Title :
A 0.5-degree error 10mW CMOS image sensor-based gaze estimation processor with logarithmic processing
Author :
Kyeongryeol Bong ; Injoon Hong ; Gyeonghoon Kim ; Hoi-Jun Yoo
Author_Institution :
Dept. of EE, KAIST, Daejeon, South Korea
Abstract :
A CMOS image sensor (CIS)-based gaze estimation processor (GEP) is proposed for the next generation user interface (UI) of the smart glasses. It integrates a functional CMOS image sensor (CIS), a RISC, and an ellipse fitting engine (EFE) on a single chip to reduce the power consumption. The functional CIS includes mixed-signal pre-processing circuits, pupil edge detection circuit (PEDC) and glint corner detection circuit (GCDC). Logarithmic processing elements (LPE) is used for EFE to reduce the power consumption further. As a result, PEDC/GCDC and LPE can reduce overall power by 55.2% and 14.1%, respectively. Implemented in 65nm CMOS technology, the 11.29mm2 chip consumes 10mW power at 30fps real-time operation, and 0.5° error is achieved in a glass-type wearable system.
Keywords :
CMOS image sensors; edge detection; glass; intelligent materials; intelligent sensors; reduced instruction set computing; user interfaces; CIS; CMOS image sensor; EFE; GCDC; GEP; LPE; PEDC; RISC; UI; ellipse fitting engine; gaze estimation processor; glint corner detection circuit; logarithmic processing element; mixed-signal pre-processing circuit; power 10 mW; power consumption; pupil edge detection circuit; size 65 nm; smart glass-type wearable system; user interface; CMOS image sensors; Estimation; Fitting; Glass; Image edge detection; Power demand; Reduced instruction set computing;
Conference_Titel :
VLSI Circuits (VLSI Circuits), 2015 Symposium on
Conference_Location :
Kyoto
Print_ISBN :
978-4-86348-502-0
DOI :
10.1109/VLSIC.2015.7231321
