Estimating steady-state response of a resonant transducer in a reverberant underwater environment

The authors examine the estimation of steady-state amplitude and phase using short, noisy records of the transient response of systems excited by a stepped sinusoid close to a resonance. Linear prediction estimation strategies are tested, and near maximum-likelihood (ML) performance is obtained by combining FIR (finite-impulse response) cancellation of the excitation poles with the truncated singular-value decomposition approach of R. Kumaresan and D.W. Tufts (1982). For the model tested, this excitation constrained estimation strategy departs from ML performance at a threshold signal-to-noise ratio that depends on the separation between the excitation and resonant frequencies