Robust Estimation of Modulation Frequency in Impulsive Acoustic Data

Estimation of the modulation frequency of a source is important in various sonar signal processing applications. The conventional estimator of modulating frequency has been formulated under the assumption that the source and noise are Gaussian distributed. This is often inaccurate in several applications such as underwater acoustics where the signal or noise can be impulsive. We formulate novel modulation frequency estimators that are robust to impulses in the data, and outperform the conventional estimator in environments where the observed data are contaminated by impulses. We demonstrate their performance using recorded data. We characterize the performance of the methods in terms of their accuracy, harmonic distortion, and robustness to knowledge of noise statistics. We also derive the Cramer-Rao lower bound for the modulation frequency estimation problem in impulsive data.