Steady-state visual evoked potential-oriented electroencephalogram feature decoding method and brain-computer interface

Abstract

This invention discloses a method for decoding EEG features for steady-state visual evoked potentials, including: the subject's fixation frequency being f n The original steady-state visual evoked potentials collected during visual stimulation are X. n Using N fb A filter bank for X n Preprocessing is performed to obtain N fb After preprocessing, the steady-state visual evoked potential pairs are filtered to obtain useful signal components and noise signal components, and then a spatial filter is obtained to obtain the steady-state visual evoked potential template signal. This is used to acquire the unknown steady-state visual evoked potential K and obtain N. fb K is a preprocessed unknown steady-state visual evoked potential. (m) By improving and K (m) The signal-to-noise ratio, calculate N fb The filtered sum K (m) The correlation coefficient of N fb The weighted summation yields ρ n ; We obtained K and steady-state visual evoked potentials [X1,…X i ,…,X n ,...,X Nf The correlation coefficient between ], ρ i If the maximum value is reached, then the visual stimulus frequency of K is f. i The technical solution in this embodiment achieves the effect of improving the accuracy of recognizing unknown visual stimuli using a small amount of training data.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the coding and decoding technology of brain-computer interface signals, in particular to a brain electrical feature decoding method for steady-state visually evoked potentials and a brain-computer interface system. BACKGROUND

[0002] In recent years, with the in-depth research of researchers on brain waves (Electroencephalogram, EEG), the non-invasive brain-computer interface (Brain Computer Interface, BCI) based on scalp EEG has developed rapidly. Among them, the BCI based on visually evoked potential (Visually Evoked Potential, VEP) has attracted widespread attention due to its high information transmission rate and decoding reliability. The steady-state visually evoked potential (Steady-State Visual Evoked Potentials, SSVEP) BCI paradigm has become one of the mainstream paradigms of BCI because of its advantages such as stable evoked characteristics, high signal-to-noise ratio, etc.

[0003] In the prior art, the SSVEP decoding algorithm represented by the task-related component analysis (Task-Related Component Analysis, TRCA) algorithm needs to rely on a large amount of training data to obtain a high recognition accuracy. Compared with using standard sine and cosine as templates, pre-training helps the algorithm to learn and generate spatial filters and EEG template signals that conform to the EEG characteristics of the subject from the individual EEG data, so as to make the algorithm obtain satisfactory recognition results. Since the principle of the TRCA algorithm itself is to maximize the repeated signal components between trials, the recognition performance will often decrease significantly when the training data is less, especially when there is only single-trial training data, the recognition performance is very poor. SUMMARY

[0004] The present application provides a brain electrical feature decoding method for steady-state visually evoked potentials and a brain-computer interface system to achieve the technical effect of improving the unknown visual stimulus recognition accuracy.

[0005] In a first aspect, the embodiment of the present application provides a brain electrical feature decoding method for steady-state visually evoked potentials, comprising:

[0006] presenting N f visual stimuli with frequencies f1,..., f i ,..., f n ,..., f Nf on a screen, collecting original steady-state visually evoked potential signals of a subject when the subject gazes at the visual stimuli as training data, and the visual frequency of the subject is f nThe primitive steady-state visual evoked potentials of the subjects' multi-channel EEG collected during visual stimulation were X. n Using N fb Each filter bank affects the original steady-state visual evoked potential X. n Preprocessing is performed to obtain N fb Preprocessed steady-state visual evoked potentials in specific frequency bands Where m = 1, 2, ..., N fb ;

[0007] The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the preprocessed steady-state visual evoked potentials. The useful signal components and noise signal components, wherein the useful signal components include the visual stimulus frequency f n The signal components of the noise signal and its harmonics, wherein the noise signal components include signal components other than the useful signal components;

[0008] A spatial filter capable of enhancing the signal-to-noise ratio is obtained based on the useful signal components and the noise signal components.

[0009] The pre-processed steady-state visual evoked potentials The steady-state visual evoked potential template signal is obtained by averaging the results across trials.

[0010] The unknown steady-state visual evoked potential K of the subject was collected, and the N was used. fb A filter bank preprocesses the unknown steady-state visual evoked potential K to obtain N. fb Unknown steady-state visual evoked potential K after preprocessing in a specific frequency band (m) Where m = 1, 2, ..., N fb ;

[0011] Through the spatial filter Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio, calculate N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient For N fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n ;

[0012] The unknown steady-state visual evoked potential K and its frequencies are obtained as f1, ... f1. i , ...f n ,,...f Nf All original steady-state visual evoked potentials [X1, ... X i , ..., X n , ..., X Nf The correlation coefficients between [ρ1, ... ρ] are [ρ1, ... ρ]. i , ..., ρ n , ..., ρ Nf The i-th correlation coefficient ρ i If the maximum value is reached, then the visual stimulus frequency of the unknown steady-state visual evoked potential K is f. i .

[0013] Secondly, embodiments of the present invention also provide a brain-computer interface system for steady-state visual evoked potentials, comprising:

[0014] One or more processors;

[0015] Memory, used to store one or more programs.

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the EEG feature decoding method for steady-state visual evoked potentials as described in any of the embodiments.

[0017] The technical solution of this embodiment involves applying a frequency of f to the subject. n Visual stimuli were used to collect the subjects' raw steady-state visual evoked potentials (X-rays). n Using N fb Each filter bank affects the original steady-state visual evoked potential X. n Preprocessing is performed to obtain N fb Preprocessed steady-state visual evoked potentials in specific frequency bands Where m = 1, 2, ..., N fb ; for the preprocessed steady-state visual evoked potentials Filtering is performed to obtain the preprocessed steady-state visual evoked potentials. Different signal components, including useful signal components and noise signal components; a spatial filter capable of enhancing the signal-to-noise ratio is obtained based on the useful signal components and the noise signal components. The unknown steady-state visual evoked potential K of the subject is collected and passed through the spatial filter. The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained. n The correlation coefficient ρ between them nThe technical problem of great decline of recognition performance when using the TRCA algorithm or the eTRCA algorithm when the training data is less, especially when only single trial training data is available, and poor recognition performance is solved, and a technical effect of improving the unknown visual stimulus recognition accuracy is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 It is a schematic diagram of the composition structure of a steady-state visual evoked potential (SSVEP) brain-computer interface system.

[0019] Figure 2 It is a schematic diagram of the signal-to-noise ratio characteristics of a typical steady-state visual evoked potential in the frequency domain.

[0020] Figure 3 It is a flowchart of a steady-state visual evoked potential-oriented electroencephalogram feature decoding method provided by the embodiment one of the application.

[0021] Figure 4 a is a flowchart for obtaining a spatial filter according to a useful signal component and a noise signal component .

[0022] Figure 4 b is a flowchart for obtaining a spatial filter according to a useful signal component a first noise signal component and a second noise signal component .

[0023] Figure 5 a is a flowchart for obtaining a steady-state visual evoked potential template signal and a correlation coefficient of a filtered preprocessed unknown steady-state visual evoked potential K by using a spatial filter (m) .

[0024] Figure 5 b is another flowchart for obtaining a steady-state visual evoked potential template signal and a correlation coefficient of a filtered preprocessed unknown steady-state visual evoked potential K (m) by using a spatial filter .

[0025] Figure 6 It is a percentage correct rate heat map of unknown visual stimulus recognition of different algorithms.

