The problem of estimating bearings of impulsive wideband acoustic signals produced by vocalizing animals and received by a compact array of synchronized sensors is addressed. The accuracy provided by the maximum-likelihood (ML), the beamformer (BF), and the time-difference-of-arrival (TDOA) based estimators is evaluated by simulations and in situ tests and compared with the Cramér–Rao bounds. Test results demonstrated that the ML estimator and BF provided similar bearing estimation accuracy for all types of signals. They are more accurate than the TDOA-based estimator for mid- and low-frequency impulsive signals. The accuracy of the TDOA-based estimates comes close to the ML- and BF-based estimates when the signal bandwidth increases. TDOA-based estimators outperformed the BFs and the ML algorithms when estimating the bearings of clicks. The empirical standard deviations provided by the ML/BF and TDOA-based estimators were $\text{0.2}^\circ \ldots \text{3}^\circ $ and $\text{2}^\circ \ldots \text{70}^\circ $ for mid- and low-frequency impulsive signals and $\text{5.0}^\circ $ and $\text{0.6}^\circ $ for clicks, respectively.

中文翻译：

---

海洋动物产生的宽带信号承载估计的准确性

解决了估计由发声动物产生并由紧凑型同步传感器阵列接收的脉冲宽带声信号的方位的问题。最大似然 (ML)、波束成形器 (BF) 和基于到达时间差 (TDOA) 的估计器提供的精度通过模拟和原位测试并与 Cramér-Rao 边界进行比较。测试结果表明，ML 估计器和 BF 为所有类型的信号提供了相似的方位估计精度。对于中低频脉冲信号，它们比基于 TDOA 的估计器更准确。当信号带宽增加时，基于 TDOA 的估计精度接近基于 ML 和 BF 的估计。在估计点击方位时，基于 TDOA 的估计器优于 BFs 和 ML 算法。ML/BF 和基于 TDOA 的估计器提供的经验标准差为$\text{0.2}^\circ \ldots \text{3}^\circ $ 和 $\text{2}^\circ \ldots \text{70}^\circ $ 用于中低频脉冲信号和 $\text{5.0}^\circ $ 和 $\text{0.6}^\circ $ 分别为点击次数。