Detection method, medium and electronic device of brain-computer interface based on electroencephalogram signal

Abstract

The present invention provides a detection method, medium and electronic device for a brain-computer interface based on EEG signals, the method comprising: calculating a template signal for each visual target; weighting and merging multi-channel EEG signals based on a spatial filter to generate a single-channel EEG signal; calculating a preset correlation coefficient between the EEG signal and each of the template signals; obtaining the minimum value of each distance metric, and determining the visual target corresponding to the minimum value as the recognition result. The present invention can improve the target recognition accuracy and information transmission rate of the brain-computer interface, and has stronger robustness.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, in particular to the field of signal detection technology. Background Art

[0002] Brain-computer interface is a new type of information output method that translates EEG signals collected from the surface of the human skull to achieve information output that is different from the normal language pathway and behavior pathway. Several main implementation methods of brain-computer interface can be divided into P300 brain-computer interface, motor imagery brain-computer interface and SSVEP (Steady-state visual evoked potential) brain-computer interface according to the type of EEG signals extracted. Among them, the SSVEP signal extracted by the SSVEP brain-computer interface has the largest signal-to-noise ratio and the highest interface communication rate that can be achieved. In addition to the EEG signal acquisition and analysis equipment, the SSVEP brain-computer interface requires additional visual stimulation equipment to evoke SSVEP signals. SSVEP signals refer to EEG signals with corresponding frequencies and high harmonics that can be detected in the visual area (occipital area) of the subject's head when the subject's eyes are fixed on a target that flickers at a fixed frequency. In a typical SSVEP brain-computer interface paradigm, a display device is placed in front of the subject to present visual stimuli. Generally, visual stimuli are flickering disks or squares with different frequencies and phases presented at different positions of the above-mentioned display device. For different application scenarios, each of the above different visual stimuli represents a different operation instruction. The subject outputs different operation instructions by looking at the corresponding visual stimulus. The background program synchronously collects and analyzes the subject's EEG signals continuously, and translates the recognition results into operation instructions and outputs them to external devices for function implementation.

[0003] The decoding and recognition algorithm used in the SSVEP brain-computer interface has a profound impact on the achievable brain-computer interface communication rate and the robustness of the corresponding control function. In general, the design goal of the decoding and recognition algorithm is to use fewer EEG acquisition channels, shorter EEG acquisition time, and smaller training data size to achieve higher target recognition accuracy and brain-computer interface communication rate. In the SSVEP brain-computer interface, the design difficulties of the decoding and recognition algorithm include the following points. First, there is currently no recognized objective signal model for SSVEP, and its occurrence and propagation mechanism is unknown; secondly, in addition to the SSVEP component, the collected EEG signal also contains a large amount of background noise and motion artifact signals, which interfere with judgment and recognition; in addition, the SSVEP signal is diffused on the surface of the head and is collected through dispersed electrode channels. How to comprehensively analyze the signals independently collected by each channel to obtain reliable decoding and recognition results is also one of the key issues that need to be solved.

[0004] In response to the above-mentioned problem of multi-channel joint decoding and recognition, the commonly used solution in the field is to calculate a set of multi-channel merging coefficients, linearly superimpose the EEG signals from multiple channels, enhance event-related components, and suppress noise components, thereby obtaining a higher signal-to-noise ratio and achieving more accurate target judgment and recognition. The above-mentioned set of calculated multi-channel merging coefficients is also called a spatial filter in the SSVEP brain-computer interface. However, there has been no consensus in the field on the selection criteria for the optimal spatial filter. The following is a brief introduction to the design and use of spatial filters, taking the CCA (canonical correlation analysis) algorithm, which is the most commonly used algorithm in the SSVEP brain-computer interface, as an example.

