Statistics Seminar

Title: DUAL-RATE SAMPLING IN WAVELET ESTIMATION

Speaker: Dr Spiro Penev
Department of Statistics
University of New South Wales

Time:4:00p.m. Wednesday 22nd September 2004

Venue:Red Centre Building Room RC-3084
near Barker Street Gate 14

We present a wavelet-based algorithm for online recovery of noisy signals,when the signal is sampled at discrete times and the estimator is transmitted after a relatively short time delay. The algorithm involves increasing the sampling rate when higher- frequency terms are incorporated into the wavelet estimator, and decreasing it when, using the size of the empirical wavelet coefficients as a guide, signal complexity is judged to have decreased.This allows to accurately recover relatively complex signals without increasing the long-run average sampling cost.

The advantages of the algorithm are seen when comparing it with a constant-rate estimator with the same long-run average sampling cost. We
illustrate these advantages and the robustness of the algorithm numerically and formulate statements related to the ability of the dual-rate sampler to consistently estimate high-frequency parts of the signal that would not be
assessible when using a constant-rate sampler with the same average sampling
cost.

This is joint work with Professor Peter Hall (Australian National University,
Canberra).

Y. Fan E-mail yanan@maths.unsw.edu.au
You are invited to join us for an afternoon tea in the staff common, Red Centre Building RC-3082 at 3:45pm.