System identification under Lebesgue sampling and its asymptotic property

In system identification, the more data is collected, the more accurate model is obtained. However, under limited communication bandwidth or computational resources, it is sometimes difficult to collect and store all the measured data, so it is desired to collect and store only useful data for improving model accuracy. This paper focuses on Lebesgue sampling, which uses thresholds on signal level for triggering sampler and proposes a system identification method with using information obtained in intervals between Lebesgue sampled data. The asymptotic variance of the estimated parameter is analyzed and effectiveness of the proposed method is examined through numerical examples.