.png){kind=icon}
.png){kind=icon}
Return
Underwater array DOA estimation method via signal-enhanced spatiotemporal convolution fusion
.png){kind=icon}
PRE
AI
A
N
J
W
L
DOI：10.1016/j.sigpro.2026.110610.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Underwater acoustic direction-of-arrival (DOA) estimation is severely degraded under low signal-to-noise ratio (SNR) and channel distortions. Conventional subspace methods like MUSIC and ESPRIT are highly noise-sensitive, while existing deep learning approaches often lack robustness to nonstationary underwater signals in low-SNR conditions. To address these limitations, this paper proposes a Signal-Enhanced Spatiotemporal Convolution Fusion (STCF) network. The method first employs the Tsallis-Optimized Spectral Enhancement (TOSE) framework with adaptive frequency scaling to suppress nonstationary noise. The covariance matrix of the enhanced signal is then processed by a dual-branch network: an attention-enhanced GraphSAGE module captures inter-sensor spatial dependencies, and a Convolution and Attention Fusion Module (CAFM) extracts multi-scale temporal features. Their fused spatiotemporal representation significantly improves estimation accuracy and robustness. Experimental results show that the STCF method consistently outperforms existing algorithms across various SNRs, achieving high estimation accuracy with reasonable computational latency, and thus offering an effective solution for high-precision underwater DOA estimation.
Keywords:
DOA Estimation
Graph sample and aggregate (graphsage) neural network
Underwater acoustics
Linear array
Wavelet packet transform
Journal
IF:
3.6
Papers:
9.8K
Citations:
1.7W
Organization
Cited Papers