Title :
A 27mW reconfigurable marker-less logarithmic camera pose estimation engine for mobile augmented reality processor
Author :
Injoon Hong ; Gyeonghoon Kim ; Youchang Kim ; Donghyun Kim ; Byeong-Gyu Nam ; Hoi-Jun Yoo
Author_Institution :
Dept. of Electr. Eng., KAIST, Daejeon, South Korea
Abstract :
A marker-less Camera Pose Estimation Engine (CPEE) with reconfigurable logarithmic processor is proposed for the view angle estimation in the low-power mobile Augmented Reality (AR) applications. The proposed CPEE is required to overcome 150x huge gap in computational cost for marker-less pose estimation including floating-point operations resulting in the bottlenecks in mobile platforms. Speculative Execution (SE) and Reconfigurable Data-arrangement Layer (RDL) are proposed to reduce the computing time of CPEE by 17% and 27%, respectively. For low-power implementation of floating-point units, Logarithmic Processing Element (LPE) is used to reduce overall power consumption by 18% with maximum conversion error under 0.49%. The proposed marker-less CPEE is fabricated in 65nm Logic CMOS technology, and successfully realizes real-time marker-less camera pose estimation with only 27mW power consumption.
Keywords :
augmented reality; mobile computing; parallel programming; pose estimation; AR; CPEE; LPE; RDL; SE; logarithmic processing element; marker-less camera pose estimation engine; mobile augmented reality; reconfigurable data-arrangement layer; reconfigurable logarithmic processor; speculative execution; Cameras; Estimation; Instruction sets; Jacobian matrices; Mobile communication; Power demand; Three-dimensional displays;
Conference_Titel :
Solid-State Circuits Conference (A-SSCC), 2014 IEEE Asian
Conference_Location :
KaoHsiung
Print_ISBN :
978-1-4799-4090-5
DOI :
10.1109/ASSCC.2014.7008897