Research Of Virtual Reality And Physiological Signal Characteristics

Virtual reality is a hot spot in the computer network world in recent years,which has attracted much attention.Emotion classification from EEG data are mature with the rapid development of dry electrode techniques and machine learing algorithms.This paper is a combination of computer science and technology and psychophysiology.The experiments illustrate the effectiveness of our proposed method.The traditional method of exposure therapy is the most effective in treating acrophobia,which is divided into real exposure and imagined exposure.Exposure therapy is a method of putting acrophobe exposed in the environment they are afraid of,in which acrophobe will gradually adapt to the process,so as to reduce the symptoms of fear.Also acrophobe will change perceiving and understanding of the fear of stimulation in the process and form a new pattern of behavior.But exposure therapy has some problems.First of all,the real exposure is dangerous in real environment and is difficult to implement.The limitations of imagined exposure is that it needs acrophobe have a very rich imagination.Acrophobe which has a poor imagination can not reach a good treatment.This paper puts forward a virtual reality exposure therapy,which uses the virtual environment instead of real environment and imagined environment.This method has the advantages of safety and repeatability.In order to improve the accuracy of classification,we combine the two algorithms of neural network and support vector machine.What's more,we combine principal component analysis,which is a algorithm of dimensionality reduction.From experimental results on our EEG data set,we found that(a)the paper puts forward the virtual reality exposure therapy which does have an effect on the treatment of acrophobia.(b)3s is the best time window of emotion.(c)higher frequency bands(alpha,beta,gamma)play a more important role in emotion activities than lower frequency bands(delta,theta).(d)detrended fluctuation analysis than is superior to other three kinds of features,namely standard deviation,approximate entropy and hurst exponent.(e)principal component analysis method is better thanlinear discriminant analysis.(f)we proposed method improves the accuracy rate of1.16 and 2.04 respectively in comparision to neural network algorithm and support vector machine algorithm.