Feature extraction for EEG based emotion prediction applications through chaotic analysis

Abstract

Electroencephalogram (EEG)-based emotion recognition is a rapidly growing field. This paper presents a research study on identifying the changes caused by human emotions in brain signals through EEG records and further examines the availability of chaotic analysis for emotion prediction applications. We designed an efficient presentation including 30 pictures from International Affective Picture System (IAPS) which could stimulate the emotions of happiness, sadness and fear. A group of 20 persons consisting of 12 males and 8 females took place in our research study as subjects voluntarily. In order to acquire the EEG signals under picture induction environment, a total of 30 pictures that might reveal the feelings of happiness, sadness and fear were shown to the volunteer subjects that participated in the study voluntarily. Data acquired from EEG records were analyzed by using chaotic analysis through MATLAB program and two different attributes consisting of Largest Lyapunov Exponent (LLE) and Correlation Dimension were obtained. The mean values of each attribute that obtained through chaotic analysis were compared by using independent t-test and dependent t-test in SPSS program with 95% confidence interval and a P value of p<0.05. The independent samples t-test was used to compare the mean values of each attribute relevant to two unrelated groups (Females-Males) and Dependent samples t-test was used to test whether the mean values of two related observations (two observations for per subject) significantly differs from the hypothesized value. Our study shows that chaotic analysis promises hope for future studies.