Nonparametric Estimation of Local Bandwidth from Level Crossings Sampling

Miroslaw Pawlak

Abstract:

We study the problem of a local bandwidth recovery for
nonstationary stochastic signals when the measured information
is given in terms of level crossings.
We propose a kernel estimate of the local bandwidth from samples generated from level
crossings of stochastic signals being the time-warped version of stationary
Gaussian processes. The positivity and bandlimited nature of the local intensity
is captured by the properly selected class of bandlimited and positive kernel functions.
The asymptotic properties of the estimator are
derived.

Submission Type: Full Paper