Instantaneous coherence as a sensible parameter for considering human information processing

It has be shown by several authors that the EEG coherence reflects mental activities. Generally, the estimation of the coherence is based on the Fourier transformation. Due to its low time resolution this estimation method is unsufficient for the examination of fast thinking processes. In the paper a new method for fitting ARMA models with time-varying parameters is introduced which combines LMS-filter techniques and a general adaptive estimation procedure for parameters of non-stationary processes. The method is realized for two-dimensional processes and enables the parametric calculation of one- and two-dimensional momentary spectral parameters, especially of the instantaneous coherence. Using this adaptive approach the sensibility of the instantaneous EEG coherence in respect to mental activities is strongly shown for elementary thinking processes