A low-power FPGA implementation of eye tracking

The use of eye-tracking in energy-constrained applications requires systems capable of tracking human eyes without consuming large amount of power. In this paper we present a low-power implementation of real-time non-intrusive eye-tracking by single camera. Unlike existing hardware designs, our eye-tracker does not restrict the user nor requires computationally expensive tools for maintaining high quality tracking. The system provides over 88% eye detection accuracy at 8f/s image processing rate. We describe the hardware and present the results of its experimental evaluation.