Online detection and parameter estimation with correlated data in wireless sensor networks
Sima Sobhiyeh, Mort Naraghi‐Pour · 2018
We present an online algorithm for hypothesis testing from correlated observations obtained from a network of heterogeneous sensors and in the presence of model uncertainty. The correlated observations are modeled using copula theory. The batch-mode expectation maximization (EM) algorithm is first developed and then extended to an online algorithm for model parameter estimation and hypothesis testing. Using real-world as well as simulation data, we compare the detection accuracy of our method with other supervised and unsupervised methods and also with a model which ignores the correlation in the data.