Underwater Delay Estimation Based on Adaptive Singular Value Decomposition Reconstruction Under Low SNR and Multipath Conditions

Abstract

Delay estimation aims to determine the distance between the signal source and the receiver by measuring the signal's arriving time, which is crucial for underwater positioning. Traditional delay estimation algorithms, such as Generalized Cross-Correlation (GCC), often perform poorly in low signal-to-noise ratio (SNR) or multipath channels. In response to this issue, this paper proposes an algorithm based on adaptive Singular Value Decomposition Reconstruction (SVDR). This method initially requires obtaining the cross-power spectrum between the transmitted and received signals. Subsequently, the inter-correlation results at different frequency bands are assembled into a Frequency-Sliding Generalized Cross-Correlation (FSGCC) matrix. Then, Singular Value Decomposition Reconstruction (SVDR) is applied to extract crucial delay information from the matrix, aiming to alleviate the impact of noise and multipath effects on delay estimation. However, the selection of singular values during the reconstruction process directly influences the degree of noise reduction in the signal. Therefore, this manuscript further calculates the matrix represented by each singular value obtained from the SVD operation. The similarity between each matrix and the low-noise FSGCC matrix is computed to select the most suitable singular values to retain. Through simulation experiments, this algorithm can overcome the influence of the multipath effects and achieve better delay estimation results compared to traditional GCC and SVD algorithms, and validates its effectiveness in low SNR multipath underwater acoustic channels.