Title :
Autoregressive and bispectral analysis techniques: EEG applications
Author :
Ning, T. ; Bronzino, J.D.
Author_Institution :
Dept. of Eng., Trinity Coll., Hartford, CT, USA
fDate :
3/1/1990 12:00:00 AM
Abstract :
Some basic properties of autoregressive (AR) modeling and bispectral analysis are reviewed, and examples of their application in electroencephalography (EEG) research are provided. A second-order AR model was used to score cortical EEGs in order. In tests performed on five adult rats to distinguish between different vigilance states such a quiet-waking (QW), rapid-eye-movement (REM), and slow-wave sleep (SWS), SWS activity was correctly identified over 96% of the time, and a 95% agreement rate was achieved in recognizing the REM sleep stage. In a bispectral analysis of the rat EEG, third-order cumulant (TOC) sequences of 32 epochs belonging to the same vigilance state were estimated and then averaged. Preliminary results have shown that bispectra of hippocampal EEGs during REM Sleep exhibit significant quadratic phase couplings between frequencies in the 6-8-Hz range, associated with the theta rhythm.<>
Keywords :
electroencephalography; spectral analysis; 6 to 8 Hz; Fourier transform; adult rats; autoregressive modelling; bispectral analysis techniques; cortical EEG; electroencephalography; hippocampal EEG; quadratic phase couplings; quiet-waking; rapid-eye-movement; slow-wave sleep; theta rhythm; third order cumulant sequences; vigilance states; Autocorrelation; Autoregressive processes; Brain modeling; Electroencephalography; Equations; Frequency; Gaussian distribution; Parametric statistics; Random processes; White noise;
Journal_Title :
Engineering in Medicine and Biology Magazine, IEEE