[0026] Figure 7 (a) is a schematic diagram of the influence of the length of the visual stimulus signal on the correct recognition rate under the condition of 1 trial and 9 electroencephalogram channels for different algorithms.​​​​

[0027] Figure 7 (b) is a schematic view of the influence of the length of visual stimulation signal on information transmission efficiency of different algorithms under the condition of 1 trial number and 9 brain electrical channel numbers;

[0028] Figure 8 It is a schematic view of the influence of the number of brain electrical channels on the correct recognition rate under the condition of 1 trial number;

[0029] Figure 9 It is an experimental result graph of the signal-to-noise ratio of the spatial filter filtered by different algorithms under the condition of 1 trial number;

[0030] Figure 10 It is a structural schematic view of a brain electrical feature decoding device for steady-state visual evoked potential provided by the second embodiment of the application;

[0031] Figure 11 It is a structural schematic view of a brain-computer interface system for steady-state visual evoked potential provided by the third embodiment of the application. DETAILED DESCRIPTION

[0032] The application will be further described below in conjunction with the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the application, and not to limit the application. In addition, it should be noted that, for the convenience of description, only the parts related to the application are shown in the drawings, not all the structures.

[0033] Embodiment one

[0034] The BCI based on scalp EEG can establish an information transmission channel between the brain and the external device non-invasively, thereby helping the patients with severe movement disorders or other healthy people in need to communicate with the outside environment. Figure 1 It is a structural schematic view of a steady-state visual evoked potential (SSVEP) brain-computer interface system, which mainly comprises a stimulation module, an acquisition module and a processing module. The stimulation module refers to the display encoding a fixed frequency visual stimulation. When the subject gazes at the visual stimulation, the steady-state visual evoked potential can be induced in the occipital region of the subject's brain. The acquisition module acquires the EEG physiological signals representing the brain thinking and cognitive activities from the subject's brain cortex or brain cavity through the electrode. The processing module extracts the corresponding EEG feature signals from the EEG physiological signals through a series of signal processing methods and converts them into machine language instructions through pattern recognition.

[0035] Figure 2A typical steady-state visual evoked potential signal in frequency domain. The steady-state visual evoked potential (SSVEP) signal is induced by a fixed frequency visual stimulus. When the visual stimulus is presented periodically at a specific frequency (such as flashing, pattern reversing, pattern scaling, etc.), the visual system of the subject will be affected to produce an induced response with stable frequency domain characteristics, which contains the frequency component of the visual stimulus and its high-order harmonic components. Figure 2 The result of averaging multiple trials of steady-state visual evoked potentials. Averaging multiple trials can offset random noise and improve the signal-to-noise ratio. In the case of a single trial, the signal-to-noise ratio may not be high enough to achieve the desired effect. Figure 2 The high signal-to-noise ratio effect shown at the visual stimulus frequency and the high-order harmonic frequency.

[0036] The steady-state visual evoked potential also has spatial characteristics. For example, different frequency components of the SSVEP signal come from different brain regions of the subject, so when the SSVEP signal is transmitted to different parts of the subject's scalp, the signal-to-noise ratio of each harmonic is not the same.

[0037] Figure 3 A flowchart of a steady-state visual evoked potential EEG feature decoding method provided by Embodiment One of the present application. This embodiment can be used to provide a spatial filter to decode the EEG features of the steady-state visual evoked potential. The training of the spatial filter can be based on experimental data from a small number of trials, or even a single trial. The method can be executed by a computer and specifically includes the following steps:

[0038] Steps S310-S330 are the process of training the spatial filter, and steps S340-S370 use the trained spatial filter to filter and analyze the unknown steady-state visual evoked potential, thereby obtaining the meaning represented by the unknown steady-state visual evoked potential, i.e., the frequency information of the visual stimulus being gazed at by the subject.

[0039] S310, present N f frequency visual stimuli with frequencies f1,... f i ,... f n ,... f Nf on the screen, collect the raw steady-state visual evoked potential signals of the subject gazing at these visual stimuli as training data, and the gazing frequency of the subject is f n The raw steady-state visual evoked potential of the subject collected when the subject is gazing at the visual stimulus with frequency f n is X fb , and N n filter banks are used to preprocess the raw steady-state visual evoked potential X fb to obtain N Wherein, m = 1, 2,..., N fb .

[0040] Assuming that the frequency of the subject's attention is f n , the subject's electroencephalogram at this time is collected, and the original steady-state visual evoked potential X is obtained, wherein n = 1, 2,..., N f ; N f is the number of frequency types of the visual stimulus, N c , N s , N t are the number of electroencephalogram channels, the number of sampling points, and the number of trials, respectively.

[0041] Optionally, before the original steady-state visual evoked potential X n is preprocessed, the collected multi-electroencephalogram X n may be band-pass filtered and power frequency notched to filter out obvious noise signals. For example, the frequency of the visual stimulus and each higher harmonic is usually between 8Hz and 80Hz, so the multi-electroencephalogram X n is band-pass filtered in a wide range to obtain signals with a frequency between 8Hz and 80Hz.

[0042] N fb filter banks are used to preprocess the original steady-state visual evoked potential X n , that is, the original steady-state visual evoked potential X n is divided into different sub-band signals to obtain N fb sub-band signals of specific frequency bands Wherein, m = 1, 2,..., N fb ; N fb is the number of filter banks, and the filters in the filter bank can be set according to actual needs. For example, there are 8 filters in a filter bank with m = 2, which are 16Hz-80Hz, 24Hz-80Hz, 32Hz-80Hz, 40Hz-80Hz, 48Hz-80Hz, 56Hz-80Hz, 64Hz-80Hz and 72Hz-80Hz. The above operation of preprocessing the original steady-state visual evoked potential X n using a filter bank is a relatively recognized signal processing method in the art, and will not be described here.

[0043] S320, filtering the preprocessed steady-state visual evoked potential X to obtain the useful signal component and the noise signal component of the preprocessed steady-state visual evoked potential X , the useful signal component includes the visual stimulus frequency f nand harmonic signal components, the noise signal components including other signal components except the useful signal components.

[0044] Due to the original steady-state visual evoked potential X n is a multi-electroencephalogram channel signal, i.e., N c ≥ 2, therefore, the filtered useful signal components and noise signal components are also multi-electroencephalogram channel signals, and have the same dimension as the original steady-state visual evoked potential X n . The filtering operation of step S320 further distinguishes the useful signal and the noise signal by utilizing the characteristics that the SSVEP signal contains the frequency components and the high-order harmonic components of the visual stimulus. For example, the frequency of the visual stimulus is 10 Hz, and the pre-processed steady-state visual evoked potential X is filtered to obtain the useful signal components 19 Hz-21 Hz, 29 Hz-31 Hz, etc.

[0045] Optionally, the pre-processed steady-state visual evoked potential X is filtered to obtain the useful signal components and the noise signal components of the pre-processed steady-state visual evoked potential X , including:

[0046] S320-11, a narrow-band filter with a passband around k×f n is used to filter the pre-processed steady-state visual evoked potential X to obtain the useful signal components, wherein k = m, m+1,..., M; M satisfies the maximum effective frequency of .

[0047] S320-12, a narrow-band filter with a stopband around k×f n is used to filter the pre-processed steady-state visual evoked potential X to obtain the noise signal components, wherein k = m, m+1,..., M; M satisfies the maximum effective frequency of .

[0048] S320-11 is a process for obtaining the useful signal components. For a visual stimulus with a frequency of f n , the useful signal in the pre-processed steady-state visual evoked potential X is the signal with a frequency of the stimulus frequency f n and the signal with a frequency of the multiple of the stimulus frequency f n , therefore, a narrow-band filter with a passband around k×f n is used to filter to obtain the useful signal components. Since the pre-processed steady-state visual evoked potential X also contains the noise signal components, the narrow-band filter with a stopband around k×f n is used to filter to obtain the noise signal components. different values of m, thus may not contain the frequency f n signal, so the value range of k is k = m, m + 1, …, M; M satisfies the maximum effective frequency of . Specifically, in the case of the narrowband filter is a frequency domain filter, where d p，s is the bandwidth coefficient of the passband, which can be determined by the optimal parameters through grid search according to the experimental setting, or the recommended parameters d p，s = 1.75 of the present application. The narrowband filter can retain the frequency components of mf n , (m + 1)f n , …, Mf n in the signal, and filter out the frequency components of other background signals.

