Design And Application Of SSVEP-based Brain-Robot Interaction Online System

Brain computer interface(BCI)is a new human-computer interaction technology.It directly controls the external devices by acquiring and analyzing EEG signals without depending on peripheral nerve pathways and muscles.Brain robot interaction(BRI)system is a human-machine fusion control system based on BCI technology in which the brain serves as the center and the robot as the control object.With the development of Brain robot interaction system,it has been widely applied in medical,education,entertainment,military and other fields.In order to improve the classification accuracy of steady-state visual evoked potential signals in BCI system,a novel classification algorithm combining Fisher and Fuzzy is developed in this dissertation.Initially,the acquired EEG signals are spatially filtered by CCA,and then the best projection plane is obtained by Fisher operation.The distance d between the sample points and the projection plane can be calculated.Finally,the fuzzy measures are obtained from the distance d by using a fuzzy algorithm and the classification result can be acquired.The classification algorithm overcomes the shortcoming that the samples in the ambiguous area cannot be accurately classified by a single Fisher classifier in multiple classification problems.And the influence of sensitivity to different frequency stimuli is greatly eliminated.The results show that the performance of proposed algorithm is much better than a single Fisher classifier or SVM by comparing average classification accuracy and variance of each algorithm.The Fisher+Fuzzy classification algorithm enables to improve the accuracy of the BRI system,but the required parameter optimization process increases the complexity of this algorithm.In order to enhance the performance of online operation,this dissertation combines limited penetrable visibility graph with customizable programmable robot,aiming to fulfill the robot obstacle avoidance task online.The system includes a stimulus interface with four flickering images.We infer complex networks from EEG signals and then extract the degree values,to generate four instructions for controlling the robot forward,turning left,turning right and back,respectively,which allow the robot to complete the obstacle avoidance task online.Our paradigm builds a bridge between complex network and BCI system and its powerfulness has been articulated.This dissertation designs an EEG analysis toolbox aiming at gathering our proposed methods as well as some commonly used methods in EEG signal preprocessing,feature extraction and pattern classification.The designed toolbox is capable of reading two commonly used file formats,displaying,filtering and down sampling EEG signals.In particular,for the extensively used SSVEP and P300 paradigms,we design specialized modules and add the corresponding analysis algorithms into these modules,which make the signal analysis of the brain control system more convenient.