Abstract
In the conventional system identification problems, it is commonly assumed that the output signal of a dynamical system is measured at regular time intervals. This paper addresses a system identification problem under the Lebesgue sampling, which is a type of irregular sampling methods and samples the output signal only when it crosses specific thresholds. In the proposed method, not only information in output samples but also that in inter-samples are utilized for the parameter estimation to efficiently improve the estimation accuracy. The asymptotic variance of the estimated parameter is also analyzed. Effectiveness of the proposed method is examined through numerical examples. In the numerical examples, systems driven by a Gaussian white signal is identified. We illustrate that the variance of the estimates by the Lebesgue sampled data is smaller than that by the Riemann sampled data.
