A survey on the maximum entropy method and parameter spectral estimation

This paper presents a survey on the maximum entropy method and parameter spectral estimation. High-quality spectral estimation of a random signal can be defined as the process of abstracting as much frequency information as possible from measurement data. Information theory provides a theoretical foundation. In order to make use of some a prior knowledge, on needs to choose a proper model in either the time domain or the frequency domain. Some real world disturbances, such as noise interference and signal fluctuation, greatly affect the resulting estimates. In this paper, the history of the development of many algorithms and methods in maximum entropy spectral estimation is reviewed. Most of the existing spectral estimation methods are systematized in terms of signal models, with emphasis on the eigenanalysis method. Also, some considerations for software and hardware performance, such as the information bank and the lattice filter, are discussed.