A brain-computer intelligent interaction system with a very large instruction set based on edge frequency modulation

Abstract

The present invention discloses a brain-computer intelligent interaction system with a very large instruction set based on edge frequency modulation, comprising: a human brain feature visual elicitation module, an EEG acquisition module, an EEG signal processing module, a frequency-domain spatial domain encoding and decoding module, and a recognition result feedback module; the human brain feature visual elicitation module is used to encode the stimulus source based on edge frequency modulation and present the encoded stimulus source to the user; the EEG acquisition module is used to collect the EEG signal of the user who receives the stimulus source; the EEG signal processing module is used to process and extract the features of the EEG signal based on edge frequency modulation to obtain the frequency domain eigenvalues ​​of the EEG signal; the frequency-domain spatial domain encoding and decoding module is used to decode the EEG signal into brain control instructions based on the frequency domain eigenvalues; and the recognition result feedback module is used to display the brain control instructions in real time and mark the stimulus source, providing instant feedback to the user. The present invention has the advantages of good real-time performance, high recognition accuracy, and good scalability.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent biometric recognition and novel human-computer interaction technology, and specifically relates to an ultra-large instruction set brain-computer intelligent interaction system based on edge frequency modulation. Background Art

[0002] In current research on brain-computer interaction, visual evoked methods are being widely explored, with one of their core goals being to maximize the number of brain-control commands that can be output. However, the traditional steady-state visual evoked potential (SSVEP) paradigm has several technical limitations, such as a limited stimulation frequency range and frequency resolution constrained by the length of the EEG signals being analyzed. Therefore, employing the traditional SSVEP paradigm alone is limited in achieving large-scale brain-control command output.

[0003] To address this issue, researchers have begun exploring methods to fuse different brain potentials to broaden the command output capabilities of intelligent brain-computer interaction systems. An early study attempted to combine SSVEP with P300 potentials, successfully increasing the number of stimulus encodings from the limited number in the traditional SSVEP paradigm to 108 commands. This fusion approach fully utilizes two different brain potentials, expanding the available command set through complementarity, thereby improving the functionality and flexibility of intelligent brain-computer interaction systems.

[0004] Further research has adopted a more comprehensive fusion strategy, combining motor evoked potentials, P300 potentials, and SSVEPs, enabling the system to output up to 216 brain-controlled commands. This fusion of multiple brain potentials demonstrates greater sensitivity and richness in activating brain-computer intelligent interaction, providing users with more actionable control options. However, despite the significant progress made by these methods, some technical challenges and limitations still exist.

[0005] The P300 paradigm itself suffers from a long recognition time. While P300 fusion can increase the number of commands, the corresponding recognition process can be relatively slow, resulting in long wait times for users and affecting the efficiency of practical applications. Furthermore, the generation of motor evoked potentials actually collects myoelectric signals, requiring the user to perform muscle movements. This is not true brain control, and for users unable to perform muscle movements, the system's usability is limited. Summary of the Invention

[0006] The present invention aims to address the shortcomings of the existing technology and proposes a brain-computer intelligent interaction system with a very large instruction set based on edge frequency modulation, which is used to achieve a significant increase in the number of brain control instructions on the basis of the traditional SSVEP paradigm, and further increase the number of brain control instructions to the square of the original number, providing users with richer and more accurate brain control options.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] A brain-computer intelligent interaction system with a large instruction set based on edge frequency modulation, comprising: a human brain feature visual evoked module, an EEG acquisition module, an EEG signal processing module, a frequency domain-spatial domain encoding and decoding module, and a recognition result feedback module;

[0009] The human brain characteristic visual elicitation module is used to encode the stimulus source based on edge frequency modulation and present the encoded stimulus source to the user;

[0010] The EEG acquisition module is used to acquire EEG signals of the user who receives the stimulation source;

[0011] The EEG signal processing module is used to process and extract features of the EEG signal based on the edge frequency modulation to obtain frequency domain feature values ​​of the EEG signal;

[0012] The frequency domain-spatial domain encoding and decoding module is used to decode the EEG signal into a brain control command based on the frequency domain eigenvalue;

[0013] The recognition result feedback module is used to display the brain control instructions in real time and mark the stimulation source to provide instant feedback to the user.

