论文部分内容阅读
为了解决观观测噪声和信道噪声概率分布不完全已知时的多传感器分布式量化估计融合问题,提出了一种期望极大化算法(EM算法)的分布式量化估计融合方法。该方法将未知的噪声参数以及局部量化器量化概率建模为EM算法中二元高斯混合模型参数,利用极大似然估计方法的估计不变性得到目标参数的估计融合结果。仿真实验结果表明:该方法在局部传感器观测样本数目大于6000和信噪比大于6 dB时与已有理想信道条件下的估计方法性能相当。本文方法对水下分布式协同探测问题提供了一种简化的估计融合实现途径。
In order to solve the problem of fusion of quantized multisensor quantization estimation when observation noise and channel noise probability distribution are not completely known, a distributed quantization estimation fusion method based on expectation maximization algorithm (EM algorithm) is proposed. In this method, the unknown noise parameters and the quantification probability of the local quantizer are modeled as the parameters of the binary Gaussian mixture model in the EM algorithm. The estimated fusion results of the target parameters are obtained by using the estimation invariance of the maximum likelihood estimation method. Simulation results show that this method is equivalent to the existing estimation method under ideal channel conditions when the number of local sensor observations is more than 6000 and the SNR is more than 6 dB. The proposed method provides a simplified approach to the estimation fusion of underwater distributed cooperative detection.