[0049] S320-12 is the process of obtaining the noise signal component. For the visual stimulus with the frequency f n , the useful signal in the preprocessed steady-state visual evoked potential is the signal with the frequency of the stimulus frequency f n and the signal with the frequency of the multiple of the stimulus frequency f n , and the rest is the noise signal. Therefore, the narrowband filter with the stop band around k x f n is used to filter , so as to obtain the noise signal component. Since the preprocessed steady-state visual evoked potential has different values of m, thus may not contain the frequency f n signal, so the value range of k is k = m, m + 1, …, M; M satisfies the maximum effective frequency of . Specifically, in the case of the narrowband filter is a frequency domain filter, where d s，r is the bandwidth coefficient of the stop band, which can be determined by the optimal parameters through grid search according to the experimental setting, or the recommended parameters d s，r = 0.25 of the present application. The narrowband filter filters out the frequency components of mf n , (m + 1)f n , …, Mf n in the signal, and retains the frequency components of other background signals.

[0050] The advantage of such setting is that the narrowband filter and obtaining a useful signal component and a noise signal component, giving a pre-processed steady-state visual evoked potential The specific scheme of filtering is given.

[0051] Specifically, the filtering method for obtaining the useful signal component in S320-11 can be frequency domain filtering, specifically including:

[0052] The pre-processed steady-state visual evoked potential is subjected to fast Fourier transform to obtain a corresponding pre-processed steady-state visual evoked potential spectrum

[0053] The pre-processed steady-state visual evoked potential spectrum is multiplied by a narrowband filter to obtain a spectrum of the useful signal component wherein the pre-processed steady-state visual evoked potential spectrum is subjected to point-by-point multiplication with the narrowband filter

[0054] The spectrum of the useful signal component is subjected to inverse fast Fourier transform to obtain the useful signal component

[0055] Correspondingly, the filtering method for obtaining the noise signal component in S320-12 can be frequency domain filtering, specifically including:

[0056] The pre-processed steady-state visual evoked potential spectrum is multiplied by a narrowband filter to obtain a spectrum of the noise signal component wherein the pre-processed steady-state visual evoked potential spectrum is subjected to point-by-point multiplication with the narrowband filter

[0057] The spectrum of the noise signal component is subjected to inverse fast Fourier transform to obtain the noise signal component

[0058] The advantage of such arrangement is that the useful signal component and the noise signal component are obtained through frequency domain filtering. It should be noted that, in addition to using the frequency domain filtering method, the time domain filtering method can also be used to obtain the useful signal component and the noise signal component.

[0059] ​​In addition to the methods described above for obtaining useful signal components and noise signal components, optionally, the preprocessed steady-state visual evoked potentials... Filtering is performed to obtain the preprocessed steady-state visual evoked potentials. The useful signal components and noise signal components may also include:

[0060] S320-21, Using a passband around k×f n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, where k = m, m+1, ..., M; M satisfies Maximum effective frequency;

[0061] S320-22, Using a passband around k×f n -Δf narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the first noise signal component, where k = m, m+1, ..., M; M satisfies Maximum effective frequency;

[0062] S320-23, Using a passband around k×f n +Δf narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the second noise signal component, where k = m, m+1, ..., M; M satisfies The maximum effective frequency.

[0063] Steps S320-21 for obtaining the useful signal component are the same as S320-11, the difference being the method for obtaining the noise signal component. S320-12 uses a band-stop filter to obtain the noise signal component, while S320-22 and S320-23 use a band-pass filter. Narrowband filters are also used. and To each Filtering is performed to obtain the following results: k×f in the spectrum n The signals on the left and right sides, specifically the resulting signal frequencies are determined by k×f n ±Δf determines.

[0064] Specifically, in narrowband filters and narrowband filters When all are frequency domain filters, Where, d p，rdetermining the bandwidth of the noise signal component, Δf determining the center frequency of the noise signal component, d p，s , d p，r and Δf can be determined by grid search to find the optimal parameters, or using the recommended parameters of the present application d p，s = 0.75, d p，r = 3, Δf = 3.

[0065] The advantage of such a setting is that the useful signal component, the first noise signal component and the second noise signal component are obtained through the narrowband filter and gives a specific scheme for filtering the preprocessed steady-state visual evoked potential.

[0066] Specifically, the filtering method for obtaining the useful signal component described in S320-21 can be frequency domain filtering, specifically including:

[0067] performing fast Fourier transform on the preprocessed steady-state visual evoked potential to obtain the corresponding preprocessed steady-state visual evoked potential spectrum

[0068] multiplying the preprocessed steady-state visual evoked potential spectrum by the narrowband filter to obtain the spectrum of the useful signal component wherein the preprocessed steady-state visual evoked potential spectrum and the narrowband filter are multiplied point by point;

[0069] performing inverse fast Fourier transform on the spectrum of the useful signal component to obtain the useful signal component

[0070] Correspondingly, the filtering method for obtaining the first noise signal component described in S320-22 can be frequency domain filtering, specifically including:

[0071] multiplying the preprocessed steady-state visual evoked potential spectrum by the narrowband filter to obtain the spectrum of the first noise signal component wherein the preprocessed steady-state visual evoked potential spectrum and the narrowband filter are multiplied point by point;

[0072] performing inverse fast Fourier transform on the spectrum of the first noise signal component ​The first noise signal component was obtained.

[0073] The filtering method for obtaining the second noise signal component described in S320-23 can be frequency domain filtering, specifically including:

[0074] The preprocessed steady-state visual evoked potential spectrum With narrowband filters Multiplying them together yields the spectrum of the second noise signal component. Among them, the preprocessed steady-state visual evoked potential spectrum With narrowband filters Perform point-by-point multiplication;

[0075] The spectrum of the second noise signal component Perform Inverse Fast Fourier Transform Obtain the second noise signal component

[0076] The advantage of this setup is that the useful signal component, the first noise signal component, and the second noise signal component can be obtained through frequency domain filtering. It should be noted that, in addition to using the frequency domain filtering method, the time domain filtering method can also be used to obtain the useful signal component, the first noise signal component, and the second noise signal component.

[0077] S330. Obtain a spatial filter capable of enhancing the signal-to-noise ratio based on the useful signal components and the noise signal components.

[0078] Due to the volume conductor effect and the influence of the enclosed electric field, the spatial resolution of the scalp EEG signals collected from the subjects is not high, only reaching the centimeter level. At the same time, the collected EEG signals are also very weak, generally at the microvolt level, while the noise signal is relatively large compared to the useful signal. Therefore, the signal-to-noise ratio of the collected scalp EEG signals is low.

[0079] The signal-to-noise ratio (SNR) of scalp EEG signals affects subsequent extraction of EEG feature signals and pattern recognition, thus impacting the performance of brain-computer interfaces. Studies have demonstrated that spatial filtering plays a crucial role in reducing the SNR of SSVEP EEG signals. The basic principle of spatial filtering is to transform the EEG signals recorded from different EEG channels, enhancing the intensity of specific signal components and reducing common noise in each EEG channel, resulting in a projection method that can better distinguish between useful and noise signals—that is, a spatial filter.

