A Kalman Smoother-Based Approach for Estimating Time-Varying Cortical Connectivity from High-Density EEG
Inferring the cortical connectivity from highdensity electroencephalogram (EEG) is attracting growing attention in the filed of cognitive neuroscience as it provides the time-varying patterns of information transfer among distributed cortical areas with high time resolution and satisfactory spatial resolution. Currently, one commonly-used approach to investigate the cortical connectivity is the Granger causality, which describes the multichannel high-density EEG data as a multivarjate autoregressive (MVAR) model. A set of measures of the dynamic cortical connectivity, such as the directed transfer function (DTF) and the partial directed coherence (PDC), can be obtained from the MVAR coefficient estimates. Identification of the MVAR model is conventionally achieved by the sliding-window approach or the recursive least squares (RLS) algorithm. However, these methods often exhibit considerably large variability when dealing with high-dimensional EEG, which implies that a large number of parameters are to be estimated from a limited number of measurements. Therefore, the sliding window and the RLS methods cannot accurately estimate the dynamic MVAR coefficients, possibly leading to a wrong interpretation of cortical connectivity.
Z. G. Zhang C. Q. Chang Y. S. Hung
Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong
国际会议
The 7th International Conference on Cognitive Science(第七届国际认知科学大会 ICCS 2010)
北京
英文
467-468
2010-08-01（万方平台首次上网日期，不代表论文的发表时间）