DYNAMIC SPECTRAL-ANALYSIS OF EVENT-RELATED POTENTIALS

This paper presents a new method for the identification of individual event related potential (ERP) components in both frequency and time do mains. Using the similar basis function (SBF) algorithm the method pro vides a time to frequency transform, representing a frequency domain e quivalent of the component waveform. Notable features of the SBF algor ithm are that it allows for unevenly spaced sampled functions in both the time and frequency domains, and estimates of spectral densities ar e obtained by numerical computation of finite Fourier integrals. Appli cation of this method to ERP data from 20 normal subjects demonstrated a similar shape of component amplitude frequency characteristics for traditional late component waveforms (N-1, P-2, N-2 and P-3). On this basis, a low-frequency band was found where the component amplitude fr equency characteristic was described by a Gaussian function, while the component phase frequency characteristic was a linear function of fre quency. These relationships are interpreted as frequency domain equiva lents of the component. Transformed to the time domain, they provided an analytical description of the ERP as the sum of positive-and negati ve-going monopolar waves. The study points to similar mechanisms under lying these component waveforms, and analytically defines dynamic prop erties for the components both in the frequency and time domains. (C) 1998 Elsevier Science Ireland Ltd.