[0005] The CCA algorithm is an algorithm commonly used in statistics to analyze the linear relationship between two sets of high-dimensional random variables. For two sets of random variables x and y (the dimensions of x and y can be different), CCA finds a merge vector for each of them, that is, ω x ,ω y (ω x is equal to the dimension of the random variable x, ω y and the dimension of the random variable y), when using ω x ,ω y After merging x and y, ω x T x and ω y T y is a one-dimensional random variable, and the traditional Pearson correlation coefficient can be calculated between the two, while CCA chooses ω x ,ω y The standard is exactly ω x T x and ω y T The Pearson correlation coefficient ρ between ·y reaches the maximum. The CCA algorithm is applied to target recognition in SSVEP brain-computer interface. The two sets of high-dimensional random variables x and y considered are actually the collected multi-channel EEG signal s and the reference signal y(f) corresponding to a certain stimulation frequency f, s∈R n×t , y(f)∈R 2m×t , where n is the number of channels of the collected EEG signal, m is the harmonic order of the reference signal used, and t is the number of time sampling points of the signal. In general,

[0006]

[0007] where f sis the temporal sampling frequency of the EEG signal. It should be noted that at this time, for the EEG signal s, the optimal merging vector matched by the CCA algorithm is the spatial filter of the multi-channel EEG signal mentioned above. For the reference signal y(f) corresponding to the EEG signal s and the stimulation frequency f, the maximum correlation coefficient between the two obtained by the CCA algorithm is called ρ(f). To a certain extent, ρ(f) reflects the possibility that the collected EEG signal s is generated under the target stimulation of frequency f. Therefore, the basic strategy for SSVEP target detection using CCA is to 1 ,f 2 ,...,f k (k is the total number of targets), and the CCA algorithm is used to calculate the CCA correlation coefficients of the corresponding EEG signals s, that is, ρ(f 1 ),ρ(f 2 ),...,ρ(f k ), then the final judgment result obtained and output by the algorithm is

[0008] However, in the practice of SSVEP brain-computer interface, the CCA algorithm has the following shortcomings. The CCA algorithm always tries to dynamically match the optimal spatial filter for the acquired EEG signal s so that the correlation coefficient between it and the reference signal under consideration is maximized. The dynamic characteristics of CCA make it easily affected by the noise components in the signal, which reduces the accuracy of decoding recognition. Summary of the invention

[0009] In view of the shortcomings of the prior art mentioned above, the purpose of the present invention is to provide a detection method, medium and electronic equipment for a brain-computer interface based on EEG signals, so as to improve the target recognition accuracy and information transmission rate of the brain-computer interface.

[0010] To achieve the above-mentioned purpose and other related purposes, the present invention provides a detection method for a brain-computer interface based on EEG signals, including: calculating a template signal for each visual target; weightedly combining multi-channel EEG signals based on a spatial filter to generate a single-channel EEG signal; calculating a distance metric between the EEG signal and each of the template signals; obtaining the minimum value of each of the preset distance metrics, and determining the visual target corresponding to the minimum value as the recognition result.

[0011] In one embodiment of the present invention, the method of calculating the template signal for each visual target is as follows: Among them, Γ i,0 is the template signal of the i-th visual target, Γ i,m is the mth training data of the ith visual stimulus target, Nt is the total number of independent training data.

[0012] In one embodiment of the present invention, the template signal Γ i,0 and training data Γ i,m Each is N c ×T-dimensional real matrix; where N c is the number of channels of the EEG signal, and T is the number of time sampling points of the collected EEG signal.

[0013] In one embodiment of the present invention, the distance metric d i The calculation method is ω i T is the spatial filter ω corresponding to the i-th target i , S is the transpose of the collected EEG signal to be classified.

[0014] In one embodiment of the present invention, the distance metric is calculated using the square of the norm of the difference between the EEG signal merged by the spatial filter and the template signal. The norm can be any norm with a standard definition in mathematics.

[0015] In one embodiment of the present invention, i represents the spatial filter of the i-th target, ω i is a real vector, R represents a real vector, N c is the number of channels of EEG signal.

[0016] In one embodiment of the present invention, the following objective function is minimized by adjusting the spatial filter:

[0017]

[0018] Where: N t is the total number of independent training data, ω i T is the spatial filter ω corresponding to the i-th target i The transpose of is the mth target corresponding to the i-th target 1 training data, is the mth target corresponding to the i-th target 2 training data, is a norm with a canonical definition, x represents the specific type of the norm, calculate and When using the spatial filter ω i The square of the norm distance after merging.

