A Sub-Milliwatt and Sub-Millisecond 3-D Gaze Estimator for Ultra Low-Power AR Applications

Sungmin Moon, Chao Zhang, Sooill Park, Hui Zhang, Woo-Shik Kim, Jong Hwan Ko

Abstract: The critical factors of real-time gaze tracking are high accuracy and user-friendliness such as low latency and no run-time calibration. Existing gaze estimator hardware designs are based on 2D regression algorithms as 2D methods have a simple computation process. However, they require multiple run-time calibration steps, and are vulnerable to head motions. On the other hand, the 3D model-based method can maintain better accuracy than the 2D method without run-time calibration steps, and is robust to head motions. In this paper, we aim to design the first 3D model-based gaze estimator hardware that consumes less than 1mW power and 1ms latency per frame. The simulation results based on the hardware synthesis show that the proposed design requires 172μW and 0.5ms per frame, while maintaining less than 0.9° error.