[0080] Optionally, for the preprocessed steady-state visual evoked potentials obtained as described in S320-11 and S320-12 The method for different signal components, based on the useful signal components and the noise signal components Obtain a spatial filter that can enhance the signal-to-noise ratio. include:

[0081] For the useful signal components Calculate the covariance to obtain the useful signal components. Useful signal covariance matrix

[0082] For the noise signal components Calculate the covariance to obtain the noise signal components. noise signal covariance matrix

[0083] The covariance matrix of the useful signal is solved using the generalized eigenvalue decomposition method. and the noise signal covariance matrix The generalized eigenvector matrix, wherein the eigenvector corresponding to the largest eigenvalue of the generalized eigenvector matrix is ​​the spatial filter. Right now

[0084] Figure 4 a is based on the useful signal components and noise signal components Obtain spatial filter The flowchart, for the sake of structural integrity, also includes the process of generating visually evoked potentials in the original steady state. Obtain useful signal components and noise signal components The process.

[0085] Optionally, for the preprocessed steady-state visual evoked potentials obtained as described in S320-21, S320-22, and S320-23, The method for different signal components, based on the useful signal components The first noise signal component and the second noise signal component Obtain a spatial filter that can enhance the signal-to-noise ratio. include:

[0086] For the useful signal components Calculate the covariance to obtain the useful signal components. Useful signal covariance matrix

[0087] For the first noise signal component and the second noise signal component The noise signal covariance matrix is ​​obtained by calculating the covariance and averaging the results.

[0088] The covariance matrix of the useful signal is solved using the generalized eigenvalue decomposition method. and the noise signal covariance matrix The generalized eigenvector matrix, wherein the eigenvector corresponding to the largest eigenvalue of the generalized eigenvector matrix is ​​the spatial filter.

[0089] Figure 4 b is based on the useful signal components First noise signal component Second noise signal component Obtain spatial filter The flowchart, for the sake of structural integrity, also includes the process of generating visually evoked potentials in the original steady state. Obtain useful signal components First noise signal component Second noise signal component The process.

[0090] The advantage of this setup is that it provides a method for obtaining spatial filters by calculating the covariance matrix and generalized eigenvalue decomposition.

[0091] S340, The preprocessed steady-state visual evoked potentials The steady-state visual evoked potential template signal is obtained by averaging the results across trials.

[0092] Template signal It is based on the original steady-state visual evoked potential X used for training. n This is obtained, rather than based on the unknown steady-state visual evoked potential K that needs to be detected. If only training data from a single trial is available... That is, the subjects only pay attention to frequencies of f. n The visual stimulus is presented once, and the EEG signal of the subject's attention is collected once, i.e., the number of trials N. t =1, then the template signal Steady-state visual evoked potentials after pretreatment for a single trial It itself; if there is training data from multiple trials. Number of trials N t >1, then Steady-state visual evoked potentials after preprocessing in multiple trials The average signal.

[0093] S350. Collect the unknown steady-state visual evoked potential K of the subject, using the N... fb A filter bank preprocesses the unknown steady-state visual evoked potential K to obtain N. fbUnknown steady-state visual evoked potential K after preprocessing in a specific frequency band (m) Where m = 1, 2, ..., N fb .

[0094] Using the training space filter to investigate the unknown steady-state visual evoked potential K The same N fb The filter banks are preprocessed to obtain the preprocessed unknown steady-state visual evoked potential K(m). For details on the meaning of this preprocessing, please refer to S310, which will not be repeated here.

[0095] S360, via the spatial filter Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio, calculate N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient For N fb indivual Weighted summation yields the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X. n The correlation coefficient ρ between them n .

[0096] Optionally, the spatial filter Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio includes:

[0097] Through the spatial filter For the steady-state visual evoked potential template signal Spatial filtering is performed to obtain the spatially filtered steady-state visual evoked potential template signal.

[0098] Through the spatial filter The preprocessed unknown steady-state visual evoked potential K (m) Spatial filtering is performed to obtain the spatially filtered unknown steady-state visual evoked potentials.

[0099] Accordingly, the calculation of N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient comprises:

[0100] computing N fb spatially filtered steady-state visual evoked potential template signals and N fb unknown steady-state visual evoked potentials

[0101] Figure 5 a is a correlation coefficient between the steady-state visual evoked potential template signal and the filtered pre-processed unknown steady-state visual evoked potential K (m) for the structural integrity of the flowchart, the process of obtaining the spatial filter is also included in the flowchart. Wherein, is a correlation coefficient between the m-th sub-band signal K(m) of the unknown steady-state visual evoked potential K and the m-th sub-band signal of the electroencephalogram signal induced by the visual stimulus with frequency f n

[0102] Correspondingly, the correlation coefficient p n between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n is obtained by N fb weighted summation, comprising:

[0103] The correlation coefficient p fb between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n is obtained by N n weighted summation, wherein a(m) is a weighting coefficient of the correlation coefficient of each specific frequency band, which can be determined according to the grid search with the final correct identification rate as the objective function.

[0104] The method for obtaining the correlation coefficient p n between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n is called the basic steady-state component analysis (SSCA) algorithm.

[0105] ​​​​​​​In addition to using the basic SSCA algorithm to obtain the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X, n The correlation coefficient ρ between them n In addition, ensemble steady-state component analysis (eSSCA) algorithms can be used.

[0106] Figure 5 b is the result of the spatial filter. Steady-state visual evoked potential template signal obtained Unknown steady-state visual evoked potential K after filtering and preprocessing (m) correlation coefficient Another flowchart. Integrated steady-state component analysis refers to integrating multiple spatial filters targeting different frequencies to obtain a combined spatial filter, which is then used for subsequent processing. Specifically, N is displayed on the screen. f One visual stimulus was collected from the subject's N. f One primitive steady-state visual evoked potential X n Using N fb Each filter bank respectively applies to N f The original steady-state visual evoked potential X n Preprocessing is performed to obtain N f ×N fb A spatial filter capable of enhancing the signal-to-noise ratio. Where, N f The number of frequency types applied to the subjects, n = 1, 2, ..., N f .

[0107] The space filter Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio includes:

[0108] N with the same m f Space filters By combining these methods, we obtain the combined spatial filter W. (m) ,

[0109] Through the combined spatial filter W (m) For the steady-state visual evoked potential template signal Spatial filtering is performed to obtain the spatially filtered steady-state visual evoked potential template signal.

[0110] Through the combined spatial filter W (m) The preprocessed unknown steady-state visual evoked potential K(m) Spatial filtering is performed to obtain the spatially filtered unknown steady-state visual evoked potential (K). (m) ) T W (m) .

[0111] Accordingly, the calculation of N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient include:

[0112] Calculate N fb Spatial filtered steady-state visual evoked potential template signal With N fb Unknown steady-state visual evoked potentials (K) after spatial filtering (m) ) T W (m) Correlation coefficient between

[0113] Accordingly, the pair N fb indivual Weighted summation yields the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X. n The correlation coefficient ρ between them n ,include:

[0114] For N fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n , Where a(m) is the correlation coefficient for each specific frequency band. The weighting coefficients.

[0115] The correlation coefficient ρ was obtained based on the eSSCA algorithm. n ρ obtained based on SSCA algorithm n The difference lies in the fact that the eSSCA algorithm preprocesses the unknown steady-state visual evoked potential K for each specific frequency band. (m) Spatial filters corresponding to all possible frequencies of visual stimuli Use in combination.

[0116] The advantage of this setting is that, under normal circumstances, the ρ obtained by the eSSCA algorithm... n It is more accurate than the SSCA algorithm.