[0014] Preferably, the feature visual induction module includes: an encoding mode setting unit, a stimulus source encoding unit and a display unit;

[0015] The coding mode setting unit is used to set the edge frequency modulation coding mode based on the minimum resolution;

[0016] The stimulus source encoding unit is used to encode the stimulus source based on the edge frequency modulation encoding mode;

[0017] The display unit is used to present the encoded stimulus source to the user.

[0018] Preferably, the workflow of the encoding mode setting unit includes:

[0019] Obtain the length of the user's EEG signal and determine the minimum frequency resolution Δf;

[0020] Based on the minimum frequency resolution Δf, several frequency values ​​are obtained between the maximum frequency and the minimum frequency of the steady-state visual evoked potential, which are recorded as the basic frequency sequence b=[f1,f2,f3,...,f n ],in:

[0021]

[0022] The basic frequency sequence b of length n is expanded into an n×n edge frequency modulation coding matrix F to complete the coding mode setting:

[0023]

[0024] The edge frequency modulation coding matrix F includes the main frequency component f i and FM component f j .

[0025] Preferably, the workflow of the stimulus source encoding unit includes:

[0026] Each of the stimulation sources is set to correspond to the output of one brain control command, and the overall size and shape of each of the stimulation sources are exactly the same;

[0027] The stimulus source includes two sub-patterns of identical shape but different sizes, wherein the sub-patterns include: a smaller first sub-pattern and a larger second sub-pattern, the geometric centers of the sub-patterns coincide, and the first sub-pattern covers the second sub-pattern;

[0028] The flickering frequency of the two sub-patterns in each stimulus source is set, and the flickering frequency of the first sub-pattern is the main frequency component f of the corresponding element in the edge frequency modulation coding matrix F. i The flashing frequency of the second sub-pattern is the frequency modulation component f of the corresponding element in the edge frequency modulation coding matrix F. j .

[0029] Preferably, the EEG acquisition module includes: an EEG cap, an amplifier and a data transmission interface;

[0030] The EEG cap is used to collect EEG signals of the user who receives the stimulation source;

[0031] The amplifier is used to amplify the EEG signal;

[0032] The data transmission interface is used to transmit the amplified EEG signal.

[0033] Preferably, the EEG signal processing module includes: a signal preprocessing unit and a feature extraction unit;

[0034] The signal preprocessing unit is used to filter the EEG signal to obtain a processed EEG signal;

[0035] The feature extraction unit is used to perform frequency domain feature extraction on the processed EEG signal based on a one-dimensional correlation matrix calculation method. The extracted eigenvalues ​​are the frequency domain eigenvalues ​​of all frequencies in the basic frequency sequence b, wherein the largest eigenvalue is the main frequency eigenvalue and the second largest eigenvalue is the frequency modulation eigenvalue.

[0036] Preferably, the frequency-space domain encoding and decoding module includes: an adaptive threshold setting unit and a brain control instruction decoding unit;

[0037] The adaptive threshold setting unit is used to construct a threshold target function and set an adaptive EEG feature threshold based on the threshold target function;

[0038] The brain control instruction decoding unit is used to determine the size relationship between the output of the main frequency characteristic value and the adaptive EEG characteristic threshold, as well as the size relationship between the output of the frequency modulation characteristic value and the adaptive EEG characteristic threshold, and output the brain control instruction based on the judgment result.

