Time-frequency spectral optimization of underwater sonar acoustic returns for the purpose of characterization using spectral descriptors

Underwater sonar acoustic returns of objects on the seafloor have many time varying components including the specular geometric backscatter, scattering from vibrational dynamics, helical, circumferential, and meridional Rayleigh waves, guided circumferential Lamb waves, and multipath from wave field reflections from both the object of interest and scene. These features of the acoustic return can reveal information about an object's shape, structure, material composition, and relationship to the background scene. Analysis of the acoustic return in the time-frequency domain is, therefore, useful for mine counter measure objectives.Optimizing the performance of various time-frequency transforms supports better understanding of the spectral behavior related to the time varying components of the acoustic return.This talk will examine the accuracy of several time-frequency representations including the short-time Fourier transform, the smoothed pseudo Wigner-Ville distribution, the empirical wavelet transform, and the chirplet transform.This work will take a detailed look at the relationships between measures of concentration, such as Renyi entropy, Stankovic concentration, frequency specificity, and resolution. Techniques are devised to improve the fidelity of the signal’s spectral content for each respective transform and to compare performance across transforms. The impact of this process is then assessed against a collection of time-frequency spectral descriptors.