ACUPUNCTURE EEG TIME SERIES DECOMPOSITION BASED ON STATE SPACE METHOD

摘要

This paper aims to decompose EEG time series by using state space modeling method. EEG is recorded when acupuncture zusanli (ST-36). Observations are regarded as made up of distinct components such as trend, period and stochastic disturbance terms. State space modeling method can extract the principal distinct components from time series observations. Firstly, modeling autoregressive moving average, the Akaike Information Criterion is used to choose the order of the model, and the method of least square is used to determine the parameters of the model. Secondly, the ARMA model is cast in a state space framework. To obtain distinct components, we transform state space expression into Jordan canonical form. Lastly, Kalman filter is used for components estimation. The results show that this method can effectively extract the EEG characteristics which can be applied to eliminating artifacts and extracting brain rhythms.