[0039] Preferably, the workflow of the adaptive threshold setting unit includes:

[0040] Based on the adaptive EEG feature threshold θ to be set, the threshold objective function Objective(θ) is constructed, and its expression is:

[0041] Objective(θ)=α·Accuracy(θ)+β·Balance(θ)-γ·Regularization(θ)

[0042] Among them, Accuracy(θ) represents the classification accuracy under the threshold θ, Balance(θ) represents the balance between positive and negative categories, Regularization(θ) is the regularization term, and α, β, and γ are weight parameters;

[0043] The adaptive EEG feature threshold is obtained by maximizing the threshold objective function.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention has opened up a new method and device for expanding the number of output instructions for brain-computer intelligent interaction. Its induction mechanism is similar to the traditional steady-state visual induced point mechanism, and has the advantages of good real-time performance and high recognition accuracy. In addition, the present invention has good scalability and can be combined with event-related potentials and motor evoked potentials in previous studies to further increase the number of brain-controlled instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1Schematic diagram of the system structure of an embodiment of the present invention;

[0048] Figure 2 This is a schematic structural diagram of a single stimulation source according to an embodiment of the present invention;

[0049] Figure 3 This is a diagram showing the results of EEG decoding using a one-dimensional correlation matrix calculation method when a user gazes at a stimulus source with a main frequency of 12.4 Hz and a frequency modulation of 13.0 Hz in an embodiment of the present invention;

[0050] Figure 4 This is a diagram showing the results of EEG decoding using a one-dimensional correlation matrix calculation method when a user gazes at a stimulus with both a main frequency and a modulated frequency of 13.0 Hz in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Example 1

[0054] In this embodiment, if Figure 1 As shown, a very large instruction set brain-computer intelligent interaction system based on edge frequency modulation includes: a human brain feature visual induction module, an EEG acquisition module, an EEG signal processing module, a frequency domain-spatial domain encoding and decoding module, and a recognition result feedback module.

[0055] The human brain feature visual evoked module is used to encode the stimulus source based on edge frequency modulation and present the encoded stimulus source to the user.

[0056] The feature visual induction module includes: a coding mode setting unit, a stimulus source coding unit and a display unit; the coding mode setting unit is used to set the edge frequency modulation coding mode based on the minimum resolution; the stimulus source coding unit is used to encode the stimulus source based on the edge frequency modulation coding mode; and the display unit is used to present the encoded stimulus source to the user.

[0057] The workflow of the encoding mode setting unit includes: obtaining the length of the user's EEG signal and determining the minimum frequency resolution Δf; based on the minimum frequency resolution Δf, obtaining several frequency values ​​between the maximum frequency and the minimum frequency of the steady-state visual evoked potential, which are recorded as the basic frequency sequence b = [f1, f2, f3, ..., f n ],in:

[0058]

[0059] Expand the basic frequency sequence b of length n into an n×n edge frequency modulation coding matrix F to complete the coding mode setting:

[0060]

[0061] The edge frequency modulation coding matrix F includes the main frequency component f i and FM component f j .

[0062] The workflow of the stimulus source encoding unit includes: setting each stimulus source to correspond to the output of one brain control command, and the overall size and shape of each stimulus source are exactly the same; the stimulus source includes two sub-patterns of the same shape but different sizes, where the sub-patterns include: a smaller first sub-pattern and a larger second sub-pattern, the geometric centers of the sub-patterns coincide, and the first sub-pattern covers the second sub-pattern; setting the flashing frequency of the two sub-patterns in each stimulus source, and the flashing frequency of the first sub-pattern is the main frequency component f of the corresponding element in the edge frequency modulation encoding matrix F i The flashing frequency of the second sub-pattern is the frequency modulation component f of the corresponding element in the edge frequency modulation coding matrix F. j .

[0063] The EEG acquisition module is used to collect EEG signals of the user who receives the stimulation source.

[0064] The EEG acquisition module includes: an EEG cap, an amplifier and a data transmission interface; the EEG cap is used to collect the EEG signals of the user who receives the stimulation source; the amplifier is used to amplify the EEG signals; and the data transmission interface is used to transmit the amplified EEG signals.

[0065] The EEG signal processing module is used to process and extract features of EEG signals based on edge frequency modulation to obtain frequency domain eigenvalues ​​of EEG signals.

[0066] The EEG signal processing module includes: a signal preprocessing unit and a feature extraction unit; the signal preprocessing unit is used to filter the EEG signal to obtain the processed EEG signal; the feature extraction unit is used to extract the frequency domain features of the processed EEG signal based on the one-dimensional correlation matrix calculation method, and the extracted eigenvalues ​​are the frequency domain eigenvalues ​​of all frequencies in the basic frequency sequence b, among which the largest eigenvalue is the main frequency eigenvalue and the second largest eigenvalue is the frequency modulation eigenvalue.

