Gravity. Most of us can't escape it. It ties us to Earth. It makes our world somewhat predictable. It keeps our feet on the ground. And it is this fact that we've tried to exploit in our quest to create a system to support two of the goals of ubiquitous computing: identifying and locating a person. A person is in contact with the floor most of the time. Why not make the floor "smart" and use it to identify and track people?
In ubiquitous computing, designers have tried to support mobile and collaborative activities by creating systems that use innovative hardware devices and that use a person's context. A person's context may include location, identity, current activity, emotional state, the context of those around him, and the history of these pieces of information. In the Smart Floor project, we are attempting to provide a reliable mechanism to identify a person and track her location. In addition, we are exploring innovative applications of this technology, including uses in the home, art and performance applications, and entertainment.
The Smart Floor aims to identify and track a user around an instrumented space. To do this, we have instrumented a floor with force measuring load cells. Each floor tile lays on four load cells (one at each corner), and each load cell has four tile corners resting on it. The load cells measure the force of the user's foot (ground reaction force, GRF) as the user walks over the floor tiles. We then train a Hidden Markov Model (HMM) system using these footfall force signatures. The HMM system can be trained with a number of users' footfall signatures. When a user then walks over the floor, the system attempts to match the user's footfall signature with the trained library of signatures. For a small group of users (on the order of twenty), the correct identity can be established with better than 90% accuracy.
Once we know who a user is, tracking her is the next step. We are exploring various signal processing techniques to track single users around a space, and to extend this to track multiple people in the instrumented space.
We are currently developing a single-user, small-scale version of the Smart Floor. Our continuing efforts to extend the system have two major components:
There are also a number of scientific questions that we are attempting to answer:
