Symbolic time-series analysis of neural data

The symbolic representation of experimental data offers a potentially power ful tool for studying dynamic behavior and for model fitting. We show how e xperimental time-series data from lamprey locomotion can be mapped onto sym bols. Spike time-series from bursting oscillators in the spinal cord are vi sualized as Poincare sections that slice through the initiation of the fast (spike) oscillations represented in a higher-dimensional phase space. This Poincare map is systematically converted to a string of symbols that captu re the dynamics in the plane of slicing. The patterns in these symbol strin gs can be used to search for repeating behaviors, or, potentially, to fit m odels. (C) 2000 Elsevier Science B.V. All rights reserved.