Parsimonious Volterra System Identification
For researchers in nonlinear system identification, this work addresses the challenge of parsimonious modeling of Volterra systems with infinite memory, but the results are preliminary and limited to academic examples.
This paper develops algorithms for sparse identification of infinite impulse response Volterra systems, aiming to find the model with the fewest exponentials in its impulse response. The approach handles fragmented data and noise, with academic examples demonstrating efficacy.
In this short paper, we aim at developing algorithms for sparse Volterra system identification when the system to be identified has infinite impulse response. Assuming that the impulse response is represented as a sum of exponentials and given input-output data, the problem of interest is to find the "simplest" nonlinear Volterra model which is compatible with the a priori information and the collected data. By simplest, we mean the model whose impulse response has the least number of exponentials. The algorithms provided are able to handle both fragmented data and measurement noise. Academic examples at the end of paper show the efficacy of proposed approach.