deposit_your_work

New Bayesian Model Averaging Framework for Wavelet-Based Signal Processing

Files in this item

Files Size Format View
Wan2000Jun5NewBayesia.PS 413.2Kb application/postscript View/Open

Show full item record

Item Metadata

Title: New Bayesian Model Averaging Framework for Wavelet-Based Signal Processing
Author: Wan, Yi; Nowak, Robert David
Type: Conference Paper
Keywords: Bayesian; signal modeling framework; wavelet-based signal processing; segmentation
Citation: Y. Wan and R. D. Nowak,"New Bayesian Model Averaging Framework for Wavelet-Based Signal Processing," in IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP),
Abstract: This paper develops a new signal modeling framework using Bayesian model averaging formulation and the redundant or translation-invariant wavelet transform. The aim of this framework is to provide a paradigm general enough to effectively treat fundamental problems arising in wavelet-based signal processing, segmentation, and modeling. Unlike many other attempts to mitigate the translation-dependent nature of wavelet analysis and processing, this framework is based on a well-defined statistical model averaging paradigm and improves over standard translation-invariant schemes for wavelet denoising. In addition to deriving new and more powerful signal modeling and denoising schemes, we demonstrate that certain existing methods are special suboptimal solutions of our proposed model averaging criterion. Experimental results demonstrate the promise of this framework.
Date Published: 2000-06-20

This item appears in the following Collection(s)

  • ECE Publications [1032 items]
    Publications by Rice University Electrical and Computer Engineering faculty and graduate students
  • DSP Publications [508 items]
    Publications by Rice Faculty and graduate students in digital signal processing.