CN-116269445-B - Accurate target identification method for SSVEP short time window signal

Abstract

The accurate target identification method for the SSVEP short time window signals comprises the steps of constructing a signal extension model DP-MAFD-SEM by using one or more SSVEP signals corresponding to a stimulation target, carrying out signal extension on the SSVEP short signals by using the signal extension model, increasing the SSVEP signal length, and then completing identification and classification of characteristic frequencies by using a typical correlation analysis method (SE-CCA) based on the signal extension; the SSVEP short time window signal length is increased through signal extension, so that the reliability of identification based on covariance matrix estimation methods such as CCA (clear channel assessment) and the like is improved, and higher identification accuracy is realized; the invention provides a brand new angle and approach for improving the identification accuracy of the special frequency of the SSVEP short signal, realizes higher identification accuracy by using the shorter signal, is beneficial to further improving the information transmission rate of the SSVEP-BCIs, promotes the development of high-speed SSVEP-BCIs and promotes the practical development of the SSVEP-BCIs.

Inventors

• XU GUANGHUA
• LI HUI
• LI ZEJIN
• HAN CHENGCHENG
• DU CHENGHANG

Assignees

• 西安交通大学

Dates

Publication Date

20260505

Application Date

20230320

Claims (3)

1. 1. A method for accurate target identification of SSVEP short time window signals, comprising: The SSVEP signal extension model DP-MAFD-SEM（Signal Extension Model Based on Multi-channel Adaptive Fourier Decomposition with Different Phase）： utilizes different phase Multi-channel adaptive Fourier decomposition (Multi-CHANNEL ADAPTIVE Fourier decomposition WITH DIFFERENT PHASE, DP-MAFD) to simultaneously decompose one or more source-stimulated identical channel signals, extracts common periodic components, reconstructs the SSVEP signal from the common components, and extends the signal by time shifting; The SSVEP characteristic frequency identification method based on Signal extension is SE-CCA (Signal-Extension Canonical Correlation Analysis) which comprises the steps of carrying out Signal extension on the Signal to be detected at each stimulation target frequency by utilizing the constructed DP-MAFD-SEM, calculating typical correlation coefficients by utilizing CCA respectively, and taking the frequency corresponding to the maximum typical correlation coefficient value as an identification result; The accurate target identification method for the SSVEP short time window signal comprises the following steps: step 1) multichannel signal acquisition, namely acquiring a tested multichannel SSVEP signal, wherein the SSVEP signal is subjected to amplification, filtering and digital-analog conversion; Step 2) signal preprocessing; Step 3) constructing a signal extension model DP-MAFD-SEM to realize signal extension; Randomly selecting one or more stimulation targets as SSVEP signals stimulated by a source, decomposing one or more signals of the same channel stimulated by the source by using DP-MAFD at the same time, extracting common periodic components, and obtaining self-adaptive bases and decomposition coefficients of corresponding decomposition layers; Step 4) SSVEP signal continuation; respectively carrying out signal extension on each channel of the SSVEP by using the signal extension model obtained in the step 3) through time offset to obtain extension part signals, and splicing the extension part signals after the signals to finish signal extension; step 5) using SE-CCA to complete SSVEP characteristic frequency identification; and 3) realizing final characteristic frequency identification by using SE-CCA, carrying out signal extension on the signal to be detected at each stimulation target frequency by using the DP-MAFD-SEM constructed in the step 3) and the extension process of the step 4), respectively calculating typical correlation coefficients by using CCA, and taking the frequency corresponding to the maximum typical correlation coefficient value as an identification result.
2. 2. The method according to claim 1, wherein 9 measuring electrodes are arranged in the occipital region of the head to be tested in the step 1), namely, pz, PO5, PO3, POz, PO4, PO6, oz, O1, O2, a reference electrode Fpz is arranged in the earlobe on one side thereof, a ground electrode Gnd is arranged at the forehead thereof, and electroencephalogram signal acquisition is performed by using an electroencephalogram acquisition device.
3. 3. The method according to claim 1, wherein in step 2), the signal is normalized and zero-averaged, and then the signal component of 6-90Hz is extracted by means of a butterworth filter bandpass filter.

