IoT Aided Smart Home Architecture for Anomaly Detection

Internet of Things (IoT), and Information and Communication Technology (ICT) enhance the cities and homes with intelligence using various types of devices that are embedded for sensing purposes. The IoT nodes can collect time-series information for a given purpose, like measuring temperature, air pollution and traffic congestion, motion tracking etc. and provide node behavior and environment interaction. Besides the normal ritual events that happen according the measured parameters, some unusual states, called anomalies, can be detected by using the same measurements. The anomalies that happen at homes, such as fire, elderly people fall etc. are life-critical and their early detection and/or prevention can save lives. In this chapter, we propose an Anomaly Detection System Architecture for smart homes. The fire detection/prediction and fall detection of elderly people are examined as anomalies from the usual, everyday activities. An experimental study, as proof of concept for the fire emergence case, is run by using LSTM neural network architecture. One of the positive features of the proposed methodology is that the faulty readings and false positive alarms on specific parameters are not enough to rise an anomaly detection if those readings are not supported by other parameters.