based on the physiological significance of the determined signal numbers and the parameter values for generating typical waveforms of delta, alpha and gamma waves, a basic neuron group model is established, as shown in Figure 2. By adjusting the excitatory and inhibitory parameters in the transfer function, a variety of rhythmic narrowband signals corresponding to the frequency range in the above steps are generated. According to the frequency selected by the subject in the frequency range, the required multi-dynamic neuron group model is determined, as shown in Figure 3; and the combination of excitatory and inhibitory parameters of its linear transfer function, and by adjusting the weight, A suitable single-channel SSVEP frequency signal is modulated.

Finally, select the SSVEP brain-computer interface channel for brain-controlled smart devices. After the channel selection is completed, adjust the coupling coefficient between channels through the model, and then distinguish different subjects. Adjusting the coupling coefficient can reflect the subject's difference.

In this embodiment one, the signal frequency adjustment process of the multi-dynamic neuron group model to the basic neuron group model is: adjust the weight W through the multi-dynamic neuron group model and set the excitation and inhibition of several different linear transfer functions parameters that adjust the relative proportions of signals in cortical regions with different dynamics, resulting in signals with richer frequencies and wider bandwidths. In addition, the multi-dynamic neuron group model can limit the required EEG signal frequency through the setting of weight coefficients.

It should be noted that, in the first embodiment, the generation process of the multi-channel EEG signal model includes: the multi-channel multi-dynamic neuron group model is used to realize the interaction between EEG signals of different frequencies by setting the coupling coefficient matrix. Coupling with different strengths, select the channel facing the SSVEP brain-computer interface. After the channel selection is completed, adjust the coupling coefficient between the channels through the model, and then modulate the EEG signals of different subjects. The differences between the subjects are adjusted Coupling coefficients are distinguished.

Embodiment two

In the second embodiment, the modeling method of the above-mentioned multi-channel multi-dynamic neuron group model is emphatically described, and the steps include:

S1. Excitatory or inhibitory dynamic linear transfer function: convert the average pulse density of the action potential into the post-synaptic average membrane potential, and the unit step response of the dynamic linear transfer function is as follows:

In the formula: t is the time constant, u(t) is the unit step function, He _e , H _i are the average excitatory synaptic gain and the average inhibitory synaptic gain respectively, a _e , a _i are the average time constants of the excitatory membrane and the reciprocal of the mean time delay of the dendrites, the mean time constant of the inhibitory membrane and the reciprocal of the mean time delay of the dendrites, _he (t) and _hi (t) are the mean membrane potential of the excitatory postsynaptic and the mean mean of the inhibitory postsynaptic Membrane potential, e is the base of the natural logarithm function.

S2. Static nonlinear function: mainly transforms the average presynaptic membrane potential into the average pulse density of the action potential, such as the formula:

In the formula: S(v) is a static nonlinear function, 2e ₀ is the maximum firing rate, v ₀ is the post-synaptic potential relative to the firing rate e ₀ , and r is the bending degree of the S-type function. v is the average presynaptic membrane potential. The static nonlinear functions in the neuron population model are all consistent.

S3. All external inputs from indeterminate and subcortical areas, represented by Gaussian distributed excitatory inputs p(t). The average number of synaptic connections between pyramidal cell populations and interneuron populations is represented by connection constants C ₁ , C ₂ , C ₃ , and C ₄ . In summary, the basic neural cell community model can be expressed by the following differential equation:

In the formula: y ₀ , y ₁ , y ₂ , y ₃ , y ₄ , y ₅ are the output signal of subgroup 1, the excitatory feedback of subgroup 2, the inhibitory feedback of subgroup 2, and the first _- order Derivative, the first derivative of y ₁ , the first derivative of y ₂ , He _{and Hi} are the average excitatory synaptic gain and the average inhibitory synaptic gain respectively, ae and ai are the average time constant of the excitatory membrane and the average dendrite Reciprocal of time delay, average time constant of inhibitory membrane and reciprocal of average time delay of dendrites, C ₁ , C ₂ , C ₃ , C ₄ are connection constants representing the average synapse between a population of pyramidal cells and a population of interneurons The number of connections, S( _y1 (t) _-y2 (t)) is the average density that converts the presynaptic average membrane voltage of subpopulation 1 into action potentials.

S4. Brain function is formed by strong coupling between distant regions, and the output of the cerebral cortex to distant targets is excitatory. Among them, q _jk represents the coupling coefficient of channel j to channel k. In summary, the multi-channel multi-dynamic neural group model can be expressed as a differential equation:

In the formula: y ₀ , y ₁ , y ₂ , y ₃ , y ₄ , y ₅ are respectively the ith output signal of j-channel subgroup 1, the i-th excitatory feedback of j-channel subgroup 2, and the j-th channel subgroup The ith inhibitory feedback of group 2, the first derivative of y ₀ , the first derivative of y ₁ , and the first derivative of y ₂ , He _{and Hi} are the average excitatory synapse gain and the average inhibitory synapse gain, respectively, a _e , a _i are the reciprocals of the average time constant of the excitatory membrane and the average time delay of the dendrites, and the reciprocals of the average time constant of the inhibitory membrane and the average time delay of the dendrites, respectively, and C ₁ , C ₂ , C ₃ , and C ₄ are connections Constant, denoting the average number of synaptic connections between the pyramidal cell population and the interneuron population. S(.) represents a static nonlinear function, specifically, S(C ₃ ∑W _ji y ₀ ^ji ) represents the conversion of the inhibitory membrane voltage obtained by combining the first to N inhibitory feedbacks of the j-channel subgroup 2 into The average density, S _kj is the coupling signal input to channel j from other channels.