Description

Accurate target identification method for SSVEP short time window signal Technical Field The invention relates to the technical fields of biomedical engineering and brain-computer interfaces, in particular to a precise target identification method for SSVEP short-time window signals. Background Brain-computer interface (BCI) is a technology which does not depend on human neuromuscular channels and can realize direct information communication between the brain and external equipment, has been developed in the last decade, and developed into systems such as brain-controlled artificial limbs, brain-controlled spelling, brain-controlled wheelchairs, brain-controlled unmanned aerial vehicles and the like. Steady State Visual Evoked Potential (SSVEP) is the result of the modulation of the potential activity of the visual cortex in the occipital region of the brain when the human visual system is subjected to visual stimuli at an external frequency (typically greater than 3 Hz). The method is widely applied to various brain-computer interface systems due to the characteristics of no need of training, high signal-to-noise ratio, good robustness and the like. Increasing the Information transfer rate (Information TRANSFER RATE, ITR) of SSVEP-BCIs is a research hotspot in this area. The ITR is mainly determined by the number of stimulus targets presented, the accuracy of target recognition, the signal window length used. Advanced stimulated target coding methods and target recognition methods have been developed in recent years to further improve the ITR of SSVEP-BCIs, but the key to improving the ITR is to improve the performance of the algorithm, i.e. to achieve higher recognition accuracy with shorter signal window lengths, and recent studies to improve the ITR have focused mainly on improving the performance of SSVEP target recognition methods. The exemplary correlation analysis (Canonical correlation analysis, CCA) is a standard target recognition method of SSVEP because it uses spatial information between multiple channels, and has better target recognition performance than conventional fourier transform, etc. Numerous researchers develop CCA from different angles to further improve the performance, such as optimizing P-CCA (Phase Constrained Canonical Correlation Analysis) from a phase angle, enhancing FBCCA (Filter bank canonical correlation analysis) by using harmonic signals, and the like, constructing a signal template with real electroencephalogram signals from the point of constructing the signal template, TRCA (Task-Related Component Analysis) extracting Task related components as the signal template, and eTRCA (Task-Related Component Analysis) integrating spatial filters obtained by different stimulus targets on the basis of TRCA to improve the target recognition performance. A trained method such as MsetCCA, mwayCCA, IT-CCA, etc. has better target recognition performance than a non-trained method such as CCA, IT-CCA, FBCCA, etc., but a trained-based method requires a large amount of calibration data for calibration, which limits ITs practical application. A user-friendly SSVEP-BCIs is more prone to use short-time window signals and untrained methods for target recognition. The current advanced identification method has obviously improved the target identification accuracy of the SSVEP, but the target identification performance of the SSVEP on short-time window signals is still poor, which severely limits the development and practical application of the high-speed SSVEP-BCIs. As the SSVEP signal length increases, the accuracy of target recognition will increase, which has become common sense. The reason for this is that longer SSVEP signals help to obtain a more accurate spatial filter. Secondly, CCA and its extension methods are based on statistical methods of covariance estimation, which become unreliable when the signal length is short. The end point effect for solving the empirical mode decomposition in the field of fault diagnosis based on the signal extension method is widely applied, but a method for improving the signal target recognition accuracy of the SSVEP short time window by the signal extension technology is not available yet. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide an accurate target recognition method for an SSVEP short time window signal, which is used for prolonging the length of the SSVEP signal from the data perspective by a signal extension technology so as to improve the target recognition accuracy of the SSVEP short time window signal. In order to achieve the above purpose, the invention adopts the following technical scheme: a method for accurate target identification of SSVEP short time window signals, comprising: the SSVEP signal extension model DP-MAFD-SEM(Signal Extension Model Based on Multi-channel Adaptive Fourier Decomposition with Different Phase)： utilizes different phase Multi-channel adaptive Fourier de