Visual Target Selection On P300-Based Brain-Computer Interface

The brain-computer interface(BCI)based on P300 event-related electrical potential can achieve a target selection task by only detecting the human brain activities,is one of the most widely used type in BCI filed.In conventional P300-based BCIs,the row-column mode was widely used to modulate the stimulation of the targets.However,when extend the P300-based BCIs to practical applications,the regular stimulation mode is insufficient to reflect the complex target information in actual environments.To address this problem,we propose a novel target selection approach by incorporating image segmentation method and two-step selection into the P300-based BCIs(IST)and improve the ability of selecting the target in the real environment by BCIs.Exploring IST paradigm design and verification.The image of the environment captured by a camera was partitioned by the Entropy Rate Super-pixel Segmentation(ERS)algorithm,a image segmentation method in the computer vision filed.Then,a random flash stimulation was embedded on each segments of the image to evoke P300 signal.A two-step mechanism was used in our BCI system,where a group contained the target was selected first and then the target was selected from this group.In the design process,the stimulus time,the stimulus type,the stimulus code,the signal processing arithmetic are designed appropriately.To verify the performance of our approach,a target selection experiment was performed in different real environments and a single step paradigm was compared with our approach.The average online accuracy in the experiment for ten subjects was 86.0% using our proposed approach which performed better than single step mechanism with the accuracy 69.0%,and the average information transfer rate was 22.1bits/min using our approach compared to the 10.29bits/min in single step mechanism.The results showed that the feasibility and practicality of the P300-based BCIs for target selection was improved by incorporating the image segment method.Experiment analyzation and optimal design on IST-based BCI.To improve the performance of the proposed paradigm,we analyze it from four aspects and propose some improvement measures.Firstly,a comfirmatory experiment was designed to showed the correlation between the image complexity and the online selection accuracy.Then we optimize the new approach by reduce the complexity of the image through image processing method.Secondly,we determine the optimal number of stimulus trials for each subject during online selection by the discussion and analysis the performance of different subject in different trials.Thirdly,after analyzing the questionnaire during the experiment we suggested optimizing the interface of the proposed approach to be more friendly.Lastly,the adjacent problem has perplexed either the traditional P300-BCI or the presented approach.To adress the problem,we design a contrast experiment between disperse block code and centralized block code,and optimize the block code mechanism and the performance of the presented approach.The analysis of the four aspects showed that although it had some shortcomings the proposed approach was feasibility and responsibility.The experiment result and the analysis of the result have verified that the P300-based BCIs for target selection was improved by incorporating the image segment method.The proposed approach will inject new vitality to the BCI field and the computer vision field at the same time,and put forward a new way to do the target selection in the real environment.What’s more,it shows great potential in the practicality of P300-BCIs because it can achieve better performance by parameter optimization and design optimization.