Kernel-based identification using Lebesgue-sampled data (Q6550245)

Lebesgue sampling records the locations of the signals' amplitude bands in at each time moment. It is promising DSP method relevant to quantization. The main contribution of this paper is the estimation of non-parametric continuous-time models from Lebesgue-sampled output data. Based on the possibly noisy and sparce/irregular data records the kernel-based estimator for continuous-time, linear and time invariant Lebesgue-sampled systems is proposed. This non-parametric estimator suitable for the bounded intersample signals behavior provides more accurate results comparing to traditional (Riemann) sampling, but use much fewer samples. The method efficiency is illustrated using a mass-spring-damper system and the continuous-time system identification benchmarks.