Tracking Conductivity Variations in The Absence of Accurate State Evolution Models in Electrical Impedance Tomography
We present results on both linear and non-linearapproaches in tracking conductivity variations in electricalimpedance tomography. Throughout this study, we use bothsynthetic and measured data. The true system dynamics isconsidered as unknown and modelled as a random walk. In thelinear reconstructions, the time evolution model is augmentedwith a Gaussian smoothness prior and results are shown usingtwo different models for the covariance matrix of the processnoise. Furthermore, we compare the reconstructions of the onestep Gauss-Newton method to the Kalman filter on measureddata from an adult human subject. In the non-linear study,we compare the performance of the extended Kalman filteragainst the particle filter on a simple test case. It is observedthat the particle filter shows superior performance in trackingnonlinear/non-Gaussian conductivity variations.
Parham Hashemzadeh Vijay Sahota Martina F.Callaghan Hussain El Dib Andrew Tizzard Lennart Svensson Richard Bayford
Department of Natural Sciences, School of Health and Social Sciences Middlesex University, Hendon NW4 4BT, London UK
国际会议
成都
英文
1-6
2010-06-18（万方平台首次上网日期，不代表论文的发表时间）