Line Spectrum Trajectory Detection Method of Underwater Acoustic Signal Based on Particle filter

Abstract

Line spectrum detection is a significant task to detect, track, and classify weak target sources for passive sonar systems. In order to solve the problem of tracking time-varying frequency trajectories, a track-before-detect (TBD) algorithm based on the two-layer particle filter structure is proposed. The detect layer deals with the potential tracks by maintaining a low-threshold dynamic detection grid, which is used to determine the birth, death, and false alarm of targets. The tracking layer utilizes the particle filter to compute the joint multitarget probability density through a Bayesian framework. Simulation results show that the proposed algorithm can effectively detect and track time-varying frequency trajectories in a low signal-to-noise ratio situation, which has certain research and application value in weak signal detection.