SSVEP Brain-Computer Interface For Life Assistance Based On Deep Learning And Augmented Reality

Many patients in the world lost the ability to communicate and move,but they still produce electrical signals that can be recorded by brain-computer interface(BCI)devices.BCI can help these patients communicate with the outside world.Steady-state visual evoked potential(SSVEP)-BCI has attracted wide attention due to its simple usage and high information transmission efficiency.However,the current level of BCI limits its further development.This paper studies its algorithm and application.For the SSVEP-EEG recognition algorithm,the traditional algorithm is often difficult to obtain a good recognition accuracy in a short time-window due to the SSVEP-EEG data containing a lot of noise,which undoubtedly limits the development of SSVEP-BCI in life application.So,a convolutional neural network(CNN)structure based on a filter bank(FBtCNN)is proposed.The filter bank in the proposed FB-tCNN can remove some noise,make the network architecture extract frequency-related features(fundamental and harmonics)better,realize effective recognition in a short time-window.It can improve the application potential of SSVEP-BCI.It lays the foundation of a recognition algorithm for constructing a faster and more flexible human-computer interaction system.For the SSVEP-BCI application,an independent monitor is used in the traditional SSVEP-BCI to display the stimulus targets,which enables the user to switch the eye between stimulus targets and the robotic arm frequently.And stimulus targets always correspond to the pre-set target objects.The mapping relationship between stimulus targets and target objects needs to be readjusted in the new environment,which undoubtedly limits the development of SSVEP-BCI in life.So,a stimulation interface displayed on the augmented reality(AR)device is proposed,which combines the stimulus target and vision information,can update the mapping between the stimulus targets and the target objects in real-time according to the work environment.The experimental results show that the proposed AR-baed SSVEP-BCI can enable users to choose target objects more naturally,and has the potential to be applied in more complex environments.