Efficient matched processing for localisation of a moving acoustic source

Abstract

In recent years matched processing techniques have emerged as a solution for the problem of target motion analysis. A target's motion parameters are estimated by determining which of a bank of replica signals best correlates with the signal received from that target. Matched processing methods have been used in low SNR environments such as the underwater acoustic environment because they outperform pseudomeasurement-based target motion analysis. However, matched processing is notoriously computationally demanding, requiring evaluation of complex equations for replica generation, and batch processing via an exhaustive search of a multidimensional parameter space. This study presents a matched processing method for localisation of a slow-moving low-frequency acoustic source by matching Fourier coefficients from a single frequency bin. A simple range-dependent propagation model is assumed, and the Doppler effect is carefully modelled to produce realistic replicas. An efficient matching algorithm is presented which computes cost values in a recursive fashion, discarding poorly scoring replicas before they are fully evaluated. Simulated examples demonstrate the ability of the technique to resolve the locus of low-Doppler targets at extremely low SNR with a fraction of the cost associated with exhaustive search batch processing.