Anomaly Detection Based on Fixed and Wearable Sensors in Assisted Living Environments

The increasing number of physical devices connected to Internet enables accelerated development of advanced Internet of Things (IoT) applications able to improve our quality of life. One of such applications is Ambient Assisted Living (AAL). Its main goal is to determine the wellness of older people, people with disabilities, or people with acute or chronic pathologies living independently in their homes in order to initiate emergency response actions if necessary. Wellness of people can be inferred by monitoring activities of daily living (ADLs) and detecting possible anomalies. This paper proposes a novel method based on artificial neural networks (ANNs) for ADL identification using fixed and wearable sensor readings. Fixed sensors' readings are suitable for user localization in AAL, while wearable sensors' readings may estimate user body position. Besides identifying ADLs, the synergy of readings from these different sensor types proved to be valuable for anomaly detection. For this purpose, identified activities based on fixed sensor readings are matched with wearable sensors' readings in order to determine anomalous situation that might indicate medical or similar emergency.