Research On The Independent Brain Computer Interface Based On Rapid Serial Visual Presentation

Compared with the traditional Subject-depend Brain-computer Interface(SDBCI),independent brain-computer interface system is non-individual dependent.Users can shorten or not undergo the calibration process when using it.However,due to the large inter-individual differences,background interference and other factors,the accuracy of independent brain-computer interface system classification is low,so that it cannot be widely used.Therefore,it is very important to explore the influencing factors and optimization methods of universal brain computer interface system for subjects.Based on the above background,this paper first designed EEG experiments based on Rapid Serial Visual Presentation(RSVP),and established EEG database of58 people.The response data of single trial were extracted.By calculating the cosine similarity between subjects,it was found that compared with the traditional matrix stimulation paradigm,the response waveforms between subjects were more similar under the RSVP paradigm,which proved that the response waveform induced by RSVP paradigm had smaller difference between subjects.Then,by extracting the difference wave to remove the influence of individual baseline,the result showed that the performance of the classifier with the difference wave as the feature was significantly improved(p<0.01).Based on this,the factors that affect the classification performance of SDBCI and independent brain-computer interface were analyzed.And SDBCI,Random Matching Algorithm(RMA)and Similar Matching Algorithm(SMA)were calculated.The average AUC values independent brain-computer interface were 0.8354,0.5948,and 0.6247,respectively.At last,we optimized the performance of independent brain-computer interface classification based on Active Learning(AL),and proposed SMA-AL algorithm.The average AUC of the independent brain-computer interface classification based on this algorithm was0.8194,which was significantly higher than that of SMA(p<0.01).The results proved the feasibility of applying the SMA-AL algorithm to independent brain-computer interface classification,as well as the promotion of Similar Matching Algorithm and the data of the subject's own data to be tested on the effect of recognition.In summary,this topic optimized the implementation scheme of independent brain-computer interface system from three aspects: experimental paradigm,feature extraction and classification methods.This topic explores the feasibility and improvement methods of the individual universal model,and provides a technical basis for establishing an individual universal brain-computer interface system and implementing a user-independent brain-computer interface technology.