[0117] S370, The unknown steady-state visual evoked potential K and its frequency are obtained as f1, ... f1, ... f2. i ,..f n , ...f Nf All original steady-state visual evoked potentials [X1, ... X i , ..., X n , ..., X Nf The correlation coefficients between [ρ1, ... ρ] are [ρ1, ... ρ]. i , ..., ρ n , ..., ρ Nf The i-th correlation coefficient ρ i If the maximum value is reached, then the visual stimulus frequency of the unknown steady-state visual evoked potential K is f. i .

[0118] The purpose of steady-state visual evoked potential (SVP) identification is essentially to determine which frequency of visual stimulus the subject was fixating on when a previously acquired, unknown SVP was identified. Therefore, after obtaining the correlation coefficient ρ... n Then, based on all N f ρ n The maximum value in [the value] determines which frequency of visual stimulus corresponds to the unknown steady-state visual evoked potential K, i.e., the recognition result is [the value]. If the unknown steady-state visual evoked potential K has the largest correlation coefficient with the original steady-state visual evoked potential of a certain frequency, then the unknown steady-state visual evoked potential K is the signal induced by the visual stimulus of the frequency that the subject focuses on, thus realizing the identification of the unknown visual stimulus.

[0119] Figure 6 A heatmap showing the percentage accuracy of different algorithms in recognizing unknown visual stimuli. (e.g.) Figure 6 As shown, the comparison results of the unknown visual stimulus recognition accuracy of the SSCA algorithm and eSSCA algorithm described in this invention with the existing ensemble TRCA (eTRCA) algorithm are as follows: When the visual stimulus signal length is 1 second, under different numbers of EEG channels and different number of trials, the recognition accuracy of the SSCA algorithm and eSSCA algorithm of this invention is consistently better than that of the eTRCA algorithm. Furthermore, as the number of trials and the number of EEG channels decrease, the recognition advantage of the SSCA algorithm and eSSCA algorithm of this invention compared to the eTRCA algorithm becomes more prominent. It is noteworthy that the SSCA algorithm and eSSCA algorithm of this invention can still achieve a high recognition accuracy even with single-trial training data. Here, the number of trials refers to the number of times a single-frequency visual stimulus is applied to the subject during the training of the spatial filter. The number of EEG channels refers to the number of electrodes worn at different positions on the subject's head.

[0120] Figure 7(a) is the visual stimulation signal length influence diagram of the recognition accuracy of different algorithms under the condition of 1 test number and 9 brain electrical channel numbers. Figure 7 (b) is the visual stimulation signal length influence diagram of the information transmission efficiency of different algorithms under the condition of 1 test number and 9 brain electrical channel numbers. Figure 7 The results show that the recognition accuracy and information transmission efficiency of the eSSCA algorithm of the present application are always higher than those of the eTRCA algorithm.

[0121] Figure 8 It is the influence diagram of the brain electrical channel number on the recognition accuracy under the condition of 1 test number. Figure 8 As shown in the figure, the recognition accuracy of the eSSCA algorithm of the present application is higher than that of the eTRCA algorithm under different brain electrical channel numbers.

[0122] Figure 9 It is the experimental result diagram of the signal-to-noise ratio of the spatial filter after filtering of different algorithms under the condition of 1 test number. Figure 9 It is shown that the brain electrical signal filtered by the spatial filter established by the eSSCA algorithm has a higher signal-to-noise ratio than that of the eTRCA algorithm.

[0123] The technical scheme of the present embodiment, by applying visual stimulation with a frequency of f n to the subject, collects the original steady-state visual evoked potential X n of the subject, uses N fb filter banks to preprocess the original steady-state visual evoked potential X n , and obtains N fb preprocessed steady-state visual evoked potentials of specific frequency bands wherein m=1, 2,..., N fb ; filtering the preprocessed steady-state visual evoked potential obtains different signal components of the preprocessed steady-state visual evoked potential , the signal components include useful signal components and noise signal components; and obtaining a spatial filter that can enhance the signal-to-noise ratio according to the useful signal components and the noise signal components. Collecting the unknown steady-state visual evoked potential K of the subject, and obtaining the correlation coefficient p n between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n through the spatial filter . The technical problem of the recognition performance greatly decreasing when the training data is less when using the TRCA algorithm or the eTRCA algorithm, especially when there is only single test training data, and the recognition performance is poor, is solved, and the technical effect of improving the recognition accuracy of unknown visual stimulation is achieved.

[0124] Embodiment Two

[0125] Figure 10 A structural schematic diagram of an electroencephalogram feature decoding device for steady-state visual evoked potential provided by Embodiment Two of the present application. The electroencephalogram feature decoding device can execute the electroencephalogram feature decoding method for steady-state visual evoked potential provided by Embodiment One of the present application, and has the function modules and beneficial effects corresponding to the execution method.

[0126] An electroencephalogram feature decoding device for steady-state visual evoked potential, comprising:

[0127] A preprocessing module 1010 is configured to present N f visual stimuli with frequencies f1,...f i ,...f n ,...f Nf on a screen, collect original steady-state visual evoked potential signals of a subject when the subject gazes at the visual stimuli as training data, and the gazing frequency of the subject is f n The original steady-state visual evoked potential of the subject collected when the subject gazes at the visual stimulus with frequency f n is X fb N n filter banks are used to preprocess the original steady-state visual evoked potential X fb to obtain N preprocessed steady-state visual evoked potentials with specific frequency bands fb , where m = 1, 2,...N

[0128] A signal component acquisition module 1020 is configured to filter the preprocessed steady-state visual evoked potential X to obtain useful signal components and noise signal components of the preprocessed steady-state visual evoked potential X The useful signal components include signal components of the visual stimulus frequency f n and its harmonics, and the noise signal components include other signal components except the useful signal components.

[0129] A spatial filter acquisition module 1030 is configured to obtain a spatial filter H

[0130] A template signal acquisition module 1040 is configured to average the preprocessed steady-state visual evoked potential X in the trial dimension to obtain a steady-state visual evoked potential template signal X

[0131] The unknown steady-state visual evoked potential acquisition module 1050 is used to acquire the unknown steady-state visual evoked potential K of the subject, using the N... fb A filter bank preprocesses the unknown steady-state visual evoked potential K to obtain N. fb Unknown steady-state visual evoked potential K after preprocessing in a specific frequency band (m) Where m = 1, 2, ..., N fb ;

[0132] The correlation coefficient calculation module 1060 is used to calculate the correlation coefficient through the spatial filter. Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio, calculate N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient For N fb indivual Weighted summation yields the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X. n The correlation coefficient ρ between them n ;

[0133] The visual stimulus frequency acquisition module 1070 is used to obtain the unknown steady-state visual evoked potential K and its frequencies f1, ... f2. i , ...f n ,,...f Nf All original steady-state visual evoked potentials [X1, ... X i , ..., X n , ..., X Nf The correlation coefficients between [ρ1, ... ρ] are [ρ1, ... ρ]. i , ..., ρ n , ..., ρ Nf The i-th correlation coefficient ρ i If the maximum value is reached, then the visual stimulus frequency of the unknown steady-state visual evoked potential K is f. i .

[0134] Optionally, the signal component acquisition module 1020 includes:

[0135] Useful signal component acquisition submodule, used to acquire signal components around k×f using a passband. n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, where k = m, m+1, ..., M; M satisfies Maximum effective frequency;

[0136] The noise signal component acquisition submodule is used to obtain noise signal components around a stopband of k×f. n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the noise signal components, where k = m, m+1, ..., M; M satisfies The maximum effective frequency.

