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 domains. Using the similar basis function (SBF) algorithm the method provides a time to frequency transform, representing a frequency domain equivalent of the component waveform. Notable features of the SBF algorithm are that it allows for unevenly spaced sampled functions in both the time and frequency domains, and estimates of spectral densities are obtained by numerical computation of finite Fourier integrals. Application of this method to ERP data from 20 normal subjects demonstrated a similar shape of component amplitude frequency characteristics for traditional late component waveforms (N1, P2, N2 and P3). On this basis, a low-frequency band was found where the component amplitude frequency characteristic was described by a Gaussian function, while the component phase frequency characteristic was a linear function of frequency. These relationships are interpreted as frequency domain equivalents of the component. Transformed to the time domain, they provided an analytical description of the ERP as the sum of positive- and negative-going monopolar waves. The study points to similar mechanisms underlying these component waveforms, and analytically defines dynamic properties for the components both in the frequency and time domains.