System identification of the vestibulo-ocular reflex: application of the recursive least-squares algorithm

A discrete-time model of the vestibulo-ocular reflex (VOR) can be obtained by step-invariant transformation of the two-pole continuous-time model frequency used to describe the system. Simulations show that for sum-of-sines input, output predictions of the two VOR models are in agreement except during an initial period. The discrete model allows for parameter estimation by the recursive-least-squares (RLS) algorithm. RLS data yields values that are easily transformed back into continuous model parameters. RLS identification of actual data from electro-oculographic recordings is complicated by noise induced by data recording and nystagmus algorithms. Valid VOR models can be obtained by RLS without the large amounts of data required by frequency response methods.<>