Knowing what devices or appliances a person operates is important in many ubiquitous computing applications, such as personalized energy disaggregation, activity tracking, and adaptive user interfaces. For example, a cookbook application can preload and progress automatically when a user begins cooking. In a nursing home, tracking an individual elder’s pattern throughout the day can help caregivers monitor each person’s activity.
A potential way to track someone’s activity throughout the day is to analyze which electronic devices they are using in their environment. By sensing the electronics being used, a system could infer their current activity (e.g., stove implies cooking, car implies commuting).
MagnifiSense is a user-worn magnetic sensing system that captures near-field electro-magnetic interference (EMI) produced by electronic devices to track user-specific device interaction and usage. An important advantage of this wearable solution is that it is both user-specific and does not require specially instrumented environments. By using the radiated EM signal inherent to many appliances and devices, no modifications are needed to the devices either. This is achieved by using a set of 3 orthogonal magneto-inductive coil to detect the radiated EMI signature from nearby appliances. A machine learning, random forest classifier is then applied to detect the type of appliance that is being used.