[0137] Optionally, the signal component acquisition module 1020 includes:

[0138] Useful signal component acquisition submodule, used to acquire signal components around k×f using a passband. n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, where k = m, m+1, ..., M; M satisfies The maximum effective frequency;

[0139] The first noise signal component acquisition submodule is used to obtain the noise signal components around k×f using a passband. n -Δf narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the first noise signal component, where k = m, m+1, ..., M; M satisfies The maximum effective frequency;

[0140] The second noise signal component acquisition submodule is used to obtain the noise signal components around k×f using a passband. n +Δf narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the second noise signal component, where k = m, m+1, ..., M; M satisfies The maximum effective frequency.

[0141] In the case where the signal component acquisition module 1020 includes a useful signal component acquisition submodule and a noise signal component acquisition submodule, optionally, the useful signal component acquisition submodule includes:

[0142] Fast Fourier Transform (FFT) unit is used to process the preprocessed steady-state visual evoked potentials. Perform a Fast Fourier Transform to obtain the corresponding preprocessed steady-state visual evoked potential spectrum.

[0143] a multiplier unit configured to multiply the pre-processed steady-state visual evoked potential spectrum by a narrow-band filter to obtain a spectrum of noise signal components

[0144] an inverse fast Fourier transform unit configured to perform an inverse fast Fourier transform on the spectrum of noise signal components to obtain noise signal components

[0145] correspondingly, the noise signal component obtaining submodule comprises:

[0146] a multiplier unit configured to multiply the pre-processed steady-state visual evoked potential spectrum by a narrow-band filter to obtain a spectrum of noise signal components

[0147] an inverse fast Fourier transform unit configured to perform an inverse fast Fourier transform on the spectrum of noise signal components to obtain noise signal components

[0148] in the case where the signal component obtaining module 1020 comprises a useful signal component obtaining submodule, a first noise signal component obtaining submodule and a second noise signal component obtaining submodule, optionally, the useful signal component obtaining submodule comprises:

[0149] a fast Fourier transform unit configured to perform a fast Fourier transform on the pre-processed steady-state visual evoked potential to obtain a corresponding pre-processed steady-state visual evoked potential spectrum a multiplier unit configured to multiply the pre-processed steady-state visual evoked potential spectrum by a narrow-band filter to obtain a spectrum of useful signal components

[0150]

[0151] an inverse fast Fourier transform unit configured to perform an inverse fast Fourier transform on the spectrum of useful signal components to obtain useful signal components

[0152] correspondingly, the first noise signal component obtaining submodule comprises:

[0153] a multiplier unit configured to multiply the pre-processed steady-state visual evoked potential spectrum by a narrow-band filter to obtain a spectrum of noise signal components ​​​​​​​​​​​multiplying the pre-processed steady-state visual evoked potential spectrum

[0154] an inverse fast Fourier transform unit configured to perform inverse fast Fourier transform on the spectrum of the first noise signal component to obtain the first noise signal component

[0155] Correspondingly, the second noise signal component obtaining submodule comprises:

[0156] a multiplier unit configured to multiply the pre-processed steady-state visual evoked potential spectrum and a narrowband filter to obtain the spectrum of the second noise signal component

[0157] an inverse fast Fourier transform unit configured to perform inverse fast Fourier transform on the spectrum of the second noise signal component to obtain the second noise signal component

[0158] In the case where the signal component obtaining module 1020 comprises a useful signal component obtaining submodule and a noise signal component obtaining submodule, and the useful signal component obtaining submodule comprises a fast Fourier transform unit, a multiplier unit and an inverse fast Fourier transform unit, and the noise signal component obtaining submodule comprises a multiplier unit and an inverse fast Fourier transform unit, optionally, the spatial filter obtaining module 1030 comprises:

[0159] a covariance matrix obtaining submodule configured to perform covariance operation on the useful signal component to obtain a useful signal covariance matrix of the useful signal component a covariance matrix obtaining submodule configured to perform covariance operation on the noise signal component to obtain a noise signal covariance matrix of the noise signal component

[0160] a maximum eigenvalue obtaining submodule configured to use a generalized eigenvalue decomposition method to solve a generalized eigenvector matrix of the useful signal covariance matrix and the noise signal covariance matrix , and the eigenvector corresponding to the maximum eigenvalue of the generalized eigenvector matrix is the spatial filter

[0161] ​​​​In the signal component obtaining module 1020 comprises a useful signal component obtaining submodule, a first noise signal component obtaining submodule and a second noise signal component obtaining submodule; and the useful signal component obtaining submodule comprises a fast Fourier transform unit, a multiplier unit and an inverse fast Fourier transform unit; and the first noise signal component obtaining submodule comprises a multiplier unit and an inverse fast Fourier transform unit; and the second noise signal component obtaining submodule comprises a multiplier unit and an inverse fast Fourier transform unit, in the case that the spatial filter obtaining module 1030 comprises:

[0162] a covariance matrix obtaining submodule for obtaining a useful signal covariance matrix of the useful signal component by calculating the covariance of the useful signal component a first noise signal component and a second noise signal component by calculating the covariance of the first noise signal component and the second noise signal component and taking the average value, to obtain a noise signal covariance matrix

[0163] a maximum eigenvalue obtaining submodule for obtaining a generalized eigenvector matrix of the useful signal covariance matrix and the noise signal covariance matrix by using a generalized eigenvalue decomposition method, wherein the eigenvector corresponding to the maximum eigenvalue of the generalized eigenvector matrix is the spatial filter

[0164] Optionally, the correlation coefficient calculation module 1060 comprises:

[0165] a spatial filtering submodule for spatially filtering the steady-state visual evoked potential template signal by using the spatial filter to obtain a spatially filtered steady-state visual evoked potential template signal a spatial filtering submodule for spatially filtering the preprocessed unknown steady-state visual evoked potential K ( m ) by using the spatial filter

[0166] a first correlation coefficient calculation submodule for calculating the correlation coefficients between N fb spatially filtered steady-state visual evoked potential template signals and N fb spatially filtered unknown steady-state visual evoked potentials ​

[0167] The second correlation coefficient calculation submodule is used to calculate N. fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n , Where a(m) is the correlation coefficient for each specific frequency band. The weighting coefficients.

[0168] N is displayed on the screen. f One visual stimulus was collected from the subject's N. f One primitive steady-state visual evoked potential X n Using N fb Each filter bank respectively applies to N f The original steady-state visual evoked potential X n Preprocessing is performed to obtain N f ×N fb A spatial filter capable of enhancing the signal-to-noise ratio. Where, N f The number of frequency types applied to the subjects, n = 1, 2, ..., N f In this case, optionally, the correlation coefficient calculation module 1060 includes:

[0169] The combined spatial filter acquisition submodule is used to combine N with the same m f Space filters By combining these methods, we obtain the combined spatial filter W. (m) ,

[0170] The spatial filtering submodule is used to filter the combined spatial filter W. (m) For the steady-state visual evoked potential template signal Spatial filtering is performed to obtain the spatially filtered steady-state visual evoked potential template signal. Through the combined spatial filter W (m) Spatially filter the preprocessed unknown steady-state visual evoked potential K(m) to obtain the spatially filtered unknown steady-state visual evoked potential (K). (m) ) T W (m) ;

[0171] The first correlation coefficient calculation submodule is used to calculate N. fb Spatial filtered steady-state visual evoked potential template signal With N fb Unknown steady-state visual evoked potentials (K) after spatial filtering (m) )T W (m) Correlation coefficient between

[0172] The second correlation coefficient calculation submodule is used to calculate N. fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n , Where a(m) is the correlation coefficient for each specific frequency band. The weighting coefficients.

[0173] The technical solution of this embodiment solves the problem that the recognition performance drops significantly when using the TRCA algorithm or eTRCA algorithm with limited training data, especially when there is only single-trial training data, resulting in very poor recognition performance. This achieves the technical effect of improving the recognition accuracy of unknown visual stimuli.

