Research On Image Recognition Method Based On Multi-channel Eeg Signal

Image recognition is to select specific target image from image library,the current recognition methods include machine vision and manual interpretation.The former can realize the rapid detection of images with the help of the powerful execution capabilities of the computer,but in some images with relatively complicated backgrounds,the recognition effect of machine vision is often unsatisfactory.Manual interpretation relies on the characteristics of high robustness and high reliability of the human visual system,which can achieve high-precision recognition of the target image.The disadvantages of manual interpretation are long time,low efficiency,fatigue and high cost.With the help of the Rapid Serial Visual Presentation(RSVP)paradigm,a humancomputer hybrid Brian computer interface(BCI)system can be realized.The system realizes the rapid recognition of the target image by detecting the P300 component corresponding to the target image in the EEG signal.This article starts with the generation mechanism,nature and characteristics of the P300 component,and combines the time,frequency,space and multi-channel synchronous coupling analysis methods of EEG signals to study how to further improve the accuracy and efficiency of target image recognition under the RSVP paradigm.The main work as follows:Firstly,aiming at the problem of uneven distribution of P300 components in different EEG rhythms,the frequency domain features are introduced in the Hierarchical Discriminant Component Analysis(HDCA)algorithm,and a Multi Domain Filtering HDCA(MDF-HDCA)algorithm based on time space frequency multi domain filtering is proposed.The algorithm uses a sliding time-frequency window to structurally weight the EEG signals in time,space,and frequency,and ultimately improves the accuracy of P300 component detection.Secondly,based on the phenomenon of synchronous coupling between EEG signals,the method of extracting local band coupling features of multi-channel EEG signals in RSVP paradigm is studied.The synchronous coupling features of multi-channel EEG signals caused by visual stimulation of target image are described from the perspectives of linear correlation,nonlinear correlation and phase synchronization.Thirdly,combining the structural discrimination algorithm with the synchronous coupling feature extraction method,the feature set of target image recognition based on EEG signal is further expanded.The information fusion strategy is designed from feature layer and decision layer respectively,and the algorithm framework of multi-level and highdimensional feature recognition target image based on multi-channel EEG signal is constructed.Finally,the RSVP experiment is designed based on the actual application scenario,and the experimental platform is built based on MATLAB environment.The EEG data is collected to verify the effectiveness of MDF-HDCA algorithm and multi-level feature fusion image recognition algorithm in improving the P300 detection performance.