Semantic anomaly detection in daily activities

We monitor activities of daily living of smart home residents to detect anomalies in their behavior. Unlike traditional anomaly detection systems, we aim to reduce false positives in anomaly detection with the help of semantic rules. Some of these rules are predefined based on expert knowledge and the rest are learned by the system with the help of resident/expert feedback. We also correlate trend of change in different activities to improve anomaly detection. In addition to monitor statistical deviation from regular behavior, we also detect deviation from healthy and social norms (defined by experts) as anomalies.