SSVEP identification method based on frequency domain signal-to-noise ratio

Through the recognition method based on the frequency domain signal-to-noise ratio, SSVEP frequency point information is extracted and the impact of EEG background noise and harmonics is considered, which solves the impact of screen refresh rate limiting and harmonics in SSVEP frequency recognition technology, and improves the classification accuracy. Expand the application scope of brain-computer interface.

CN118965196APending Publication Date: 2024-11-15NANJING RES INST OF ELECTRONICS TECH
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410955527.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing SSVEP frequency recognition technology is affected by screen refresh rate limits and harmonics, and has failed to effectively consider EEG background noise, resulting in insufficient classification accuracy, limiting the widespread application of brain-computer interfaces.

Method used

Using the frequency domain signal-to-noise ratio identification method, through frequency domain transformation, peak search, noise calculation and signal-to-noise ratio calculation, the frequency domain peak information of the EEG signal is extracted, the SSVEP frequency points are automatically obtained, and the signal-to-noise ratio calculation is calculated. Consider the fundamental frequency and harmonics, and output the frequency identification result.

Benefits of technology

The classification accuracy of the SSVEP recognition method is improved, the frequency offset caused by screen refresh rate limitation and individual status differences is solved, the scope of SSVEP application is expanded, and the accuracy is achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965196A_ABST
    Figure CN118965196A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of brain-computer interfaces, and discloses an SSVEP (Steady-State Visual Evoked Potential) identification method based on a frequency domain signal-to-noise ratio. The SSVEP identification method relates to frequency domain transformation, superposition average, peak value search, noise calculation, signal-to-noise ratio calculation and result output. Compared with the prior art, the method has the advantages that the induced SSVEP frequency point information is automatically obtained by extracting the frequency domain peak value information, so that the problem of frequency deviation caused by screen refresh rate limitation and individual state difference is solved to a certain extent; based on the solution of the optimal signal-to-noise ratio of the fundamental frequency and the harmonic waves, the influence of background noise and harmonic waves in the electroencephalogram is taken into consideration, the application range of the SSVEP recognition method is widened to a certain extent, and the SSVEP recognition method has high accuracy in practical application.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention mainly relates to the technical field of brain-computer interface, and in particular to a SSVEP recognition method based on frequency domain signal-to-noise ratio. Background Art

[0002] As a system that detects neural activity in the brain and converts it into output, the brain-computer interface (BCI) can directly detect a person's actions and thoughts from the brain and transmit the information directly to the machine. The brain-computer interface system is mainly composed of EEG signal acquisition, signal processing and decoding (preprocessing, feature extraction, classification and recognition), command output and other units. The acquisition device first collects the EEG signals of the subject in the brain area, amplifies, filters, and converts the signals into analog-to-digital signals, and transmits them to the processing system; the processing system first preprocesses the signals, filters out the noise and artifacts contained in the EEG signals, and then extracts the required signal features for classification; finally, the classified signals are converted into control command outputs, thereby achieving the purpose of human-computer interaction.

[0003] Steady-state Visual Evoked Potential (SSVEP) is mainly distributed in the occipital region of the brain. It is an EEG signal generated when the subject is subjected to continuous visual stimulation of a fixed frequency from the outside world. The SSVEP paradigm has become one of the most popular brain-computer interface paradigms due to its excellent electroencephalographic noise, relative immunity to eye artifacts, and the fact that the subject does not need to undergo prior training. However, factors such as the stimulus source and the classification accuracy of frequency recognition technology limit the large-scale application of this paradigm.

[0004] The frequency recognition technology for SSVEP determines the coding frequency of the target visual stimulus that the subject is looking at by analyzing and processing the EEG signal, thereby realizing the selection of the corresponding character or the execution of the instruction. At present, the frequency recognition technology of SSVEP is mainly divided into supervised methods and unsupervised methods. Among them, the supervised method can achieve higher accuracy, but it requires a lot of training data support. The commonly used method is TRCA; the unsupervised method can be used directly without training data, which improves the applicability of the BCI system. Common methods include CCA and frequency domain spectrum energy extraction method.

