MEM (Maximum Entropy Method) Resolution of Line Spectra

摘要

This report summarizes a study of the use of the Maximum Entropy Method (MEM) in processing synthetic and laboratory obtained interferograms. The capabilities and limitations of the Burg and Yule-Walker MEM algorithms were evaluated and compared to the conventional Fast Fourier Transform (FFT) technique of recovering spectrum from data containing discrete spectral lines. The results showed that the Burg MEM algorithm provided better spectral resolution than the Y-W MEM algorithm. In the absence of unresolved multiple lines, the recovered spectral line width obtained with the Burg MEM was dependent on the line signal-to-noise ratio, the separation of the neighboring lines, and the number of coefficients used. The line shape was consistently a Lorentzian function. The MEM spectral resolution was approximately a factor of two better than the conventional FFT when both correct line strength and line resolution were considered. The time required to compute the coefficient for the MEM algorithm may be prohibitive where near real time recovery of the spectrum is required. The importance of data measurement time versus spectral recovery computational time must be considered for each application of the MEM technique in spectral recovery.