EEG brain imaging based on Kalman filtering and subspace identification

The non-invasive neural activity estimation has a wide number of possible applications, from localization of pathologies inside the brain to the control of devices with the mind. But it is still an open research area because the limited number of sensors and the thousands of possible sources make it an ill-posed inverse problem. Minimum norm algorithms are widely used to estimate neuronal activity, but they do not include enough information to effectively reconstruct the sources. With the advent of more powerful computers it has been possible to add temporal information on the EEG inverse problem, but as the neuronal behavior is still under study there are problems to define a not too complex but useful temporal model. In this paper the use of subspace identification to include the temporal information available on the data on the temporal model of the brain is proposed. With this model a Kalman filter is used to locate the activation regions.