Title :
Forward and backward autoregressive modeling of EEG
Author_Institution :
Dept. of Electr. Eng., Northern Illinois Univ., DeKalb, IL, USA
fDate :
30 Oct-2 Nov 1997
Abstract :
Causal autoregressive (forward prediction) process is the most popular model used to parameterize an EEG segment. In this paper, we attempt to increase the modeling accuracy by removing the causality constraint. Two important observations can be made from the analysis of a set of EEG data collected during an animal experiment with induced brain injury. It was found that the residual error decreases when a forward and backward prediction model is used. The amount of the residual error decrease is minimal for those segments of the EEG data corresponding to severe brain injury
Keywords :
autoregressive processes; brain models; electroencephalography; mean square error methods; medical signal processing; prediction theory; signal sampling; time series; EEG segment parameterization; animal experiment; backward autoregressive modeling; causal autoregressive process; causality constraint removal; forward autoregressive modeling; forward prediction process; induced brain injury; mean square prediction error; minimum energy; modeling accuracy; residual error; severe brain injury; white noise; Algorithm design and analysis; Animals; Brain injuries; Brain modeling; Computerized monitoring; Context modeling; Electroencephalography; Predictive models; Signal processing; Time series analysis;
Conference_Titel :
Engineering in Medicine and Biology Society, 1997. Proceedings of the 19th Annual International Conference of the IEEE
Conference_Location :
Chicago, IL
Print_ISBN :
0-7803-4262-3
DOI :
10.1109/IEMBS.1997.756582
