Sparse time-delay and amplitude estimation of multipath arrivals in ocean acoustics

We estimate multipath arrival time delays and amplitudes with high resolution for accurate characterization of ocean acoustic channels. Although matched filtering (MF) is robust and widely used, its output represents a superposition of contributions from multiple arrivals, which can bias delay and amplitude estimates. More advanced high-resolution (HR) methods exist but are usually computationally expensive. To mitigate those limitations, we propose a sparse parametric framework that models the received signal as a superposition of a sparse number of time-delayed replicas of a known source waveform. Time delays are treated as continuous-valued parameters, and a variable projection approach eliminates the amplitude parameters in closed form, reducing the estimation to a low-dimensional optimization over delays only. The formulation is implemented in the time domain, enabling direct waveform-level comparison between reconstructed and measured signals. Simulations demonstrate that the proposed method mitigates the delay and amplitude biases observed in MF and achieves improved resolution of closely spaced arrivals compared to HR methods, at lower computational cost. Experimental results using airgun data from the New England Seamount Acoustics (NESMA) experiment validate its recovery of physically interpretable channel impulse responses and provide consistent arrival estimates across multiple measurements, under practical ocean acoustic conditions with closely spaced arrivals.