Passive tracking of underwater acoustic sources with sparse innovations

Tracking acoustic sources via passive sonar is a challenging task common to several underwater monitoring and surveillance systems. Classical tracking approaches based on matched-field tracking and Kalman filtering techniques are impractical due to the their large computational and storage requirements. This work uses sparse-signal modeling tools to develop a computationally-affordable broadband tracking algorithm for an entire source-location map (SLM). Spectral data from multiple frequency bands are processed coherently so as to unambiguously agree on the source locations across frequencies. The tracking problem is cast so that the sparsity inherent in the SLM and in the SLM-innovations can be exploited. A numerical solver based on the proximal gradient method and the alternating directions method of multipliers is developed for SLM estimation. Numerical tests on real data illustrate their performance.