[0067] The frequency domain-spatial domain encoding and decoding module is used to decode EEG signals into brain control commands based on frequency domain eigenvalues.

[0068] The frequency domain-spatial domain encoding and decoding module includes: an adaptive threshold setting unit and a brain control instruction decoding unit; the adaptive threshold setting unit is used to construct a threshold target function and set an adaptive EEG feature threshold based on the threshold target function; the brain control instruction decoding unit is used to judge the size relationship between the output of the main frequency eigenvalue and the adaptive EEG feature threshold, as well as the size relationship between the output of the frequency modulation eigenvalue and the adaptive EEG feature threshold, and output the brain control instruction based on the judgment result.

[0069] The workflow of the adaptive threshold setting unit includes: constructing a threshold objective function Objective(θ) based on the adaptive EEG feature threshold θ to be set, which is expressed as:

[0070] Objective(θ)=α·Accuracy(θ)+β·Balance(θ)-γ·Regularization(θ)

[0071] Among them, Accuracy(θ) represents the classification accuracy under the threshold θ, Balance(θ) represents the balance between positive and negative categories, Regularization(θ) is the regularization term, and α, β, and γ are weight parameters. The adaptive EEG feature threshold is obtained by maximizing the threshold objective function.

[0072] The recognition result feedback module is used to display brain control instructions in real time and mark the stimulation source to provide users with instant feedback.

[0073] Example 2

[0074] In this embodiment, the human brain characteristic visual elicitation module is used to encode the stimulus source based on edge frequency modulation and present the encoded stimulus source to the user.

[0075] The feature visual induction module includes: a coding mode setting unit, a stimulus source coding unit and a display unit; the coding mode setting unit is used to set the edge frequency modulation coding mode based on the minimum resolution; the stimulus source coding unit is used to encode the stimulus source based on the edge frequency modulation coding mode; and the display unit is used to present the encoded stimulus source to the user.

[0076] The workflow of the encoding mode setting unit includes: determining the minimum frequency resolution Δf based on the length of the acquired EEG signal to determine the step size of the basic frequency sequence b. In this embodiment, 0.2 Hz is selected as the step size; determining the maximum frequency f that can be induced based on the steady-state visual evoked potential. max and minimum frequency f min , in [f min ,f max ) interval, with Δf as the step size, obtain a series of frequency values ​​[f1,f2,f3,...,f n ], recorded as the basic frequency sequence b, where:

[0077]

[0078] In the basic frequency sequence b, there cannot be any two elements f i With f j , so that f i is f j If an integer multiple exists, a very small constant term (preferably 0.1 Hz) should be added to all elements in b. This can effectively avoid the problem of overlap between different features when encoding a large number of instructions. Specifically, the constant term should ensure that there is no integer multiple relationship between any two elements in the generated basic frequency sequence. In this embodiment, the minimum frequency is selected as 12.4 Hz, the maximum frequency is 14.6 Hz, and the step size is 0.2 Hz. There are 12 elements in b. The basic frequency sequence b of length n is expanded into an n×n matrix F:

[0079]

[0080] The elements in the matrix are a two-digit array: F ij =[f i , f j ]. Among them, the first element of the array f i It is called the main frequency component, and the second element f j It is called the frequency modulation component; the matrix F is called the edge frequency modulation coding matrix, and the number of elements in the matrix is ​​the maximum number of instructions that can be output by the brain-computer intelligent interaction system. Therefore, in this embodiment, a traditional brain-computer intelligent interaction system with 12 instructions is expanded into a brain-computer intelligent interaction system with 144 instructions output.

[0081] In the stimulus encoding unit, the design structure of the stimulus should include the following principles:

[0082] (1) Each stimulus source corresponds to the output of a brain control command, and the overall size and shape of each stimulus source are exactly the same;

[0083] (2) Each stimulus source contains two sub-patterns of the same shape but different sizes. The geometric centers of the two sub-patterns coincide, and the smaller sub-pattern covers the larger sub-pattern.

