Anomaly event detection in temporal sensor network data of intelligent environments

Abstract

Intelligent environments enhanced the interactions between human and computers. People can seamlessly communicate with the system via some event, such as gesture, voice, motion and context. Anomaly event detection in the temporal data, which collected in sensor network of intelligent environments, is a challenging problem, particularly there have no direct priori knowledge of the anomaly events and no prominent patterns are known. In this paper, we propose a technique which can extract patterns in the temporal sensor data and identify the anomaly events efficiently. This method is based on the covariance information of temporal data, and T2 test of Mahalanobis distance is used to detect the outliers. The experiment results show that the propose method can detect the anomaly and uncommon events in temporal data. It can be of great use in intelligent environments.