Statistical Property for Processing Gain of Lofargram Smoothing using Morphological Filter.

Abstract

The methods for analyzing the temporal structures of passive sonar signals employ spectral analyses of the received signals. One of these methods is called the lofargram (low-frequency analysis record-gram). Fine-quality lofargrams are generally obtained by using averaging procedure, which constitutes a data smoothing method. However, the procedure causes the impulse noise to spread across the lofargram's time-axis. This makes it difficult to distinguish a target signal from a received signal on the lofargram. Morphological filtering is a nonlinear signal transformation that locally modifies the geometric features of signals. Morphological opening filter can be used effectively to suppress positive impulse noise in signals and images. The authors have proposed a method for smoothing a lofargram using morphological opening filter instead of averaging procedure. The effective improvements in processing gain and time resolution obtained in the experiment are compared with the results obtained using averaging procedure. This paper describes some statistical properties of the processing gain of a lofargram obtained using morphological opening filter. The processing gain is defined in the case of the line structuring element, and its theoretical equation is derived when the source signal is Gaussian noise. The results of the computer simulation agree approximately with the theoretical value obtained by this equation.