[0084] (3) Both sub-patterns flicker at a certain frequency. The flickering frequency of the smaller sub-pattern corresponds to the main frequency component of the corresponding element in the edge frequency modulation coding matrix F, and the flickering frequency of the larger sub-pattern corresponds to the frequency modulation component of the corresponding element in the edge frequency modulation coding matrix F.

[0085] In this embodiment, if Figure 2 As shown, a pair of concentric circles is used as the sub-pattern, with the smaller circle overlaying the larger one. Because the smaller circle flickers in the center of the visual field when the user is looking at the stimulus, the smaller circle flickers at the main frequency, while the larger circle flickers at the edge of the smaller circle, resulting in a modulated frequency.

[0086] The EEG acquisition module is used to collect EEG signals from the user receiving the stimulation source. It includes an EEG cap, an amplifier, and a data transmission interface. The EEG cap collects EEG signals from the user receiving the stimulation source; the amplifier amplifies the EEG signals; and the data transmission interface transmits the amplified EEG signals.

[0087] The EEG signal processing module is used to process and extract features of the EEG signal based on edge frequency modulation to obtain the frequency domain eigenvalues ​​of the EEG signal. The EEG signal processing module includes: a signal preprocessing unit and a feature extraction unit. The signal preprocessing unit is used to filter the EEG signal to obtain the processed EEG signal; the feature extraction unit is used to extract frequency domain features of the processed EEG signal based on a one-dimensional correlation matrix calculation method. The extracted eigenvalues ​​are the frequency domain eigenvalues ​​of all frequencies in the basic frequency sequence b, where the largest eigenvalue is the main frequency eigenvalue and the second largest eigenvalue is the frequency modulation eigenvalue.

[0088] Specifically, the one-dimensional correlation matrix calculation method includes the following steps:

[0089] Step 1: Define input and output variables

[0090] Let X be the EEG signal to be tested, {Y f1 ,Y f2 ,Y f3 ,...,Y fn} is a series of frequencies [f1,f2,f3,...,f n ] template EEG signal, the goal of the algorithm is to calculate the difference between the test signal X and each template signal {Y f1 ,Y f2 ,Y f3,...,Yfn}. The final output is a one-dimensional matrix R, where each element represents the correlation between X and the corresponding frequency template signal.

[0091] Step 2: Apply filter banks

[0092] Define a series of bandpass filters F1, F2, F3, ..., F m , each filter targets a specific frequency band. For the test signal X and each template signal Y f1 , respectively processed by these filters to extract relevant frequency band information. Fj Indicates that the signal X passes through the filter F j The processed result, Yfi, Fj Indicates that the frequency template signal Yfi passes through the filter F j The result after processing.

[0093] Step 3: Apply canonical correlation analysis to each filtered signal pair

[0094] For each filter F j , calculate X Fj With all Yfi, Fj Canonical correlation analysis is applied between them to obtain the correlation coefficient ρi j Among them, ρi j is signal X Fj With the template signal Yfi,F j After filter F j Correlation coefficient after processing.

[0095] Step 4: Calculate the final correlation coefficient

[0096] For each frequency template fi, the sum of all filters F j The final correlation measure R is calculated using the canonical correlation analysis results. i This is achieved by a weighted average method, where the weights can be determined based on the contribution of each frequency band or other optimization criteria, defined as:

[0097]

[0098] Among them, wj is the weight of the j-th filter, satisfying

[0099] The final one-dimensional matrix R can be expressed as:

[0100] R=[R1,R2,R3,…,R n ]

[0101] Among them, R iRepresents the signal X and the i-th frequency template Y fi The comprehensive correlation measure between them is used as the eigenvalue for subsequent analysis; the mean of all elements in the matrix is ​​recorded as μ r , which is used for the subsequent calculation of the regularization term.

