Power analysis of a mobile EEG system with compressed sensing

Abstract

We analyze the power tradeoffs for computation and transmission in a mobile electroencephalography (EEG) system. The EEG system comprises an analog front end, microcontroller, and wireless transceiver. We measured the power consumption of the system under a variety of conditions in order to estimate the power attributable to each component separately. We developed simple models for power consumption that incorporate transient power behavior of the devices and estimated parameters by fitting experimental data to the models. We found that the costs of transmission and computation were similar, with transmission power decreasing and computation power increasing with the compression ratio and all costs increasing with the number of channels. For the system configuration reported, the transmission costs dominated, leading to the conclusion that the system should be operated with: a) the lowest clock rate for the microcontroller; and b) the highest data compression consistent with system fidelity requirements. A discussion of tradeoffs in alternate system configurations is provided.