Accuracy enhancement of underwater target detection with time-frequency analysis techniques

Abstract

Underwater target detection is mainly based on acoustic emissions generated by the target of interest (TOI) as it propels itself through the ocean, and operates non-propulsion-related onboard systems. The spectral signature of these acoustic emissions is used for identification and localization of TOI. This paper focuses on underwater target detection using passive sonobuoys. Discrete Fourier transform (DFT) is used in most sonobuoy processing systems to provide spectral analysis of the received signals. Due to the relatively high noise level and the several sources of interference that may exist underwater, the DFT may not determine the spectral signature of TOI with adequate accuracy. The low signal-to-noise ratio (SNR) and the presence of strong interference sources with frequencies close to the TOI frequency jeopardize the detection accuracy, the bearing estimation performance, and target tracking capabilities. The aim of this paper is to: (1) examine the performance of both wavelet packet analysis and high resolution spectral estimation techniques, and (2) provide a comparative study between both methods. Based on underwater acoustic simulation developed in this research, results showed that the proposed methods can achieve robust target detection with low levels of SNR and interferences of nearby signatures that cannot be detected by DFT.