[0102] The frequency-domain-spatial-domain encoding and decoding module is used to decode EEG signals into brain control commands based on frequency-domain eigenvalues. The frequency-domain-spatial-domain encoding and decoding module includes an adaptive threshold setting unit and a brain control command decoding unit. The adaptive threshold setting unit is used to construct a threshold target function and set the adaptive EEG feature threshold based on the threshold target function. The brain control command decoding unit is used to determine the relationship between the output of the main frequency eigenvalue and the adaptive EEG feature threshold, as well as the relationship between the output of the frequency modulation eigenvalue and the adaptive EEG feature threshold, and outputs the brain control command based on the judgment results.

[0103] The workflow of the adaptive threshold setting unit includes: constructing a threshold objective function Objective(θ) based on the adaptive EEG feature threshold θ to be set, which is expressed as:

[0104] Objective(θ)=α·Accuracy(θ)+β·Balance(θ)-γ·Regularization(θ)

[0105] Among them, Accuracy(θ) indicates the classification accuracy under the threshold θ, which can be calculated by the weighted sum of the true positive rate (TPR) and the true negative rate (TNR); Balance(θ) indicates the balance between positive and negative categories, which can be calculated using Balance(θ)=1-|TPR(θ)-TNR(θ)|, which encourages the model to consider the performance of both positive and negative classes, especially when the data categories are unbalanced; Regularization(θ) is a regularization term used to reduce model complexity and prevent overfitting. It can be expressed as Regularization(θ)=(θ-μ r ) 2 , μ r is the mean of the correlation coefficient; α, β, and γ are weight parameters used to adjust the accuracy, balance, and relative importance of the regularization term in the objective function; the adaptive EEG feature threshold is obtained by maximizing the threshold objective function.

[0106] The judgment method in the brain control instruction decoding unit includes: based on the adaptive EEG characteristic threshold obtained above, comparing whether the main frequency characteristic value is higher than the trigger threshold to determine whether to output the brain control instruction; judging whether the frequency modulation characteristic value is higher than the judgment threshold based on the linear discrimination method to determine whether the main frequency and frequency modulation of the stimulus source corresponding to the output instruction are the same, assuming that the frequency corresponding to the main frequency characteristic value is f1 and the frequency corresponding to the frequency modulation characteristic value is f2. If it is higher than the threshold, it is judged that the output is the main frequency f1 and the frequency modulation f2; otherwise, it is judged that the output is the main frequency and the frequency modulation f1. In this embodiment, Figure 3 The figure shows the result of EEG decoding using the one-dimensional correlation matrix calculation method when the user is looking at a stimulus source with a main frequency of 12.4Hz and a frequency modulation of 13.0Hz. It can be seen that the main frequency eigenvalue corresponding to the frequency f1 is 12.4Hz, and the frequency modulation eigenvalue corresponding to the frequency f2 is 13.0Hz. The frequency modulation eigenvalue exceeds the threshold, so the output is the brain control command corresponding to the element [12.4,13.0] in the edge frequency modulation encoding matrix. Similarly, Figure 4 This figure shows the results of EEG decoding using a one-dimensional correlation matrix calculation method when the user is gazing at a stimulus source with both a main frequency and a modulated frequency of 13.0 Hz. The decoded main frequency eigenvalue corresponds to frequency f1 of 13.0 Hz, and the FM eigenvalue corresponds to frequency f2 of 14.2 Hz. However, the FM eigenvalue does not exceed the threshold, so the output is the brain control command corresponding to the element [13.0, 13.0] in the edge FM encoding matrix.

