Underwater DOA Estimator Based on Variational Bayesian Inference with Non-uniform Noise

作者：Yongfeng Huang, Zhendong Chen, Dingzhao Li, Haixin Sun · 年份：2024 · DOI：10.1109/icdsca63855.2024.10859977 · 被引用次数：1 · 研究领域：Blind Source Separation Techniques、Underwater Acoustics Research、Speech and Audio Processing

In complex underwater communication environments, Gaussian white noise models often fail to capture the true characteristics of the ocean environment accurately. Especially for low-frequency signals, the relative increase in array aperture leads to differences in the noise variance received between array elements, resulting in a decrease in the precision of the direction of arrival (DOA) estimator. Therefore, we innovatively propose a non-uniform noise Bayesian framework to address this challenge. Within this framework, the received signal is regarded as a synthesis of the expected signal and non-uniform noise. Subsequently, we employ the variational Bayesian inference (VBI) to learn the hyperparameters and use the first-order Taylor linear expansion to address the modeling error. The results of simulation experiments confirm that the proposed method demonstrates excellent performance compared to the existing algorithm.