Remote Identification of Electrical Devices Using Dynamic Time Warping

摘要

Common household devices unintentionally inject signals onto the mains electricity supply. These signals can be measured using commercially-available equipment designed for electromagnetic compatibility testing, allowing the identification of devices from their current time series when switched on or off. Since measurements can be made outside buildings, it is possible to obtain remote intelligence of building interiors. This paper describes measurements of several devices, details a dynamic time warping algorithm to classify signals, and outlines a process for simulating large amounts of annotated data, which can be used to test end-to-end detection and classification systems.