计算机科学
噪音(视频)
频道(广播)
水下
水声通信
贝叶斯概率
水声学
噪声测量
语音识别
声学
人工智能
算法
信噪比(成像)
稳健性(进化)
最大似然
估计理论
背景噪声
稀疏逼近
估计
贝叶斯推理
高斯噪声
作者
Zheming Zhang,Xiao Han,Weizhe Li,Li Wei,Jingwei Yin
标识
DOI:10.1109/lwc.2026.3693448
摘要
Impulsive noise and channel time variation are two major factors that severely degrade the performance of underwater acoustic communication (UAC) systems, especially in short sub-block processing with limited observations. This letter proposes a variational Bayesian inference (VBI)-based time-correlated sparse Bayesian learning (SBL) framework for reliable channel estimation and tracking in short-block single-carrier UAC systems. Within the variational Bayesian framework, a first-order autoregressive prior is introduced to exploit temporal correlation across adjacent data blocks. Furthermore, an identification-and-rejection mechanism is developed, which leverages a Beta–Bernoulli formulation to adaptively exclude severely corrupted observations during likelihood evaluation. The joint action of these mechanisms mitigates estimator degradation and inter-block error propagation caused by unreliable observations under short-block conditions. Simulation results and under-ice experimental data collected in Songhua Lake demonstrate that the proposed method achieves significantly superior performance over existing approaches in impulsive noise environments.
科研通智能强力驱动
Strongly Powered by AbleSci AI