Time-series estimation from randomly time-warped observations

Abstract

• We consider the problem of estimating a signal given randomly time warped observations of the signal. • Many existing methods rely on estimating the unknown time warps, a computationally challenging problem. • We propose a scalable method based on interpreting the time warped observations as samples of a manifold. • We search for the ’center’ of the manifold, by resorting to techniques from graph/network processing. • We demonstrate that the number of time warped observations required is not exceedingly high for practical scenarios. We consider the problem of estimating a signal from its warped observations. Such estimation is commonly performed by altering the observations through some inverse-warping, or solving a computationally demanding optimization formulation . While these may be unavoidable if observations are few, when large amounts of warped observations are available, the cost of running such algorithms can be prohibitive. We consider the scenario where we have many observations, and propose a computationally simple algorithm for estimating the function of interest. We demonstrate the utility of the algorithm on streaming biomedical signals.