Mallet, Yvette Lelia (1997) Wavelet based feature extraction methods for the discrimination and regression of spectral data. PhD thesis, James Cook University.

	PDF (Thesis front) Download (624kB)
	PDF (Thesis whole) Download (6MB)

Abstract

This thesis is concerned with the application of statistical methods to spectral data. A major concern which arises from spectral data is that the number of variables or dimensionality usually exceeds the number of available spectra. This leads to a degradation in performance of traditional statistical methods. There are basically two strategies which can be implemented for overcoming such situations. It is common practice to first reduce the dimensionality of the data by some feature extraction preprocessing method, and then use an appropriate low dimensional statistical procedure. An alternative procedure is to use a high dimensional statistical procedure which is capable of handling a large number of variables. This thesis considers both approaches, and investigates the applicability of wavelets as features for statistical analyses, as well as other feature extraction procedures. The particular statistical analyses investigated are discriminant and regression analysis. It is shown that, the wavelet based methods, particularly wavelets which have been designed to suit a particular task, perform quite adequately when compared to traditional approaches.

Item ID:	17437
Item Type:	Thesis (PhD)
Keywords:	statistical methods, statistical analysis,discriminant analysis, regression analysis, spectral data, variables, wavelets, feature extraction, dimensionality
Date Deposited:	23 Jun 2011 22:36
FoR Codes:	01 MATHEMATICAL SCIENCES > 0104 Statistics > 010401 Applied Statistics @ 100%
SEO Codes:	97 EXPANDING KNOWLEDGE > 970101 Expanding Knowledge in the Mathematical Sciences @ 100%
Downloads:	Total: 651 Last 12 Months: 12
	More Statistics