[0019] In one embodiment of the present invention, it further includes: configuring a regularization constraint condition for the objective function, wherein the regularization constraint condition is ω i T ·Γ i ·Γ i T ·ω i =1, where Γ i is all N that will correspond to the i-th target t The matrix obtained by concatenating training data, where Represent the 1st, 2nd, ..., Nth corresponding to the i-th target respectively t training data, Γ i T is Γ i The transpose of i is the spatial filter corresponding to the i-th target.

[0020] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for detecting a brain-computer interface based on electroencephalogram signals.

[0021] An embodiment of the present invention also provides an electronic device, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-mentioned brain-computer interface detection method based on electroencephalogram signals.

[0022] As described above, the detection method of the brain-computer interface based on EEG signals of the present invention has the following beneficial effects:

[0023] 1. The present invention can improve the target recognition accuracy and information transmission rate of the brain-computer interface, and compared with the traditional algorithm, the present invention requires fewer EEG signal acquisition channels, shorter EEG signal acquisition time, and smaller training data size.

[0024] 2. Compared with the traditional dynamic filter algorithm, the present invention can better capture the static characteristics of the SSVEP response, such as its intensity distribution and the statistical parameters of the brain background noise, and therefore has stronger robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Shown is a flow chart of the detection method of the brain-computer interface based on EEG signals in the present invention.

[0026] Figure 2 Shown is a schematic diagram of the basic paradigm of the SSVEP brain-computer interface in the detection method of the brain-computer interface based on EEG signals in the present invention.

[0027] Figure 3 Shown is a schematic diagram of the decoding and recognition process of the electroencephalogram signal-based brain-computer interface detection method of the present invention based on a static spatial filter.

[0028] Figure 4 Shown is a schematic diagram of the principle of the electronic device in the present invention.

[0029] Component number description

[0030] 101 Electronic Equipment

[0031] 1001 Processor

[0032] 1002 Memory

[0033] S100~S400 steps DETAILED DESCRIPTION

[0034] The following is a description of the implementation of the present invention by means of specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.

[0035] See also Figures 1 to 4 . It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings in this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention, so they have no substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size should still fall within the scope of the technical content disclosed by the present invention without affecting the effects and purposes that can be achieved by the present invention. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" quoted in this specification are only for the convenience of description, and are not used to limit the scope of the implementation of the present invention. Changes or adjustments in their relative relationships should also be regarded as the scope of the implementation of the present invention without substantially changing the technical content.

[0036] In order to continue to improve the target recognition accuracy and information transmission rate of the SSVEP brain-computer interface, as well as the technical robustness of the corresponding brain control function. This embodiment aims at the instability of the dynamic spatial filter used in the decoding and recognition algorithm of the SSVEP brain-computer interface. This embodiment proposes a new SSVEP brain-computer interface decoding and recognition scheme based on a static spatial filter, as well as a specific design algorithm for the spatial filter.

[0037] The purpose of this embodiment is to provide a detection method, medium and electronic equipment of a brain-computer interface based on EEG signals, which are used to improve the target recognition accuracy and information transmission rate of the brain-computer interface.

[0038] The following will describe in detail the detection method, medium, and electronic device principles and implementation methods of a brain-computer interface based on EEG signals in this embodiment, so that those skilled in the art can understand the detection method, medium, and electronic device of a brain-computer interface based on EEG signals in this embodiment without creative work.

[0039] Example 1

[0040] like Figure 1 As shown, this embodiment provides a method for detecting a brain-computer interface based on an electroencephalogram signal, comprising the following steps:

[0041] Step S100, calculating a template signal for each visual target;

[0042] Step S200, weighted merging of multi-channel EEG signals based on a spatial filter to generate a single-channel EEG signal;

[0043] Step S300, calculating the distance metric between the EEG signal and each of the template signals;

[0044] Step S400, obtaining the minimum value of each distance metric, and determining the visual target corresponding to the minimum value as the recognition result.

