Quantitative Search for Stimulus-Specific Patterns in the Human Electroencephalogram (EEG) During a Somatosensory Task

Abstract

How does the brain recognize and classify sensory input? We hypothesize that the required perceptual information is contained in the spatial pattern of amplitudes of a common carrier waveform in the gamma range (20–60 Hz) of the EEG. To test this hypothesis, human subjects fitted with a 64-electrode array over the somatosensory cortex are asked to discriminate among 3 different intensities of electrical stimuli by responding with 3 different pressures on a force transducer button. EEG segments 200 msec long from each of the 3 trial types are submitted to a computer for classification by stimulus type. The analysis uses either the Fast Fourier Transform (FFT) or Principal Components Analysis (PCA), followed by a Euclidean distance classification procedure; various filters and procedures for reducing outliers are also being tested. Preliminary results using a small, exploratory data set are described.