In this paper, new classes of lower bounds on the outage error probability and on the mean-square-error (MSE) in Bayesian parameter estimation are proposed. The minima of the h-outage error probability and the MSE are obtained by the generalized maximum a-posteriori probability and the minimum MSE (MMSE) estimators, respectively. However, computation of these estimators and their corresponding performance is usually not tractable and thus, lower bounds on these terms can be very useful for performance analysis. The proposed class of lower bounds on the outage error probability is derived using Holder's inequality. This class is utilized to derive a new class of Bayesian MSE bounds. It is shown that for unimodal symmetric conditional probability density functions (pdf) the tightest probability of outage error lower bound in the proposed class attains the minimum probability of outage error and the tightest MSE bound coincides with the MMSE performance. In addition, it is shown that the proposed MSE bounds are always tighter than the Ziv-Zakai lower bound (ZZLB). The proposed bounds are compared with other existing performance lower bounds via some examples.
On Bayesian Fisher Information Maximization for Distributed Vector Estimation
本文研究了线性观测模型下高斯向量的分布估计问题。每个传感器对未知向量进行标量噪声观测,量化其观测,将其映射到数字调制符号,并通过正交功率约束衰落信道将符号发送到融合中心(FC)。FC的任务是融合来自传感器的接收信号并估计未知向量。我们推导出三种接收机的贝叶斯Fisher信息矩阵(Bayes Fisher Information Matrix,简称Bayes):(i)相干接收机(ii)信道包络已知的非相干接收机(iii)信道统计仅已知的非相干接收机。我们还推导了Weiss-Weinstein界(WWB)。以传感器发射功率为优化变量,提出了两个约束优化问题,即网络发射功率约束下的最大化贝叶斯网络迹和贝叶斯网络对数行列式。我们表明,相干接收机,这些问题是凹的。然而,对于非相干接收器,它们不一定是凹的。贝叶斯最大化问题的迹的解决方案可以以分布式的方式实现。我们数值研究如何在传感器之间的FIM-max功率分配取决于传感器的观测质量和物理层参数,以及网络的发射功率约束。此外,我们评估的系统性能方面的MSE使用FIM-max计划的解决方案,并将其与最小化的LMMSE估计(MSE-最小计划)的MSE,和均匀的功率分配的解决方案。这些比较表明,虽然WWB是严格的比逆贝叶斯网络,它仍然是合适的使用FIM-max计划,因为在MSE的LMMSE估计的性能损失是不显着的。