Resource Adaptive Periodicity Estimation of Streaming Data

Streaming environments typically dictate incomplete or approximate algorithm execution, in order to cope with sudden surges in the data rate. Such limitations are even more accentuated in mobile environments (such as sensor networks) where computational and memory resources are typically limited. This paper introduces the first “resource adaptive” algorithm for periodicity estimation on a continuous stream of data. Our formulation is based on the derivation of a closed-form incremental computation of the spectrum, augmented by an intelligent load-shedding scheme that can adapt to available CPU resources. Our experiments indicate that the proposed technique can be a viable and resource efficient solution for real-time spectrum estimation.