EEG-based Virtual Smart Home Control And Tense Emotion Recognition

In recent years,brain-computer interface(BCI)and virtual reality(VR)technology have become the hot topics in research and application in domestic and overseas.However,the current common VR interaction method relies on external devices which are applicable only to a limited part of people.Also,emotion recognition with BCI usually only uses the form of poor immersion such as video,picture or music,which results in poor recognition accuracy.Combining BCI with VR into a BCI-VR system can realize complementary advantages.On the one hand,BCI provides a new way of interaction for VR.On the other hand,VR provides BCI with highly immersive and rich scenarios.Therefore,this dissertation mainly studies the interaction method and emotion recognition based on electroencephalogram(EEG).A virtual smart home control system with P300 as interactive method and a highly immersive emotional recognition system are designed.The main work of this dissertation is as follows:1)A P300-based virtual smart home control system is designed and built.Seriously paralyzed patients cannot interact with virtual reality through keyboards,mice,handles,etc.Accordingly,the proposed system uses P300 as the control signal,which allows patients to interact with VR directly through EEG,and provides them with a new way of entertainment.The system embeds the P300 stimulus interface to stimulate the user to generate P300 signals.Signal-tonoise ratio can be improved by superposition of P300 signal by repetitive stimulation.In order to solve the low information transmission rate problem caused by excess repetitions,this dissertation designs a algorithm framework based on Bayesian linear discriminant analysis(BLDA)which ensures high information transmission rate and classification accuracy.The experimental results of 8 subjects show that the average accuracy of the proposed system achieves 95.85%,and the false trigger rate is low.Moreover,the practicability of the system is also validated by the mental state evaluation of the subjects.2)An EEG-based tense emotion recognition system is designed and built.In the field of emotion recognition,it is often difficult to achieve the desired effect due to different cognitive situations and cultural backgrounds.The proposed system uses an immersive virtual drifting scene as a material for emotional stimulation,which combines multiple sensory stimuli such as sight,hearing and perception.The proposed way of emotional stimulation is able to stimulate users' tense emotion effectively.The system filters the EEG signals into five frequency bands.Comparisons of the three characteristics including power spectral density(PSD),rational asymmetry(RASM)and differential asymmetry(DASM)in each frequency band are made.Features are selected by principal component analysis(PCA)algorithm and support vector machine(SVM)is used to classify and recognize tense emotions.In order to improve the accuracy,feature fusion is carried out in the five frequency bands.The experiment results of 8 subjects show that the average recognition accuracy of the system reaches 88.13%.Also,it is verified that the high immersion stimulus can achieve satisfactory emotional recognition effect.