Firefighter' chief reason for entering a burning structure is to search for and rescue potential victims. Currently, their primary method for communication is an often-congested two-way radio -which the firefighters use from within a burning building to relay their activities (and other information) to an external battalion chief. In response to discussions with firefighters in the field, we introduce an approach for automatically segmenting and classifying a select set of activities using wireless accelerom-eters attached to the human body. The activities we focus on are the ones that are most commonly conducted by firefighters and that are important to the battalion chief for understanding the ongoing search and rescue. In our implementation, sensors continuously measure the acceleration of a small number of body segments and transmit data back to a central base station. At runtime, our system classifies data for short intervals, relying on training examples of the activities of interest. We show that our approach can appropriately detect motions in real-time without significant latency using as few as two accelerometers.
展开▼
