Cheap noisy sensors can improve activity monitoring under stringent energy constraints

This paper proposes a low-power acoustic sensor built using off-the-shelf components. To reduce energy consumption, the ADC and the microphone's signal-conditioning circuit are replaced by a low-power analog comparator with adjustable thresholds. Although the SNR of the proposed sensor is reduced, we demonstrate how recent advancements in adaptive sensor scheduling can utilize this sensing modality at the right times to deliver high estimation performance in the presence of extremely stringent device-energy constraints. In a long-term acoustic wildlife monitoring application, energy consumption is reduced by a factor of up to 15× over the scheduling policy that does not utilize the proposed sensor. Furthermore, when device energy is extremely scarce, optimal sensor management can reduce error rate from nearly 40% to 8%.