Software-controlled processor speed setting for low-power streaming multimedia

In this paper, we describe a software-controlled approach for adaptively mi nimizing energy in embedded systems for real-time multimedia processing. En ergy is optimized by clock speed setting: the software controller dynamical ly adjusts processor clock speed to the frame rate (FR) requirements of the incoming multimedia stream. The speed-setting policy is based on a system model that correlates clock speed with best case, average case, and worst c ase sustainable FRs, accounting for data dependency in multimedia streams. The technique has been implemented in a energy-efficient MPEG3 real-time de coder algorithm designed for wearable devices as a case study. The target s ystem is the Hewlett-Packard SmartBadgeIII prototype system based on the St rongARM1100 processor. Hardware measurements show that computational energy can be drastically reduced (up to 40%) with respect to fixed-frequency ope ration.