Abstract: Aimed at analyzing human activities accurately, this paper proposes a new approach combing with real-time water and electrical events. Applying infrastructure-mediated sensing technology, a system of sensing residential water and power consumption is designed to obtain the water and electrical usage data. Applying the power spectrum vector chasing algorithm and k-Nearest Neighbor algorithm, working devices in the home can be recognized and classified according to the abundant of high electromagnetic interference noises generated during switched mode power supplies devices operation. The fixtures in the home can be classified by using sliding window technique and pattern recognition algorithm based on inner product of vector. Via being deployed in real-houses, this approach successfully classifies consumption events in device-level with 95% accuracy. The analysis and post-process results can be transmitted to a clouding storage via internet. Disaggregated residential water and power consumption is the fundamental of energy monitoring and human activities analysis in a home.

Key words: Information Processing Technology, Human Activities Analysis, Water and Electrical Events Sensing, Energy Monitoring, Infrastructure-mediated Sensing