[0045] The detection method of the brain-computer interface based on EEG signals in this embodiment is a decoding and recognition scheme based on a static spatial filter. The decoding and recognition task is to detect a series of newly collected EEG signals. Determine which visual target stimulus produced the phenomenon.

[0046] The following is a detailed description of steps S100 to S400 in the method for detecting a brain-computer interface based on EEG signals of this embodiment.

[0047] Step S100, calculating a template signal for each visual target.

[0048] The usage scenario of this embodiment, i.e., a typical paradigm of SSVEP brain-computer interface, is as follows Figure 2 As shown, a monitor is placed in front of the subject, and different visual stimulus patterns are displayed at different positions on the monitor. Figure 2The squares in the figure flash at different frequencies and phases. Each visual stimulus (also called a visual target) represents a control instruction, and the translation of the control instruction is related to the specific application scenario. When the subject wants to output a certain control instruction, he or she completes the operation by staring at the corresponding visual stimulus and maintaining it for a certain period of time. When the subject stares at the visual stimulus, an SSVEP signal corresponding to the frequency and phase of the visual stimulus is generated in the occipital area of ​​the subject's brain. The signal is collected by having the subject wear an electrode cap, and subsequent judgment and recognition are performed to finally obtain the recognition result, and the instruction is translated and output.

[0049] The detection method of the brain-computer interface based on EEG signals of this embodiment is based on training, that is, each subject needs to go through a training phase before using the SSVEP brain-computer interface to perform tasks. In this training phase, the subject is required to gaze at each target in the visual stimulation provided by the SSVEP brain-computer interface, and the gaze is maintained for a certain period of time (generally within 5 seconds), and the gaze may be repeated several times (generally within 10 times). The EEG signals of the subjects in the training phase are synchronously collected and used as training data after cutting.

[0050] In this embodiment, let Γ i,m It represents the mth training data corresponding to the i-th visual stimulus target of the subject, considering a total of N f visual targets, and a total of N t independent training data, that is, 1≤i≤N f , and 1≤m≤N t . Each training data Γ i,m is an N c ×T-dimensional real number matrix, representing the multi-channel EEG signals collected, where N c is the number of channels of the EEG signal, and T is the number of time sampling points of the collected EEG signal.

[0051] Specifically, in this embodiment, firstly, a template signal is calculated for each visual target. The template signal is calculated by averaging all training data of the target. The template signal is calculated for each visual target in the following manner:

[0052]

[0053] Among them, Γ i,0 is the template signal of the i-th visual target, Γ i,m is the mth training data of the ith visual stimulus target, N t is the total number of independent training data.

[0054] In this embodiment, the template signal Γ i,0 For an Nc ×T-dimensional real matrix; where N c is the number of channels of the EEG signal, T is the number of time sampling points of the collected EEG signal, and the training data Γ i,m For an N c ×T-dimensional real matrix; where N c is the number of channels of the EEG signal, and T is the number of time sampling points of the collected EEG signal.

[0055] Step S200: weightedly combine multi-channel EEG signals based on a spatial filter to generate a single-channel EEG signal.

[0056] The study showed that the choice of spatial filter should be closely related to the intensity distribution of SSVEP response and the local noise statistics of the brain. The static properties of the intensity distribution of SSVEP response and the local noise statistics of the brain indicate that static spatial filters will be able to achieve better decoding and recognition performance.

[0057] In this embodiment, ω i represents the spatial filter of the i-th target, ω i is a real vector, R is a real vector, N c is the number of channels of EEG signal.

[0058] It is worth noting that when using the training data to obtain the spatial filter ω i After that, the spatial filter ω in the subsequent recognition task i It no longer changes, so the spatial filter of this embodiment is a static spatial filter.

