Abstract It is known that the underwater acoustic channels (UAC) exhibit non-uniform cluster-sparse characteristics, meaning that the channel impulse response (CIR) typically consists of a large number of near-zero taps and with only a few none-zero ones assemble in clusters in a structured manner, moreover, the cluster structure is non-uniformly distributed on the time domain. In this paper, an improved proportionate affine projection algorithm (IPAPA) with the non-uniform l 21 -norm constraint is proposed for non-uniform cluster-sparse UAC estimation. Firstly, the auxiliary channel information is obtained via the correlation of training sequence, whereby the priori cluster-sparse structure positioning of UAC is realized. Then the non-uniform l 21 -norm is added on the IPAPA for the final accurate channel estimation: it encourages correlation among coefficients inside each group via the l 2 norm and facilitates sparsity across all groups using the l 1 norm. The results of numerical simulations and sea trial show that the proposed UAC estimation algorithm can achieve a better performance in terms of mean square error (MSE) compared to existing sparse channel estimation methods.

中文翻译：

---

基于非均匀l21范数约束的水声信道自适应估计

摘要 众所周知，水声信道 (UAC) 表现出非均匀簇稀疏特性，这意味着信道脉冲响应 (CIR) 通常由大量接近零的抽头组成，并且只有少数非零抽头。以结构化的方式组装成簇，而且簇结构在时域上分布不均匀。在本文中，提出了一种具有非均匀l 21 -范数约束的改进比例仿射投影算法（IPAPA）用于非均匀簇稀疏UAC估计。首先通过训练序列的相关性获得辅助信道信息，从而实现UAC的先验聚类-稀疏结构定位。然后在 IPAPA 上添加非均匀 l 21 -norm 用于最终准确的信道估计：它通过 l 2 范数促进每个组内系数之间的相关性，并使用 l 1 范数促进所有组的稀疏性。数值模拟和海试结果表明，与现有的稀疏信道估计方法相比，所提出的 UAC 估计算法在均方误差 (MSE) 方面可以获得更好的性能。