Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical and Computer Engineering

First Advisor

Jorge L. Aravena


This research focuses on the analysis and classification of multicomponent non-stationary signals of arbitrary duration. The proposed classification approach has potential applications in areas like moving target detection, object recognition, oil exploration, and speech processing. The wavelet transform is used as the basis for the analysis. The classification technique is based on novel scale energy density functions, called pseudo power signatures , which are independent of signal length, and which can be used to characterize the time-scale energy distribution of the signal. These signatures allow for fast classification of signals regardless of their length. Two approaches to determine pseudo power signatures are presented in this work. The first approach is based on a singular value principal component analysis technique, which, though computationally simple, is not very sensitive to signal characteristics. The second is a more sophisticated approach, and is optimal in a weighted least mean squares sense. The latter technique involves solving an inverse projection problem arising from a nonlinear infinite dimensional minimization, and generates good quality signatures with excellent discriminating capability. An algorithm, with fast convergence, for application to discrete data sets is developed, and a complete analysis of the computational complexity is obtained. Several simulation examples are presented to illustrate the methodology, and its application to practical classification problems. Finally, suggestions for further work in the area are given.