[0005] Since SSVEP is a visual evoked signal, it is very sensitive to the physical properties of the stimulus source in practical applications, including the frequency, color, brightness of the stimulus source and the hardware device used for stimulation. The range of stimulus frequency selection is affected by three factors. First, the frequency components of the signal induced by SSVEP include both the stimulus frequency signal and the harmonic components of the fundamental frequency signal. Different human brains respond differently to each stimulus frequency and have different target frequencies. Therefore, the selected frequency cannot contain odd / even harmonic frequencies with the fundamental frequency. Secondly, the selected frequency is also limited by the refresh rate of the display, which is an integer multiple of the screen refresh rate. Finally, only within a certain frequency range can a strong SSVEP signal be induced in the brain.

[0006] Currently, there is a technology in SSVEP that uses fast Fourier transform to extract spectral features for frequency recognition. The shortcoming is that this method is based on EEG energy for processing, does not take into account the impact of EEG background noise, and ignores the impact of harmonics. Summary of the invention

[0007] The purpose of the present invention is to provide a SSVEP recognition method based on frequency domain signal-to-noise ratio, which can solve the frequency limitation problem caused by screen refresh and the influence of harmonics in the prior art to a certain extent, improve the classification accuracy of frequency recognition technology, and take background noise into consideration, thereby improving the application of steady-state visual evoked potential (SSVEP).

[0008] To achieve the above object, the present invention provides a SSVEP identification method based on frequency domain signal-to-noise ratio, comprising the following steps:

[0009] Step 1: Frequency domain transformation, intercept the data within a specific time window of the EEG data signal after EEG preprocessing, set the number of Nfft points, and obtain the frequency domain signal of each channel;

[0010] Step 2: Superposition and averaging: square the amplitude of the frequency domain signal obtained after FFT of each channel and then calculate the superposition and averaging to obtain a set of frequency domain signals of the overall lead for each trial;

[0011] Step 3: Peak search, take the midpoint of the peak width, find the peak value of the frequency domain signal stimulation frequency of a group of overall leads in each trial within the left half peak width and the right half peak width, and obtain the actual stimulation frequency and peak amplitude;

[0012] Step 4: Noise calculation, set the parameters of the noise window length and the protection window length, and calculate the median value within the noise window length at the peak width of the right half of the stimulation frequency plus the protection window length as the noise signal amplitude;

[0013] Step 5: Calculate the signal-to-noise ratio, calculate the signal-to-noise ratio of each trial at the fundamental frequency of the stimulation frequency and its harmonics;

[0014] Step 6: Output the results, compare the signal-to-noise ratio parameters, and output the frequency recognition results.

[0015] Furthermore, the specific time window length is 2s and the number of Nfft points is 8192.

[0016] Furthermore, the midpoint of the peak width is taken, and both the left half peak width and the right half peak width are set to 0.35 Hz.

[0017] Furthermore, the noise window length is set to 2 Hz and the protection window length is set to 0.2 Hz.

[0018] Furthermore, the value of the minimum frequency difference between the stimulation frequencies and the value of the specific time window length are reciprocals of each other.

[0019] Furthermore, the number of Nfft points is an integer multiple of 2 greater than the number of points in the time window length.

[0020] Beneficial effects:

[0021] The present invention provides an SSVEP recognition method based on frequency domain signal-to-noise ratio. Compared with the prior art, the induced SSVEP frequency point information is automatically obtained by extracting frequency domain peak information, which solves the screen refresh rate limitation and the frequency offset problem caused by individual state differences to a certain extent. The solution based on the optimal signal-to-noise ratio of fundamental frequency and harmonics takes the influence of background noise and harmonics in EEG into consideration, which improves the application scope of the SSVEP recognition method to a certain extent and has a higher accuracy in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of the present invention;

[0023] Figure 2 This is a comparison chart of the accuracy of this method and CCA for multiple binary classification experimental data of two subjects;

[0024] Figure 3 This is a comparison chart of the accuracy of this method and CCA for four eight-category experimental data of a subject.

