Soft-decision detection of weak tonals for passive sonar using track-before-detect method

Abstract

Tonal detection represents a significant task to detect and classify target sources for passive sonar in marine environments. However, in low signal-to-noise ratio (SNR) situations, tonals are difficult to detect and are readily immersed in the time-varying marine noise. The birth and death of tonals are also difficult to resolve due to bad detection performances. This paper utilizes the particle filter track-before-detect (PF-TBD) method to detect weak tonals with soft-decision results. The statistical properties of the tonal and noise are derived to build dynamic model, measurement model and likelihood function for tonal detection with TBD. The maximum likelihood estimation (MLE) method is used to estimate the noise variance which varies with time. A soft-decision detection result between 0 and 1 is estimated over each frequency bin to obtain an intuitive time–frequency spectrogram. Both the computer simulations and real sea trial data analyses (detect tonals from a deterministic source and a surface vessel) results validate the feasibility of the proposed method.