Anomaly detection based on probability density function with Kullback-Leibler divergence

Anomaly detection is a popular problem in many fields. We investigate an anomaly detection method based on probability density function (PDF) of different status. The constructed PDF only require few training data based on Kullback–Leibler Divergence method and small signal assumption. The measurement matrix was deduced according to principal component analysis (PCA). And the statistical detection indicator was set up under iid Gaussian Noise background. The performance of the proposed anomaly detection method was tested with through wall human detection experiments. The results showed that the proposed method could detection human being for brick wall and gypsum wall, but had unremarkable results for concrete wall. & 2016 Elsevier B.V. All rights reserved.