Untitled Page

Spectral Power Estimation is a principal component in many practical systems. Whenever there is indecision in spectral power localization and (or) amplitude, Spectral Power Estimation theory helps to counter this indecision, and produces a reasonable estimation or an image of better quality. This indecision got it sources at the following situations:

From the previous reasons, the domain(s), in which the Spectral Power Estimation is in use, do include many practical applications. From these applications we mention: Radar Systems, Magnetic Resonance Systems (MRI), Astrophysics and Geophysics.
In my thesis, with cooperation with my thesis director, A. Seghier*, I did develop a new discreet Maximum of Entropy Multidimensional Spectral Power Estimation.

*Laboratoire de Statistique, CNRS, Université de Paris Sud, bat. 425, 91405 Orsay, France.
Mail: Abdelatif.Seghier@math.u-psud.fr.

·	Discrete signal problem: in contemporary times, discrete representation of the signal is a common practice. Of course, by the Shannon theory, there is a theoretical connection between the discrete form and the continuous form of the signal, but in many situations we have to decide the spectral power starting from a finite number of signal samples. This can be interpreted as an incomplete information situation. Spectral Power Estimation aims to compensate this lack of information by finding a “best guess” of the actual Spectral Power, for the entire signal and not only a finite part of it.
·	Low SNR: for different reasons, the signal after acquisition may contain a high degree of additive noise. In situations where we are looking to correctly estimate the Spectral Power of the signal, Spectral Power Estimation is an ideal solution to the problem.
·	Practical constraints: in many practical systems, as in SAR (Synthetic Aperture Radar) Radar Systems and MRI Systems, due to practical constraints we have to limit the number of signal sensors, which results in a 2D signal with a very limited number of signal samples in the direction of the sensors. For this, Multidimensional Spectral Power Estimation is also necessary.