[0107] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A brain-computer intelligent interaction system with a large instruction set based on edge frequency modulation, characterized in that: include: Human brain feature visual evoked module, EEG acquisition module, EEG signal processing module, frequency domain-spatial domain encoding and decoding module, and recognition result feedback module; The human brain characteristic visual elicitation module is used to encode the stimulus source based on edge frequency modulation and present the encoded stimulus source to the user; The EEG acquisition module is used to acquire EEG signals of the user who receives the stimulation source; The EEG signal processing module is used to process and extract features of the EEG signal based on the edge frequency modulation to obtain frequency domain feature values ​​of the EEG signal; The frequency domain-spatial domain encoding and decoding module is used to decode the EEG signal into a brain control command based on the frequency domain eigenvalue; The recognition result feedback module is used to display the brain control command in real time and mark the stimulation source to provide instant feedback to the user; The workflow of the encoding mode setting unit in the human brain feature visual elicitation module includes: Obtain the length of the user's EEG signal and determine the minimum frequency resolution Δf; Based on the minimum frequency resolution Δf, several frequency values ​​are obtained between the maximum frequency and the minimum frequency of the steady-state visual evoked potential, which are recorded as the basic frequency sequence b=[f1, f2, f3, ...,f n ],in: ； The basic frequency sequence b of length n is expanded into an n×n edge frequency modulation coding matrix F to complete the coding mode setting: ， The edge frequency modulation coding matrix F includes the main frequency component f i and FM component f j ; The workflow of the stimulus source encoding unit in the human brain feature visual elicitation module includes: The stimulus source includes two sub-patterns of identical shape but different sizes, wherein the sub-patterns include: a smaller first sub-pattern and a larger second sub-pattern, the geometric centers of the sub-patterns coincide, and the first sub-pattern covers the second sub-pattern; The flashing frequency of the first sub-pattern is the main frequency component f of the corresponding element in the edge frequency modulation coding matrix F. i The flashing frequency of the second sub-pattern is the frequency modulation component f of the corresponding element in the edge frequency modulation coding matrix F. j ; The EEG signal processing module includes: a signal preprocessing unit and a feature extraction unit; The signal preprocessing unit is used to filter the EEG signal to obtain a processed EEG signal; The feature extraction unit is used to extract frequency domain features from the processed EEG signal based on a one-dimensional correlation matrix calculation method, and the extracted eigenvalues ​​are the frequency domain eigenvalues ​​of all frequencies in the basic frequency sequence b, wherein the largest eigenvalue is the main frequency eigenvalue and the second largest eigenvalue is the frequency modulation eigenvalue; The frequency-space domain encoding and decoding module includes: an adaptive threshold setting unit and a brain control instruction decoding unit; The adaptive threshold setting unit is used to construct a threshold target function and set an adaptive EEG feature threshold based on the threshold target function; The brain control instruction decoding unit is used to determine the size relationship between the output of the main frequency characteristic value and the adaptive EEG characteristic threshold, as well as the size relationship between the output of the frequency modulation characteristic value and the adaptive EEG characteristic threshold, and output the brain control instruction based on the judgment result.

2. According to claim 1, a very large instruction set brain-computer intelligent interaction system based on edge frequency modulation is characterized in that: The human brain feature visual elicitation module includes: an encoding mode setting unit, a stimulus source encoding unit and a display unit; The coding mode setting unit is used to set the edge frequency modulation coding mode based on the minimum resolution; The stimulus source encoding unit is used to encode the stimulus source based on the edge frequency modulation encoding mode; The display unit is used to present the encoded stimulus source to the user.

3. The ultra-large instruction set brain-computer intelligent interaction system based on edge frequency modulation according to claim 1, characterized in that: In the stimulus source encoding unit: It is set that each of the stimulation sources corresponds to the output of one brain control command, and the overall size and shape of each of the stimulation sources are exactly the same.

4. The ultra-large instruction set brain-computer intelligent interaction system based on edge frequency modulation according to claim 1, characterized in that: The EEG acquisition module includes: an EEG cap, an amplifier and a data transmission interface; The EEG cap is used to collect EEG signals of the user who receives the stimulation source; The amplifier is used to amplify the EEG signal; The data transmission interface is used to transmit the amplified EEG signal.

5. The ultra-large instruction set brain-computer intelligent interaction system based on edge frequency modulation according to claim 1, characterized in that: The workflow of the adaptive threshold setting unit includes: Based on the adaptive EEG feature threshold θ to be set, the threshold objective function Objective(θ) is constructed, and its expression is: Among them, Accuracy(θ) represents the classification accuracy under the threshold θ, Balance(θ) represents the balance between positive and negative categories, Regularization(θ) is the regularization term, and α, β, and γ are weight parameters; The adaptive EEG feature threshold is obtained by maximizing the threshold objective function.

Images