Anomalies Detection in Smart-Home Activities

Anomalies are the instances of an activity class that deviate from the normal or expected sequence of events in their performance. In this paper we propose an anomaly detection approach for activities performed in a smart home. In order to detect anomalies, we exploit the information of number of events and the time duration involved in performing an activity instance. The information is obtained through a network of wireless sensors deployed at multiple objects and locations within a smart home. We apply a density based clustering algorithm on the recognized activity instances to separate the normal from the anomalous. Evaluation of the proposed approach on two publicly available smart home datasets demonstrates its effectiveness in identifying the anomalous activity instances.