[0025] Description of reference numerals:

[0026] Wherein: FFT is fast Fourier transform; Nfft points is the number of FFT points. DETAILED DESCRIPTION

[0027] The preferred mechanism and implementation method of the present invention are further described below in conjunction with the accompanying drawings and specific implementation methods.

[0028] like Figures 1 to 3 As shown, an embodiment of the present invention discloses a technical solution of a SSVEP identification method based on frequency domain signal-to-noise ratio.

[0029] Figure 1 It is a flow chart of the present invention.

[0030] Figure 2 This is a comparison chart of the accuracy of this method and CCA for multiple binary classification experimental data of two subjects.

[0031] Figure 3 The following is a comparison chart of the accuracy of this method and CCA for four eight-category experimental data of a subject.

[0032] Example 1

[0033] The SSVEP recognition method based on the frequency domain signal-to-noise ratio in this embodiment has a total of 50 trials. In each trial, a circular red spot with a single frequency flashing will appear. Each red spot lasts for about 5 seconds. There is a 2-second interval between different trials. The whole experiment takes about 6 minutes. The subject only needs to sit in a chair and stare at the red spot on the screen when the red spot appears. Figure 2 (a) The subjects underwent a total of 7 binary classification experiments, with stimulation frequencies of 8 Hz and 9 Hz; Figure 2 (b) The subjects had a total of 4 binary classification experiments. Except for the third experiment with frequencies of 17 Hz and 17.8 Hz, the other 3 experiments had frequencies of 17 Hz and 19 Hz. Figure 3 The subjects were tested in 4 eight-category experiments. Except for the third experiment with 10Hz, 10.8Hz, 11.6Hz, 12.4Hz, 7Hz, 13.2Hz, 14.9Hz and 8Hz, the stimulation frequencies of the other three experiments were 10.1Hz, 10.9Hz, 11.7Hz, 12.5Hz, 13.3Hz, 14.1Hz, 14.9Hz and 15.7Hz. The number of trials in the three experiments was different, which was 70 trials, 60 trials and 50 trials in order. The stimulus source used in the experiment was a laptop display screen with a refresh rate of 60Hz.

[0034] The experimental EEG equipment is the Boruikang EEG acquisition equipment. In the experiment, the Boruikang wireless EEG acquisition equipment was used with a 32-lead EEG cap, and the sampling frequency was set to 4000Hz. In this experiment, due to the study of steady-state visual evoked, only 5 electrodes, 'PO4', 'PO3', 'O2', 'Oz', 'O1' occipital area and parietal occipital area, were selected for the experiment.

[0035] The acquired EEG data is preprocessed, and the preprocessing process includes 1-45Hz bandpass filtering, 50Hz notch, data segmentation, artifact removal and other routine operations. Figure 1 The flowchart method is further processed as follows:

[0036] according to Figure 1 Flowchart step 1: Frequency domain transformation, FFT processing is performed under a fixed time window length of 2s, the number of Nfft points is 8192, and the frequency domain signals of five channels are obtained.

[0037] according to Figure 1 In step 2 of the flowchart, the FFT values ​​of the five channels are squared and then averaged to obtain the frequency domain signal of a group of overall leads in a single trial.

[0038] according to Figure 1 Step 3 of the flowchart is peak search. Take the midpoint of the peak width, set half of the peak width to 0.35 Hz, find the peak value of the stimulation frequency within the left half peak width and the right half peak width [f-0.35 Hz, f+0.35 Hz], and obtain the actual stimulation frequency and peak amplitude.

[0039] according to Figure 1 Flowchart step 4: Noise calculation, set the protection window length to 0.2Hz, the noise window length to 2Hz, and calculate the median value of the noise window length [f+0.55Hz, f+2.55Hz] at the peak width of the right half of the stimulation frequency plus the protection window length as the noise signal amplitude

[0040] according to Figure 1 Step 5 of the flowchart is to calculate the signal-to-noise ratio, and calculate the signal-to-noise ratio of the fundamental frequency and harmonics based on the peak amplitude and the noise signal amplitude.