[0059] The following example describes how to use the above training data to obtain the spatial filter ω i To illustrate. The role of the spatial filter is to enhance the event-related potential component and suppress the noise component by merging the multi-channel EEG signals. Ideally, the noise in the single-channel signal obtained after merging with the spatial filter is completely filtered out, and the signal is the event-related potential itself. In this case, the multi-channel EEG signal processed by the spatial filter has an infinite signal-to-noise ratio, and the target recognition accuracy and information transmission rate can reach arbitrarily high. Although the above situation is an ideal situation, it can still inspire the design of the calculation of the spatial filter. In this embodiment, the EEG signals processed by the spatial filter are as similar as possible to each other in the training set, because the similarity reflects the effectiveness of the spatial filter in extracting the event-related potential component to a certain extent.

[0060] Specifically, this embodiment minimizes the following objective function by adjusting the spatial filter, namely:

[0061]

[0062] Where: N t is the total number of independent training data, ω i T is the spatial filter ω corresponding to the i-th target i The transpose of is the mth target corresponding to the i-th target 1 training data, is the mth target corresponding to the i-th target 2 training data, is a norm with a canonical definition, x represents the specific type of the norm, calculate and When using the spatial filter ω i The square of the norm distance after merging.

[0063] In this embodiment, is any well-defined norm, calculate and When using the spatial filter ω i The square of the norm distance after merging. The smaller the norm distance, the and The more similar the ingredients are after merging.

[0064] Without any constraints, each element of the optimal spatial filter generated by the above objective function is zero, which has no practical significance. To solve this problem, this implementation adds a regularization constraint.

[0065] Specifically, in this embodiment, it also includes: configuring a regularization constraint condition for the objective function, and the regularization constraint condition is ω i T ·Γ i ·Γ i T ·ω i =1, where Γ i is all N that will correspond to the i-th target t The matrix obtained by concatenating training data, where Represent the 1st, 2nd, ..., Nth corresponding to the i-th target respectively t training data, Γ i T is Γ i The transpose of i is the spatial filter corresponding to the i-th target.

[0066] make represents the concatenation of all training data corresponding to the i-th target, then the regularization constraint can be expressed as ω i T ·Γ i ·Γ i T ·ω i =1, that is, using the spatial filter ω i Γ i The total energy of the signal obtained after multi-channel merging is 1. The optimal ω i That is, the spatial filter found by the framework for the i-th target.

[0067] Step S300, calculating the distance metric d between the EEG signal and each of the template signals i .

[0068] In this embodiment, the distance metric is the square of the norm of the difference between the EEG signal and the template signal after the spatial filter is used to merge. The norm can be any norm type with a standard definition in mathematics.

[0069] Specifically, ω i T is the spatial filter ω corresponding to the i-th target i , S is the transpose of the collected EEG signal to be classified.

[0070] Step S400, obtaining the minimum value of each distance metric, and determining the visual target corresponding to the minimum value as the recognition result.

[0071] like Figure 3 As shown, the specific implementation process of the detection method of the brain-computer interface based on EEG signals in this embodiment is as follows:

[0072] Before the subjects actually use the SSVEP brain-computer interface to realize the corresponding control function, the subjects are asked to look at each target several times in turn to generate SSVEP response signals, which are collected through the electrode cap as the basis for algorithm design. In order to calculate the optimal spatial filter, the proposed framework proposes that it can be solved by solving a constrained optimization problem. Specifically, the objective function of the optimization problem is the sum of the square of the norm distance between the training data in the training set after merging using the spatial filter, and the constraint is that the signal energy after the training data is concatenated and then merged using the spatial filter is 1. By solving the above optimization problem, a matching spatial filter can be found for each visual target in the SSVEP brain-computer interface. It should be noted that the framework does not specify the specific norm type used in the objective function here. Any norm with a standardized definition can be used as a new version of the objective function to guide the design of the spatial filter. Of course, there will be differences in the specific optimization problems, solution methods, and performance levels formed.

[0073] Here, this embodiment selects the Euclidean norm which is commonly used in practice to make a more specific introduction to the calculation of the proposed spatial filter. After selecting the Euclidean norm, the optimization problem to be solved is rewritten as follows:

[0074]

[0075] stω i T ·Γ i ·Γ i T ·ω i =1,

[0076] make in And Q = Γ i ·Γ i T , then the above problem can be simplified as:

[0077]

[0078] stω i T ·Q·ω i =1.

