Estimation of a coupling feature for target classification using passive sonar

This paper presents a novel feature and its estimation method, for the classification of marine vessels using passive sonar. Conventionally, the classification of such targets is based on spectral estimation methods that identify the machinery and propeller spectral lines and other spectral features. We propose a classification feature, namely the coefficient of quadratic nonlinearity (CQNL), which quantifies the quadratic coupling that exists between a pair of running machinery on the target platform. The classification is based on the premise that the degree of nonlinear coupling between the various machineries on a target platform is unique for a particular class of targets and depends on the types of machinery and their placements. The estimation of the CQNL feature uses higher-order spectral analyses in conjunction with a method analogous to matched field processing. The CQNL feature provides unique information about the target platform that is not present in spectrum based features. The performance of the algorithm has been studied numerically using an extensive simulated target signal database at the receiver. It is observed that the algorithm performs well even in low SNR conditions. The CQNL feature has also been estimated from recorded ship-radiated signals.