[0041] according to Figure 1 Flowchart step 6 outputs the result, and outputs the frequency recognition result based on the signal-to-noise ratio.

[0042] Example 2

[0043] The method steps of this embodiment are basically the same as those of embodiment 1, except that Figure 3 The experiment discussed the classification performance under different time window lengths in detail, in which the number of Nfft points was adjusted to be an integer multiple of 2 greater than the number of points in the time window length. Figure 2 (a) shows that in most experiments this method performs better than the CCA algorithm. Figure 2 In (b), it can be seen that this method is close to the CCA accuracy and will not be affected by the subjects changing the stimulation frequency; Figure 3 It can be seen that in the eight-classification experiment, when the time window is longer than 2s, the performance of this method is generally better than the CCA algorithm.

[0044] The present invention provides an SSVEP recognition method based on frequency domain signal-to-noise ratio. Compared with the prior art, the induced SSVEP frequency point information is automatically obtained by extracting frequency domain peak information, which solves the screen refresh rate limitation and the frequency offset problem caused by individual state differences to a certain extent. The solution based on the optimal signal-to-noise ratio of fundamental frequency and harmonics takes the influence of background noise and harmonics in EEG into consideration, which improves the application scope of the SSVEP recognition method to a certain extent and has a higher accuracy in practical applications.

[0045] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. However, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A SSVEP identification method based on frequency domain signal-to-noise ratio, characterized in that: The following steps are involved: Step 1: Frequency domain transformation, intercept the data within a specific time window of the EEG data signal after EEG preprocessing, set the number of Nfft points, and obtain the frequency domain signal of each channel; Step 2: Superposition and averaging: square the amplitude of the frequency domain signal obtained after FFT of each channel and then calculate the superposition and averaging to obtain a set of frequency domain signals of the overall lead for each trial; Step 3: Peak search, take the midpoint of the peak width, find the peak value of the frequency domain signal stimulation frequency of a group of overall leads in each trial within the left half peak width and the right half peak width, and obtain the actual stimulation frequency and peak amplitude; Step 4: Noise calculation, set the parameters of the noise window length and the protection window length, and calculate the median value within the noise window length at the peak width of the right half of the stimulation frequency plus the protection window length as the noise signal amplitude; Step 5: Calculate the signal-to-noise ratio, calculate the signal-to-noise ratio of each trial at the fundamental frequency of the stimulation frequency and its harmonics; Step 6: Output the results, compare the signal-to-noise ratio parameters, and output the frequency recognition results.

2. The SSVEP identification method based on frequency domain signal-to-noise ratio according to claim 1, characterized in that: The specific time window length is 2s and the number of Nfft points is 8192.

3. The SSVEP identification method based on frequency domain signal-to-noise ratio according to claim 1, characterized in that: The midpoint of the peak width was taken, and both the left and right half peak widths were set to 0.35 Hz.

4. The SSVEP identification method based on frequency domain signal-to-noise ratio according to claim 1, characterized in that: Set the noise window length to 2Hz and the protection window length to 0.2Hz.

5. The SSVEP identification method based on frequency domain signal-to-noise ratio according to claim 1, characterized in that: The value of the minimum frequency difference between the stimulation frequencies and the value of the specific time window length are reciprocals of each other.

6. The SSVEP identification method based on frequency domain signal-to-noise ratio according to claim 1 or 2, characterized in that: The number of Nfft points is an integer multiple of 2 greater than the number of points in the time window length.

Citation Information

Patent Citations

  • HHT-based high-frequency combined coding steady state visual evoked potential brain-computer interface method

    CN102063180A

  • SSVEP classification method based on convolutional neural network

    CN113052099A

  • SSVEP characteristic frequency extraction method based on FHN stochastic resonance

    CN113180706A

  • Electroencephalogram feature recognition method based on FBCCA and USSR

    CN117251768A