[0079] The simplified problem above is a classic constrained Rayleigh quotient optimization problem, and its optimal solution can be expressed analytically, corresponding to the matrix Q -1The eigenvector of the smallest eigenvalue of P. After the spatial filter calculation is completed, the newly collected EEG test signal s can be judged and identified according to the process described below. The steps include template signal calculation, test signal spatial filtering, correlation coefficient calculation, and recognition result output. The details are as follows:

[0080] First, the template signal is calculated for each visual target. The template signal is calculated by averaging all the training data of the target, that is, Γ i,0 is the template signal corresponding to the i-th visual target, Γ i,0 The same is an N c Next, this embodiment calculates the distance metric d between the EEG signal s and each template signal i , and before calculating the correlation coefficient, the corresponding spatial filter is used to weight the multi-channel EEG signals to generate a single-channel EEG signal, as follows, in Calculate the square value of the Euclidean norm. Finally, the target with the smallest distance metric is output as the recognition result, that is, π is the result of recognition judgment, that is, it is believed that the EEG signal s to be judged is generated by the πth visual target stimulus.

[0081] The detection method of the brain-computer interface based on EEG signals in this embodiment can be applied to any SSVEP brain-computer interface system to undertake the tasks of decoding analysis and target recognition. Different from the traditional decoding algorithm based on spatial filters, this embodiment proposes a new optimization logic to solve the definition and calculation problems of the optimal spatial filter. Specifically, the method of this embodiment is based on training. The objective function of the optimization problem is the sum of the square of the norm distance between the training data in the training set after being merged using the spatial filter, and the constraint is that the signal energy after the training data is concatenated and then merged using the spatial filter is 1. It should be noted that this embodiment does not specifically specify the type of norm used in the objective function. Any norm with a standard definition in mathematics can be applied here, which will accordingly generate a new optimization problem, a new spatial filter solution, and different SSVEP decoding and recognition performance. After obtaining the spatial filter, the spatial filter will no longer change during the subsequent use of the SSVEP brain-computer interface. The decoding and recognition process is shown in the attached figure. Figure 3As shown, it can be summarized as template signal calculation, merging of signals to be identified, correlation coefficient calculation, and output of judgment results. Specifically, template signal calculation refers to averaging the training data corresponding to each visual target to obtain a multi-channel template EEG signal corresponding to each visual target; merging of signals to be identified refers to using the corresponding spatial filter to merge the newly collected multi-channel EEG signals to obtain a single-channel signal; correlation coefficient calculation refers to calculating the correlation coefficient between the merged signal to be identified and the merged template signal (it can be a general Pearson correlation coefficient or any other correlation coefficient with a standard mathematical definition); output of judgment results refers to outputting the target corresponding to the maximum value obtained in the above correlation coefficient calculation as the result, and translating it into specific control instructions to interact with the outside world.

[0082] Compared with the traditional dynamic filter algorithm (such as the above-mentioned CCA algorithm, etc.), this embodiment can better capture the static characteristics of the SSVEP response, such as its intensity distribution, and the statistical parameters of the brain background noise, so it has stronger robustness. In addition, compared with the traditional recognition algorithm based on static filters, the proposed framework proposes a new design logic, and the optimization objective function is selected as the sum of the square of the norm distance between the two pairs on the training set. In order to ensure the existence of a non-zero solution of the spatial filter, a regularization condition is added, requiring that the energy of the single channel signal obtained after the spatial filter is combined is 1. The proposed decoding recognition framework and design algorithm have been verified on the SSVEP data set recognized in the field. The results show that the proposed method can effectively extract the SSVEP signal hidden in the noise. Under different signal lengths, different training data scales, and different channel numbers, the proposed method can achieve a higher target recognition accuracy than the traditional algorithm, which means that in order to achieve the same level of target recognition accuracy, the proposed method requires fewer EEG signal acquisition channels, shorter EEG signal acquisition time length, and smaller training data scale than the traditional algorithm.

