New Approaches to Non-Stationary Signal Analysis and Applications.

Abstract : We addressed the problem of resolving signals that are closely spaced in frequency using short data segments. Such problems are of interest in electronic warfare and for locating the sources of electromagnetic radiation such as radar and communication devices. Effective, computationally efficient methods of high resolution parameter estimation were developed. The performance of the algorithms were predicted by theoretical results and verified by computer simulations. We also addressed the problem of tracking signals whose spectrum was varying with time. Signals were assumed to consist of multiple time varying signal components and improved methods of residual signal analysis were developed to isolate and track the individual components. The methods were applied to real world signals such as voiced speech to decompose them into their harmonic components. During this period two PhD theses were completed. The research results were published in the form of ten conference publications and six journal publications.