On-Device Eye Tracking System with Dual Lightweight AI Processor

With the growth of personalized devices, in various field, eye tracking research is conducted for enhancing user experience. Also, the recent research focuses on the method based on artificial intelligence (AI) algorithm for improving the performance. However, the eye tracking through AI algorithm needs significant resources. For this reason, the AI-based methods have limitations applying to embedded system. For solving the problems, in this paper, we propose an eye tracking system with two AI processors based on light weight algorithm, k-nearest neighbor (k-NN). The processor was implemented on a field programmable gate array (FPGA), and the eye tracking system was verified through visible light eye images. The results demonstrate the feasibility of the proposed eye tracking system.