Optimized wavelets for classification : application to hyperspectralimage and ECG signal

Abstract

The contribution of the present work is twofold. First, Wavelets are known to be a valuable tool for analyzing hyperspectral images. In this thesis, we propose to improve further their performance by means of a novel classification-driven design scheme which aims at deriving a wavelet that best represents in terms of between-class discrimination capability the spectral signatures conveyed by a given hyperspectral image. This is achieved by adopting a polyphase representation of the wavelet filter bank and formulating the wavelet optimization problem within a particle swarm optimization (PSO) framework. Experimental results show that the proposed wavelet design method outperforms the popular Daubechies wavelets whatever the classifier type adopted in the classification process. Second, Wavelets have proved particularly effective for classifying ECG signals. In this thesis, we show that wavelet performances in terms of classification accuracy can be pushed further by customizing them for the classification task. A novel approach for designing ECG signal classification-driven wavelets is proposed. It makes use of the polyphase representation of the wavelet filter bank and formulates the design problem within a particle swarm optimization (PSO) framework. Experimental results conducted on the benchmark MIT/BIH arrhythmia database with the state-of-the-art support vector machine (SVM) classifier confirm the superiority in terms of classification accuracy and stability of the proposed method over standard wavelets (i.e., Daubechies and Symlet wavelets)