Research On Multimodal Brain-Computer Interaction System And Application

Brain-computer interaction system,as an important way for the communication between the brain and the external device,has obtained active attention and research in various fields.The traditional single-mode brain-computer interaction system can not meet the actual needs because of its fewer task categories,so a multi-mode system emerged.At the same time,other types of interactions are gradually applied to brain-computer interaction system.Based on those,this paper combines alpha-wave,P300 signal and eye-movement signal to explore the working mode of multi-mode system.Constructing the hardware and software platform and solving the signal synchronization,an effective signal processing method is presented as well.The specific works are as follows:(1)Propose an alpha-wave detection algorithm based on canonical correlation analysis to ensure the effective use of alpha wave for system suspension.The core idea of the method is to use the CCA method to extract characteristic frequencies from the concentrated frequency band.These characteristic frequencies can accurately represent the differences between alpha-wave and non-alpha-wave of different experimenters.The correlation coefficient corresponding to the feature frequency is combined as the feature that is ultimately used for classification.The experimental results show that using 1s EEG data,the method can accurately analyze whether it belongs to alpha-wave,and the average detection accuracy rate is above 95%.At the same time,extending the data time and increasing the number of characteristic frequencies can make the accuracy higher.Compared with other methods,the algorithm has obvious advantages in terms of timeliness and accuracy.In addition,the algorithm is easy to use and does not require repeated training.(2)Propose a P300 signal detection method based on segmented spacetime reduction.This method is used to detect and identify target images.The core idea of the algorithm is to use FLD and PCA to perform dimensionality reduction on a set of segmented EEG data in both time and space,which can not only distinguish between P300 signals and non-P300 signals,but also can effectively reduce signal redundancy features.The experimental results show that under the segmentation of 16 cases,the area under the ROC curve(AUC)of the ten experimenters is 0.922,and the average detection accuracy rate is 86.8%.When the segmentation time is 8,the AUC is highest,which is 0.9375,and the average detection accuracy rate was 88.53%.Compared with other algorithms,the method has a certain advantage in the accuracy,so it can be used to detect image target in the multi-mode brain-computer interaction system.In addition,this paper uses eye movement information to achieve the target location calibration.The main idea is to use the eye movement signal which is collected during training to determine the response time of the experimenter after the target appeared,and to fit the response time to a normal distribution.In the actual system test stage,the system randomly generates a number that obeys the experimenter's reaction time distribution,and takes the position of the eye gaze point after the time as the target appearance position,and then frames the label.The experimental results show that the reaction time varies from person to person,but this method can complete the target calibration task.