A posteriori digital filtering to reduce signal-averaging of steady-state evoked potentials

This paper describes a signal extraction technique that involves taking the time series representation of signals, transforming them into the frequency domain, determining the chance occurrence of power at each frequency, and filtering accordingly. An inverse Fourier transform is then used to recreate the new time domain representation, which has been appropriately filtered for extraneous noise. Thus, this technique involves a posteriori digital filtering based on a statistical criterion for component inclusion. Computer simulation indicated that, at poor signal-to-noise ratios and with fewer samples, this technique is 5 to 10 times better at signal extraction than conventional signal averaging. Examples of pattern-elicited electroretinograms (PERG's) are used to illustrate the efficacy of this method.