Time-frequency analysis of event-related brain potentials

The event related potentials recorded before, during and after the voluntary movement of a finger, contain relevant information regarding the nervous mechanism involved at central level. As the phenomena of interest take place in short time, the frequency information was obtained from the signal by means of a recursive implementation of a bivariate autoregressive (AR) model. The spectral and cross-spectral parameters have been calculated in correspondence of each sample in the signal and for each recording electrode in the 10-20 system. The time course of the power in different frequency bands was obtained together with coherence and phase relationships. The algorithm allowed the evaluation of the time instant in which the events of interest take place, evidencing the beginning and the end of event related desynchronization and event related synchronization giving a global insight into the observed phenomena.