Wavelet network with convex wavelets: applications to nonlinear system modeling and feature extraction in vectorcardiography

This chapter describes several applications of wavelet networks in biomedic al signal processing. A specific wavelet network with convex wavelets is in troduced and its distinctive features are presented. The main characteristi cs of the network are good designability (the number of network connection weights can be determined by the designer) and improved generalization capa bility. Three applications of the wavelet networks with convex wavelets are present ed, namely, modeling of the nonlinear dynamical systems, feature extraction from vectorcardiograms, and data compression. We demonstrate how the extra cted features can be used for visualization for medical diagnoses. The netw ork is then used in data compression for electrocardiograms (ECGs). We intr oduce an improved (faster) learning algorithm for data compression, based o n wavelets. Conclusions regarding the use of wavelet networks with convex w avelets in cardiography and in other medical applications are presented in the final section of the chapter. For the basic theory of wavelet transform and for other applications in the field of medicine, the reader is referred to the textbooks in the Referenc e list and to the titles in the Further rending section at the end of the c hapter.