An estimation of the correlation dimension for the EEG in emotional states

Abstract

To study the underlying dynamics of emotional states in the human brain by a nonlinear dynamical analysis, we estimate the correlation dimension for the electroencephalogram (EEG) in positive and negative emotional states. The time delay for the reconstruction of the EEG in the embedding space is not determined from a single function, such as the autocorrelation function, but is determined to yield the largest plateau in the local slope of the correlation integral for each embedding dimension. In our case, the absence of an abrupt change in magnitudes of singular values makes singular value decomposition useless, which implies the absence of a dominant process in emotional states. Our results show that the correlation exponent for the emotional state is higher than that for the rest state even if the variation between subjects is considered. This may be a first step to discriminate between "Yes" and "No" for communication via EEG, particularly for severely disabled people.