Font Size: a A A

Topics on selected nonlinear problems in signal and image processing

Posted on:2008-09-11Degree:Ph.DType:Dissertation
University:University of DelawareCandidate:Weng, BinweiFull Text:PDF
GTID:1448390005478026Subject:Engineering
Abstract/Summary:
Many signal and image processing problems involve nonlinear systems. There are different nonlinear systems used for various purposes. No unifying theory exists for all types of nonlinear systems and each nonlinear system must be analyzed differently. We mainly focus on two nonlinear signal processing methods in this dissertation.; In the first topic, Volterra systems and filters are introduced as a method for nonlinear modelling and filtering in different applications. In particular, identification and estimation of Volterra filters are addressed. We provide several non-linear system identification methods under different scenarios. The first subtopic is nonlinear system identification in impulsive environments. The second subtopic is second-order Wiener system identification under non-Gaussian excitation. The third subtopic is time-varying Volterra system identification. Besides system identification and parameter estimation, an interesting property possessed by a class of Volterra filters is proposed and analyzed. As an extension of Volterra filter to cope with impulsive noise, the polynomial weighted median filter is applied to the image sequence prediction where noise effects are considered in predictive coding.; In the second topic, the empirical mode decomposition (EMD) is introduced. We propose an optimal empirical mode decomposition scheme designed to estimate a given signal under certain optimality criterion. Following the theoretical developments, several applications of the EMD are presented. ECG denoising and baseline wander correction are firstly addressed. A new method based on the empirical mode decomposition (EMD) is proposed to remove both artifacts without introducing severe distortions. The proposed scheme can be used in real stress ECG tests. The EMD is also used in the DNA sequence analysis and clustering based on Terahertz spectroscopy data.
Keywords/Search Tags:Nonlinear, Signal, Image, Processing, EMD, Used, Empirical mode decomposition
Related items