[0174] Example 3

[0175] Figure 11 This is a schematic diagram of a brain-computer interface system for steady-state visual evoked potentials provided in Embodiment 3 of the present invention, as shown below. Figure 11 As shown, the system includes a processor 1110, a memory 1120, an input device 1130, and an output device 1140; the number of processors 1110 in the system can be one or more. Figure 11 Taking a processor 1110 as an example; the processor 1110, memory 1120, input device 1130 and output device 1140 in the system can be connected via a bus or other means. Figure 11 Taking the example of a connection between China and Israel via a bus.

[0176] The memory 1120, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the EEG feature decoding method for steady-state visual evoked potentials in this embodiment of the invention (e.g., the preprocessing module 1010, signal component acquisition module 1020, spatial filter acquisition module 1030, template signal acquisition module 1040, unknown steady-state visual evoked potential acquisition module 1050, correlation coefficient calculation module 1060, and visual stimulus frequency acquisition module 1070 in an EEG feature decoding device for steady-state visual evoked potentials). The processor 1110 executes various functional applications and data processing of the system by running the software programs, instructions, and modules stored in the memory 1120, thereby realizing the aforementioned EEG feature decoding method for steady-state visual evoked potentials.

[0177] The memory 1120 can include a program storage area and a data storage area, where the program storage area can store an operating system, application programs required for at least one function, and the data storage area can store data created according to the use of the terminal, etc. In addition, the memory 1120 can include a high-speed random access memory, and can further include a non-volatile memory such as at least one of a magnetic disk storage device, a flash memory device, or other non-volatile solid state memory device. In some examples, the memory 1120 can further include a memory disposed remotely with respect to the processor 1110, which can be connected to the system through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] The input device 1130 can be used to receive input digital or character information, and to generate key signal input related to user settings of the system and function control. The output device 1140 can include a display device such as a display screen.

[0179] Embodiment Four

[0180] The embodiment four of the present application also provides a computer readable storage medium containing a computer program, which when executed by a processor is used to execute an electroencephalogram feature decoding method for steady-state visual evoked potential, the method comprising:

[0181] presenting N f visual stimuli with frequencies f1,...f i ,...f n ,...f Nf on a screen, collecting raw steady-state visual evoked potential signals of a subject gazing at the visual stimuli as training data, the raw steady-state visual evoked potential of the subject gazing at a visual stimulus with a frequency f n is X n , using N fb filter banks to preprocess the raw steady-state visual evoked potential X n to obtain N fb preprocessed steady-state visual evoked potentials of specific frequency bands wherein m = 1, 2,...N fb ;

[0182] filtering the preprocessed steady-state visual evoked potentials to obtain useful signal components and noise signal components of the preprocessed steady-state visual evoked potentials , the useful signal components including signal components of the visual stimulus frequency f n and its harmonics, and the noise signal components including other signal components other than the useful signal components;

[0183] A spatial filter capable of enhancing the signal-to-noise ratio is obtained based on the useful signal components and the noise signal components.

[0184] The pre-processed steady-state visual evoked potentials The steady-state visual evoked potential template signal is obtained by averaging the results across trials.

[0185] The unknown steady-state visual evoked potential K of the subject was collected, and the N was used. fb A filter bank preprocesses the unknown steady-state visual evoked potential K to obtain N. fb Unknown steady-state visual evoked potential K after preprocessing in a specific frequency band (m) Where m = 1, 2, ..., N fb ;

[0186] Through the spatial filter Improve the steady-state visual evoked potential template signal and the preprocessed unknown steady-state visual evoked potential K (m) The signal-to-noise ratio, calculate N fb The filtered steady-state visual evoked potential template signal for a specific frequency band The preprocessed unknown steady-state visual evoked potential K after filtering (m) correlation coefficient For N fb indivual Weighted summation yields the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X. n The correlation coefficient ρ between them n ;

[0187] The unknown steady-state visual evoked potential K and its frequencies are obtained as f1, ... f1. i , ...f n ,,...f Nf All original steady-state visual evoked potentials [X1, ... X i , ..., X n , ..., X Nf The correlation coefficients between [ρ1, ... ρ] are [ρ1, ... ρ]. i , ..., ρ n , ..., ρ Nf The i-th correlation coefficient ρ i If the maximum value is reached, then the visual stimulus frequency of the unknown steady-state visual evoked potential K is f. i .

[0188] Of course, the computer readable storage medium provided by the embodiment of the present application includes a computer program, and the computer program is not limited to the method operation described above, but can also perform the related operation in the steady-state visual evoked potential-oriented electroencephalogram feature decoding method provided by any embodiment of the present application.

[0189] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by software and necessary general hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk or an optical disk, etc., including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in the embodiment of the present application.

[0190] It is worth noting that the above embodiment of the steady-state visual evoked potential-oriented electroencephalogram feature decoding device includes various units and modules, which are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific name of each functional unit is only for easy distinction, and does not limit the protection scope of the present application.

[0191] Note that the above is only the preferred embodiment of the present application and the technical principle applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and those skilled in the art can make various obvious changes, readjustments and substitutions without departing from the scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A steady-state visual evoked potential oriented electroencephalogram feature decoding method, comprising: N is displayed on the screen. f The frequencies are f1, ... f2. i ,…f n ,…f Nf Visual stimuli were used to collect raw steady-state visual evoked potential signals from subjects as they gazed at these stimuli, with a gaze frequency of f. n The primitive steady-state visual evoked potentials of the subjects' multi-channel EEG collected during visual stimulation were X. n Using N fb Each filter bank affects the original steady-state visual evoked potential X. n Preprocessing is performed to obtain N fb Preprocessed steady-state visual evoked potentials in specific frequency bands Where m = 1, 2, ..., N fb ; The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the preprocessed steady-state visual evoked potentials. The useful signal components and noise signal components, wherein the useful signal components include the visual stimulus frequency f n The signal components of the noise signal and its harmonics, wherein the noise signal components include signal components other than the useful signal components; a spatial filter capable of enhancing signal-to-noise ratio is obtained from the useful signal component and the noise signal component the pre-processed steady-state visual evoked potential averaging over trials to obtain a steady-state visual evoked potential template signal Collecting an unknown steady-state visual evoked potential K of a subject, adopting the N fb filter banks to pre-process the unknown steady-state visual evoked potential K, obtaining N fb pre-processed unknown steady-state visual evoked potentials K of specific frequency bands (m) , wherein m = 1, 2, …, N fb ; through the spatial filter improving the signal-to-noise ratio of the steady-state visual evoked potential template signal and the pre-processed unknown steady-state visual evoked potential K (m) , calculating N fb specific frequency band filtered steady-state visual evoked potential template signals and the correlation coefficient of the filtered pre-processed unknown steady-state visual evoked potential K (m) weighted sum of N fb to obtain the correlation coefficient p n between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n ;​​ The unknown steady-state visual evoked potentials K and frequencies f1, ... f2 are obtained. i ,…f n ,…f Nf All primitive steady-state visual evoked potentials [X1,…X i ,…,X n ,...,X Nf The correlation coefficients between [ρ1,…ρ] i ,…,ρ n ,…,ρ Nf The i-th correlation coefficient ρ i If the maximum value is reached, then the visual stimulus frequency of the unknown steady-state visual evoked potential K is f. i .

2. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 1, characterized in that, said pre-processed steady-state visual evoked potentials filtering said pre-processed steady-state visual evoked potentials to obtain a useful signal component and a noise signal component of said pre-processed steady-state visual evoked potentials, comprising: Using a passband around k×f n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, where k = m, m+1, ..., M; M satisfies M×f n ≤ Maximum effective frequency; A narrow-band filter with a stop band around k x f n The pre-processed steady-state visual evoked potential is filtered to obtain noise signal components, where k = m, m+1, …, M; M satisfies M x f n ≤ the maximum effective frequency.​ 3.The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 1, characterized in that, said pre-processed steady-state visual evoked potentials filtering said pre-processed steady-state visual evoked potentials to obtain a useful signal component and a noise signal component of said pre-processed steady-state visual evoked potentials, comprising: Using a passband around k×f n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, where k = m, m+1, ..., M; M satisfies M×f n ≤ Maximum effective frequency; A passband is used which is centered around k x f n - a narrowband filter with a passband centered around k x f The pre-processed steady-state visual evoked potentials are filtered to obtain a first noise signal component, wherein k = m, m+1, …, M; M satisfies the maximum effective frequency of the signal. A narrow band filter having a pass band centered around k x f n + Δf The pre-processed steady-state visual evoked potential is filtered to obtain a second noise signal component, wherein k = m, m + 1, …, M; M satisfies the maximum effective frequency.

4. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 2, characterized in that, The narrow-band filter has a passband that is a narrow band around kxf n Filtering the pre-processed steady-state visual evoked potentials to obtain useful signal components includes: Filtering the pre-processed steady-state visual evoked potentials to obtain useful signal components includes:​ the pre-processed steady-state visual evoked potential performing a fast Fourier transform to obtain a corresponding pre-processed steady-state visual evoked potential spectrum The pre-processed steady-state visual evoked potential spectrum is multiplied with a narrow band filter to obtain the spectrum of the useful signal component spectrum of the useful signal component performing an inverse fast Fourier transform to obtain a useful signal component Accordingly, the stopband used is around k×f n Narrowband filter The pre-processed steady-state visual evoked potentials After filtering, the noise signal components are obtained, including: The pre-processed steady-state visual evoked potential spectrum is multiplied with a narrow band filter to obtain a spectrum of noise signal components a spectrum of the noise signal component performing an inverse fast fourier transform to obtain a noise signal component 5. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 3, characterized in that, The passband used is around k×f n Narrowband filter The pre-processed steady-state visual evoked potentials Filtering is performed to obtain the useful signal components, including: the pre-processed steady-state visual evoked potential performing a fast Fourier transform to obtain a corresponding pre-processed steady-state visual evoked potential spectrum The pre-processed steady-state visual evoked potential spectrum is multiplied with a narrow band filter to obtain the spectrum of the useful signal components spectrum of the useful signal component performing an inverse fast Fourier transform to obtain a useful signal component Accordingly, the narrowband filter with passband around k x f n -Δf filtering the pre-processed steady-state visual evoked potentials filtering the pre-processed steady-state visual evoked potentials the pre-processed steady-state visual evoked potential spectrum multiplied with a narrow band filter to obtain a spectrum of a first noise signal component a spectrum of the first noise signal component performing an inverse fast Fourier transform to obtain a first noise signal component Accordingly, the narrow-band filter has a passband that is centered around k x f n + Δf The pre-processed steady-state visual evoked potential is filtered to obtain a second noise signal component, comprising: The pre-processed steady-state visual evoked potential spectrum is multiplied with a narrow band filter to obtain a spectrum of a second noise signal component a spectrum of the second noise signal component performing an inverse fast Fourier transform to obtain a second noise signal component 6. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 4, characterized in that, the spatial filter capable of enhancing signal-to-noise ratio is obtained according to the useful signal component and the noise signal component comprising: to the useful signal component covariance, obtaining a useful signal covariance matrix of the useful signal component to the noise signal component covariance, obtaining a noise signal covariance matrix of the noise signal component solving the useful signal covariance matrix using a generalized eigenvalue decomposition method and the noise signal covariance matrix a generalized eigenvector matrix whose largest eigenvalue corresponds to an eigenvector that is the spatial filter 7. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to claim 5, characterized in that, the spatial filter capable of enhancing signal-to-noise ratio is obtained according to the useful signal component and the noise signal component comprising: to the useful signal component covariance, obtaining a useful signal covariance matrix of the useful signal component for the first noise signal component and for the second noise signal component ​ solving the useful signal covariance matrix using a generalized eigenvalue decomposition method and the noise signal covariance matrix a generalized eigenvector matrix whose largest eigenvalue corresponds to an eigenvector that is the spatial filter 8. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to any one of claims 1-7, characterized in that, said through said spatial filter improving the signal-to-noise ratio of said steady-state visual evoked potential template signal and said pre-processed unknown steady-state visual evoked potential K (m) of interest, comprising: through the spatial filter the steady-state visual evoked potential template signal spatially filtering to obtain a spatially filtered steady-state visual evoked potential template signal through the spatial filter the pre-processed unknown steady-state visual evoked potential K (m) spatial filtering to obtain a spatially filtered unknown steady-state visual evoked potential Correspondingly, the calculation N fb of the filtered steady-state visual evoked potential template signal of the specific frequency band and the correlation coefficient of the filtered pre-processed unknown steady-state visual evoked potential K (m) includes:​ N fb spatially filtered steady-state visual evoked potential template signals N fb spatially filtered unknown steady-state visual evoked potentials correlation coefficient between Correspondingly, the pair of N fb Weighted summation obtains the correlation coefficient p n between the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X n , comprising:​ For N fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n , Where a(m) is the correlation coefficient for each specific frequency band. The weighting coefficients.

9. The steady-state visually evoked potential-oriented electroencephalogram feature decoding method according to any one of claims 1-7, characterized in that, N is displayed on the screen. f One visual stimulus was collected from the subject's N. f One primitive steady-state visual evoked potential X n Using N fb Each filter bank respectively applies to N f The original steady-state visual evoked potential X n Preprocessing is performed to obtain N f ×N fb A spatial filter capable of enhancing the signal-to-noise ratio. Where, N f The number of frequency types applied to the subjects, n = 1, 2, ..., N f ; said through said spatial filter improving the signal-to-noise ratio of said steady-state visual evoked potential template signal and said pre-processed unknown steady-state visual evoked potential K(m) N spatial filters with the same m f (m) , ​​ through the combination spatial filter W (m) to the steady-state visual evoked potential template signal spatially filtering to obtain a spatially filtered steady-state visual evoked potential template signal through the combination spatial filter W (m) on the pre-processed unknown steady-state visual evoked potential K (m) spatial filtering to obtain a spatially filtered unknown steady-state visual evoked potential (K (m) ) T W (m) ; Correspondingly, the calculation N fb of the filtered steady-state visual evoked potential template signal of the specific frequency band and the correlation coefficient of the filtered pre-processed unknown steady-state visual evoked potential K (m) of the specific frequency band includes: Compute N fb spatially filtered steady-state visual evoked potential template signals correlation coefficient between N fb spatially filtered unknown steady-state visual evoked potential (K (m) ) T W (m) ​ Accordingly, the pair N fb indivual Weighted summation yields the unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X. n The correlation coefficient ρ between them n ,include: For N fb indivual The unknown steady-state visual evoked potential K and the original steady-state visual evoked potential X are obtained by weighted summation. n The correlation coefficient ρ between them n , Where a(m) is the correlation coefficient for each specific frequency band. The weighting coefficients.

10. A steady-state visual evoked potential oriented brain-computer interface system, comprising: one or more processors (1110); a memory (1120) for storing one or more programs, the one or more programs, when executed by the one or more processors (1110), cause the one or more processors (1110) to implement the steady-state visual evoked potential oriented electroencephalogram feature decoding method of any one of claims 1-9.

Images