S5. The brain-controlled operator controls the brain-controlled intelligent device according to his own control intention, and focuses on the corresponding stimulation interface (through SSVEP) to generate the corresponding EEG signal. The brain-computer interface obtains the operator's information by analyzing the EEG signal. control intention, and output the corresponding recognition control command. Record the frequencies required by the subjects, adjust the corresponding frequencies through the models described in S ₁ , S ₂ , S ₃ , and S ₄ , and finally distinguish different subjects by adjusting the coupling coefficient.

As shown in Figure 4, a coupling method between different channels is also provided. By adjusting the coupling coefficient, the S _jx coupled signals output by different channels are coupled to different degrees, and the subjects are distinguished.

The above-mentioned embodiments are only to describe the preferred mode of the present invention, not to limit the scope of the present invention. Without departing from the design spirit of the present invention, those skilled in the art may make various Variations and improvements should fall within the scope of protection defined by the claims of the present invention.

Claims (8)

1. a kind of multi-channel EEG signal modeling method facing SSVEP brain-computer interface, it is characterized in that, the step comprises: Determine the SSVEP signal frequency range and signal number of the subject; Based on the number of signals, a basic neuron group model is established for generating a variety of rhythmic narrowband signals in the frequency range; Based on the frequency range with a variety of rhythmic narrowband signals, determine a multi-dynamic neuron group model for modulating a single-channel SSVEP frequency signal suitable for the subject; Construct a multi-channel multi-dynamic neuron group model based on the multi-dynamic neuron group model and the single-channel SSVEP frequency signal, for setting the coupling coefficient matrix, and then by adjusting the coupling coefficients between different basic neuron groups , to construct multi-channel EEG signals, and to reflect the differences among the subjects. 2. the multi-channel EEG signal modeling method facing SSVEP brain-computer interface according to claim 1, is characterized in that, described multi-dynamic neuron group model can pass through the setting of weight coefficient to required EEG signal Frequency is limited. 3. The multi-channel EEG signal modeling method facing the SSVEP brain-computer interface according to claim 1, wherein the method for obtaining the frequency range of the SSVEP signal and the number of signals comprises: eliciting by SSVEP stimulation to obtain the measured The EEG signal of the subject; select the SSVEP frequency required by the subject according to the EEG signal; modulate the frequency corresponding to the SSVEP stimulation, and determine the frequency range and signal number of the SSVEP signal suitable for controlling the brain-controlled intelligent device . 4. The multi-channel EEG signal modeling method facing the SSVEP brain-computer interface according to claim 1, wherein the method for generating the rhythmic narrowband signal comprises: using the basic neuron group model to adjust the excitability and the parameters of the suppressive linear transformation function, and generate said narrowband signals at different frequencies. 5. The multi-channel EEG signal modeling method facing SSVEP brain-computer interface according to claim 1, wherein the method of modulating the single-channel SSVEP frequency signal comprises: according to the subject in the frequency range The selected frequency determines the desired combination of excitatory and inhibitory parameters of the multi-dynamic neuron group model and its linear transfer function, and adjusts the weight to modulate a suitable single-channel SSVEP frequency signal. 6. The multi-channel EEG signal modeling method facing the SSVEP brain-computer interface according to claim 5, wherein the method for modulating the single-channel SSVEP frequency signal comprises: adjusting by the multi-dynamic neuron group model Weighting and setting excitatory and inhibitory parameters of several different linear transfer functions, adjusting the relative ratio of cortical area signals with different dynamic characteristics, thereby generating the single-channel SSVEP frequency signal with richer frequency and wider frequency band. 7. The multi-channel EEG signal modeling method facing the SSVEP brain-computer interface according to claim 1, wherein the method for distinguishing the difference comprises: the multi-channel multi-dynamic neuron group model is set by setting The coupling coefficient matrix realizes the coupling of different intensities between EEG signals of different frequencies, and selects the channel facing the SSVEP brain-computer interface. The coupling coefficient is adjusted to modulate the EEG signals of different subjects, and the difference is distinguished by adjusting the coupling coefficient. 8. The multi-channel EEG signal modeling method facing the SSVEP brain-computer interface according to claim 1, wherein the parameters of the multi-channel multi-dynamic neuron group model include: average excitatory synaptic gain, average inhibition Synaptic gain, mean time constant of excitatory membrane and mean time delay of dendrites, mean time constant of inhibitory membrane, mean time delay of dendrites, mean number of synaptic connections on excitatory feedback loop, mean number of synaptic connections on inhibitory feedback loop, Nonlinear S-function parameters, several pairs of weighting coefficients and coupling coefficients to other channels.

Images