Abstract:
The multi-sensor tracker will be able to track several people’s poses in an office environment. The tracker’s output will be accurate enough for an office application to operate properly. For example, the tracker’s output is will be used by an auto stereoscopic screen, which displays 3D images according to the viewer’s pose. 

The tracker will accomplish its accurate tracking of an object’s pose with a network of sensors spread throughout the tracking environment. Each sensor will be connected to a processor. The sensor data is used to compute the tracker’s output with a tracking algorithm that executes, distributed across the processors connected to the sensors. The tracking algorithm will apply ideas from the SCAAT method to distribute the tracking computation. A network connecting the sensors will enable the tracking algorithm to exchange partial data to complete the tracker’s output. The network will be designed with ideas from directed diffusion. The design of the multi-sensor tracker will enable robust tracking in a large volume.