SAMMPLE: Detecting Semantic Indoor Activities in Practical Settings using Locomotive Signatures

We analyze mobile phone-generated sensor feeds to de- termine the high-level (i.e., at the semantic level), indoor, lifestyle activities of individuals, such as cooking/dining at home and working/having-lunch at the workplace. We pro- pose and evaluate a 2-Tier activity extraction framework (called SAMMPLE) where features of the accelerometer data are first used to identify individual locomotive micro- activities (e.g., sitting or standing), and the micro-activity sequence is subsequently used to identify the discrimina- tory characteristics of individual semantic activities. Using 152 days of real-life behavioral traces from users to detect semantic activities, our approach achieves an average ac- curacy of 77.14%, an improvement of 16.37% from the traditional 1-Tier approach that directly uses statistical features of the accelerometer stream.