Detection of EEG changes via a generalized Itakura distance

Changes in electroencephalograms (EEGs) carry important clinical information. The accurate detection and characterization of such changes in EEG can be a valuable tool for the clinical assessment of the neurological system condition. Autoregressive (AR) modes have previously been used to detect changes in the EEG signal. Based on the AR model parameters, an off-line distance measure called the Itakura distance (F. Itakura, 1975) has been used to effectively quantify changes in the EEG signal related to brain injury. The ordinary Itakura distance measure used for such quantification requires the same order for the AR models in all EEG segments. In this paper, a generalized Itakura distance measure is proposed without this constraint. The generalized measure is applied to the analysis of EEG signals. Preliminary results suggest that the generalized Itakura distance measure performs better for detecting injury-related changes in EEG