Spatial filtering and its application to EEG source localization

In sensor array processing, an array of sensors is used to discriminate sources based on their location. The principal goal of spatial filtering or beamforming is to achieve spatial selectivity among different sources of interest by using a filter at the output of the array. Optimum beamformers, also known as minimum variance distortionless response (MVDR) beamformers, can be designed for this purpose. These filter weights can also be used to localize sources by looking at the filter response to different locations. Subspace techniques like multiple signal classification (MUSIC) have been shown to have high resolution characteristics for localization of sources. An unified formulation called the enhance minimum variance (EMV) technique can be used for both these techniques. The performance of these techniques for localization of narrow-band sources is analyzed. Expressions are derived for the variance and bias of the estimates due to different types of errors.; In electroencephalogram (EEG) an array of sensors is used to record the potentials due to electric activity within the brain at the surface of the scalp. Using this data, multiple current dipole localization is currently being used to provide information about the functional brain activity. In this thesis, spatial filtering techniques are used to localize the dipole source within the brain. In particular, the EMV technique is used to estimate the location of the dipole source. The performance of this method is also analyzed and expressions are derived for the variance of the estimates due to different types of errors.; Another important problem is the placement of electrodes for recording EEG signals. While the most commonly used electrode configuration uses 21 sensors, it has been shown to be inadequate for accurate evoked potential localization. New standards for electrode placement are needed and this need is addressed in this thesis. A general framework based on sampling a continuous potential is presented for selecting the number of electrodes. Conditions are also derived for placement of these electrodes.