[0083] Example 2

[0084] like Figure 4 As shown, this embodiment provides an electronic device 101, and the electronic device 101 includes: a processor 1001 and a memory 1002; the memory 1002 is used to store computer programs; the processor 1001 is used to execute the computer programs stored in the memory 1002, so that the electronic device 101 performs the steps of the detection method of the brain-computer interface based on EEG signals in Example 1. Since the specific implementation process of the steps of the detection method of the brain-computer interface based on EEG signals has been described in detail in Example 1, it will not be repeated here.

[0085] The processor 1001 is a central processing unit (CPU). The memory 1002 is connected to the processor 1001 through a system bus and communicates with each other. The memory 1002 is used to store computer programs, and the processor 1001 is used to run computer programs so that the processor 1001 executes the detection method of the brain-computer interface based on EEG signals. The memory 1002 may include a random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage.

[0086] In addition, this embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by the processor 1001, the steps in the method for detecting a brain-computer interface based on an EEG signal described in Example 1 are implemented. Example 1 has already described the method for detecting a brain-computer interface based on an EEG signal in detail, which will not be repeated here.

[0087] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

[0088] In summary, the present invention can improve the target recognition accuracy and information transmission rate of the brain-computer interface, and compared with the traditional algorithm, the present invention requires fewer EEG signal acquisition channels, shorter EEG signal acquisition time, and smaller training data size; compared with the traditional dynamic filter algorithm, the present invention can better capture the static characteristics of the SSVEP response, such as its intensity distribution and statistical parameters of the brain background noise, and therefore has stronger robustness. Therefore, the present invention effectively overcomes the shortcomings of the prior art and has a high industrial utilization value.

[0089] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for detecting a brain-computer interface based on electroencephalogram signals, characterized in that: include: Calculate the template signal for each visual target; The method for calculating the template signal for each visual target is as follows: Among them, Γ i，0 is the template signal of the i-th visual target, Γ i，m is the mth training data of the ith visual stimulus target, N t is the total number of independent training data; Based on the spatial filter, multi-channel EEG signals are weighted and combined to generate a single-channel EEG signal; i represents the spatial filter of the i-th target, ω i is a real vector, R represents a real vector, N c is the number of channels of EEG signal; Calculating a distance metric between the EEG signal and each of the template signals; the distance metric is the square of the x-norm of the difference between the EEG signal and the template signal after merging using the spatial filter, where the x-norm is any norm with a standard definition in mathematics; Obtaining the minimum value of each distance metric, and determining the visual target corresponding to the minimum value as the recognition result; The distance metric d i The calculation method is where ω i T is the spatial filter ω corresponding to the i-th target i The transpose of , S is the collected EEG signal to be classified; The spatial filter is adjusted to minimize the following objective function: Where: N t is the total number of independent training data, ω i T is the spatial filter corresponding to the i-th target, is the m1th training data corresponding to the i-th target, is the m2th training data corresponding to the i-th target, is a norm with a canonical definition, x represents the specific type of the norm, calculate and When using the spatial filter ω i The square of the norm distance after merging; A regularization constraint condition is configured for the objective function, and the regularization constraint condition is ω i T ·Γ i ·Γ i T ·ω i =1, where Γ i is all N that will correspond to the i-th target t The matrix obtained by concatenating training data, where Represent the 1st, 2nd, ..., Nth corresponding to the i-th target respectively t training data Γ i T is Γ i The transpose of i is the spatial filter corresponding to the i-th target.

2. The method for detecting a brain-computer interface based on an electroencephalogram signal according to claim 1, characterized in that: Template signal Γ i，0 and training data Γ i，m Each is N c ×T-dimensional real matrix; where N c is the number of channels of the EEG signal, and T is the number of time sampling points of the collected EEG signal.

3. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for detecting a brain-computer interface based on electroencephalogram signals as described in any one of claims 1 to 2 is implemented.

4. An electronic device, characterized in that: include: Processor and memory; The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory, so that the electronic device executes the method for detecting a brain-computer interface based on electroencephalogram signals as described in any one of claims 1 to 2.

Images