Autocorrelation and Crosscorrelation Analysis in Electroencephalography

Abstract

Autocorrelation and crosscorrelation analysis, which have been used extensively in statistical communication theory in the past few years, can be applied, with certain limitations, to the study of the EEG (electroencephalograph). Autocorrelograms for normal subjects can be classified in several categories, according to the dominant frequency, or frequencies, present, and other parameters. Crosscorrelograms of EEG recordings from different locations on the head permit a comparison of the electrical activity at the two locations. Correlation functions and power-density spectra contain equivalent information because the one may be obtained from the other by Fourier transformation; but, because of the squaring and multiplication that appear in the computation process, the data so obtained are not exact equivalents of the frquency spectra derived from tuned resonators. A special case of crosscorrelation analysis (crosscorrelation of a repetitive signal with a synchronously occurring brief pulse) can be applied to the detection of electric responses evoked by sensory stimulation. This process is equivalent to averaging a large number of individual responses. Illustrative examples, obtained from semi-automatic computers especially designed for the purpose, are given.