Some detection and resolution properties of maximum entropy spectrum analysis

An experimental comparison between conventional spectral estimation techniques and a Maximum Entropy Spectral Analysis (MESA) algorithm is made. Three factors in the experimentation make the results of considerable interest to workers in acoustic signal processing, especially sonar and surveillance. These are the range of signal-to-noise ratio (SNR) studied, the comparisons based equal length observation intervals and the use of ensemble averaging after maximum entropy analysis. Results are presented, for both resolution and peak signal response, which tend to indicate that the Maximum Entropy Method (MEM) offers considerable promise in achieving the detection performance of long observation interval discrete Fourier transform (DFT) analysis at a much reduced length of observation time.