Improve long-range source localization in the South China Sea by suppressing frequency-difference autoproduct cross-term

Abstract

Match field processing (MFP), combining with underwater acoustics and physics, is a signal processing technology and is popular in passive source localization. Unfortunately, in many situations, especially when long-range sources are involved, incomplete understanding of the actual propagation environment hinders accurate propagation modeling, leading to the failure of source localization via MFP. Recently, low-frequency MFP using the frequency-difference autoproduct achieved some long-range source localization success fully. While this method has been proved more robust than conventional methods, many of the metrics, such as Peak-to-Background Ratio (PBR) and ambiguity surface peak values, are lower than commonly observed levels. This performance degradation is related to the cross-term of frequency-difference autoproduct. In this paper, we combine low rank matrix representation (LRR) to suppress the influence of cross-term. This method is used to improve source localization metrics in the South China Sea (deep ocean) environment. The experimental results show that the PBR and ambiguity surface peak values are improved by about 2.5dB and 4.5dB respectively.