Toward single-trial measurement of vibrotactile driving responses in EEG: New approaches based on multi-channel adaptive filtering

Human somatosensory cortical neurons are entrained by vibratactile stimuli applied to the skin, and these population-level “driving” responses are reflected in the power spectrum of concurrently-recorded EEG. Our long term goal is to extract and measure these responses on a trial-by-trial basis, which will enable us to analyze relationships between neurophysiological and psychophysical responses to vibration. For this purpose we are exploring new approaches to measuring the neurophysiological response trial by trial, using multichannel adaptive filtering algorithms: Adaptive Line Enhancer (ALE) and Kalman filtering.; The EEG response to a vibrotactile stimulus is typically small relative to the background “noise”, so that filtering is necessary, and for this, the key is identify and exploit systematic differences between signal and noise. For example, if a signal is present on a single channel it will be correlated over a longer stretch of the time record than the noise in which it is embedded. The algorithms of Adaptive Line Enhancer and Kalman filter exploit this difference in two ways. The ALE exploits this by fitting the input signal to a copy of itself time-shifted by an amount which exceeds the correlation length of the noise. The Kalman filter estimates a process by using a form of feedback control: the filter estimates the process state at some time and then obtains feedback in the form of noisy measurements. Kalman filter is very powerful in several aspects: it supports estimations of past, present, and even future states, and it can do so even when the precise nature of the modeled system is unknown.; Multichannel ALE and multichannel Kalman are direct generalizations of single channel ALE and single channel Kalman, which exploits differences in both time and space. That is, it combines each channel being analyzed with as many as possible additional channels that contain correlated signal components but uncorrelated noise components. We evaluated this approach by embedding the filter algorithm within a control program that systematically generated combinations of its parameters and writes the results into SYSTAT filter, permitting statistical evaluation.; The research results show that Kalman filter performs uniformly superior to ALE. It has better filtering ability than Adaptive Line Enhancer. And it has better frequency tracking ability than Adaptive Line Enhancer. (Abstract shortened by UMI.)...