非線形予測を用いた脳波の解析

概要

論文の詳細を見る
Recently, electroencephalagraphic potentials(EEG) have been analysed form the view point of nonlinear determinisity, or deterministic chaos. As for evaluating deterministic chaos, novel methods are proposed, for example, correlation dimensional analysis and Lyapunov spectral analysis, which quantify the important characteristics of chaos, self-similarity and orbital instability. Although these approaches are very effective for nonlinear time series, it is also pointed out that a careless application would lead to spurious estimates of these nonlinear invariants, especially in case of analysing EEG. In order to avoid such kinds of misinterpretation, in this paper we analyse EEC by nonlinear prediction in order to quantify long-term unpredictabilities and short-term predictabilities, which are also fundamental characteristics of deterministic chaso.In this paper, we introduce a local linear prediction algorithm that estimates Jacobian matrices, and two correlation coefficients are evaluated. As comparative studies, time series data from macroscopic observation of internal states of neurons implemented in asynchronous chaotic neural network(ACNN),and random time series are also analysed.
社団法人電子情報通信学会の論文
1996-03-11