Dynamic time warp distances as feedback for EEG feature density

摘要

This work presents a feature detection method built around a dynamic time-warping (DTW) -based confusion matrix. It can be used to discern potential features with minimal data manipulation and minimal prior knowledge. This technique provides a robust distance measurement between sample electroencephalogram (EEG) signals that form the basis of a confusion matrix indexed against events carried out as part of shared data from the PhysioNet Imagined Motion database. DTW matches signals by reconstructing the common time axis to match the amplitudes of signals as closely as possible. The resulting confusion matrices present visual patterns, or motifs, useful for distinguishing artifacts and potential features of interest in each motion trial. The results suggest this technique could be used as a tool to find areas of interest within EEG recordings and then to map them to similar occurrences.