Research On Augmented Reality And Visual Decoding Based Brain-Computer Interface

As a new way of human-computer interaction,Brain-Computer Interface(BCI)is regarded as one of the disruptive technologies to change the future of the human beings for the reason that it can directly read human mind.After years of development,various control paradigms have been developed in BCI.According to the degree of system's dependence on external environment,these paradigms can be divided into independent BCI and dependent BCI.The dependent BCI can be further divided into tactile,visual and auditory evoked BCI.This study mainly focuses on the research on visual evoked BCI systems.In this paper,the method of improving the visual evoked BCI is studied from the perspective of interactive method and signal reverse coding algorithm.Firstly,the method of dynamic interaction is developed and validated in the practical system.Secondly,the proposed eLSTM(ensemble LSTM)network in this paper has a high accuracy in reverse coding the content of 40 kinds of images from Electroencephalography(EEG).The proposed method is expected to be applied in the actual system,and the signal process method of judging whether there is visual evoked signal can be promoted to multi-judgment of stimulus content,so as to further improve the efficiency of interaction.As the signal to noise ratio of the EEG is relatively low and a specific signal induction method is required,the information transmission rate is not very high.It means that the output of one command tends to have a large time delay.Aiming at this problem,this paper designs a dynamic interaction method based on the Augmented Reality technology and applies it to a BCI environment control system.By detecting the objects to be controlled in the field environment and excluding the irrelevant instruction options,the dynamic interaction method dynamically presents the interactive content between human and system.Five subjects were tested by the system.The results show that under the same condition,the dynamic interaction method can reduce the command delay time by 17.4% compared with the static interaction method.At the same time,due to the design of redundant instructions in the dynamic interface,the number of the wrong operation was maintained at a low level.At present,the basic principle of P300-based BCI is to judge whether there is evoked potential in the signal.Physically,people are more sensitive to visual information.Visual-evoked signals contain rich information about the subjects' sensory and cognitive processes.If the content of visual stimuli can be decoded from EEG,it will be helpful to improve the current interactive methods.Based on picture-evoked EEG,this paper proposes a neural network-based visual information reverse coding algorithm.The proposed eLSTM network has achieved an accuracy of 98.14% for 40 kinds of signals.The average time consumption is less than 0.01s?As an integrated form of multiple networks,the proposed e LSTM network integrates the reverse coding information of multiple Stacked LSTM networks,so it has a higher accuracy.The paper also analyzed the effects of different brain region's data on reverse coding accuracy.The results show that under the condition of limited channels data,the reverse coding accuracy of utilizing the electrode signals in occipital and parietal lobes region is better than that in frontal lobe region or whole brain region.