Some New Tools for EEG Modeling and Analysis
Several tools for the modeling and analysis of scalar and multivariate EEG data that are not necessarily stationary (and may in fact be subject to abrupt changes in covariance structure), and not necessarily Gaussian distributed are introduced.
First we show a “smoothness priors” quasi Bayesian method of time series analysis that is applicable to the modeling of scalar non stationary covariance EEG data. That methodology is exploited here to realize the power spectral density of both slowly varying nonstationary covariance EEGs as well as EEGs whose covariance structure changes abruptly. (The latter is achieved via a state space non-Gaussian smoothness priors analysis, that does not require data segmentation.) Secondly we introduce a one channel at-a-time″ paradigm which yields the autoregressive (AR) modeling of multivariate stationary and nonstationary covariance EEG time series by successive scalar autoregressive time series modeling An application of that paradigm to the parsimonious modeling of multivariate stationary EEGs is shown. Such relatively statistically efficient parsimonious modeling can potentially enhance stationary multivariate EEG classification performance. Finally, the smoothness priors modeling of scalar nonstationary covariance time series and the one channel at-a-time paradigms are both exploited to achieve the modeling and analysis of multivariate nonstationary covariance data. An application of the latter methodology to the identification of the epileptic focus in a human epileptic event is shown.
KeywordsStationary Time Series Stochastic Trend Partial Coherence Nonstationary Time Series General State Space
Unable to display preview. Download preview PDF.
- Akaike, H., 1971 Information theory and an extension of the maximum likelihood principle. 2nd Int. Svm. On Information Theory, Petrov, B. N., and Csaki, F., eds.. Akademiai Kiado, Budapest, pp. 267–281.Google Scholar
- Akaike, H., 1980, Likelihood and the Bayes procedure, In: Bayesian Statistics. Bernado, J. J., De Groot, M.H., Lindley, D. V., and Smith, A. F. M., eds., University Press, Valencia, Spain, PP. 143–166.Google Scholar
- Brillinger, D. R., 1975, Time Series Data Analysis and Theory, Holt, Rinehart and Winston Inc.. New York.Google Scholar
- Cover, T. M., Hart, P. G., 1967, Nearest neighbor pattern classification, IEEE Trans. lnformation Theory IT-19:827–833.Google Scholar
- Gersch, W., 1981, Nearest neighbor rule classification of stationary and nonstationary time series, In: Applied Time Series Analysis II, Findley, D. F. ed, Academic Press, New York, pp. 221–270.Google Scholar
- Gersch, W., 1987, Non-stationary multichannel time series analysis. In: EEG Handbook vol. II: Methods of Analysis of Brain Electrical and Magnetic Signals, Gevins. A. S.. and Remond, A.. Eds. Academic Press, New York, pp. 261–296.Google Scholar
- Gersch, W., 1992, Smoothness Priors. In: New Directions in Time Series Part I1, Brillinger, D., Caines, P., Geweke, J., Parzen, E., Rosenblatt. M., and Taqqu, M. S., eds.. The GL1.4 Volumes in Math, and its Applications, volume 46, Springer-Verlag, pp. 113–146.Google Scholar
- Gersch, W., and Kitagawa, G.. 1983, A multivariate time varying autoregressive modeling of nonstationary econometric time series, Proc. Bus. and Econ. Star. Section,.4m. Star. Assoc., Toronto, Canada, August 1983.Google Scholar
- Gersch, W. and Kitagawa, G.. 1988, Smoothness priors in time series, In: Bayesian Anal psis of Time Series and Dynamic Systems. Spall, J. C., ed., Marcel Dekker. New York, pp. 431–476.Google Scholar
- Gersch, W., and Stone, D.. 1990, Multichannel time varying autoregressive modeling: A circular lattice-smoothness priors realization. Proc. 29th IEEE Conk on Decision and Control, Honolulu, Hawaii.Google Scholar
- Gevins, A. S., 1980, Pattern recognition of human brain electrical potentials. IEEE Trans. Pattern Analysis and Machine Intelligence PAMI-2: 383–404.Google Scholar
- Grenier, Y., 1982, Time varying lattices and autoregressive models: Parameter estimation, Proc. ofICASSP 1982, Paris, France.Google Scholar
- Hurvich, C. M., and Tsai, C. L., 1992, A corrected Akaike information criterion for vector autoregressive model selection, Report SOR-92–10, New York University, L. N. Stern School of Business.Google Scholar
- Kitagawa, G. and Akaike, H., 1981, On TIMSAC-78, In: Applied Time Series Analysis II, Findley, D. F., ed.. Academic Press, New York, pp. 499–547.Google Scholar
- Kitagawa. G., and Gersch. W., 1985, A smoothness priors time varying AR coefficient modeling of nonstationary time series. IEEE Trans. Automatic Control AC-30: 48–56.Google Scholar
- Kitagawa, G., and Gersch, W., 1995, Smoothness Priors Analysis of Time Series,Springer-Verlag, New York, in press.Google Scholar
- Kozin, F., 1977, Estimation and modeling of nonstationary time series, Proc. Symp. Appl. Comp. Methods in Eng., Univ. Southern California, Los Angeles, pp. 603–612.Google Scholar
- Stone. D., 1993, One channel at-a-time multichannel autoregressive modeling: Applications to stationary and nonstationary covariance time series, Ph. D. Thesis in Communication and Information Sciences. University of Hawaii.Google Scholar
- Whittaker. E. T., 1923, On a new method of graduation, Proc. Edinborough Math. Assoc. 78: 81–89.Google Scholar