Matrix Information Geometry for Passive Sonar Signal Detection in a Non-Stationary Environment

Abstract

In this paper, a passive sonar signal detection based on the matrix information geometry theory in a non-stationary environment is proposed. Particularly, the power spectral density (PSD) matrix is the feature which is used for the signal detection. The noise power is estimated using the Kullback-Leibler divergence mean of the noise PSD matrices. The presence and absence of a target is determined by the distance between the PSD matrix of the received signal and the mean matrix of noise. Simulation results show that the performance of our proposed method outperforms the energy detector in a non-stationary noise.