Hidden Markov models for online classification of single trial EEG data
Pattern Recognition Letters
| Hi-index | 0.00 |
Chaos in nonlinear dynamical systems has become a widely-known phenomenon and its presence has been identified in many different systems in virtually all the fields of science. In medical world, analyzing chaos in the brain and explore its dynamics is a challenging task to every individual. In this paper, an effective and a practical method for exploring such brain activities are studied. This paper relates a method to analyze an Electroencephalogram (EEG) using Correlation Dimension (CD) for drowsiness estimation in sleep onset and epilepsy. Dimension is a critical property since, it indicates how many independent state variables are required to reproduce the system dynamics in state space and this in turn indicates how many state variables should be included in a mathematical model of the system. Aside from this practical issue, the dimension is an indicator of the degree of "complexity" of a system and tracking any changes in dimension due to pathology or other manipulations to the system is a useful diagnostic criterion. For many chaotic systems, accurate calculation of the CD from measured data is difficult because of very slow convergence as the scale size is reduced. This paper proposes a method for collecting data at large scales, creating the time series and determining the possibility of constructing an attractor for establishing the deterministic character of dynamics of the underlying system.