Buffering approach for energy saving in video sensors

Networked video sensors need to execute two dependent periodic tasks: video encoding and transmission. The dependency and periodicity often result in small idle intervals of CPU and wireless network interface card (WNIC). In this paper, we present a sender-buffering approach to exploit such idle intervals for energy saving. Specifically, a video sensor encodes frames in a timely fashion, but buffers encoded frames and transmits them in bursts at longer intervals. In doing so, it (1) accumulates short WNIC idle intervals into longer ones, during which the WNIC can enter the lower-power sleep mode, and (2) slows down the CPU by avoiding CPU idle intervals, which are resulted from both early completion of frame encoding and waiting for frame transmission. Our experimental results show that the buffering approach can save 32-80% CPU energy and 35- 54% WNIC energy, while increasing the overall end-to-end transmission delay by at most 2 frames.