Architectural support for detection of sleepy eye behavior

Detecting human sleep is an area of active research. At Siemens Corporate Research Inc., a method to track human eyes in a sequence of images was developed. The method is used to detect when the subject is on the verge of sleep. A key component of the methodology is a filter used to isolate the eyes. In this paper, the algorithm, architecture and implementation of this filter is presented. The implementation of the filter is based on field programmable gate arrays from Xilinx.